{ "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": [ { "KL/chosen_KL_mean": 0.02867889404296875, "KL/mean": 0.029354453086853027, "KL/rejected_KL_mean": 0.030029296875, "KL/std": 0.2071000635623932, "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": 28.21779441833496, "learning_rate": 0.0, "logits/chosen": 0.13337239623069763, "logits/rejected": 0.12492949515581131, "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 }, { "KL/chosen_KL_mean": -0.00289154052734375, "KL/mean": -0.021616414189338684, "KL/rejected_KL_mean": -0.04033660888671875, "KL/std": 0.19624735414981842, "epoch": 0.0030234315948601664, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.037450045347213745, "fcm_dpo/q_t": 0.4990639090538025, "grad_norm": 27.82764434814453, "learning_rate": 7.462686567164179e-09, "logits/chosen": 0.09414851665496826, "logits/rejected": 0.07363267242908478, "logps/chosen": -56.101890563964844, "logps/ref_chosen": -56.0989990234375, "logps/ref_rejected": -66.59971618652344, "logps/rejected": -66.64006042480469, "loss": 1.3828, "margin_dpo/margin_mean": 0.03744968771934509, "margin_dpo/margin_std": 0.27811938524246216, "step": 2 }, { "KL/chosen_KL_mean": 0.017974853515625, "KL/mean": 0.01571372151374817, "KL/rejected_KL_mean": 0.013454437255859375, "KL/std": 0.2552720606327057, "epoch": 0.0045351473922902496, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.004527091979980469, "fcm_dpo/q_t": 0.4998869001865387, "grad_norm": 31.184932708740234, "learning_rate": 1.4925373134328357e-08, "logits/chosen": 0.09944942593574524, "logits/rejected": 0.06144290789961815, "logps/chosen": -65.43928527832031, "logps/ref_chosen": -65.45726013183594, "logps/ref_rejected": -90.82853698730469, "logps/rejected": -90.81507873535156, "loss": 1.3861, "margin_dpo/margin_mean": 0.004526615142822266, "margin_dpo/margin_std": 0.3228065073490143, "step": 3 }, { "KL/chosen_KL_mean": 0.021240234375, "KL/mean": 0.024831712245941162, "KL/rejected_KL_mean": 0.028415679931640625, "KL/std": 0.23518507182598114, "epoch": 0.006046863189720333, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.007175266742706299, "fcm_dpo/q_t": 0.5001801252365112, "grad_norm": 34.333335876464844, "learning_rate": 2.2388059701492534e-08, "logits/chosen": 0.10089954733848572, "logits/rejected": 0.08478045463562012, "logps/chosen": -76.83894348144531, "logps/ref_chosen": -76.86018371582031, "logps/ref_rejected": -79.91523742675781, "logps/rejected": -79.88682556152344, "loss": 1.3873, "margin_dpo/margin_mean": -0.007175803184509277, "margin_dpo/margin_std": 0.3401098847389221, "step": 4 }, { "KL/chosen_KL_mean": 0.015611648559570312, "KL/mean": 0.014620736241340637, "KL/rejected_KL_mean": 0.01363372802734375, "KL/std": 0.22797781229019165, "epoch": 0.007558578987150416, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0019778311252593994, "fcm_dpo/q_t": 0.499950647354126, "grad_norm": 29.629215240478516, "learning_rate": 2.9850746268656714e-08, "logits/chosen": 0.08228473365306854, "logits/rejected": 0.04336842894554138, "logps/chosen": -62.955726623535156, "logps/ref_chosen": -62.97134017944336, "logps/ref_rejected": -79.9192123413086, "logps/rejected": -79.90557861328125, "loss": 1.3863, "margin_dpo/margin_mean": 0.001978069543838501, "margin_dpo/margin_std": 0.30833548307418823, "step": 5 }, { "KL/chosen_KL_mean": -0.02263641357421875, "KL/mean": 0.01097261905670166, "KL/rejected_KL_mean": 0.04458427429199219, "KL/std": 0.2150484025478363, "epoch": 0.009070294784580499, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06721735000610352, "fcm_dpo/q_t": 0.5016798377037048, "grad_norm": 29.89048194885254, "learning_rate": 3.731343283582089e-08, "logits/chosen": 0.1416800618171692, "logits/rejected": 0.10234207659959793, "logps/chosen": -51.33000183105469, "logps/ref_chosen": -51.30736541748047, "logps/ref_rejected": -82.77239227294922, "logps/rejected": -82.72781372070312, "loss": 1.3932, "margin_dpo/margin_mean": -0.06721735000610352, "margin_dpo/margin_std": 0.28968173265457153, "step": 6 }, { "KL/chosen_KL_mean": 0.03011322021484375, "KL/mean": 0.0288875550031662, "KL/rejected_KL_mean": 0.027660369873046875, "KL/std": 0.21546083688735962, "epoch": 0.010582010582010581, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0024529993534088135, "fcm_dpo/q_t": 0.4999385476112366, "grad_norm": 27.15511131286621, "learning_rate": 4.477611940298507e-08, "logits/chosen": 0.01981586217880249, "logits/rejected": -0.024220559746026993, "logps/chosen": -51.429298400878906, "logps/ref_chosen": -51.45941162109375, "logps/ref_rejected": -66.3828125, "logps/rejected": -66.35514831542969, "loss": 1.3862, "margin_dpo/margin_mean": 0.0024527907371520996, "margin_dpo/margin_std": 0.26524245738983154, "step": 7 }, { "KL/chosen_KL_mean": 0.02264404296875, "KL/mean": -0.0015153884887695312, "KL/rejected_KL_mean": -0.025671005249023438, "KL/std": 0.23731042444705963, "epoch": 0.012093726379440665, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04831033945083618, "fcm_dpo/q_t": 0.49879276752471924, "grad_norm": 28.646804809570312, "learning_rate": 5.223880597014925e-08, "logits/chosen": 0.07039310038089752, "logits/rejected": 0.04824473336338997, "logps/chosen": -62.174903869628906, "logps/ref_chosen": -62.197547912597656, "logps/ref_rejected": -74.66180419921875, "logps/rejected": -74.6874771118164, "loss": 1.3818, "margin_dpo/margin_mean": 0.04831072688102722, "margin_dpo/margin_std": 0.3311997950077057, "step": 8 }, { "KL/chosen_KL_mean": -0.04302787780761719, "KL/mean": -0.051038309931755066, "KL/rejected_KL_mean": -0.059051513671875, "KL/std": 0.2540060877799988, "epoch": 0.013605442176870748, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.016026049852371216, "fcm_dpo/q_t": 0.4995996654033661, "grad_norm": 31.397911071777344, "learning_rate": 5.970149253731343e-08, "logits/chosen": 0.15405794978141785, "logits/rejected": 0.09594576060771942, "logps/chosen": -55.67274856567383, "logps/ref_chosen": -55.629722595214844, "logps/ref_rejected": -86.21221923828125, "logps/rejected": -86.27127075195312, "loss": 1.385, "margin_dpo/margin_mean": 0.01602587103843689, "margin_dpo/margin_std": 0.33587515354156494, "step": 9 }, { "KL/chosen_KL_mean": -0.009862899780273438, "KL/mean": 0.005285277962684631, "KL/rejected_KL_mean": 0.020427703857421875, "KL/std": 0.22252368927001953, "epoch": 0.015117157974300832, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.030300527811050415, "fcm_dpo/q_t": 0.5007571578025818, "grad_norm": 29.809185028076172, "learning_rate": 6.71641791044776e-08, "logits/chosen": 0.13334016501903534, "logits/rejected": 0.10230215638875961, "logps/chosen": -62.700462341308594, "logps/ref_chosen": -62.69060134887695, "logps/ref_rejected": -90.610107421875, "logps/rejected": -90.58968353271484, "loss": 1.3896, "margin_dpo/margin_mean": -0.03030034899711609, "margin_dpo/margin_std": 0.34233033657073975, "step": 10 }, { "KL/chosen_KL_mean": 0.016397476196289062, "KL/mean": 0.004511818289756775, "KL/rejected_KL_mean": -0.00738525390625, "KL/std": 0.20762629806995392, "epoch": 0.016628873771730914, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.023783236742019653, "fcm_dpo/q_t": 0.4994049668312073, "grad_norm": 29.40945816040039, "learning_rate": 7.462686567164178e-08, "logits/chosen": 0.10654198378324509, "logits/rejected": 0.09964478015899658, "logps/chosen": -65.75071716308594, "logps/ref_chosen": -65.76712036132812, "logps/ref_rejected": -72.4764633178711, "logps/rejected": -72.48384857177734, "loss": 1.3842, "margin_dpo/margin_mean": 0.023783445358276367, "margin_dpo/margin_std": 0.3237120509147644, "step": 11 }, { "KL/chosen_KL_mean": -0.018245697021484375, "KL/mean": -0.011372998356819153, "KL/rejected_KL_mean": -0.0045013427734375, "KL/std": 0.23007062077522278, "epoch": 0.018140589569160998, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.01374778151512146, "fcm_dpo/q_t": 0.5003437995910645, "grad_norm": 28.194849014282227, "learning_rate": 8.208955223880596e-08, "logits/chosen": 0.029350608587265015, "logits/rejected": 0.013005432672798634, "logps/chosen": -60.72313690185547, "logps/ref_chosen": -60.704891204833984, "logps/ref_rejected": -69.41564178466797, "logps/rejected": -69.4201431274414, "loss": 1.388, "margin_dpo/margin_mean": -0.013747721910476685, "margin_dpo/margin_std": 0.33297550678253174, "step": 12 }, { "KL/chosen_KL_mean": -0.0060176849365234375, "KL/mean": 0.006569623947143555, "KL/rejected_KL_mean": 0.019161224365234375, "KL/std": 0.20004111528396606, "epoch": 0.019652305366591082, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02517753839492798, "fcm_dpo/q_t": 0.5006293654441833, "grad_norm": 29.047027587890625, "learning_rate": 8.955223880597014e-08, "logits/chosen": 0.10595913231372833, "logits/rejected": 0.04388178139925003, "logps/chosen": -49.91527557373047, "logps/ref_chosen": -49.90925598144531, "logps/ref_rejected": -92.37818145751953, "logps/rejected": -92.35902404785156, "loss": 1.389, "margin_dpo/margin_mean": -0.025177836418151855, "margin_dpo/margin_std": 0.27132922410964966, "step": 13 }, { "KL/chosen_KL_mean": 0.0030002593994140625, "KL/mean": -0.004290089011192322, "KL/rejected_KL_mean": -0.0115814208984375, "KL/std": 0.1868174970149994, "epoch": 0.021164021164021163, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.014579087495803833, "fcm_dpo/q_t": 0.4996357560157776, "grad_norm": 29.573348999023438, "learning_rate": 9.701492537313432e-08, "logits/chosen": 0.09998003393411636, "logits/rejected": 0.08181308954954147, "logps/chosen": -60.61579132080078, "logps/ref_chosen": -60.61879348754883, "logps/ref_rejected": -71.79306030273438, "logps/rejected": -71.80464172363281, "loss": 1.385, "margin_dpo/margin_mean": 0.014579474925994873, "margin_dpo/margin_std": 0.2635040581226349, "step": 14 }, { "KL/chosen_KL_mean": -0.02704620361328125, "KL/mean": -0.035765036940574646, "KL/rejected_KL_mean": -0.0444793701171875, "KL/std": 0.24218779802322388, "epoch": 0.022675736961451247, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.017432361841201782, "fcm_dpo/q_t": 0.4995632767677307, "grad_norm": 33.17366027832031, "learning_rate": 1.044776119402985e-07, "logits/chosen": 0.05620576813817024, "logits/rejected": 0.013878353871405125, "logps/chosen": -63.49658203125, "logps/ref_chosen": -63.46953582763672, "logps/ref_rejected": -88.88951110839844, "logps/rejected": -88.93399047851562, "loss": 1.3849, "margin_dpo/margin_mean": 0.017432481050491333, "margin_dpo/margin_std": 0.349947452545166, "step": 15 }, { "KL/chosen_KL_mean": -0.0045413970947265625, "KL/mean": -0.01623951643705368, "KL/rejected_KL_mean": -0.027942657470703125, "KL/std": 0.2111794650554657, "epoch": 0.02418745275888133, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.023399069905281067, "fcm_dpo/q_t": 0.49941548705101013, "grad_norm": 26.882421493530273, "learning_rate": 1.1194029850746268e-07, "logits/chosen": 0.12233875691890717, "logits/rejected": 0.08475068211555481, "logps/chosen": -46.536842346191406, "logps/ref_chosen": -46.53229904174805, "logps/ref_rejected": -74.27533721923828, "logps/rejected": -74.30328369140625, "loss": 1.3842, "margin_dpo/margin_mean": 0.02339915931224823, "margin_dpo/margin_std": 0.29788079857826233, "step": 16 }, { "KL/chosen_KL_mean": -0.008623123168945312, "KL/mean": -0.014116168022155762, "KL/rejected_KL_mean": -0.019603729248046875, "KL/std": 0.23058179020881653, "epoch": 0.025699168556311415, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.010984718799591064, "fcm_dpo/q_t": 0.49972665309906006, "grad_norm": 32.859825134277344, "learning_rate": 1.1940298507462686e-07, "logits/chosen": 0.059348031878471375, "logits/rejected": 0.04041973873972893, "logps/chosen": -64.08645629882812, "logps/ref_chosen": -64.07783508300781, "logps/ref_rejected": -86.40876770019531, "logps/rejected": -86.42837524414062, "loss": 1.3855, "margin_dpo/margin_mean": 0.010984480381011963, "margin_dpo/margin_std": 0.3624400794506073, "step": 17 }, { "KL/chosen_KL_mean": 0.042514801025390625, "KL/mean": 0.020855039358139038, "KL/rejected_KL_mean": -0.0008106231689453125, "KL/std": 0.22396549582481384, "epoch": 0.027210884353741496, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0433235764503479, "fcm_dpo/q_t": 0.4989171028137207, "grad_norm": 27.754545211791992, "learning_rate": 1.2686567164179106e-07, "logits/chosen": 0.08723866194486618, "logits/rejected": 0.042168259620666504, "logps/chosen": -44.83182144165039, "logps/ref_chosen": -44.87433624267578, "logps/ref_rejected": -70.97604370117188, "logps/rejected": -70.97686004638672, "loss": 1.3822, "margin_dpo/margin_mean": 0.04332330822944641, "margin_dpo/margin_std": 0.31900152564048767, "step": 18 }, { "KL/chosen_KL_mean": 0.01389312744140625, "KL/mean": -0.016233228147029877, "KL/rejected_KL_mean": -0.04636383056640625, "KL/std": 0.2271110713481903, "epoch": 0.02872260015117158, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06026040017604828, "fcm_dpo/q_t": 0.4984937906265259, "grad_norm": 30.69112777709961, "learning_rate": 1.343283582089552e-07, "logits/chosen": 0.058250539004802704, "logits/rejected": 0.044945910573005676, "logps/chosen": -68.14591979980469, "logps/ref_chosen": -68.1598129272461, "logps/ref_rejected": -81.17138671875, "logps/rejected": -81.21775817871094, "loss": 1.3805, "margin_dpo/margin_mean": 0.06026081740856171, "margin_dpo/margin_std": 0.2684270441532135, "step": 19 }, { "KL/chosen_KL_mean": 0.018886566162109375, "KL/mean": 0.013330504298210144, "KL/rejected_KL_mean": 0.007778167724609375, "KL/std": 0.21853777766227722, "epoch": 0.030234315948601664, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.011106997728347778, "fcm_dpo/q_t": 0.49972179532051086, "grad_norm": 29.30516815185547, "learning_rate": 1.4179104477611938e-07, "logits/chosen": 0.1435462087392807, "logits/rejected": 0.12017589807510376, "logps/chosen": -53.65967559814453, "logps/ref_chosen": -53.67856216430664, "logps/ref_rejected": -74.16911315917969, "logps/rejected": -74.16133880615234, "loss": 1.3854, "margin_dpo/margin_mean": 0.011106759309768677, "margin_dpo/margin_std": 0.2792432904243469, "step": 20 }, { "KL/chosen_KL_mean": 0.0058422088623046875, "KL/mean": 0.0030414238572120667, "KL/rejected_KL_mean": 0.000247955322265625, "KL/std": 0.22705534100532532, "epoch": 0.031746031746031744, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.005594924092292786, "fcm_dpo/q_t": 0.4998602867126465, "grad_norm": 29.399574279785156, "learning_rate": 1.4925373134328355e-07, "logits/chosen": 0.11058309674263, "logits/rejected": 0.08474206924438477, "logps/chosen": -64.69570922851562, "logps/ref_chosen": -64.70155334472656, "logps/ref_rejected": -81.02095031738281, "logps/rejected": -81.02069854736328, "loss": 1.386, "margin_dpo/margin_mean": 0.005594193935394287, "margin_dpo/margin_std": 0.3037152886390686, "step": 21 }, { "KL/chosen_KL_mean": -0.030012130737304688, "KL/mean": -0.02553868293762207, "KL/rejected_KL_mean": -0.02106475830078125, "KL/std": 0.23182955384254456, "epoch": 0.03325774754346183, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.008944094181060791, "fcm_dpo/q_t": 0.5002233386039734, "grad_norm": 29.012664794921875, "learning_rate": 1.5671641791044775e-07, "logits/chosen": -0.010908234864473343, "logits/rejected": -0.03109363466501236, "logps/chosen": -58.066001892089844, "logps/ref_chosen": -58.03599166870117, "logps/ref_rejected": -80.72721862792969, "logps/rejected": -80.74828338623047, "loss": 1.3875, "margin_dpo/margin_mean": -0.008943915367126465, "margin_dpo/margin_std": 0.3591557741165161, "step": 22 }, { "KL/chosen_KL_mean": -0.003086090087890625, "KL/mean": -0.006322100758552551, "KL/rejected_KL_mean": -0.009563446044921875, "KL/std": 0.22721782326698303, "epoch": 0.03476946334089191, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.006478816270828247, "fcm_dpo/q_t": 0.4998379647731781, "grad_norm": 33.13698959350586, "learning_rate": 1.6417910447761193e-07, "logits/chosen": 0.12168412655591965, "logits/rejected": 0.09707251936197281, "logps/chosen": -66.35917663574219, "logps/ref_chosen": -66.35608673095703, "logps/ref_rejected": -93.02769470214844, "logps/rejected": -93.0372543334961, "loss": 1.3859, "margin_dpo/margin_mean": 0.0064778923988342285, "margin_dpo/margin_std": 0.31190019845962524, "step": 23 }, { "KL/chosen_KL_mean": -0.03143882751464844, "KL/mean": 0.0080418661236763, "KL/rejected_KL_mean": 0.047527313232421875, "KL/std": 0.231191486120224, "epoch": 0.036281179138321996, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.07896275818347931, "fcm_dpo/q_t": 0.5019730925559998, "grad_norm": 26.388242721557617, "learning_rate": 1.716417910447761e-07, "logits/chosen": 0.15333206951618195, "logits/rejected": 0.11985106766223907, "logps/chosen": -54.49267578125, "logps/ref_chosen": -54.461238861083984, "logps/ref_rejected": -68.33817291259766, "logps/rejected": -68.2906494140625, "loss": 1.3945, "margin_dpo/margin_mean": -0.0789632499217987, "margin_dpo/margin_std": 0.3113078773021698, "step": 24 }, { "KL/chosen_KL_mean": -0.02866363525390625, "KL/mean": -0.041737914085388184, "KL/rejected_KL_mean": -0.0548095703125, "KL/std": 0.24598127603530884, "epoch": 0.03779289493575208, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.026146292686462402, "fcm_dpo/q_t": 0.4993473291397095, "grad_norm": 29.624902725219727, "learning_rate": 1.7910447761194027e-07, "logits/chosen": 0.1234542727470398, "logits/rejected": 0.07062897086143494, "logps/chosen": -60.032867431640625, "logps/ref_chosen": -60.00420379638672, "logps/ref_rejected": -90.47376251220703, "logps/rejected": -90.52857208251953, "loss": 1.384, "margin_dpo/margin_mean": 0.0261460542678833, "margin_dpo/margin_std": 0.33622923493385315, "step": 25 }, { "KL/chosen_KL_mean": -0.029012680053710938, "KL/mean": -0.014886699616909027, "KL/rejected_KL_mean": -0.000759124755859375, "KL/std": 0.2382136434316635, "epoch": 0.039304610733182165, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02825431525707245, "fcm_dpo/q_t": 0.5007061958312988, "grad_norm": 29.823959350585938, "learning_rate": 1.8656716417910447e-07, "logits/chosen": 0.12276704609394073, "logits/rejected": 0.10391613841056824, "logps/chosen": -56.848167419433594, "logps/ref_chosen": -56.81915283203125, "logps/ref_rejected": -77.84333038330078, "logps/rejected": -77.84408569335938, "loss": 1.3894, "margin_dpo/margin_mean": -0.028254270553588867, "margin_dpo/margin_std": 0.33982354402542114, "step": 26 }, { "KL/chosen_KL_mean": 0.036243438720703125, "KL/mean": 0.017910704016685486, "KL/rejected_KL_mean": -0.0004253387451171875, "KL/std": 0.24093255400657654, "epoch": 0.04081632653061224, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03667256236076355, "fcm_dpo/q_t": 0.49908387660980225, "grad_norm": 28.84993553161621, "learning_rate": 1.9402985074626865e-07, "logits/chosen": 0.09245912730693817, "logits/rejected": 0.06795473396778107, "logps/chosen": -62.840782165527344, "logps/ref_chosen": -62.87702560424805, "logps/ref_rejected": -71.34437561035156, "logps/rejected": -71.34480285644531, "loss": 1.3829, "margin_dpo/margin_mean": 0.0366722047328949, "margin_dpo/margin_std": 0.32028186321258545, "step": 27 }, { "KL/chosen_KL_mean": -0.02478790283203125, "KL/mean": -0.04612994194030762, "KL/rejected_KL_mean": -0.06747055053710938, "KL/std": 0.22664371132850647, "epoch": 0.042328042328042326, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04268306493759155, "fcm_dpo/q_t": 0.4989331364631653, "grad_norm": 27.578332901000977, "learning_rate": 2.0149253731343282e-07, "logits/chosen": 0.06610069423913956, "logits/rejected": 0.05711779743432999, "logps/chosen": -59.85816192626953, "logps/ref_chosen": -59.8333740234375, "logps/ref_rejected": -70.39804077148438, "logps/rejected": -70.46550750732422, "loss": 1.3822, "margin_dpo/margin_mean": 0.04268324375152588, "margin_dpo/margin_std": 0.27869123220443726, "step": 28 }, { "KL/chosen_KL_mean": -0.0189361572265625, "KL/mean": -0.04419657588005066, "KL/rejected_KL_mean": -0.0694580078125, "KL/std": 0.19601967930793762, "epoch": 0.04383975812547241, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05052053928375244, "fcm_dpo/q_t": 0.4987371861934662, "grad_norm": 32.570743560791016, "learning_rate": 2.08955223880597e-07, "logits/chosen": 0.1362171769142151, "logits/rejected": 0.11860013008117676, "logps/chosen": -74.1391372680664, "logps/ref_chosen": -74.12020111083984, "logps/ref_rejected": -83.33099365234375, "logps/rejected": -83.40045166015625, "loss": 1.3814, "margin_dpo/margin_mean": 0.05052059888839722, "margin_dpo/margin_std": 0.26012739539146423, "step": 29 }, { "KL/chosen_KL_mean": -0.040557861328125, "KL/mean": -0.06584726274013519, "KL/rejected_KL_mean": -0.09113693237304688, "KL/std": 0.2571483552455902, "epoch": 0.045351473922902494, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05058065056800842, "fcm_dpo/q_t": 0.49873781204223633, "grad_norm": 30.18358612060547, "learning_rate": 2.1641791044776117e-07, "logits/chosen": 0.09700868278741837, "logits/rejected": 0.04471509903669357, "logps/chosen": -50.791847229003906, "logps/ref_chosen": -50.75128936767578, "logps/ref_rejected": -89.29063415527344, "logps/rejected": -89.38177490234375, "loss": 1.3816, "margin_dpo/margin_mean": 0.05057990550994873, "margin_dpo/margin_std": 0.3860338032245636, "step": 30 }, { "KL/chosen_KL_mean": -0.060115814208984375, "KL/mean": -0.05566447973251343, "KL/rejected_KL_mean": -0.05120849609375, "KL/std": 0.2631988823413849, "epoch": 0.04686318972033258, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.008907794952392578, "fcm_dpo/q_t": 0.5002219676971436, "grad_norm": 34.45195388793945, "learning_rate": 2.2388059701492537e-07, "logits/chosen": 0.12011389434337616, "logits/rejected": 0.07308140397071838, "logps/chosen": -65.39686584472656, "logps/ref_chosen": -65.33675384521484, "logps/ref_rejected": -100.76666259765625, "logps/rejected": -100.81787109375, "loss": 1.3875, "margin_dpo/margin_mean": -0.008907884359359741, "margin_dpo/margin_std": 0.37602365016937256, "step": 31 }, { "KL/chosen_KL_mean": -0.0176239013671875, "KL/mean": -0.02053038775920868, "KL/rejected_KL_mean": -0.023433685302734375, "KL/std": 0.2514227628707886, "epoch": 0.04837490551776266, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0058145225048065186, "fcm_dpo/q_t": 0.4998547434806824, "grad_norm": 30.088085174560547, "learning_rate": 2.3134328358208954e-07, "logits/chosen": 0.0881882905960083, "logits/rejected": 0.08028921484947205, "logps/chosen": -67.20095825195312, "logps/ref_chosen": -67.18333435058594, "logps/ref_rejected": -82.80763244628906, "logps/rejected": -82.83106994628906, "loss": 1.3859, "margin_dpo/margin_mean": 0.005814671516418457, "margin_dpo/margin_std": 0.274589866399765, "step": 32 }, { "KL/chosen_KL_mean": -0.0117340087890625, "KL/mean": -0.026652701199054718, "KL/rejected_KL_mean": -0.041568756103515625, "KL/std": 0.25098901987075806, "epoch": 0.049886621315192746, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.029830649495124817, "fcm_dpo/q_t": 0.49925386905670166, "grad_norm": 30.986589431762695, "learning_rate": 2.388059701492537e-07, "logits/chosen": 0.021548572927713394, "logits/rejected": -0.0038994159549474716, "logps/chosen": -64.05121612548828, "logps/ref_chosen": -64.03948211669922, "logps/ref_rejected": -75.68357849121094, "logps/rejected": -75.72515106201172, "loss": 1.3836, "margin_dpo/margin_mean": 0.029831483960151672, "margin_dpo/margin_std": 0.33458536863327026, "step": 33 }, { "KL/chosen_KL_mean": -0.012973785400390625, "KL/mean": -0.04527957737445831, "KL/rejected_KL_mean": -0.07758522033691406, "KL/std": 0.2115345597267151, "epoch": 0.05139833711262283, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06461545825004578, "fcm_dpo/q_t": 0.4983849823474884, "grad_norm": 28.43404769897461, "learning_rate": 2.4626865671641786e-07, "logits/chosen": 0.067808136343956, "logits/rejected": 0.03934960439801216, "logps/chosen": -53.677268981933594, "logps/ref_chosen": -53.6642951965332, "logps/ref_rejected": -65.77989959716797, "logps/rejected": -65.85748291015625, "loss": 1.3801, "margin_dpo/margin_mean": 0.06461548805236816, "margin_dpo/margin_std": 0.3139703869819641, "step": 34 }, { "KL/chosen_KL_mean": -0.08212852478027344, "KL/mean": -0.07559476792812347, "KL/rejected_KL_mean": -0.06905746459960938, "KL/std": 0.21326494216918945, "epoch": 0.05291005291005291, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.013070791959762573, "fcm_dpo/q_t": 0.5003267526626587, "grad_norm": 27.895584106445312, "learning_rate": 2.537313432835821e-07, "logits/chosen": 0.06519872695207596, "logits/rejected": 0.0424288734793663, "logps/chosen": -61.09899139404297, "logps/ref_chosen": -61.01686096191406, "logps/ref_rejected": -72.78598022460938, "logps/rejected": -72.85504150390625, "loss": 1.3878, "margin_dpo/margin_mean": -0.013071000576019287, "margin_dpo/margin_std": 0.27590203285217285, "step": 35 }, { "KL/chosen_KL_mean": -0.08506011962890625, "KL/mean": -0.0908443033695221, "KL/rejected_KL_mean": -0.09663009643554688, "KL/std": 0.21762478351593018, "epoch": 0.05442176870748299, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.011570632457733154, "fcm_dpo/q_t": 0.4997110962867737, "grad_norm": 28.65784454345703, "learning_rate": 2.611940298507462e-07, "logits/chosen": 0.12134528160095215, "logits/rejected": 0.06698712706565857, "logps/chosen": -50.62242126464844, "logps/ref_chosen": -50.53736114501953, "logps/ref_rejected": -78.11678314208984, "logps/rejected": -78.21340942382812, "loss": 1.3854, "margin_dpo/margin_mean": 0.011571049690246582, "margin_dpo/margin_std": 0.31866317987442017, "step": 36 }, { "KL/chosen_KL_mean": -0.04827880859375, "KL/mean": -0.10174515843391418, "KL/rejected_KL_mean": -0.15521621704101562, "KL/std": 0.3025971055030823, "epoch": 0.055933484504913075, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10692453384399414, "fcm_dpo/q_t": 0.4973316192626953, "grad_norm": 37.73421096801758, "learning_rate": 2.686567164179104e-07, "logits/chosen": 0.10500356554985046, "logits/rejected": 0.02512005902826786, "logps/chosen": -59.60222625732422, "logps/ref_chosen": -59.55394744873047, "logps/ref_rejected": -108.27702331542969, "logps/rejected": -108.43224334716797, "loss": 1.3761, "margin_dpo/margin_mean": 0.10692387819290161, "margin_dpo/margin_std": 0.4470498561859131, "step": 37 }, { "KL/chosen_KL_mean": -0.05261421203613281, "KL/mean": -0.06032392382621765, "KL/rejected_KL_mean": -0.06803512573242188, "KL/std": 0.2631487548351288, "epoch": 0.05744520030234316, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.015420258045196533, "fcm_dpo/q_t": 0.4996144771575928, "grad_norm": 29.76226806640625, "learning_rate": 2.761194029850746e-07, "logits/chosen": 0.06649500131607056, "logits/rejected": 0.05258432775735855, "logps/chosen": -65.84097290039062, "logps/ref_chosen": -65.78836059570312, "logps/ref_rejected": -76.1619873046875, "logps/rejected": -76.23002624511719, "loss": 1.3851, "margin_dpo/margin_mean": 0.015420526266098022, "margin_dpo/margin_std": 0.3901858329772949, "step": 38 }, { "KL/chosen_KL_mean": -0.056888580322265625, "KL/mean": -0.08803270757198334, "KL/rejected_KL_mean": -0.119171142578125, "KL/std": 0.255765825510025, "epoch": 0.05895691609977324, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06227746605873108, "fcm_dpo/q_t": 0.4984431862831116, "grad_norm": 29.257625579833984, "learning_rate": 2.8358208955223876e-07, "logits/chosen": 0.1489247977733612, "logits/rejected": 0.12244988977909088, "logps/chosen": -57.23369598388672, "logps/ref_chosen": -57.17681121826172, "logps/ref_rejected": -79.486328125, "logps/rejected": -79.60549926757812, "loss": 1.3804, "margin_dpo/margin_mean": 0.06227770447731018, "margin_dpo/margin_std": 0.36542096734046936, "step": 39 }, { "KL/chosen_KL_mean": -0.08425140380859375, "KL/mean": -0.1100742444396019, "KL/rejected_KL_mean": -0.13589859008789062, "KL/std": 0.2669486403465271, "epoch": 0.06046863189720333, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05164425075054169, "fcm_dpo/q_t": 0.49870917201042175, "grad_norm": 31.260446548461914, "learning_rate": 2.9104477611940296e-07, "logits/chosen": 0.13100269436836243, "logits/rejected": 0.0802154690027237, "logps/chosen": -61.418418884277344, "logps/ref_chosen": -61.33416748046875, "logps/ref_rejected": -79.10697174072266, "logps/rejected": -79.24287414550781, "loss": 1.3815, "margin_dpo/margin_mean": 0.05164472758769989, "margin_dpo/margin_std": 0.3377959430217743, "step": 40 }, { "KL/chosen_KL_mean": -0.14268875122070312, "KL/mean": -0.15808598697185516, "KL/rejected_KL_mean": -0.1734790802001953, "KL/std": 0.2604654133319855, "epoch": 0.06198034769463341, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.030783504247665405, "fcm_dpo/q_t": 0.49923068284988403, "grad_norm": 30.55925750732422, "learning_rate": 2.985074626865671e-07, "logits/chosen": 0.06479241698980331, "logits/rejected": 0.044267427176237106, "logps/chosen": -67.68942260742188, "logps/ref_chosen": -67.5467300415039, "logps/ref_rejected": -83.87788391113281, "logps/rejected": -84.0513687133789, "loss": 1.3835, "margin_dpo/margin_mean": 0.03078368306159973, "margin_dpo/margin_std": 0.3066967725753784, "step": 41 }, { "KL/chosen_KL_mean": -0.127410888671875, "KL/mean": -0.15839272737503052, "KL/rejected_KL_mean": -0.18938064575195312, "KL/std": 0.28839781880378723, "epoch": 0.06349206349206349, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06197001039981842, "fcm_dpo/q_t": 0.49845150113105774, "grad_norm": 29.170963287353516, "learning_rate": 3.059701492537313e-07, "logits/chosen": 0.06191250681877136, "logits/rejected": 0.04002637043595314, "logps/chosen": -61.392269134521484, "logps/ref_chosen": -61.26485824584961, "logps/ref_rejected": -76.3629150390625, "logps/rejected": -76.55229187011719, "loss": 1.3805, "margin_dpo/margin_mean": 0.06196974217891693, "margin_dpo/margin_std": 0.3553955554962158, "step": 42 }, { "KL/chosen_KL_mean": -0.09712600708007812, "KL/mean": -0.12862205505371094, "KL/rejected_KL_mean": -0.16011428833007812, "KL/std": 0.2960539758205414, "epoch": 0.06500377928949358, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06299221515655518, "fcm_dpo/q_t": 0.4984261393547058, "grad_norm": 34.61469268798828, "learning_rate": 3.134328358208955e-07, "logits/chosen": 0.09886477142572403, "logits/rejected": 0.08766770362854004, "logps/chosen": -71.90615844726562, "logps/ref_chosen": -71.80902862548828, "logps/ref_rejected": -81.12464141845703, "logps/rejected": -81.28475952148438, "loss": 1.3805, "margin_dpo/margin_mean": 0.06299227476119995, "margin_dpo/margin_std": 0.4519028663635254, "step": 43 }, { "KL/chosen_KL_mean": -0.1671600341796875, "KL/mean": -0.1565622091293335, "KL/rejected_KL_mean": -0.14596939086914062, "KL/std": 0.26872149109840393, "epoch": 0.06651549508692366, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02118849754333496, "fcm_dpo/q_t": 0.5005298256874084, "grad_norm": 32.638671875, "learning_rate": 3.2089552238805965e-07, "logits/chosen": 0.055923182517290115, "logits/rejected": 0.025120796635746956, "logps/chosen": -66.71759033203125, "logps/ref_chosen": -66.55043029785156, "logps/ref_rejected": -85.06198120117188, "logps/rejected": -85.20794677734375, "loss": 1.3888, "margin_dpo/margin_mean": -0.02118849754333496, "margin_dpo/margin_std": 0.39402931928634644, "step": 44 }, { "KL/chosen_KL_mean": -0.1441822052001953, "KL/mean": -0.19521594047546387, "KL/rejected_KL_mean": -0.24625015258789062, "KL/std": 0.28586143255233765, "epoch": 0.06802721088435375, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10206963121891022, "fcm_dpo/q_t": 0.49744898080825806, "grad_norm": 31.806509017944336, "learning_rate": 3.2835820895522385e-07, "logits/chosen": 0.12358801066875458, "logits/rejected": 0.07059746980667114, "logps/chosen": -62.388038635253906, "logps/ref_chosen": -62.24385452270508, "logps/ref_rejected": -92.96665954589844, "logps/rejected": -93.2129135131836, "loss": 1.3765, "margin_dpo/margin_mean": 0.10207107663154602, "margin_dpo/margin_std": 0.39976954460144043, "step": 45 }, { "KL/chosen_KL_mean": -0.10940742492675781, "KL/mean": -0.17646794021129608, "KL/rejected_KL_mean": -0.2435302734375, "KL/std": 0.31598347425460815, "epoch": 0.06953892668178382, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.13412192463874817, "fcm_dpo/q_t": 0.49664801359176636, "grad_norm": 31.140886306762695, "learning_rate": 3.3582089552238805e-07, "logits/chosen": 0.12344063818454742, "logits/rejected": 0.07757961750030518, "logps/chosen": -61.608314514160156, "logps/ref_chosen": -61.498905181884766, "logps/ref_rejected": -78.91172790527344, "logps/rejected": -79.15525817871094, "loss": 1.3733, "margin_dpo/margin_mean": 0.13412132859230042, "margin_dpo/margin_std": 0.38167721033096313, "step": 46 }, { "KL/chosen_KL_mean": -0.12112808227539062, "KL/mean": -0.21411283314228058, "KL/rejected_KL_mean": -0.307098388671875, "KL/std": 0.2846325635910034, "epoch": 0.0710506424792139, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.18596699833869934, "fcm_dpo/q_t": 0.4953530430793762, "grad_norm": 28.34612464904785, "learning_rate": 3.432835820895522e-07, "logits/chosen": 0.03395320102572441, "logits/rejected": -0.008879796601831913, "logps/chosen": -51.6994743347168, "logps/ref_chosen": -51.578346252441406, "logps/ref_rejected": -68.2215576171875, "logps/rejected": -68.52865600585938, "loss": 1.3682, "margin_dpo/margin_mean": 0.18596723675727844, "margin_dpo/margin_std": 0.3997984528541565, "step": 47 }, { "KL/chosen_KL_mean": -0.23851966857910156, "KL/mean": -0.25627630949020386, "KL/rejected_KL_mean": -0.2740345001220703, "KL/std": 0.2957780957221985, "epoch": 0.07256235827664399, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03551402688026428, "fcm_dpo/q_t": 0.49911266565322876, "grad_norm": 26.784517288208008, "learning_rate": 3.507462686567164e-07, "logits/chosen": 0.14234454929828644, "logits/rejected": 0.11283800005912781, "logps/chosen": -52.03217315673828, "logps/ref_chosen": -51.79365158081055, "logps/ref_rejected": -64.22503662109375, "logps/rejected": -64.49906921386719, "loss": 1.3832, "margin_dpo/margin_mean": 0.03551414608955383, "margin_dpo/margin_std": 0.3996293544769287, "step": 48 }, { "KL/chosen_KL_mean": -0.2007770538330078, "KL/mean": -0.25442013144493103, "KL/rejected_KL_mean": -0.3080596923828125, "KL/std": 0.3600367307662964, "epoch": 0.07407407407407407, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1072855144739151, "fcm_dpo/q_t": 0.4973185062408447, "grad_norm": 27.355039596557617, "learning_rate": 3.5820895522388055e-07, "logits/chosen": 0.009659804403781891, "logits/rejected": -0.011214490979909897, "logps/chosen": -58.33538055419922, "logps/ref_chosen": -58.13460159301758, "logps/ref_rejected": -64.63206481933594, "logps/rejected": -64.94012451171875, "loss": 1.3762, "margin_dpo/margin_mean": 0.10728609561920166, "margin_dpo/margin_std": 0.48972445726394653, "step": 49 }, { "KL/chosen_KL_mean": -0.2860908508300781, "KL/mean": -0.32769298553466797, "KL/rejected_KL_mean": -0.3692970275878906, "KL/std": 0.34533897042274475, "epoch": 0.07558578987150416, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08320507407188416, "fcm_dpo/q_t": 0.4979206919670105, "grad_norm": 28.02340316772461, "learning_rate": 3.6567164179104475e-07, "logits/chosen": 0.10904943943023682, "logits/rejected": 0.07938901335000992, "logps/chosen": -53.14253234863281, "logps/ref_chosen": -52.85643768310547, "logps/ref_rejected": -72.17460632324219, "logps/rejected": -72.54389953613281, "loss": 1.3784, "margin_dpo/margin_mean": 0.08320528268814087, "margin_dpo/margin_std": 0.41352373361587524, "step": 50 }, { "KL/chosen_KL_mean": -0.30113792419433594, "KL/mean": -0.38790732622146606, "KL/rejected_KL_mean": -0.47467803955078125, "KL/std": 0.3827604651451111, "epoch": 0.07709750566893424, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.17354324460029602, "fcm_dpo/q_t": 0.4956646263599396, "grad_norm": 31.038726806640625, "learning_rate": 3.7313432835820895e-07, "logits/chosen": 0.0878702700138092, "logits/rejected": 0.06008681282401085, "logps/chosen": -63.95758056640625, "logps/ref_chosen": -63.65644073486328, "logps/ref_rejected": -86.13229370117188, "logps/rejected": -86.60697937011719, "loss": 1.3696, "margin_dpo/margin_mean": 0.173543781042099, "margin_dpo/margin_std": 0.4818759560585022, "step": 51 }, { "KL/chosen_KL_mean": -0.2929725646972656, "KL/mean": -0.40666550397872925, "KL/rejected_KL_mean": -0.520355224609375, "KL/std": 0.4088834524154663, "epoch": 0.07860922146636433, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.22738106548786163, "fcm_dpo/q_t": 0.4943210184574127, "grad_norm": 32.647708892822266, "learning_rate": 3.805970149253731e-07, "logits/chosen": 0.1057782769203186, "logits/rejected": 0.05464206635951996, "logps/chosen": -68.13318634033203, "logps/ref_chosen": -67.8402099609375, "logps/ref_rejected": -96.97090911865234, "logps/rejected": -97.49126434326172, "loss": 1.3644, "margin_dpo/margin_mean": 0.2273811250925064, "margin_dpo/margin_std": 0.5355072021484375, "step": 52 }, { "KL/chosen_KL_mean": -0.38391876220703125, "KL/mean": -0.4516035318374634, "KL/rejected_KL_mean": -0.519287109375, "KL/std": 0.3600003123283386, "epoch": 0.0801209372637944, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.13536784052848816, "fcm_dpo/q_t": 0.4966163635253906, "grad_norm": 27.55713653564453, "learning_rate": 3.880597014925373e-07, "logits/chosen": 0.09621694684028625, "logits/rejected": 0.08538716286420822, "logps/chosen": -57.26205825805664, "logps/ref_chosen": -56.87813949584961, "logps/ref_rejected": -60.75569152832031, "logps/rejected": -61.27497863769531, "loss": 1.3734, "margin_dpo/margin_mean": 0.1353682279586792, "margin_dpo/margin_std": 0.47499287128448486, "step": 53 }, { "KL/chosen_KL_mean": -0.37387847900390625, "KL/mean": -0.44698044657707214, "KL/rejected_KL_mean": -0.5200824737548828, "KL/std": 0.37565189599990845, "epoch": 0.08163265306122448, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.14620435237884521, "fcm_dpo/q_t": 0.4963470697402954, "grad_norm": 26.632923126220703, "learning_rate": 3.9552238805970144e-07, "logits/chosen": 0.042465463280677795, "logits/rejected": 0.02737961709499359, "logps/chosen": -47.64080047607422, "logps/ref_chosen": -47.26692199707031, "logps/ref_rejected": -62.19426727294922, "logps/rejected": -62.71434783935547, "loss": 1.3723, "margin_dpo/margin_mean": 0.14620399475097656, "margin_dpo/margin_std": 0.4815361499786377, "step": 54 }, { "KL/chosen_KL_mean": -0.37229156494140625, "KL/mean": -0.5114315152168274, "KL/rejected_KL_mean": -0.65057373046875, "KL/std": 0.49368512630462646, "epoch": 0.08314436885865457, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2782818675041199, "fcm_dpo/q_t": 0.493056058883667, "grad_norm": 31.412118911743164, "learning_rate": 4.0298507462686564e-07, "logits/chosen": 0.03549818694591522, "logits/rejected": -0.04075536131858826, "logps/chosen": -50.698482513427734, "logps/ref_chosen": -50.32619094848633, "logps/ref_rejected": -92.44389343261719, "logps/rejected": -93.09446716308594, "loss": 1.3598, "margin_dpo/margin_mean": 0.27828145027160645, "margin_dpo/margin_std": 0.6260495185852051, "step": 55 }, { "KL/chosen_KL_mean": -0.3755607604980469, "KL/mean": -0.4541851282119751, "KL/rejected_KL_mean": -0.5328025817871094, "KL/std": 0.4594634771347046, "epoch": 0.08465608465608465, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1572437584400177, "fcm_dpo/q_t": 0.49606892466545105, "grad_norm": 27.55105972290039, "learning_rate": 4.1044776119402984e-07, "logits/chosen": 0.12995021045207977, "logits/rejected": 0.10750794410705566, "logps/chosen": -57.14253234863281, "logps/ref_chosen": -56.766971588134766, "logps/ref_rejected": -66.30504608154297, "logps/rejected": -66.83784484863281, "loss": 1.3715, "margin_dpo/margin_mean": 0.15724411606788635, "margin_dpo/margin_std": 0.583028256893158, "step": 56 }, { "KL/chosen_KL_mean": -0.5324649810791016, "KL/mean": -0.6625369787216187, "KL/rejected_KL_mean": -0.7926063537597656, "KL/std": 0.5992856025695801, "epoch": 0.08616780045351474, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2601401209831238, "fcm_dpo/q_t": 0.4935043454170227, "grad_norm": 30.670913696289062, "learning_rate": 4.17910447761194e-07, "logits/chosen": 0.14049838483333588, "logits/rejected": 0.07349002361297607, "logps/chosen": -58.300209045410156, "logps/ref_chosen": -57.76774597167969, "logps/ref_rejected": -82.75698852539062, "logps/rejected": -83.54959106445312, "loss": 1.3617, "margin_dpo/margin_mean": 0.26013994216918945, "margin_dpo/margin_std": 0.6875753402709961, "step": 57 }, { "KL/chosen_KL_mean": -0.6022109985351562, "KL/mean": -0.6994897723197937, "KL/rejected_KL_mean": -0.7967681884765625, "KL/std": 0.6317234635353088, "epoch": 0.08767951625094482, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1945592164993286, "fcm_dpo/q_t": 0.4951819181442261, "grad_norm": 30.232378005981445, "learning_rate": 4.253731343283582e-07, "logits/chosen": 0.08454892039299011, "logits/rejected": 0.0682353526353836, "logps/chosen": -73.36630249023438, "logps/ref_chosen": -72.76408386230469, "logps/ref_rejected": -84.49275207519531, "logps/rejected": -85.28952026367188, "loss": 1.3693, "margin_dpo/margin_mean": 0.19455870985984802, "margin_dpo/margin_std": 0.9174859523773193, "step": 58 }, { "KL/chosen_KL_mean": -0.5136947631835938, "KL/mean": -0.6560180187225342, "KL/rejected_KL_mean": -0.7983455657958984, "KL/std": 0.5859674215316772, "epoch": 0.08919123204837491, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.28464627265930176, "fcm_dpo/q_t": 0.49290651082992554, "grad_norm": 26.60697364807129, "learning_rate": 4.3283582089552234e-07, "logits/chosen": 0.13293355703353882, "logits/rejected": 0.06686300039291382, "logps/chosen": -50.33447265625, "logps/ref_chosen": -49.820777893066406, "logps/ref_rejected": -77.14368438720703, "logps/rejected": -77.94203186035156, "loss": 1.3597, "margin_dpo/margin_mean": 0.28464648127555847, "margin_dpo/margin_std": 0.7471631169319153, "step": 59 }, { "KL/chosen_KL_mean": -0.6310577392578125, "KL/mean": -0.6649550199508667, "KL/rejected_KL_mean": -0.6988525390625, "KL/std": 0.6437369585037231, "epoch": 0.09070294784580499, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0677938461303711, "fcm_dpo/q_t": 0.4983041286468506, "grad_norm": 29.712261199951172, "learning_rate": 4.4029850746268654e-07, "logits/chosen": 0.1172294169664383, "logits/rejected": 0.1158132255077362, "logps/chosen": -63.855831146240234, "logps/ref_chosen": -63.22477340698242, "logps/ref_rejected": -61.360477447509766, "logps/rejected": -62.059329986572266, "loss": 1.3814, "margin_dpo/margin_mean": 0.0677938461303711, "margin_dpo/margin_std": 0.8459521532058716, "step": 60 }, { "KL/chosen_KL_mean": -0.7134819030761719, "KL/mean": -0.7604533433914185, "KL/rejected_KL_mean": -0.8074264526367188, "KL/std": 0.7058064341545105, "epoch": 0.09221466364323508, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0939413458108902, "fcm_dpo/q_t": 0.4976520538330078, "grad_norm": 27.93014907836914, "learning_rate": 4.4776119402985074e-07, "logits/chosen": 0.14893561601638794, "logits/rejected": 0.11584499478340149, "logps/chosen": -49.73027801513672, "logps/ref_chosen": -49.01679992675781, "logps/ref_rejected": -74.90817260742188, "logps/rejected": -75.71559143066406, "loss": 1.3793, "margin_dpo/margin_mean": 0.09394136071205139, "margin_dpo/margin_std": 0.9773008823394775, "step": 61 }, { "KL/chosen_KL_mean": -0.7257556915283203, "KL/mean": -0.8533656001091003, "KL/rejected_KL_mean": -0.9809722900390625, "KL/std": 0.6654196977615356, "epoch": 0.09372637944066516, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.25521424412727356, "fcm_dpo/q_t": 0.4936321973800659, "grad_norm": 28.620866775512695, "learning_rate": 4.552238805970149e-07, "logits/chosen": 0.11153241246938705, "logits/rejected": 0.07246184349060059, "logps/chosen": -63.477622985839844, "logps/ref_chosen": -62.751869201660156, "logps/ref_rejected": -78.93360900878906, "logps/rejected": -79.91458129882812, "loss": 1.3632, "margin_dpo/margin_mean": 0.2552138566970825, "margin_dpo/margin_std": 0.9336051940917969, "step": 62 }, { "KL/chosen_KL_mean": -0.5593242645263672, "KL/mean": -0.8448190689086914, "KL/rejected_KL_mean": -1.1303176879882812, "KL/std": 0.6812584400177002, "epoch": 0.09523809523809523, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5709925889968872, "fcm_dpo/q_t": 0.4857600927352905, "grad_norm": 31.421058654785156, "learning_rate": 4.626865671641791e-07, "logits/chosen": 0.16452708840370178, "logits/rejected": 0.1406918168067932, "logps/chosen": -61.0745735168457, "logps/ref_chosen": -60.51525115966797, "logps/ref_rejected": -85.11021423339844, "logps/rejected": -86.24053192138672, "loss": 1.3317, "margin_dpo/margin_mean": 0.5709917545318604, "margin_dpo/margin_std": 0.8313959836959839, "step": 63 }, { "KL/chosen_KL_mean": -0.8067245483398438, "KL/mean": -0.849244236946106, "KL/rejected_KL_mean": -0.8917655944824219, "KL/std": 0.7516355514526367, "epoch": 0.09674981103552532, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08503898978233337, "fcm_dpo/q_t": 0.49786609411239624, "grad_norm": 26.175302505493164, "learning_rate": 4.701492537313433e-07, "logits/chosen": 0.07753998041152954, "logits/rejected": 0.05309564620256424, "logps/chosen": -52.013572692871094, "logps/ref_chosen": -51.20684814453125, "logps/ref_rejected": -66.93081665039062, "logps/rejected": -67.82258605957031, "loss": 1.3799, "margin_dpo/margin_mean": 0.08503925800323486, "margin_dpo/margin_std": 0.9009600281715393, "step": 64 }, { "KL/chosen_KL_mean": -0.8779640197753906, "KL/mean": -1.1214277744293213, "KL/rejected_KL_mean": -1.3648948669433594, "KL/std": 0.9085813760757446, "epoch": 0.0982615268329554, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4869285821914673, "fcm_dpo/q_t": 0.48788049817085266, "grad_norm": 30.840694427490234, "learning_rate": 4.776119402985074e-07, "logits/chosen": 0.20080530643463135, "logits/rejected": 0.17102986574172974, "logps/chosen": -68.16665649414062, "logps/ref_chosen": -67.2886962890625, "logps/ref_rejected": -74.44281005859375, "logps/rejected": -75.80770111083984, "loss": 1.3418, "margin_dpo/margin_mean": 0.48692822456359863, "margin_dpo/margin_std": 1.1947252750396729, "step": 65 }, { "KL/chosen_KL_mean": -0.9242038726806641, "KL/mean": -1.0837883949279785, "KL/rejected_KL_mean": -1.243377685546875, "KL/std": 0.8412731885910034, "epoch": 0.09977324263038549, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.31917569041252136, "fcm_dpo/q_t": 0.4920502305030823, "grad_norm": 29.339235305786133, "learning_rate": 4.850746268656717e-07, "logits/chosen": 0.09757672995328903, "logits/rejected": 0.07372169941663742, "logps/chosen": -71.66761779785156, "logps/ref_chosen": -70.743408203125, "logps/ref_rejected": -77.26499938964844, "logps/rejected": -78.50837707519531, "loss": 1.3578, "margin_dpo/margin_mean": 0.31917527318000793, "margin_dpo/margin_std": 1.1150388717651367, "step": 66 }, { "KL/chosen_KL_mean": -0.7873554229736328, "KL/mean": -0.9470667839050293, "KL/rejected_KL_mean": -1.106781005859375, "KL/std": 0.7722550630569458, "epoch": 0.10128495842781557, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3194220960140228, "fcm_dpo/q_t": 0.49203306436538696, "grad_norm": 29.272001266479492, "learning_rate": 4.925373134328357e-07, "logits/chosen": 0.0839434489607811, "logits/rejected": 0.028710132464766502, "logps/chosen": -61.389957427978516, "logps/ref_chosen": -60.60260009765625, "logps/ref_rejected": -75.22235870361328, "logps/rejected": -76.32913970947266, "loss": 1.3569, "margin_dpo/margin_mean": 0.31942254304885864, "margin_dpo/margin_std": 0.9561138153076172, "step": 67 }, { "KL/chosen_KL_mean": -1.2437705993652344, "KL/mean": -1.405458927154541, "KL/rejected_KL_mean": -1.5671539306640625, "KL/std": 0.949453592300415, "epoch": 0.10279667422524566, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3233872652053833, "fcm_dpo/q_t": 0.49194926023483276, "grad_norm": 31.313793182373047, "learning_rate": 5e-07, "logits/chosen": 0.04201754555106163, "logits/rejected": 0.013082325458526611, "logps/chosen": -78.77214050292969, "logps/ref_chosen": -77.52836608886719, "logps/ref_rejected": -93.17778015136719, "logps/rejected": -94.74493408203125, "loss": 1.3583, "margin_dpo/margin_mean": 0.32338738441467285, "margin_dpo/margin_std": 1.2540662288665771, "step": 68 }, { "KL/chosen_KL_mean": -1.1451244354248047, "KL/mean": -1.4814937114715576, "KL/rejected_KL_mean": -1.8178558349609375, "KL/std": 1.0125997066497803, "epoch": 0.10430839002267574, "fcm_dpo/beta": 0.10156106948852539, "fcm_dpo/delta": 0.15371915698051453, "fcm_dpo/margin": 0.6727360486984253, "fcm_dpo/q_t": 0.4832885265350342, "grad_norm": 31.382776260375977, "learning_rate": 4.999965034812934e-07, "logits/chosen": 0.11837112903594971, "logits/rejected": 0.07452549785375595, "logps/chosen": -67.08818054199219, "logps/ref_chosen": -65.94305419921875, "logps/ref_rejected": -89.7735595703125, "logps/rejected": -91.59141540527344, "loss": 1.3227, "margin_dpo/margin_mean": 0.672736644744873, "margin_dpo/margin_std": 1.206894874572754, "step": 69 }, { "KL/chosen_KL_mean": -1.2427654266357422, "KL/mean": -1.400121808052063, "KL/rejected_KL_mean": -1.5574722290039062, "KL/std": 0.8860043883323669, "epoch": 0.10582010582010581, "fcm_dpo/beta": 0.10312213003635406, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3147037923336029, "fcm_dpo/q_t": 0.4919206500053406, "grad_norm": 29.573619842529297, "learning_rate": 4.999860140229787e-07, "logits/chosen": 0.1305362433195114, "logits/rejected": 0.10781893879175186, "logps/chosen": -63.20067596435547, "logps/ref_chosen": -61.95791244506836, "logps/ref_rejected": -75.80945587158203, "logps/rejected": -77.36692810058594, "loss": 1.3577, "margin_dpo/margin_mean": 0.3147040009498596, "margin_dpo/margin_std": 1.1622765064239502, "step": 70 }, { "KL/chosen_KL_mean": -1.4353160858154297, "KL/mean": -1.5389230251312256, "KL/rejected_KL_mean": -1.6425323486328125, "KL/std": 1.0423591136932373, "epoch": 0.1073318216175359, "fcm_dpo/beta": 0.10312213003635406, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2072136104106903, "fcm_dpo/q_t": 0.49464237689971924, "grad_norm": 29.52690315246582, "learning_rate": 4.999685319184688e-07, "logits/chosen": 0.08314691483974457, "logits/rejected": 0.0680352970957756, "logps/chosen": -64.78289794921875, "logps/ref_chosen": -63.34757995605469, "logps/ref_rejected": -67.49658203125, "logps/rejected": -69.13911437988281, "loss": 1.3712, "margin_dpo/margin_mean": 0.2072138488292694, "margin_dpo/margin_std": 1.4971086978912354, "step": 71 }, { "KL/chosen_KL_mean": -1.2093849182128906, "KL/mean": -1.5765079259872437, "KL/rejected_KL_mean": -1.943634033203125, "KL/std": 1.095198154449463, "epoch": 0.10884353741496598, "fcm_dpo/beta": 0.10635857284069061, "fcm_dpo/delta": 0.15451082587242126, "fcm_dpo/margin": 0.7342473268508911, "fcm_dpo/q_t": 0.4809272885322571, "grad_norm": 31.995201110839844, "learning_rate": 4.999440576567755e-07, "logits/chosen": 0.10724207758903503, "logits/rejected": 0.044822730123996735, "logps/chosen": -57.06868362426758, "logps/ref_chosen": -55.85929870605469, "logps/ref_rejected": -68.45423889160156, "logps/rejected": -70.39787292480469, "loss": 1.3149, "margin_dpo/margin_mean": 0.7342471480369568, "margin_dpo/margin_std": 1.330765962600708, "step": 72 }, { "KL/chosen_KL_mean": -1.806976318359375, "KL/mean": -1.8687390089035034, "KL/rejected_KL_mean": -1.9305057525634766, "KL/std": 1.2052420377731323, "epoch": 0.11035525321239607, "fcm_dpo/beta": 0.10635857284069061, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.12353459000587463, "fcm_dpo/q_t": 0.4968770742416382, "grad_norm": 34.1959342956543, "learning_rate": 4.999125919224965e-07, "logits/chosen": 0.12511396408081055, "logits/rejected": 0.11073576658964157, "logps/chosen": -70.94578552246094, "logps/ref_chosen": -69.13880920410156, "logps/ref_rejected": -79.04586791992188, "logps/rejected": -80.97637939453125, "loss": 1.3805, "margin_dpo/margin_mean": 0.12353500723838806, "margin_dpo/margin_std": 1.5384290218353271, "step": 73 }, { "KL/chosen_KL_mean": -1.2401103973388672, "KL/mean": -1.5870051383972168, "KL/rejected_KL_mean": -1.9338951110839844, "KL/std": 0.9800302982330322, "epoch": 0.11186696900982615, "fcm_dpo/beta": 0.10799010097980499, "fcm_dpo/delta": 0.15109321475028992, "fcm_dpo/margin": 0.6937872171401978, "fcm_dpo/q_t": 0.4816681742668152, "grad_norm": 29.913087844848633, "learning_rate": 4.998741355957963e-07, "logits/chosen": 0.13613180816173553, "logits/rejected": 0.08324501663446426, "logps/chosen": -51.163848876953125, "logps/ref_chosen": -49.923736572265625, "logps/ref_rejected": -81.73213958740234, "logps/rejected": -83.66603088378906, "loss": 1.3172, "margin_dpo/margin_mean": 0.6937870383262634, "margin_dpo/margin_std": 1.2566819190979004, "step": 74 }, { "KL/chosen_KL_mean": -1.4187679290771484, "KL/mean": -1.7926573753356934, "KL/rejected_KL_mean": -2.166543960571289, "KL/std": 1.1323838233947754, "epoch": 0.11337868480725624, "fcm_dpo/beta": 0.11300066858530045, "fcm_dpo/delta": 0.1517949253320694, "fcm_dpo/margin": 0.7477781772613525, "fcm_dpo/q_t": 0.4794332981109619, "grad_norm": 29.2504940032959, "learning_rate": 4.998286897523808e-07, "logits/chosen": 0.10926586389541626, "logits/rejected": 0.07744550704956055, "logps/chosen": -47.487518310546875, "logps/ref_chosen": -46.06875228881836, "logps/ref_rejected": -66.1181411743164, "logps/rejected": -68.28468322753906, "loss": 1.3102, "margin_dpo/margin_mean": 0.7477779388427734, "margin_dpo/margin_std": 1.4347933530807495, "step": 75 }, { "KL/chosen_KL_mean": -1.6109638214111328, "KL/mean": -1.770002007484436, "KL/rejected_KL_mean": -1.9290390014648438, "KL/std": 1.1657469272613525, "epoch": 0.11489040060468632, "fcm_dpo/beta": 0.11300066858530045, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3180782198905945, "fcm_dpo/q_t": 0.4910624921321869, "grad_norm": 33.03213119506836, "learning_rate": 4.997762556634679e-07, "logits/chosen": 0.1351306140422821, "logits/rejected": 0.09061160683631897, "logps/chosen": -55.67371368408203, "logps/ref_chosen": -54.06275177001953, "logps/ref_rejected": -74.87464141845703, "logps/rejected": -76.80368041992188, "loss": 1.3583, "margin_dpo/margin_mean": 0.31807827949523926, "margin_dpo/margin_std": 1.5470020771026611, "step": 76 }, { "KL/chosen_KL_mean": -1.6516189575195312, "KL/mean": -2.0099892616271973, "KL/rejected_KL_mean": -2.368358612060547, "KL/std": 1.238326072692871, "epoch": 0.1164021164021164, "fcm_dpo/beta": 0.11631464213132858, "fcm_dpo/delta": 0.1445261687040329, "fcm_dpo/margin": 0.7167394161224365, "fcm_dpo/q_t": 0.47974491119384766, "grad_norm": 33.77786636352539, "learning_rate": 4.99716834795752e-07, "logits/chosen": 0.15451249480247498, "logits/rejected": 0.1131986528635025, "logps/chosen": -54.72771072387695, "logps/ref_chosen": -53.07609176635742, "logps/ref_rejected": -74.45601654052734, "logps/rejected": -76.82437133789062, "loss": 1.3121, "margin_dpo/margin_mean": 0.7167395353317261, "margin_dpo/margin_std": 1.4588513374328613, "step": 77 }, { "KL/chosen_KL_mean": -1.8221549987792969, "KL/mean": -2.0393378734588623, "KL/rejected_KL_mean": -2.256519317626953, "KL/std": 1.2632840871810913, "epoch": 0.11791383219954649, "fcm_dpo/beta": 0.11631464213132858, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4343580901622772, "fcm_dpo/q_t": 0.48744601011276245, "grad_norm": 34.19994354248047, "learning_rate": 4.996504288113623e-07, "logits/chosen": 0.11588499695062637, "logits/rejected": 0.09554223716259003, "logps/chosen": -69.54757690429688, "logps/ref_chosen": -67.72541809082031, "logps/ref_rejected": -79.03926849365234, "logps/rejected": -81.29579162597656, "loss": 1.3433, "margin_dpo/margin_mean": 0.4343578517436981, "margin_dpo/margin_std": 1.4269845485687256, "step": 78 }, { "KL/chosen_KL_mean": -1.7330493927001953, "KL/mean": -2.276017189025879, "KL/rejected_KL_mean": -2.8189849853515625, "KL/std": 1.5086700916290283, "epoch": 0.11942554799697656, "fcm_dpo/beta": 0.12106219679117203, "fcm_dpo/delta": 0.27233055233955383, "fcm_dpo/margin": 1.0859367847442627, "fcm_dpo/q_t": 0.46888160705566406, "grad_norm": 36.362300872802734, "learning_rate": 4.995770395678171e-07, "logits/chosen": 0.19241735339164734, "logits/rejected": 0.1304788440465927, "logps/chosen": -53.89369201660156, "logps/ref_chosen": -52.16064453125, "logps/ref_rejected": -83.31062316894531, "logps/rejected": -86.12960815429688, "loss": 1.2734, "margin_dpo/margin_mean": 1.0859365463256836, "margin_dpo/margin_std": 1.9711155891418457, "step": 79 }, { "KL/chosen_KL_mean": -2.051755905151367, "KL/mean": -2.335285186767578, "KL/rejected_KL_mean": -2.6188125610351562, "KL/std": 1.3859307765960693, "epoch": 0.12093726379440665, "fcm_dpo/beta": 0.12457603216171265, "fcm_dpo/delta": 0.14052413403987885, "fcm_dpo/margin": 0.567061722278595, "fcm_dpo/q_t": 0.48296597599983215, "grad_norm": 35.15907669067383, "learning_rate": 4.994966691179711e-07, "logits/chosen": 0.14136065542697906, "logits/rejected": 0.0816778764128685, "logps/chosen": -63.462318420410156, "logps/ref_chosen": -61.410560607910156, "logps/ref_rejected": -78.66004943847656, "logps/rejected": -81.27886199951172, "loss": 1.3314, "margin_dpo/margin_mean": 0.5670617818832397, "margin_dpo/margin_std": 1.904353141784668, "step": 80 }, { "KL/chosen_KL_mean": -2.106597900390625, "KL/mean": -2.5513858795166016, "KL/rejected_KL_mean": -2.9961700439453125, "KL/std": 1.4156627655029297, "epoch": 0.12244897959183673, "fcm_dpo/beta": 0.12924838066101074, "fcm_dpo/delta": 0.11433006823062897, "fcm_dpo/margin": 0.8895747661590576, "fcm_dpo/q_t": 0.4723474383354187, "grad_norm": 36.73268508911133, "learning_rate": 4.994093197099587e-07, "logits/chosen": 0.13766348361968994, "logits/rejected": 0.10336862504482269, "logps/chosen": -65.91096496582031, "logps/ref_chosen": -63.80437088012695, "logps/ref_rejected": -79.3484115600586, "logps/rejected": -82.3445816040039, "loss": 1.2874, "margin_dpo/margin_mean": 0.8895748853683472, "margin_dpo/margin_std": 1.6920909881591797, "step": 81 }, { "KL/chosen_KL_mean": -1.7257709503173828, "KL/mean": -2.2757372856140137, "KL/rejected_KL_mean": -2.8257064819335938, "KL/std": 1.3814198970794678, "epoch": 0.12396069538926682, "fcm_dpo/beta": 0.13408434391021729, "fcm_dpo/delta": 0.25668978691101074, "fcm_dpo/margin": 1.0999317169189453, "fcm_dpo/q_t": 0.46460217237472534, "grad_norm": 34.45591354370117, "learning_rate": 4.993149937871306e-07, "logits/chosen": 0.124167300760746, "logits/rejected": 0.06017389893531799, "logps/chosen": -50.543663024902344, "logps/ref_chosen": -48.817893981933594, "logps/ref_rejected": -70.31497955322266, "logps/rejected": -73.14068603515625, "loss": 1.2565, "margin_dpo/margin_mean": 1.0999321937561035, "margin_dpo/margin_std": 1.6014728546142578, "step": 82 }, { "KL/chosen_KL_mean": -2.0304203033447266, "KL/mean": -2.6201720237731934, "KL/rejected_KL_mean": -3.2099266052246094, "KL/std": 1.4840151071548462, "epoch": 0.1254724111866969, "fcm_dpo/beta": 0.14089030027389526, "fcm_dpo/delta": 0.23794013261795044, "fcm_dpo/margin": 1.1795077323913574, "fcm_dpo/q_t": 0.46056699752807617, "grad_norm": 39.07538986206055, "learning_rate": 4.992136939879856e-07, "logits/chosen": 0.1855728030204773, "logits/rejected": 0.13604822754859924, "logps/chosen": -59.18119430541992, "logps/ref_chosen": -57.15077209472656, "logps/ref_rejected": -75.1710205078125, "logps/rejected": -78.38095092773438, "loss": 1.2443, "margin_dpo/margin_mean": 1.1795073747634888, "margin_dpo/margin_std": 1.8788938522338867, "step": 83 }, { "KL/chosen_KL_mean": -2.489429473876953, "KL/mean": -2.9202234745025635, "KL/rejected_KL_mean": -3.351024627685547, "KL/std": 1.8019747734069824, "epoch": 0.12698412698412698, "fcm_dpo/beta": 0.14852453768253326, "fcm_dpo/delta": 0.2758869230747223, "fcm_dpo/margin": 0.8615986108779907, "fcm_dpo/q_t": 0.4696010649204254, "grad_norm": 44.80130386352539, "learning_rate": 4.991054231460969e-07, "logits/chosen": 0.16139745712280273, "logits/rejected": 0.11974325776100159, "logps/chosen": -67.26673126220703, "logps/ref_chosen": -64.77729797363281, "logps/ref_rejected": -84.71949768066406, "logps/rejected": -88.07052612304688, "loss": 1.2825, "margin_dpo/margin_mean": 0.8615986108779907, "margin_dpo/margin_std": 1.8895180225372314, "step": 84 }, { "KL/chosen_KL_mean": -2.2373600006103516, "KL/mean": -2.8613109588623047, "KL/rejected_KL_mean": -3.485260009765625, "KL/std": 1.7679171562194824, "epoch": 0.12849584278155707, "fcm_dpo/beta": 0.15582939982414246, "fcm_dpo/delta": 0.20922988653182983, "fcm_dpo/margin": 1.247900128364563, "fcm_dpo/q_t": 0.45365017652511597, "grad_norm": 39.90945053100586, "learning_rate": 4.989901842900325e-07, "logits/chosen": 0.1625782549381256, "logits/rejected": 0.11865755915641785, "logps/chosen": -52.48905563354492, "logps/ref_chosen": -50.25169372558594, "logps/ref_rejected": -66.55439758300781, "logps/rejected": -70.03965759277344, "loss": 1.2256, "margin_dpo/margin_mean": 1.2479002475738525, "margin_dpo/margin_std": 1.9751851558685303, "step": 85 }, { "KL/chosen_KL_mean": -2.76788330078125, "KL/mean": -3.2456436157226562, "KL/rejected_KL_mean": -3.7234020233154297, "KL/std": 1.8651580810546875, "epoch": 0.13000755857898716, "fcm_dpo/beta": 0.16335007548332214, "fcm_dpo/delta": 0.2474798560142517, "fcm_dpo/margin": 0.9555222988128662, "fcm_dpo/q_t": 0.4628984332084656, "grad_norm": 42.055274963378906, "learning_rate": 4.988679806432711e-07, "logits/chosen": 0.18018244206905365, "logits/rejected": 0.16184869408607483, "logps/chosen": -63.49706268310547, "logps/ref_chosen": -60.72917938232422, "logps/ref_rejected": -72.30961608886719, "logps/rejected": -76.03301239013672, "loss": 1.2629, "margin_dpo/margin_mean": 0.955522894859314, "margin_dpo/margin_std": 2.0158450603485107, "step": 86 }, { "KL/chosen_KL_mean": -2.952251434326172, "KL/mean": -3.472977638244629, "KL/rejected_KL_mean": -3.9937057495117188, "KL/std": 1.9867116212844849, "epoch": 0.13151927437641722, "fcm_dpo/beta": 0.17066358029842377, "fcm_dpo/delta": 0.2265702486038208, "fcm_dpo/margin": 1.0414538383483887, "fcm_dpo/q_t": 0.458102822303772, "grad_norm": 51.16208267211914, "learning_rate": 4.987388156241114e-07, "logits/chosen": 0.17016345262527466, "logits/rejected": 0.1106840968132019, "logps/chosen": -68.71022033691406, "logps/ref_chosen": -65.75796508789062, "logps/ref_rejected": -84.81159973144531, "logps/rejected": -88.8052978515625, "loss": 1.2572, "margin_dpo/margin_mean": 1.0414538383483887, "margin_dpo/margin_std": 2.364213466644287, "step": 87 }, { "KL/chosen_KL_mean": -2.958810806274414, "KL/mean": -3.3786325454711914, "KL/rejected_KL_mean": -3.7984542846679688, "KL/std": 2.2283806800842285, "epoch": 0.1330309901738473, "fcm_dpo/beta": 0.17520461976528168, "fcm_dpo/delta": 0.06729626655578613, "fcm_dpo/margin": 0.8396420478820801, "fcm_dpo/q_t": 0.46605074405670166, "grad_norm": 54.265357971191406, "learning_rate": 4.986026928455767e-07, "logits/chosen": 0.24361751973628998, "logits/rejected": 0.2162962704896927, "logps/chosen": -65.7828369140625, "logps/ref_chosen": -62.82402801513672, "logps/ref_rejected": -74.9607162475586, "logps/rejected": -78.75917053222656, "loss": 1.3089, "margin_dpo/margin_mean": 0.8396422863006592, "margin_dpo/margin_std": 2.872804641723633, "step": 88 }, { "KL/chosen_KL_mean": -2.457714080810547, "KL/mean": -3.142911672592163, "KL/rejected_KL_mean": -3.8281097412109375, "KL/std": 2.082172393798828, "epoch": 0.1345427059712774, "fcm_dpo/beta": 0.18018333613872528, "fcm_dpo/delta": 0.1557571142911911, "fcm_dpo/margin": 1.370391845703125, "fcm_dpo/q_t": 0.44298940896987915, "grad_norm": 48.1218376159668, "learning_rate": 4.984596161153135e-07, "logits/chosen": 0.2617233097553253, "logits/rejected": 0.17753329873085022, "logps/chosen": -43.649147033691406, "logps/ref_chosen": -41.191436767578125, "logps/ref_rejected": -85.44769287109375, "logps/rejected": -89.27580261230469, "loss": 1.2025, "margin_dpo/margin_mean": 1.370391845703125, "margin_dpo/margin_std": 2.4326417446136475, "step": 89 }, { "KL/chosen_KL_mean": -2.9885406494140625, "KL/mean": -3.592991828918457, "KL/rejected_KL_mean": -4.19744873046875, "KL/std": 2.275628089904785, "epoch": 0.1360544217687075, "fcm_dpo/beta": 0.18507075309753418, "fcm_dpo/delta": 0.18068405985832214, "fcm_dpo/margin": 1.208900809288025, "fcm_dpo/q_t": 0.4476126432418823, "grad_norm": 50.888099670410156, "learning_rate": 4.983095894354857e-07, "logits/chosen": 0.13500599563121796, "logits/rejected": 0.08118491619825363, "logps/chosen": -59.57244873046875, "logps/ref_chosen": -56.58390808105469, "logps/ref_rejected": -86.86978149414062, "logps/rejected": -91.06723022460938, "loss": 1.2211, "margin_dpo/margin_mean": 1.2089004516601562, "margin_dpo/margin_std": 2.3021106719970703, "step": 90 }, { "KL/chosen_KL_mean": -2.9596328735351562, "KL/mean": -3.510220527648926, "KL/rejected_KL_mean": -4.0608062744140625, "KL/std": 2.125333309173584, "epoch": 0.13756613756613756, "fcm_dpo/beta": 0.19327741861343384, "fcm_dpo/delta": 0.1906760036945343, "fcm_dpo/margin": 1.1011730432510376, "fcm_dpo/q_t": 0.45179492235183716, "grad_norm": 48.79199981689453, "learning_rate": 4.98152617002662e-07, "logits/chosen": 0.16606345772743225, "logits/rejected": 0.12146725505590439, "logps/chosen": -55.341976165771484, "logps/ref_chosen": -52.38234329223633, "logps/ref_rejected": -72.17642211914062, "logps/rejected": -76.23722839355469, "loss": 1.2445, "margin_dpo/margin_mean": 1.101172924041748, "margin_dpo/margin_std": 2.5953307151794434, "step": 91 }, { "KL/chosen_KL_mean": -3.0597572326660156, "KL/mean": -3.790630340576172, "KL/rejected_KL_mean": -4.521511077880859, "KL/std": 2.433997631072998, "epoch": 0.13907785336356765, "fcm_dpo/beta": 0.19871735572814941, "fcm_dpo/delta": 0.11208958178758621, "fcm_dpo/margin": 1.4617501497268677, "fcm_dpo/q_t": 0.43508481979370117, "grad_norm": 47.62337875366211, "learning_rate": 4.979887032076988e-07, "logits/chosen": 0.23759877681732178, "logits/rejected": 0.195323646068573, "logps/chosen": -56.068458557128906, "logps/ref_chosen": -53.00870132446289, "logps/ref_rejected": -79.77812957763672, "logps/rejected": -84.29963684082031, "loss": 1.1943, "margin_dpo/margin_mean": 1.4617503881454468, "margin_dpo/margin_std": 2.8890533447265625, "step": 92 }, { "KL/chosen_KL_mean": -3.3475303649902344, "KL/mean": -3.8043527603149414, "KL/rejected_KL_mean": -4.26116943359375, "KL/std": 2.362985849380493, "epoch": 0.14058956916099774, "fcm_dpo/beta": 0.2060101330280304, "fcm_dpo/delta": 0.21536466479301453, "fcm_dpo/margin": 0.9136420488357544, "fcm_dpo/q_t": 0.45910483598709106, "grad_norm": 49.81450271606445, "learning_rate": 4.978178526356172e-07, "logits/chosen": 0.20445819199085236, "logits/rejected": 0.17469534277915955, "logps/chosen": -48.25458526611328, "logps/ref_chosen": -44.90705108642578, "logps/ref_rejected": -58.7879524230957, "logps/rejected": -63.04912185668945, "loss": 1.2869, "margin_dpo/margin_mean": 0.9136421084403992, "margin_dpo/margin_std": 2.857240676879883, "step": 93 }, { "KL/chosen_KL_mean": -2.9746627807617188, "KL/mean": -3.8127119541168213, "KL/rejected_KL_mean": -4.650768280029297, "KL/std": 2.772942543029785, "epoch": 0.1421012849584278, "fcm_dpo/beta": 0.20925289392471313, "fcm_dpo/delta": 0.051092229783535004, "fcm_dpo/margin": 1.6761001348495483, "fcm_dpo/q_t": 0.4200799763202667, "grad_norm": 50.38736343383789, "learning_rate": 4.976400700654751e-07, "logits/chosen": 0.2535303235054016, "logits/rejected": 0.21268761157989502, "logps/chosen": -62.91243362426758, "logps/ref_chosen": -59.93777084350586, "logps/ref_rejected": -79.3138427734375, "logps/rejected": -83.96460723876953, "loss": 1.1828, "margin_dpo/margin_mean": 1.6761000156402588, "margin_dpo/margin_std": 3.414187431335449, "step": 94 }, { "KL/chosen_KL_mean": -3.832956314086914, "KL/mean": -4.543184757232666, "KL/rejected_KL_mean": -5.253414154052734, "KL/std": 2.3406383991241455, "epoch": 0.1436130007558579, "fcm_dpo/beta": 0.21277236938476562, "fcm_dpo/delta": 0.10093221068382263, "fcm_dpo/margin": 1.4204583168029785, "fcm_dpo/q_t": 0.4322485625743866, "grad_norm": 59.44841384887695, "learning_rate": 4.974553604702332e-07, "logits/chosen": 0.13966067135334015, "logits/rejected": 0.07558946311473846, "logps/chosen": -64.00144958496094, "logps/ref_chosen": -60.168487548828125, "logps/ref_rejected": -90.73665618896484, "logps/rejected": -95.99006652832031, "loss": 1.198, "margin_dpo/margin_mean": 1.4204577207565308, "margin_dpo/margin_std": 2.919649124145508, "step": 95 }, { "KL/chosen_KL_mean": -3.674968719482422, "KL/mean": -4.511168479919434, "KL/rejected_KL_mean": -5.3473663330078125, "KL/std": 2.884733200073242, "epoch": 0.14512471655328799, "fcm_dpo/beta": 0.21607878804206848, "fcm_dpo/delta": 0.03986040875315666, "fcm_dpo/margin": 1.672404408454895, "fcm_dpo/q_t": 0.42547816038131714, "grad_norm": 57.87812042236328, "learning_rate": 4.972637290166157e-07, "logits/chosen": 0.16100308299064636, "logits/rejected": 0.11939145624637604, "logps/chosen": -64.34375, "logps/ref_chosen": -60.66877746582031, "logps/ref_rejected": -88.30673217773438, "logps/rejected": -93.65409851074219, "loss": 1.2024, "margin_dpo/margin_mean": 1.6724035739898682, "margin_dpo/margin_std": 3.720932722091675, "step": 96 }, { "KL/chosen_KL_mean": -4.3227691650390625, "KL/mean": -4.781896114349365, "KL/rejected_KL_mean": -5.24102783203125, "KL/std": 2.8743977546691895, "epoch": 0.14663643235071808, "fcm_dpo/beta": 0.22158443927764893, "fcm_dpo/delta": 0.20110613107681274, "fcm_dpo/margin": 0.9182647466659546, "fcm_dpo/q_t": 0.45814549922943115, "grad_norm": 86.01407623291016, "learning_rate": 4.970651810649666e-07, "logits/chosen": 0.11800533533096313, "logits/rejected": 0.07440978288650513, "logps/chosen": -69.36688995361328, "logps/ref_chosen": -65.04412078857422, "logps/ref_rejected": -78.42092895507812, "logps/rejected": -83.66195678710938, "loss": 1.3629, "margin_dpo/margin_mean": 0.9182642698287964, "margin_dpo/margin_std": 3.942337989807129, "step": 97 }, { "KL/chosen_KL_mean": -3.8495330810546875, "KL/mean": -4.211657524108887, "KL/rejected_KL_mean": -4.573783874511719, "KL/std": 2.5712876319885254, "epoch": 0.14814814814814814, "fcm_dpo/beta": 0.2320966124534607, "fcm_dpo/delta": 0.23647648096084595, "fcm_dpo/margin": 0.7242507934570312, "fcm_dpo/q_t": 0.4608962833881378, "grad_norm": 72.66352081298828, "learning_rate": 4.968597221690985e-07, "logits/chosen": 0.20253103971481323, "logits/rejected": 0.175521582365036, "logps/chosen": -59.35276412963867, "logps/ref_chosen": -55.503231048583984, "logps/ref_rejected": -72.81553649902344, "logps/rejected": -77.38932800292969, "loss": 1.3091, "margin_dpo/margin_mean": 0.7242506742477417, "margin_dpo/margin_std": 2.5094985961914062, "step": 98 }, { "KL/chosen_KL_mean": -3.8535900115966797, "KL/mean": -4.35680627822876, "KL/rejected_KL_mean": -4.860027313232422, "KL/std": 2.769084930419922, "epoch": 0.14965986394557823, "fcm_dpo/beta": 0.24122494459152222, "fcm_dpo/delta": 0.16129615902900696, "fcm_dpo/margin": 1.0064336061477661, "fcm_dpo/q_t": 0.44870877265930176, "grad_norm": 85.34445190429688, "learning_rate": 4.966473580761389e-07, "logits/chosen": 0.2370331585407257, "logits/rejected": 0.1967862844467163, "logps/chosen": -62.429229736328125, "logps/ref_chosen": -58.57563781738281, "logps/ref_rejected": -78.693603515625, "logps/rejected": -83.55363464355469, "loss": 1.3245, "margin_dpo/margin_mean": 1.006433129310608, "margin_dpo/margin_std": 3.5495963096618652, "step": 99 }, { "KL/chosen_KL_mean": -4.311008453369141, "KL/mean": -4.901121139526367, "KL/rejected_KL_mean": -5.491233825683594, "KL/std": 3.0689799785614014, "epoch": 0.15117157974300832, "fcm_dpo/beta": 0.2495937943458557, "fcm_dpo/delta": 0.10603199899196625, "fcm_dpo/margin": 1.180228352546692, "fcm_dpo/q_t": 0.4410404562950134, "grad_norm": 86.84514617919922, "learning_rate": 4.964280947263676e-07, "logits/chosen": 0.21110297739505768, "logits/rejected": 0.2032468616962433, "logps/chosen": -83.89443969726562, "logps/ref_chosen": -79.58343505859375, "logps/ref_rejected": -92.152587890625, "logps/rejected": -97.64381408691406, "loss": 1.3697, "margin_dpo/margin_mean": 1.1802282333374023, "margin_dpo/margin_std": 4.390231132507324, "step": 100 }, { "KL/chosen_KL_mean": -3.517528533935547, "KL/mean": -4.3843183517456055, "KL/rejected_KL_mean": -5.251110076904297, "KL/std": 2.818021059036255, "epoch": 0.15268329554043839, "fcm_dpo/beta": 0.24841973185539246, "fcm_dpo/delta": -0.03219292312860489, "fcm_dpo/margin": 1.7335816621780396, "fcm_dpo/q_t": 0.40388786792755127, "grad_norm": 62.59505844116211, "learning_rate": 4.96201938253052e-07, "logits/chosen": 0.21590465307235718, "logits/rejected": 0.1761970967054367, "logps/chosen": -55.85031509399414, "logps/ref_chosen": -52.332786560058594, "logps/ref_rejected": -69.55589294433594, "logps/rejected": -74.8070068359375, "loss": 1.1389, "margin_dpo/margin_mean": 1.7335816621780396, "margin_dpo/margin_std": 3.1004440784454346, "step": 101 }, { "KL/chosen_KL_mean": -4.192995071411133, "KL/mean": -4.606147766113281, "KL/rejected_KL_mean": -5.019306182861328, "KL/std": 2.706145763397217, "epoch": 0.15419501133786848, "fcm_dpo/beta": 0.2551325857639313, "fcm_dpo/delta": 0.19242900609970093, "fcm_dpo/margin": 0.8263107538223267, "fcm_dpo/q_t": 0.46237632632255554, "grad_norm": 89.90907287597656, "learning_rate": 4.959688949822748e-07, "logits/chosen": 0.12861177325248718, "logits/rejected": 0.08954156935214996, "logps/chosen": -68.93647766113281, "logps/ref_chosen": -64.74348449707031, "logps/ref_rejected": -69.06132507324219, "logps/rejected": -74.08063507080078, "loss": 1.3995, "margin_dpo/margin_mean": 0.8263108730316162, "margin_dpo/margin_std": 3.9148406982421875, "step": 102 }, { "KL/chosen_KL_mean": -3.756814956665039, "KL/mean": -4.423976898193359, "KL/rejected_KL_mean": -5.0911407470703125, "KL/std": 2.8079519271850586, "epoch": 0.15570672713529857, "fcm_dpo/beta": 0.2602066993713379, "fcm_dpo/delta": 0.053878165781497955, "fcm_dpo/margin": 1.3343229293823242, "fcm_dpo/q_t": 0.4253695607185364, "grad_norm": 76.40733337402344, "learning_rate": 4.957289714327572e-07, "logits/chosen": 0.23914015293121338, "logits/rejected": 0.20504045486450195, "logps/chosen": -67.59346008300781, "logps/ref_chosen": -63.83664321899414, "logps/ref_rejected": -79.32362365722656, "logps/rejected": -84.41476440429688, "loss": 1.1951, "margin_dpo/margin_mean": 1.3343226909637451, "margin_dpo/margin_std": 2.8255200386047363, "step": 103 }, { "KL/chosen_KL_mean": -3.706247329711914, "KL/mean": -4.552823066711426, "KL/rejected_KL_mean": -5.399406433105469, "KL/std": 3.520082950592041, "epoch": 0.15721844293272866, "fcm_dpo/beta": 0.25901108980178833, "fcm_dpo/delta": -0.04045912250876427, "fcm_dpo/margin": 1.6931569576263428, "fcm_dpo/q_t": 0.41955292224884033, "grad_norm": 84.37599182128906, "learning_rate": 4.954821743156767e-07, "logits/chosen": 0.23270666599273682, "logits/rejected": 0.14419865608215332, "logps/chosen": -64.70545959472656, "logps/ref_chosen": -60.99920654296875, "logps/ref_rejected": -98.84645080566406, "logps/rejected": -104.24586486816406, "loss": 1.243, "margin_dpo/margin_mean": 1.6931579113006592, "margin_dpo/margin_std": 4.295382499694824, "step": 104 }, { "KL/chosen_KL_mean": -3.97216796875, "KL/mean": -4.591570854187012, "KL/rejected_KL_mean": -5.210971832275391, "KL/std": 2.5960726737976074, "epoch": 0.15873015873015872, "fcm_dpo/beta": 0.26095521450042725, "fcm_dpo/delta": 0.07911492139101028, "fcm_dpo/margin": 1.238802194595337, "fcm_dpo/q_t": 0.4329737424850464, "grad_norm": 80.13673400878906, "learning_rate": 4.952285105344791e-07, "logits/chosen": 0.15176521241664886, "logits/rejected": 0.09743764251470566, "logps/chosen": -74.92243957519531, "logps/ref_chosen": -70.95027160644531, "logps/ref_rejected": -87.88340759277344, "logps/rejected": -93.0943832397461, "loss": 1.2338, "margin_dpo/margin_mean": 1.2388018369674683, "margin_dpo/margin_std": 3.063159227371216, "step": 105 }, { "KL/chosen_KL_mean": -3.693246841430664, "KL/mean": -4.331310749053955, "KL/rejected_KL_mean": -4.969371795654297, "KL/std": 2.6151795387268066, "epoch": 0.1602418745275888, "fcm_dpo/beta": 0.2641947269439697, "fcm_dpo/delta": 0.06503366678953171, "fcm_dpo/margin": 1.276125192642212, "fcm_dpo/q_t": 0.4248006343841553, "grad_norm": 84.72764587402344, "learning_rate": 4.949679871846857e-07, "logits/chosen": 0.2076093852519989, "logits/rejected": 0.193533256649971, "logps/chosen": -66.15258026123047, "logps/ref_chosen": -62.45933151245117, "logps/ref_rejected": -67.00595092773438, "logps/rejected": -71.9753189086914, "loss": 1.2583, "margin_dpo/margin_mean": 1.276125192642212, "margin_dpo/margin_std": 3.409266710281372, "step": 106 }, { "KL/chosen_KL_mean": -4.450431823730469, "KL/mean": -4.831617832183838, "KL/rejected_KL_mean": -5.212799072265625, "KL/std": 3.1901092529296875, "epoch": 0.1617535903250189, "fcm_dpo/beta": 0.2652558982372284, "fcm_dpo/delta": 0.004532136023044586, "fcm_dpo/margin": 0.7623693943023682, "fcm_dpo/q_t": 0.45172134041786194, "grad_norm": 115.32234191894531, "learning_rate": 4.947006115536947e-07, "logits/chosen": 0.12762564420700073, "logits/rejected": 0.10460235178470612, "logps/chosen": -80.28839874267578, "logps/ref_chosen": -75.83796691894531, "logps/ref_rejected": -87.74038696289062, "logps/rejected": -92.95318603515625, "loss": 1.434, "margin_dpo/margin_mean": 0.7623691558837891, "margin_dpo/margin_std": 3.885159492492676, "step": 107 }, { "KL/chosen_KL_mean": -3.832916259765625, "KL/mean": -4.541226863861084, "KL/rejected_KL_mean": -5.249538421630859, "KL/std": 2.5288538932800293, "epoch": 0.16326530612244897, "fcm_dpo/beta": 0.2664978504180908, "fcm_dpo/delta": 0.02329789660871029, "fcm_dpo/margin": 1.4166216850280762, "fcm_dpo/q_t": 0.42314743995666504, "grad_norm": 75.23424530029297, "learning_rate": 4.944263911205772e-07, "logits/chosen": 0.11665268242359161, "logits/rejected": 0.08691516518592834, "logps/chosen": -72.22615051269531, "logps/ref_chosen": -68.39323425292969, "logps/ref_rejected": -83.24267578125, "logps/rejected": -88.4922103881836, "loss": 1.2263, "margin_dpo/margin_mean": 1.416621446609497, "margin_dpo/margin_std": 3.3912062644958496, "step": 108 }, { "KL/chosen_KL_mean": -3.450235366821289, "KL/mean": -4.449607849121094, "KL/rejected_KL_mean": -5.448982238769531, "KL/std": 2.9472644329071045, "epoch": 0.16477702191987906, "fcm_dpo/beta": 0.26176708936691284, "fcm_dpo/delta": -0.12986820936203003, "fcm_dpo/margin": 1.9987459182739258, "fcm_dpo/q_t": 0.3933427035808563, "grad_norm": 65.35698699951172, "learning_rate": 4.941453335558681e-07, "logits/chosen": 0.15471996366977692, "logits/rejected": 0.09994255751371384, "logps/chosen": -58.97771453857422, "logps/ref_chosen": -55.52748107910156, "logps/ref_rejected": -83.55218505859375, "logps/rejected": -89.00117492675781, "loss": 1.0639, "margin_dpo/margin_mean": 1.9987448453903198, "margin_dpo/margin_std": 3.154825210571289, "step": 109 }, { "KL/chosen_KL_mean": -4.1198272705078125, "KL/mean": -4.419821739196777, "KL/rejected_KL_mean": -4.719810485839844, "KL/std": 2.4698705673217773, "epoch": 0.16628873771730915, "fcm_dpo/beta": 0.26964741945266724, "fcm_dpo/delta": 0.24171972274780273, "fcm_dpo/margin": 0.5999836921691895, "fcm_dpo/q_t": 0.46288520097732544, "grad_norm": 109.37680053710938, "learning_rate": 4.938574467213517e-07, "logits/chosen": 0.13163237273693085, "logits/rejected": 0.14082229137420654, "logps/chosen": -85.27857208251953, "logps/ref_chosen": -81.15874481201172, "logps/ref_rejected": -72.56021118164062, "logps/rejected": -77.28002166748047, "loss": 1.3957, "margin_dpo/margin_mean": 0.5999835729598999, "margin_dpo/margin_std": 3.180210590362549, "step": 110 }, { "KL/chosen_KL_mean": -3.9667015075683594, "KL/mean": -4.667611122131348, "KL/rejected_KL_mean": -5.368522644042969, "KL/std": 2.5095343589782715, "epoch": 0.16780045351473924, "fcm_dpo/beta": 0.27373117208480835, "fcm_dpo/delta": 0.016820773482322693, "fcm_dpo/margin": 1.4018200635910034, "fcm_dpo/q_t": 0.4187893867492676, "grad_norm": 69.44522857666016, "learning_rate": 4.935627386698418e-07, "logits/chosen": 0.24480414390563965, "logits/rejected": 0.20698174834251404, "logps/chosen": -56.32568359375, "logps/ref_chosen": -52.358985900878906, "logps/ref_rejected": -77.06150817871094, "logps/rejected": -82.43002319335938, "loss": 1.1883, "margin_dpo/margin_mean": 1.4018198251724243, "margin_dpo/margin_std": 2.955274820327759, "step": 111 }, { "KL/chosen_KL_mean": -3.5427474975585938, "KL/mean": -4.380864143371582, "KL/rejected_KL_mean": -5.218982696533203, "KL/std": 2.5112671852111816, "epoch": 0.1693121693121693, "fcm_dpo/beta": 0.2712618112564087, "fcm_dpo/delta": -0.05728024244308472, "fcm_dpo/margin": 1.6762354373931885, "fcm_dpo/q_t": 0.39894920587539673, "grad_norm": 79.7928237915039, "learning_rate": 4.932612176449559e-07, "logits/chosen": 0.15141016244888306, "logits/rejected": 0.08643672615289688, "logps/chosen": -66.56281280517578, "logps/ref_chosen": -63.02006530761719, "logps/ref_rejected": -111.36941528320312, "logps/rejected": -116.58839416503906, "loss": 1.1195, "margin_dpo/margin_mean": 1.6762347221374512, "margin_dpo/margin_std": 2.8349223136901855, "step": 112 }, { "KL/chosen_KL_mean": -3.744342803955078, "KL/mean": -4.343097686767578, "KL/rejected_KL_mean": -4.941852569580078, "KL/std": 2.3790249824523926, "epoch": 0.1708238851095994, "fcm_dpo/beta": 0.27505138516426086, "fcm_dpo/delta": 0.07165899127721786, "fcm_dpo/margin": 1.1975092887878418, "fcm_dpo/q_t": 0.42461806535720825, "grad_norm": 87.63443756103516, "learning_rate": 4.929528920808854e-07, "logits/chosen": 0.16694726049900055, "logits/rejected": 0.12760739028453827, "logps/chosen": -59.552005767822266, "logps/ref_chosen": -55.80766296386719, "logps/ref_rejected": -69.84014129638672, "logps/rejected": -74.78199768066406, "loss": 1.2267, "margin_dpo/margin_mean": 1.1975088119506836, "margin_dpo/margin_std": 2.7592105865478516, "step": 113 }, { "KL/chosen_KL_mean": -3.324331283569336, "KL/mean": -4.2867326736450195, "KL/rejected_KL_mean": -5.249141693115234, "KL/std": 2.531193733215332, "epoch": 0.17233560090702948, "fcm_dpo/beta": 0.2680574655532837, "fcm_dpo/delta": -0.12321934103965759, "fcm_dpo/margin": 1.9248015880584717, "fcm_dpo/q_t": 0.3910922408103943, "grad_norm": 68.17572784423828, "learning_rate": 4.92637770602159e-07, "logits/chosen": 0.2011614441871643, "logits/rejected": 0.14068567752838135, "logps/chosen": -69.6571044921875, "logps/ref_chosen": -66.33277130126953, "logps/ref_rejected": -71.61489868164062, "logps/rejected": -76.8640365600586, "loss": 1.0673, "margin_dpo/margin_mean": 1.9248019456863403, "margin_dpo/margin_std": 2.92665958404541, "step": 114 }, { "KL/chosen_KL_mean": -3.7208175659179688, "KL/mean": -4.421115875244141, "KL/rejected_KL_mean": -5.121416091918945, "KL/std": 2.2531988620758057, "epoch": 0.17384731670445955, "fcm_dpo/beta": 0.2685314416885376, "fcm_dpo/delta": 0.024830270558595657, "fcm_dpo/margin": 1.4005990028381348, "fcm_dpo/q_t": 0.4187403619289398, "grad_norm": 74.76184844970703, "learning_rate": 4.923158620234019e-07, "logits/chosen": 0.18886233866214752, "logits/rejected": 0.12659035623073578, "logps/chosen": -59.46985626220703, "logps/ref_chosen": -55.74903869628906, "logps/ref_rejected": -79.59849548339844, "logps/rejected": -84.71990966796875, "loss": 1.1603, "margin_dpo/margin_mean": 1.4005998373031616, "margin_dpo/margin_std": 2.6480507850646973, "step": 115 }, { "KL/chosen_KL_mean": -3.4787425994873047, "KL/mean": -4.41444206237793, "KL/rejected_KL_mean": -5.350147247314453, "KL/std": 2.367363452911377, "epoch": 0.17535903250188964, "fcm_dpo/beta": 0.2657136917114258, "fcm_dpo/delta": -0.10223374515771866, "fcm_dpo/margin": 1.8714066743850708, "fcm_dpo/q_t": 0.38647329807281494, "grad_norm": 58.41557312011719, "learning_rate": 4.91987175349089e-07, "logits/chosen": 0.19088421761989594, "logits/rejected": 0.12268730252981186, "logps/chosen": -52.843910217285156, "logps/ref_chosen": -49.36516571044922, "logps/ref_rejected": -72.84671020507812, "logps/rejected": -78.19685363769531, "loss": 1.0349, "margin_dpo/margin_mean": 1.8714067935943604, "margin_dpo/margin_std": 2.270500659942627, "step": 116 }, { "KL/chosen_KL_mean": -3.357513427734375, "KL/mean": -4.0659332275390625, "KL/rejected_KL_mean": -4.77435302734375, "KL/std": 2.205228328704834, "epoch": 0.17687074829931973, "fcm_dpo/beta": 0.261690229177475, "fcm_dpo/delta": 0.02821500599384308, "fcm_dpo/margin": 1.4168295860290527, "fcm_dpo/q_t": 0.41594117879867554, "grad_norm": 67.27315521240234, "learning_rate": 4.916517197732933e-07, "logits/chosen": 0.16953285038471222, "logits/rejected": 0.1338912546634674, "logps/chosen": -61.06841278076172, "logps/ref_chosen": -57.710899353027344, "logps/ref_rejected": -69.77253723144531, "logps/rejected": -74.54689025878906, "loss": 1.1649, "margin_dpo/margin_mean": 1.4168298244476318, "margin_dpo/margin_std": 2.5413122177124023, "step": 117 }, { "KL/chosen_KL_mean": -3.2425918579101562, "KL/mean": -4.178920745849609, "KL/rejected_KL_mean": -5.115245819091797, "KL/std": 2.3707337379455566, "epoch": 0.17838246409674982, "fcm_dpo/beta": 0.2604265809059143, "fcm_dpo/delta": -0.09273022413253784, "fcm_dpo/margin": 1.8726553916931152, "fcm_dpo/q_t": 0.3897615075111389, "grad_norm": 65.98107147216797, "learning_rate": 4.913095046794281e-07, "logits/chosen": 0.24106436967849731, "logits/rejected": 0.20028874278068542, "logps/chosen": -55.72248840332031, "logps/ref_chosen": -52.479896545410156, "logps/ref_rejected": -81.359130859375, "logps/rejected": -86.47437286376953, "loss": 1.0587, "margin_dpo/margin_mean": 1.872654914855957, "margin_dpo/margin_std": 2.531564712524414, "step": 118 }, { "KL/chosen_KL_mean": -4.147886276245117, "KL/mean": -4.786754608154297, "KL/rejected_KL_mean": -5.425624847412109, "KL/std": 2.504222869873047, "epoch": 0.17989417989417988, "fcm_dpo/beta": 0.26322388648986816, "fcm_dpo/delta": 0.06562946736812592, "fcm_dpo/margin": 1.2777378559112549, "fcm_dpo/q_t": 0.4259907603263855, "grad_norm": 72.88529968261719, "learning_rate": 4.909605396399855e-07, "logits/chosen": 0.16665619611740112, "logits/rejected": 0.12804852426052094, "logps/chosen": -65.50556182861328, "logps/ref_chosen": -61.35767364501953, "logps/ref_rejected": -75.71510314941406, "logps/rejected": -81.1407241821289, "loss": 1.2262, "margin_dpo/margin_mean": 1.2777388095855713, "margin_dpo/margin_std": 3.0291521549224854, "step": 119 }, { "KL/chosen_KL_mean": -3.4492645263671875, "KL/mean": -4.388361930847168, "KL/rejected_KL_mean": -5.327457427978516, "KL/std": 2.3779149055480957, "epoch": 0.18140589569160998, "fcm_dpo/beta": 0.2602246403694153, "fcm_dpo/delta": -0.09322664141654968, "fcm_dpo/margin": 1.8781991004943848, "fcm_dpo/q_t": 0.39164602756500244, "grad_norm": 64.43611907958984, "learning_rate": 4.906048344162676e-07, "logits/chosen": 0.17029550671577454, "logits/rejected": 0.11013634502887726, "logps/chosen": -63.356834411621094, "logps/ref_chosen": -59.907569885253906, "logps/ref_rejected": -79.6910629272461, "logps/rejected": -85.01852416992188, "loss": 1.0494, "margin_dpo/margin_mean": 1.8781991004943848, "margin_dpo/margin_std": 2.4651622772216797, "step": 120 }, { "KL/chosen_KL_mean": -3.789203643798828, "KL/mean": -4.682255268096924, "KL/rejected_KL_mean": -5.575309753417969, "KL/std": 2.5316545963287354, "epoch": 0.18291761148904007, "fcm_dpo/beta": 0.2558312714099884, "fcm_dpo/delta": -0.05986927077174187, "fcm_dpo/margin": 1.786109209060669, "fcm_dpo/q_t": 0.401405930519104, "grad_norm": 59.51321792602539, "learning_rate": 4.902423989581143e-07, "logits/chosen": 0.2673395872116089, "logits/rejected": 0.18181562423706055, "logps/chosen": -59.45524597167969, "logps/ref_chosen": -55.66604232788086, "logps/ref_rejected": -101.56233978271484, "logps/rejected": -107.13764953613281, "loss": 1.0761, "margin_dpo/margin_mean": 1.7861084938049316, "margin_dpo/margin_std": 2.6160924434661865, "step": 121 }, { "KL/chosen_KL_mean": -3.9761314392089844, "KL/mean": -4.964360237121582, "KL/rejected_KL_mean": -5.9525909423828125, "KL/std": 2.7434768676757812, "epoch": 0.18442932728647016, "fcm_dpo/beta": 0.25225332379341125, "fcm_dpo/delta": -0.10360904037952423, "fcm_dpo/margin": 1.976454734802246, "fcm_dpo/q_t": 0.394494891166687, "grad_norm": 70.2988510131836, "learning_rate": 4.898732434036243e-07, "logits/chosen": 0.17670899629592896, "logits/rejected": 0.1416880190372467, "logps/chosen": -67.31050109863281, "logps/ref_chosen": -63.334373474121094, "logps/ref_rejected": -73.67523193359375, "logps/rejected": -79.62782287597656, "loss": 1.0731, "margin_dpo/margin_mean": 1.9764546155929565, "margin_dpo/margin_std": 3.0363006591796875, "step": 122 }, { "KL/chosen_KL_mean": -3.6747913360595703, "KL/mean": -4.590383529663086, "KL/rejected_KL_mean": -5.505973815917969, "KL/std": 2.4913928508758545, "epoch": 0.18594104308390022, "fcm_dpo/beta": 0.2495909333229065, "fcm_dpo/delta": -0.0601581409573555, "fcm_dpo/margin": 1.8311854600906372, "fcm_dpo/q_t": 0.39790278673171997, "grad_norm": 61.7016487121582, "learning_rate": 4.894973780788722e-07, "logits/chosen": 0.18975844979286194, "logits/rejected": 0.14888188242912292, "logps/chosen": -60.57353210449219, "logps/ref_chosen": -56.89874267578125, "logps/ref_rejected": -78.97028350830078, "logps/rejected": -84.47625732421875, "loss": 1.0911, "margin_dpo/margin_mean": 1.8311853408813477, "margin_dpo/margin_std": 2.7468996047973633, "step": 123 }, { "KL/chosen_KL_mean": -3.9783401489257812, "KL/mean": -5.216902256011963, "KL/rejected_KL_mean": -6.455467224121094, "KL/std": 2.7113471031188965, "epoch": 0.1874527588813303, "fcm_dpo/beta": 0.2392091453075409, "fcm_dpo/delta": -0.20544171333312988, "fcm_dpo/margin": 2.4771270751953125, "fcm_dpo/q_t": 0.3674274981021881, "grad_norm": 53.000423431396484, "learning_rate": 4.89114813497619e-07, "logits/chosen": 0.21945105493068695, "logits/rejected": 0.15901131927967072, "logps/chosen": -61.094425201416016, "logps/ref_chosen": -57.116085052490234, "logps/ref_rejected": -87.93074035644531, "logps/rejected": -94.38621520996094, "loss": 0.9766, "margin_dpo/margin_mean": 2.4771275520324707, "margin_dpo/margin_std": 2.7539939880371094, "step": 124 }, { "KL/chosen_KL_mean": -4.249895095825195, "KL/mean": -5.264259338378906, "KL/rejected_KL_mean": -6.278621673583984, "KL/std": 2.9378597736358643, "epoch": 0.1889644746787604, "fcm_dpo/beta": 0.2334146797657013, "fcm_dpo/delta": -0.07813793420791626, "fcm_dpo/margin": 2.028723955154419, "fcm_dpo/q_t": 0.39546871185302734, "grad_norm": 61.0987548828125, "learning_rate": 4.887255603610184e-07, "logits/chosen": 0.23440095782279968, "logits/rejected": 0.17516063153743744, "logps/chosen": -69.95606994628906, "logps/ref_chosen": -65.7061767578125, "logps/ref_rejected": -91.72711944580078, "logps/rejected": -98.0057373046875, "loss": 1.0701, "margin_dpo/margin_mean": 2.0287246704101562, "margin_dpo/margin_std": 2.8646774291992188, "step": 125 }, { "KL/chosen_KL_mean": -3.9711647033691406, "KL/mean": -4.703935623168945, "KL/rejected_KL_mean": -5.436708450317383, "KL/std": 3.10610032081604, "epoch": 0.19047619047619047, "fcm_dpo/beta": 0.23558691143989563, "fcm_dpo/delta": 0.05671154335141182, "fcm_dpo/margin": 1.4655463695526123, "fcm_dpo/q_t": 0.4278152883052826, "grad_norm": 60.16349792480469, "learning_rate": 4.883296295573176e-07, "logits/chosen": 0.04985351115465164, "logits/rejected": 0.0437270849943161, "logps/chosen": -72.14725494384766, "logps/ref_chosen": -68.17608642578125, "logps/ref_rejected": -65.1175537109375, "logps/rejected": -70.55426025390625, "loss": 1.2067, "margin_dpo/margin_mean": 1.4655463695526123, "margin_dpo/margin_std": 3.381613254547119, "step": 126 }, { "KL/chosen_KL_mean": -4.2151336669921875, "KL/mean": -5.367134094238281, "KL/rejected_KL_mean": -6.519136428833008, "KL/std": 2.6317625045776367, "epoch": 0.19198790627362056, "fcm_dpo/beta": 0.2297043651342392, "fcm_dpo/delta": -0.13883370161056519, "fcm_dpo/margin": 2.3039979934692383, "fcm_dpo/q_t": 0.3804101347923279, "grad_norm": 55.27705001831055, "learning_rate": 4.87927032161552e-07, "logits/chosen": 0.15548180043697357, "logits/rejected": 0.12521973252296448, "logps/chosen": -66.09536743164062, "logps/ref_chosen": -61.88023376464844, "logps/ref_rejected": -68.46012878417969, "logps/rejected": -74.97926330566406, "loss": 1.0161, "margin_dpo/margin_mean": 2.3039979934692383, "margin_dpo/margin_std": 2.670165538787842, "step": 127 }, { "KL/chosen_KL_mean": -4.582956314086914, "KL/mean": -5.595683574676514, "KL/rejected_KL_mean": -6.6084136962890625, "KL/std": 3.187983512878418, "epoch": 0.19349962207105065, "fcm_dpo/beta": 0.22704312205314636, "fcm_dpo/delta": -0.06312703341245651, "fcm_dpo/margin": 2.02545166015625, "fcm_dpo/q_t": 0.40157029032707214, "grad_norm": 59.90847396850586, "learning_rate": 4.875177794352363e-07, "logits/chosen": 0.2170131355524063, "logits/rejected": 0.15855154395103455, "logps/chosen": -71.29193878173828, "logps/ref_chosen": -66.708984375, "logps/ref_rejected": -94.97969055175781, "logps/rejected": -101.58810424804688, "loss": 1.1385, "margin_dpo/margin_mean": 2.025451421737671, "margin_dpo/margin_std": 3.7224786281585693, "step": 128 }, { "KL/chosen_KL_mean": -4.919008255004883, "KL/mean": -5.729151248931885, "KL/rejected_KL_mean": -6.539299011230469, "KL/std": 2.8172764778137207, "epoch": 0.19501133786848074, "fcm_dpo/beta": 0.2280907928943634, "fcm_dpo/delta": 0.031591035425662994, "fcm_dpo/margin": 1.6202887296676636, "fcm_dpo/q_t": 0.4201734662055969, "grad_norm": 67.37427520751953, "learning_rate": 4.871018828260491e-07, "logits/chosen": 0.1873515546321869, "logits/rejected": 0.18076498806476593, "logps/chosen": -70.2578353881836, "logps/ref_chosen": -65.33882904052734, "logps/ref_rejected": -68.06109619140625, "logps/rejected": -74.60038757324219, "loss": 1.1711, "margin_dpo/margin_mean": 1.6202887296676636, "margin_dpo/margin_std": 3.2078969478607178, "step": 129 }, { "KL/chosen_KL_mean": -4.648164749145508, "KL/mean": -5.663567543029785, "KL/rejected_KL_mean": -6.6789703369140625, "KL/std": 2.797410011291504, "epoch": 0.1965230536659108, "fcm_dpo/beta": 0.22417128086090088, "fcm_dpo/delta": -0.06046818196773529, "fcm_dpo/margin": 2.0308070182800293, "fcm_dpo/q_t": 0.39721450209617615, "grad_norm": 64.58387756347656, "learning_rate": 4.866793539675126e-07, "logits/chosen": 0.13751313090324402, "logits/rejected": 0.08786692470312119, "logps/chosen": -63.30891036987305, "logps/ref_chosen": -58.660743713378906, "logps/ref_rejected": -79.24510192871094, "logps/rejected": -85.924072265625, "loss": 1.0734, "margin_dpo/margin_mean": 2.0308070182800293, "margin_dpo/margin_std": 2.682551383972168, "step": 130 }, { "KL/chosen_KL_mean": -4.569456100463867, "KL/mean": -5.825868606567383, "KL/rejected_KL_mean": -7.082279205322266, "KL/std": 3.4104576110839844, "epoch": 0.1980347694633409, "fcm_dpo/beta": 0.2204248309135437, "fcm_dpo/delta": -0.16282838582992554, "fcm_dpo/margin": 2.5128283500671387, "fcm_dpo/q_t": 0.38101017475128174, "grad_norm": 53.8778190612793, "learning_rate": 4.86250204678667e-07, "logits/chosen": 0.1606750786304474, "logits/rejected": 0.09683164209127426, "logps/chosen": -57.08399200439453, "logps/ref_chosen": -52.51453399658203, "logps/ref_rejected": -85.18299865722656, "logps/rejected": -92.26527404785156, "loss": 1.0654, "margin_dpo/margin_mean": 2.5128278732299805, "margin_dpo/margin_std": 3.8156890869140625, "step": 131 }, { "KL/chosen_KL_mean": -4.946624755859375, "KL/mean": -6.04380464553833, "KL/rejected_KL_mean": -7.140985488891602, "KL/std": 3.199896812438965, "epoch": 0.19954648526077098, "fcm_dpo/beta": 0.215684711933136, "fcm_dpo/delta": -0.07706592977046967, "fcm_dpo/margin": 2.194359064102173, "fcm_dpo/q_t": 0.39275842905044556, "grad_norm": 56.11330795288086, "learning_rate": 4.858144469637408e-07, "logits/chosen": 0.2194550335407257, "logits/rejected": 0.18725933134555817, "logps/chosen": -70.63175964355469, "logps/ref_chosen": -65.68513488769531, "logps/ref_rejected": -69.54120635986328, "logps/rejected": -76.68218994140625, "loss": 1.0823, "margin_dpo/margin_mean": 2.1943588256835938, "margin_dpo/margin_std": 3.272285223007202, "step": 132 }, { "KL/chosen_KL_mean": -4.965156555175781, "KL/mean": -6.00357723236084, "KL/rejected_KL_mean": -7.041996002197266, "KL/std": 3.207258701324463, "epoch": 0.20105820105820105, "fcm_dpo/beta": 0.21359238028526306, "fcm_dpo/delta": -0.045637097209692, "fcm_dpo/margin": 2.076840400695801, "fcm_dpo/q_t": 0.40116482973098755, "grad_norm": 58.3845100402832, "learning_rate": 4.853720930118138e-07, "logits/chosen": 0.15236756205558777, "logits/rejected": 0.14272542297840118, "logps/chosen": -68.56327056884766, "logps/ref_chosen": -63.598114013671875, "logps/ref_rejected": -73.72798156738281, "logps/rejected": -80.76997375488281, "loss": 1.1045, "margin_dpo/margin_mean": 2.0768399238586426, "margin_dpo/margin_std": 3.3041024208068848, "step": 133 }, { "KL/chosen_KL_mean": -4.748861312866211, "KL/mean": -6.207903861999512, "KL/rejected_KL_mean": -7.6669464111328125, "KL/std": 3.6748008728027344, "epoch": 0.20256991685563114, "fcm_dpo/beta": 0.20595017075538635, "fcm_dpo/delta": -0.21452711522579193, "fcm_dpo/margin": 2.918086290359497, "fcm_dpo/q_t": 0.36868858337402344, "grad_norm": 47.815277099609375, "learning_rate": 4.849231551964771e-07, "logits/chosen": 0.2514156401157379, "logits/rejected": 0.19562683999538422, "logps/chosen": -58.543434143066406, "logps/ref_chosen": -53.79457092285156, "logps/ref_rejected": -74.16741943359375, "logps/rejected": -81.83436584472656, "loss": 0.9909, "margin_dpo/margin_mean": 2.9180867671966553, "margin_dpo/margin_std": 3.5976271629333496, "step": 134 }, { "KL/chosen_KL_mean": -4.821933746337891, "KL/mean": -5.825990676879883, "KL/rejected_KL_mean": -6.830051422119141, "KL/std": 3.4957146644592285, "epoch": 0.20408163265306123, "fcm_dpo/beta": 0.2038605511188507, "fcm_dpo/delta": -0.00977998971939087, "fcm_dpo/margin": 2.0081186294555664, "fcm_dpo/q_t": 0.41021621227264404, "grad_norm": 47.532005310058594, "learning_rate": 4.844676460754862e-07, "logits/chosen": 0.1996041089296341, "logits/rejected": 0.16683346033096313, "logps/chosen": -54.26300811767578, "logps/ref_chosen": -49.441078186035156, "logps/ref_rejected": -65.96878051757812, "logps/rejected": -72.798828125, "loss": 1.1397, "margin_dpo/margin_mean": 2.008118152618408, "margin_dpo/margin_std": 3.5981740951538086, "step": 135 }, { "KL/chosen_KL_mean": -5.761157989501953, "KL/mean": -6.998971939086914, "KL/rejected_KL_mean": -8.23678970336914, "KL/std": 3.8275370597839355, "epoch": 0.20559334845049132, "fcm_dpo/beta": 0.20050784945487976, "fcm_dpo/delta": -0.10157489776611328, "fcm_dpo/margin": 2.475637912750244, "fcm_dpo/q_t": 0.40083396434783936, "grad_norm": 59.76097869873047, "learning_rate": 4.840055783904106e-07, "logits/chosen": 0.1858748495578766, "logits/rejected": 0.1138669028878212, "logps/chosen": -72.52041625976562, "logps/ref_chosen": -66.75926208496094, "logps/ref_rejected": -94.61787414550781, "logps/rejected": -102.85466003417969, "loss": 1.1558, "margin_dpo/margin_mean": 2.4756383895874023, "margin_dpo/margin_std": 4.930904388427734, "step": 136 }, { "KL/chosen_KL_mean": -5.439434051513672, "KL/mean": -6.702852249145508, "KL/rejected_KL_mean": -7.966270446777344, "KL/std": 3.578794002532959, "epoch": 0.20710506424792138, "fcm_dpo/beta": 0.19725792109966278, "fcm_dpo/delta": -0.10347731411457062, "fcm_dpo/margin": 2.5268397331237793, "fcm_dpo/q_t": 0.3930894136428833, "grad_norm": 47.14102554321289, "learning_rate": 4.835369650662767e-07, "logits/chosen": 0.20675092935562134, "logits/rejected": 0.17818483710289001, "logps/chosen": -62.22323226928711, "logps/ref_chosen": -56.78379821777344, "logps/ref_rejected": -69.89952087402344, "logps/rejected": -77.86579132080078, "loss": 1.0813, "margin_dpo/margin_mean": 2.5268397331237793, "margin_dpo/margin_std": 3.943085193634033, "step": 137 }, { "KL/chosen_KL_mean": -6.369855880737305, "KL/mean": -7.403163909912109, "KL/rejected_KL_mean": -8.436473846435547, "KL/std": 3.7773261070251465, "epoch": 0.20861678004535147, "fcm_dpo/beta": 0.19536003470420837, "fcm_dpo/delta": -0.0039053638465702534, "fcm_dpo/margin": 2.066615104675293, "fcm_dpo/q_t": 0.41097143292427063, "grad_norm": 50.57745361328125, "learning_rate": 4.830618192112065e-07, "logits/chosen": 0.20486241579055786, "logits/rejected": 0.16861580312252045, "logps/chosen": -65.1358642578125, "logps/ref_chosen": -58.766014099121094, "logps/ref_rejected": -68.12371826171875, "logps/rejected": -76.56019592285156, "loss": 1.1632, "margin_dpo/margin_mean": 2.066614866256714, "margin_dpo/margin_std": 4.009616374969482, "step": 138 }, { "KL/chosen_KL_mean": -6.473136901855469, "KL/mean": -7.848593235015869, "KL/rejected_KL_mean": -9.224044799804688, "KL/std": 3.7480998039245605, "epoch": 0.21012849584278157, "fcm_dpo/beta": 0.19235269725322723, "fcm_dpo/delta": -0.13614793121814728, "fcm_dpo/margin": 2.7509093284606934, "fcm_dpo/q_t": 0.38339757919311523, "grad_norm": 51.38783264160156, "learning_rate": 4.825801541160509e-07, "logits/chosen": 0.15835514664649963, "logits/rejected": 0.12959414720535278, "logps/chosen": -77.69873046875, "logps/ref_chosen": -71.2255859375, "logps/ref_rejected": -82.1834716796875, "logps/rejected": -91.40751647949219, "loss": 1.0374, "margin_dpo/margin_mean": 2.7509093284606934, "margin_dpo/margin_std": 3.6548454761505127, "step": 139 }, { "KL/chosen_KL_mean": -6.044340133666992, "KL/mean": -7.782115459442139, "KL/rejected_KL_mean": -9.519889831542969, "KL/std": 4.871986389160156, "epoch": 0.21164021164021163, "fcm_dpo/beta": 0.1829332858324051, "fcm_dpo/delta": -0.2522822618484497, "fcm_dpo/margin": 3.4755468368530273, "fcm_dpo/q_t": 0.36386656761169434, "grad_norm": 53.137977600097656, "learning_rate": 4.820919832540181e-07, "logits/chosen": 0.16880294680595398, "logits/rejected": 0.12310120463371277, "logps/chosen": -69.32200622558594, "logps/ref_chosen": -63.27766418457031, "logps/ref_rejected": -83.30647277832031, "logps/rejected": -92.82636260986328, "loss": 1.0477, "margin_dpo/margin_mean": 3.4755468368530273, "margin_dpo/margin_std": 5.251531600952148, "step": 140 }, { "KL/chosen_KL_mean": -6.465095520019531, "KL/mean": -8.143421173095703, "KL/rejected_KL_mean": -9.821746826171875, "KL/std": 4.110562324523926, "epoch": 0.21315192743764172, "fcm_dpo/beta": 0.17414629459381104, "fcm_dpo/delta": -0.199050173163414, "fcm_dpo/margin": 3.3566455841064453, "fcm_dpo/q_t": 0.37942415475845337, "grad_norm": 47.3781852722168, "learning_rate": 4.815973202802966e-07, "logits/chosen": 0.2222299873828888, "logits/rejected": 0.17720073461532593, "logps/chosen": -68.23186492919922, "logps/ref_chosen": -61.76676940917969, "logps/ref_rejected": -88.60601806640625, "logps/rejected": -98.42776489257812, "loss": 1.0593, "margin_dpo/margin_mean": 3.356645345687866, "margin_dpo/margin_std": 5.109691619873047, "step": 141 }, { "KL/chosen_KL_mean": -6.469383239746094, "KL/mean": -7.785819053649902, "KL/rejected_KL_mean": -9.102256774902344, "KL/std": 4.037965774536133, "epoch": 0.2146636432350718, "fcm_dpo/beta": 0.17250174283981323, "fcm_dpo/delta": -0.05671734735369682, "fcm_dpo/margin": 2.6328725814819336, "fcm_dpo/q_t": 0.3999059796333313, "grad_norm": 45.11027526855469, "learning_rate": 4.810961790316729e-07, "logits/chosen": 0.231331467628479, "logits/rejected": 0.20561900734901428, "logps/chosen": -71.74415588378906, "logps/ref_chosen": -65.2747802734375, "logps/ref_rejected": -81.1378173828125, "logps/rejected": -90.24007415771484, "loss": 1.0967, "margin_dpo/margin_mean": 2.6328725814819336, "margin_dpo/margin_std": 4.121742248535156, "step": 142 }, { "KL/chosen_KL_mean": -6.877973556518555, "KL/mean": -8.075854301452637, "KL/rejected_KL_mean": -9.273731231689453, "KL/std": 4.413012981414795, "epoch": 0.2161753590325019, "fcm_dpo/beta": 0.17143571376800537, "fcm_dpo/delta": -0.011194366030395031, "fcm_dpo/margin": 2.3957576751708984, "fcm_dpo/q_t": 0.4077395796775818, "grad_norm": 55.14004135131836, "learning_rate": 4.805885735261454e-07, "logits/chosen": 0.21375364065170288, "logits/rejected": 0.1969037652015686, "logps/chosen": -69.49580383300781, "logps/ref_chosen": -62.617828369140625, "logps/ref_rejected": -70.39239501953125, "logps/rejected": -79.66612243652344, "loss": 1.19, "margin_dpo/margin_mean": 2.3957576751708984, "margin_dpo/margin_std": 5.123311996459961, "step": 143 }, { "KL/chosen_KL_mean": -7.226917266845703, "KL/mean": -8.757612228393555, "KL/rejected_KL_mean": -10.28830337524414, "KL/std": 4.669466972351074, "epoch": 0.21768707482993196, "fcm_dpo/beta": 0.16772940754890442, "fcm_dpo/delta": -0.12045253813266754, "fcm_dpo/margin": 3.0613865852355957, "fcm_dpo/q_t": 0.39098232984542847, "grad_norm": 46.48661422729492, "learning_rate": 4.800745179625307e-07, "logits/chosen": 0.21085073053836823, "logits/rejected": 0.18012139201164246, "logps/chosen": -68.02960205078125, "logps/ref_chosen": -60.80268859863281, "logps/ref_rejected": -79.07284545898438, "logps/rejected": -89.36114501953125, "loss": 1.1012, "margin_dpo/margin_mean": 3.061386823654175, "margin_dpo/margin_std": 5.1614203453063965, "step": 144 }, { "KL/chosen_KL_mean": -6.772495269775391, "KL/mean": -8.340176582336426, "KL/rejected_KL_mean": -9.907852172851562, "KL/std": 4.697710037231445, "epoch": 0.21919879062736206, "fcm_dpo/beta": 0.16429109871387482, "fcm_dpo/delta": -0.1214357241988182, "fcm_dpo/margin": 3.1353633403778076, "fcm_dpo/q_t": 0.3892659544944763, "grad_norm": 49.442745208740234, "learning_rate": 4.795540267200686e-07, "logits/chosen": 0.15911483764648438, "logits/rejected": 0.1758701205253601, "logps/chosen": -81.38395690917969, "logps/ref_chosen": -74.61146545410156, "logps/ref_rejected": -83.24461364746094, "logps/rejected": -93.1524658203125, "loss": 1.103, "margin_dpo/margin_mean": 3.135362148284912, "margin_dpo/margin_std": 5.358179092407227, "step": 145 }, { "KL/chosen_KL_mean": -6.513679504394531, "KL/mean": -8.18224048614502, "KL/rejected_KL_mean": -9.85080337524414, "KL/std": 4.615451812744141, "epoch": 0.22071050642479215, "fcm_dpo/beta": 0.16036675870418549, "fcm_dpo/delta": -0.14257195591926575, "fcm_dpo/margin": 3.3371238708496094, "fcm_dpo/q_t": 0.38525545597076416, "grad_norm": 42.53998565673828, "learning_rate": 4.790271143580173e-07, "logits/chosen": 0.16977742314338684, "logits/rejected": 0.15491747856140137, "logps/chosen": -64.35466003417969, "logps/ref_chosen": -57.84098434448242, "logps/ref_rejected": -67.47422790527344, "logps/rejected": -77.32502746582031, "loss": 1.0618, "margin_dpo/margin_mean": 3.3371238708496094, "margin_dpo/margin_std": 4.9862470626831055, "step": 146 }, { "KL/chosen_KL_mean": -7.826589584350586, "KL/mean": -9.213622093200684, "KL/rejected_KL_mean": -10.600662231445312, "KL/std": 5.016104698181152, "epoch": 0.2222222222222222, "fcm_dpo/beta": 0.15783822536468506, "fcm_dpo/delta": -0.03958672657608986, "fcm_dpo/margin": 2.7740743160247803, "fcm_dpo/q_t": 0.4058781862258911, "grad_norm": 52.86528778076172, "learning_rate": 4.784937956152489e-07, "logits/chosen": 0.1791575849056244, "logits/rejected": 0.13689681887626648, "logps/chosen": -74.64006042480469, "logps/ref_chosen": -66.81346893310547, "logps/ref_rejected": -81.1796875, "logps/rejected": -91.78034973144531, "loss": 1.1557, "margin_dpo/margin_mean": 2.774074077606201, "margin_dpo/margin_std": 5.396799564361572, "step": 147 }, { "KL/chosen_KL_mean": -7.366302490234375, "KL/mean": -9.069774627685547, "KL/rejected_KL_mean": -10.773239135742188, "KL/std": 4.651217460632324, "epoch": 0.2237339380196523, "fcm_dpo/beta": 0.15506836771965027, "fcm_dpo/delta": -0.13526862859725952, "fcm_dpo/margin": 3.40694260597229, "fcm_dpo/q_t": 0.38584333658218384, "grad_norm": 35.693355560302734, "learning_rate": 4.779540854098347e-07, "logits/chosen": 0.3457987308502197, "logits/rejected": 0.27079546451568604, "logps/chosen": -56.05405807495117, "logps/ref_chosen": -48.6877555847168, "logps/ref_rejected": -67.50503540039062, "logps/rejected": -78.27827453613281, "loss": 1.0782, "margin_dpo/margin_mean": 3.406942367553711, "margin_dpo/margin_std": 5.257372856140137, "step": 148 }, { "KL/chosen_KL_mean": -9.045269012451172, "KL/mean": -11.011503219604492, "KL/rejected_KL_mean": -12.977741241455078, "KL/std": 5.339559555053711, "epoch": 0.2252456538170824, "fcm_dpo/beta": 0.1490006297826767, "fcm_dpo/delta": -0.1973770558834076, "fcm_dpo/margin": 3.9324727058410645, "fcm_dpo/q_t": 0.3704410195350647, "grad_norm": 36.19922637939453, "learning_rate": 4.774079988386296e-07, "logits/chosen": 0.1623860001564026, "logits/rejected": 0.11575397104024887, "logps/chosen": -64.18904876708984, "logps/ref_chosen": -55.143775939941406, "logps/ref_rejected": -64.79888916015625, "logps/rejected": -77.7766342163086, "loss": 1.0031, "margin_dpo/margin_mean": 3.9324722290039062, "margin_dpo/margin_std": 4.933730125427246, "step": 149 }, { "KL/chosen_KL_mean": -7.397439956665039, "KL/mean": -9.875222206115723, "KL/rejected_KL_mean": -12.35300064086914, "KL/std": 5.53492546081543, "epoch": 0.22675736961451248, "fcm_dpo/beta": 0.14036396145820618, "fcm_dpo/delta": -0.31827008724212646, "fcm_dpo/margin": 4.955565929412842, "fcm_dpo/q_t": 0.3498973846435547, "grad_norm": 34.423675537109375, "learning_rate": 4.768555511768486e-07, "logits/chosen": 0.21100425720214844, "logits/rejected": 0.1672147810459137, "logps/chosen": -74.8681869506836, "logps/ref_chosen": -67.47074890136719, "logps/ref_rejected": -89.21170806884766, "logps/rejected": -101.56471252441406, "loss": 0.9365, "margin_dpo/margin_mean": 4.95556640625, "margin_dpo/margin_std": 5.431443691253662, "step": 150 }, { "KL/chosen_KL_mean": -6.885496139526367, "KL/mean": -9.645720481872559, "KL/rejected_KL_mean": -12.405941009521484, "KL/std": 5.8176798820495605, "epoch": 0.22826908541194255, "fcm_dpo/beta": 0.1314392387866974, "fcm_dpo/delta": -0.3507463037967682, "fcm_dpo/margin": 5.520442962646484, "fcm_dpo/q_t": 0.3430783152580261, "grad_norm": 30.11147689819336, "learning_rate": 4.762967578776406e-07, "logits/chosen": 0.2235882431268692, "logits/rejected": 0.16875801980495453, "logps/chosen": -59.34503936767578, "logps/ref_chosen": -52.45954132080078, "logps/ref_rejected": -79.0630111694336, "logps/rejected": -91.46894836425781, "loss": 0.9129, "margin_dpo/margin_mean": 5.520442962646484, "margin_dpo/margin_std": 5.8820390701293945, "step": 151 }, { "KL/chosen_KL_mean": -8.967247009277344, "KL/mean": -10.84051513671875, "KL/rejected_KL_mean": -12.713790893554688, "KL/std": 5.6841278076171875, "epoch": 0.22978080120937264, "fcm_dpo/beta": 0.1281639039516449, "fcm_dpo/delta": -0.08456161618232727, "fcm_dpo/margin": 3.746544361114502, "fcm_dpo/q_t": 0.3974611461162567, "grad_norm": 35.68574905395508, "learning_rate": 4.757316345716553e-07, "logits/chosen": 0.25782787799835205, "logits/rejected": 0.20684176683425903, "logps/chosen": -65.52108001708984, "logps/ref_chosen": -56.5538330078125, "logps/ref_rejected": -76.55074310302734, "logps/rejected": -89.26453399658203, "loss": 1.1185, "margin_dpo/margin_mean": 3.746544122695923, "margin_dpo/margin_std": 6.411558151245117, "step": 152 }, { "KL/chosen_KL_mean": -8.4833984375, "KL/mean": -10.582864761352539, "KL/rejected_KL_mean": -12.682331085205078, "KL/std": 5.9477057456970215, "epoch": 0.23129251700680273, "fcm_dpo/beta": 0.12414808571338654, "fcm_dpo/delta": -0.12824231386184692, "fcm_dpo/margin": 4.198927402496338, "fcm_dpo/q_t": 0.3833044469356537, "grad_norm": 32.4552001953125, "learning_rate": 4.751601970666064e-07, "logits/chosen": 0.18484419584274292, "logits/rejected": 0.14812716841697693, "logps/chosen": -76.49029541015625, "logps/ref_chosen": -68.00689697265625, "logps/ref_rejected": -74.83482360839844, "logps/rejected": -87.51715087890625, "loss": 1.033, "margin_dpo/margin_mean": 4.198927402496338, "margin_dpo/margin_std": 5.387975692749023, "step": 153 }, { "KL/chosen_KL_mean": -9.598503112792969, "KL/mean": -11.168582916259766, "KL/rejected_KL_mean": -12.738666534423828, "KL/std": 6.069890975952148, "epoch": 0.2328042328042328, "fcm_dpo/beta": 0.12375746667385101, "fcm_dpo/delta": 0.011842047795653343, "fcm_dpo/margin": 3.1401634216308594, "fcm_dpo/q_t": 0.4137424826622009, "grad_norm": 33.3686637878418, "learning_rate": 4.745824613468292e-07, "logits/chosen": 0.2683183252811432, "logits/rejected": 0.26516079902648926, "logps/chosen": -68.821044921875, "logps/ref_chosen": -59.222537994384766, "logps/ref_rejected": -64.19131469726562, "logps/rejected": -76.92998504638672, "loss": 1.1731, "margin_dpo/margin_mean": 3.1401634216308594, "margin_dpo/margin_std": 6.250120162963867, "step": 154 }, { "KL/chosen_KL_mean": -9.947959899902344, "KL/mean": -12.126672744750977, "KL/rejected_KL_mean": -14.305377960205078, "KL/std": 6.030064582824707, "epoch": 0.23431594860166288, "fcm_dpo/beta": 0.12072409689426422, "fcm_dpo/delta": -0.13440188765525818, "fcm_dpo/margin": 4.357418537139893, "fcm_dpo/q_t": 0.389107346534729, "grad_norm": 33.98651885986328, "learning_rate": 4.7399844357283393e-07, "logits/chosen": 0.24776160717010498, "logits/rejected": 0.22814488410949707, "logps/chosen": -78.40265655517578, "logps/ref_chosen": -68.45469665527344, "logps/ref_rejected": -77.91763305664062, "logps/rejected": -92.22301483154297, "loss": 1.1193, "margin_dpo/margin_mean": 4.357418060302734, "margin_dpo/margin_std": 7.67659854888916, "step": 155 }, { "KL/chosen_KL_mean": -10.01881217956543, "KL/mean": -12.622465133666992, "KL/rejected_KL_mean": -15.226116180419922, "KL/std": 6.6650285720825195, "epoch": 0.23582766439909297, "fcm_dpo/beta": 0.11669893562793732, "fcm_dpo/delta": -0.2210705578327179, "fcm_dpo/margin": 5.207304954528809, "fcm_dpo/q_t": 0.3674676716327667, "grad_norm": 32.13863754272461, "learning_rate": 4.7340816008085305e-07, "logits/chosen": 0.22416064143180847, "logits/rejected": 0.17890284955501556, "logps/chosen": -77.28840637207031, "logps/ref_chosen": -67.26959991455078, "logps/ref_rejected": -86.95914459228516, "logps/rejected": -102.18525695800781, "loss": 0.9856, "margin_dpo/margin_mean": 5.207304954528809, "margin_dpo/margin_std": 6.2471513748168945, "step": 156 }, { "KL/chosen_KL_mean": -9.303003311157227, "KL/mean": -11.44937515258789, "KL/rejected_KL_mean": -13.595748901367188, "KL/std": 6.676199436187744, "epoch": 0.23733938019652306, "fcm_dpo/beta": 0.1125432625412941, "fcm_dpo/delta": -0.09084660559892654, "fcm_dpo/margin": 4.292743682861328, "fcm_dpo/q_t": 0.39273664355278015, "grad_norm": 29.455463409423828, "learning_rate": 4.728116273823847e-07, "logits/chosen": 0.21543173491954803, "logits/rejected": 0.19620484113693237, "logps/chosen": -64.07588195800781, "logps/ref_chosen": -54.77287292480469, "logps/ref_rejected": -63.87866973876953, "logps/rejected": -77.47441864013672, "loss": 1.0827, "margin_dpo/margin_mean": 4.292743682861328, "margin_dpo/margin_std": 6.249413967132568, "step": 157 }, { "KL/chosen_KL_mean": -10.352668762207031, "KL/mean": -12.572793960571289, "KL/rejected_KL_mean": -14.792915344238281, "KL/std": 6.630293846130371, "epoch": 0.23885109599395313, "fcm_dpo/beta": 0.11155691742897034, "fcm_dpo/delta": -0.1003374308347702, "fcm_dpo/margin": 4.440248489379883, "fcm_dpo/q_t": 0.3917720317840576, "grad_norm": 31.280303955078125, "learning_rate": 4.7220886216373085e-07, "logits/chosen": 0.27400046586990356, "logits/rejected": 0.2366476058959961, "logps/chosen": -75.275390625, "logps/ref_chosen": -64.92271423339844, "logps/ref_rejected": -82.23789978027344, "logps/rejected": -97.03080749511719, "loss": 1.0824, "margin_dpo/margin_mean": 4.440248489379883, "margin_dpo/margin_std": 6.843707084655762, "step": 158 }, { "KL/chosen_KL_mean": -10.615558624267578, "KL/mean": -13.190221786499023, "KL/rejected_KL_mean": -15.764892578125, "KL/std": 7.167794704437256, "epoch": 0.24036281179138322, "fcm_dpo/beta": 0.1089281514286995, "fcm_dpo/delta": -0.1702059954404831, "fcm_dpo/margin": 5.149332046508789, "fcm_dpo/q_t": 0.3772723078727722, "grad_norm": 33.383506774902344, "learning_rate": 4.715998812855304e-07, "logits/chosen": 0.2682814598083496, "logits/rejected": 0.22989416122436523, "logps/chosen": -67.66255187988281, "logps/ref_chosen": -57.046993255615234, "logps/ref_rejected": -73.32441711425781, "logps/rejected": -89.08930969238281, "loss": 1.0526, "margin_dpo/margin_mean": 5.149332046508789, "margin_dpo/margin_std": 7.4767351150512695, "step": 159 }, { "KL/chosen_KL_mean": -11.970067977905273, "KL/mean": -14.173985481262207, "KL/rejected_KL_mean": -16.377899169921875, "KL/std": 6.7857513427734375, "epoch": 0.2418745275888133, "fcm_dpo/beta": 0.10626037418842316, "fcm_dpo/delta": -0.07168857008218765, "fcm_dpo/margin": 4.407832622528076, "fcm_dpo/q_t": 0.40258103609085083, "grad_norm": 27.9632625579834, "learning_rate": 4.7098470178228755e-07, "logits/chosen": 0.1273048371076584, "logits/rejected": 0.0839085727930069, "logps/chosen": -61.776981353759766, "logps/ref_chosen": -49.806915283203125, "logps/ref_rejected": -68.3370132446289, "logps/rejected": -84.71491241455078, "loss": 1.1236, "margin_dpo/margin_mean": 4.407833099365234, "margin_dpo/margin_std": 7.865055084228516, "step": 160 }, { "KL/chosen_KL_mean": -11.96878433227539, "KL/mean": -14.375494003295898, "KL/rejected_KL_mean": -16.782207489013672, "KL/std": 6.705003261566162, "epoch": 0.24338624338624337, "fcm_dpo/beta": 0.10418327152729034, "fcm_dpo/delta": -0.1067676991224289, "fcm_dpo/margin": 4.813423156738281, "fcm_dpo/q_t": 0.3899020552635193, "grad_norm": 26.78690528869629, "learning_rate": 4.703633408618955e-07, "logits/chosen": 0.27834972739219666, "logits/rejected": 0.23936110734939575, "logps/chosen": -64.46926879882812, "logps/ref_chosen": -52.50048828125, "logps/ref_rejected": -66.04540252685547, "logps/rejected": -82.82760620117188, "loss": 1.0822, "margin_dpo/margin_mean": 4.813423156738281, "margin_dpo/margin_std": 7.380710601806641, "step": 161 }, { "KL/chosen_KL_mean": -12.701950073242188, "KL/mean": -15.949088096618652, "KL/rejected_KL_mean": -19.19622802734375, "KL/std": 7.516228675842285, "epoch": 0.24489795918367346, "fcm_dpo/beta": 0.09950388967990875, "fcm_dpo/delta": -0.26341742277145386, "fcm_dpo/margin": 6.494287014007568, "fcm_dpo/q_t": 0.3561316728591919, "grad_norm": 27.878267288208008, "learning_rate": 4.697358159051549e-07, "logits/chosen": 0.3174731135368347, "logits/rejected": 0.267278790473938, "logps/chosen": -82.171142578125, "logps/ref_chosen": -69.46919250488281, "logps/ref_rejected": -92.00952911376953, "logps/rejected": -111.20575714111328, "loss": 0.9581, "margin_dpo/margin_mean": 6.494287490844727, "margin_dpo/margin_std": 7.305610656738281, "step": 162 }, { "KL/chosen_KL_mean": -11.61453628540039, "KL/mean": -14.817399978637695, "KL/rejected_KL_mean": -18.020267486572266, "KL/std": 7.376770973205566, "epoch": 0.24640967498110355, "fcm_dpo/beta": 0.09595300257205963, "fcm_dpo/delta": -0.22854195535182953, "fcm_dpo/margin": 6.405729293823242, "fcm_dpo/q_t": 0.36570775508880615, "grad_norm": 25.580001831054688, "learning_rate": 4.691021444652876e-07, "logits/chosen": 0.2480703890323639, "logits/rejected": 0.2000638246536255, "logps/chosen": -62.228370666503906, "logps/ref_chosen": -50.613834381103516, "logps/ref_rejected": -74.62033081054688, "logps/rejected": -92.64059448242188, "loss": 1.004, "margin_dpo/margin_mean": 6.405730247497559, "margin_dpo/margin_std": 7.875078201293945, "step": 163 }, { "KL/chosen_KL_mean": -12.40074348449707, "KL/mean": -15.593101501464844, "KL/rejected_KL_mean": -18.78546142578125, "KL/std": 7.785859107971191, "epoch": 0.24792139077853365, "fcm_dpo/beta": 0.09157611429691315, "fcm_dpo/delta": -0.1956927627325058, "fcm_dpo/margin": 6.384723663330078, "fcm_dpo/q_t": 0.3727573752403259, "grad_norm": 24.026826858520508, "learning_rate": 4.6846234426744624e-07, "logits/chosen": 0.2387039214372635, "logits/rejected": 0.17490170896053314, "logps/chosen": -67.24885559082031, "logps/ref_chosen": -54.848114013671875, "logps/ref_rejected": -79.0630111694336, "logps/rejected": -97.84847259521484, "loss": 1.0249, "margin_dpo/margin_mean": 6.384724140167236, "margin_dpo/margin_std": 8.432074546813965, "step": 164 }, { "KL/chosen_KL_mean": -13.576322555541992, "KL/mean": -16.508556365966797, "KL/rejected_KL_mean": -19.44078826904297, "KL/std": 7.981330871582031, "epoch": 0.2494331065759637, "fcm_dpo/beta": 0.0890810489654541, "fcm_dpo/delta": -0.12915945053100586, "fcm_dpo/margin": 5.864466667175293, "fcm_dpo/q_t": 0.3837765157222748, "grad_norm": 25.287494659423828, "learning_rate": 4.678164332082175e-07, "logits/chosen": 0.3062145709991455, "logits/rejected": 0.25015687942504883, "logps/chosen": -64.66553497314453, "logps/ref_chosen": -51.089210510253906, "logps/ref_rejected": -71.23370361328125, "logps/rejected": -90.67448425292969, "loss": 1.0654, "margin_dpo/margin_mean": 5.864466667175293, "margin_dpo/margin_std": 8.354823112487793, "step": 165 }, { "KL/chosen_KL_mean": -13.206365585327148, "KL/mean": -15.502958297729492, "KL/rejected_KL_mean": -17.799549102783203, "KL/std": 7.9299445152282715, "epoch": 0.2509448223733938, "fcm_dpo/beta": 0.08768243342638016, "fcm_dpo/delta": -0.002980044111609459, "fcm_dpo/margin": 4.593182563781738, "fcm_dpo/q_t": 0.41203340888023376, "grad_norm": 28.28586196899414, "learning_rate": 4.6716442935512214e-07, "logits/chosen": 0.2627357840538025, "logits/rejected": 0.1715720295906067, "logps/chosen": -76.39717102050781, "logps/ref_chosen": -63.19081115722656, "logps/ref_rejected": -93.8402099609375, "logps/rejected": -111.63975524902344, "loss": 1.1261, "margin_dpo/margin_mean": 4.593182563781738, "margin_dpo/margin_std": 7.637331962585449, "step": 166 }, { "KL/chosen_KL_mean": -11.713014602661133, "KL/mean": -15.292600631713867, "KL/rejected_KL_mean": -18.872188568115234, "KL/std": 8.240446090698242, "epoch": 0.25245653817082386, "fcm_dpo/beta": 0.08448252081871033, "fcm_dpo/delta": -0.22041411697864532, "fcm_dpo/margin": 7.1591691970825195, "fcm_dpo/q_t": 0.36489737033843994, "grad_norm": 22.19241714477539, "learning_rate": 4.6650635094610966e-07, "logits/chosen": 0.2110971212387085, "logits/rejected": 0.1757899820804596, "logps/chosen": -70.63728332519531, "logps/ref_chosen": -58.92427062988281, "logps/ref_rejected": -72.97377014160156, "logps/rejected": -91.84596252441406, "loss": 0.9721, "margin_dpo/margin_mean": 7.1591691970825195, "margin_dpo/margin_std": 7.9116530418396, "step": 167 }, { "KL/chosen_KL_mean": -13.643762588500977, "KL/mean": -16.358333587646484, "KL/rejected_KL_mean": -19.07291030883789, "KL/std": 8.718094825744629, "epoch": 0.25396825396825395, "fcm_dpo/beta": 0.08331099897623062, "fcm_dpo/delta": -0.05487321317195892, "fcm_dpo/margin": 5.429144859313965, "fcm_dpo/q_t": 0.40054255723953247, "grad_norm": 25.996004104614258, "learning_rate": 4.6584221638904767e-07, "logits/chosen": 0.2337023764848709, "logits/rejected": 0.19847044348716736, "logps/chosen": -79.2951431274414, "logps/ref_chosen": -65.65138244628906, "logps/ref_rejected": -79.71418762207031, "logps/rejected": -98.78709411621094, "loss": 1.085, "margin_dpo/margin_mean": 5.429144859313965, "margin_dpo/margin_std": 7.938119411468506, "step": 168 }, { "KL/chosen_KL_mean": -11.592302322387695, "KL/mean": -14.968731880187988, "KL/rejected_KL_mean": -18.34516143798828, "KL/std": 8.936497688293457, "epoch": 0.25547996976568405, "fcm_dpo/beta": 0.08121495693922043, "fcm_dpo/delta": -0.15712596476078033, "fcm_dpo/margin": 6.7528581619262695, "fcm_dpo/q_t": 0.38491493463516235, "grad_norm": 24.72463607788086, "learning_rate": 4.651720442612075e-07, "logits/chosen": 0.3097034692764282, "logits/rejected": 0.27537956833839417, "logps/chosen": -73.01817321777344, "logps/ref_chosen": -61.425865173339844, "logps/ref_rejected": -76.09590148925781, "logps/rejected": -94.44105529785156, "loss": 1.0647, "margin_dpo/margin_mean": 6.7528581619262695, "margin_dpo/margin_std": 10.517898559570312, "step": 169 }, { "KL/chosen_KL_mean": -11.574005126953125, "KL/mean": -14.787891387939453, "KL/rejected_KL_mean": -18.001773834228516, "KL/std": 8.793878555297852, "epoch": 0.25699168556311414, "fcm_dpo/beta": 0.07942621409893036, "fcm_dpo/delta": -0.11631622910499573, "fcm_dpo/margin": 6.427769184112549, "fcm_dpo/q_t": 0.3889446258544922, "grad_norm": 20.754453659057617, "learning_rate": 4.6449585330874425e-07, "logits/chosen": 0.24759146571159363, "logits/rejected": 0.24484430253505707, "logps/chosen": -68.2271957397461, "logps/ref_chosen": -56.65319061279297, "logps/ref_rejected": -63.45965576171875, "logps/rejected": -81.46142578125, "loss": 1.0987, "margin_dpo/margin_mean": 6.427768707275391, "margin_dpo/margin_std": 10.505061149597168, "step": 170 }, { "KL/chosen_KL_mean": -12.49212646484375, "KL/mean": -16.14303970336914, "KL/rejected_KL_mean": -19.793956756591797, "KL/std": 9.996442794799805, "epoch": 0.2585034013605442, "fcm_dpo/beta": 0.07601982355117798, "fcm_dpo/delta": -0.16807925701141357, "fcm_dpo/margin": 7.301830768585205, "fcm_dpo/q_t": 0.3793303966522217, "grad_norm": 23.09961700439453, "learning_rate": 4.6381366244617224e-07, "logits/chosen": 0.3099227845668793, "logits/rejected": 0.2590000331401825, "logps/chosen": -76.22688293457031, "logps/ref_chosen": -63.73476028442383, "logps/ref_rejected": -78.50328063964844, "logps/rejected": -98.29723358154297, "loss": 1.075, "margin_dpo/margin_mean": 7.3018317222595215, "margin_dpo/margin_std": 11.233263969421387, "step": 171 }, { "KL/chosen_KL_mean": -13.715980529785156, "KL/mean": -17.212726593017578, "KL/rejected_KL_mean": -20.709468841552734, "KL/std": 10.043166160583496, "epoch": 0.2600151171579743, "fcm_dpo/beta": 0.07525793462991714, "fcm_dpo/delta": -0.13359381258487701, "fcm_dpo/margin": 6.993487358093262, "fcm_dpo/q_t": 0.38472551107406616, "grad_norm": 22.1481876373291, "learning_rate": 4.631254907558365e-07, "logits/chosen": 0.3433062434196472, "logits/rejected": 0.2836194634437561, "logps/chosen": -65.91773986816406, "logps/ref_chosen": -52.201759338378906, "logps/ref_rejected": -82.85285949707031, "logps/rejected": -103.56233215332031, "loss": 1.0655, "margin_dpo/margin_mean": 6.9934868812561035, "margin_dpo/margin_std": 10.196330070495605, "step": 172 }, { "KL/chosen_KL_mean": -13.024921417236328, "KL/mean": -16.835899353027344, "KL/rejected_KL_mean": -20.64687728881836, "KL/std": 10.502202987670898, "epoch": 0.2615268329554044, "fcm_dpo/beta": 0.07153315842151642, "fcm_dpo/delta": -0.15856504440307617, "fcm_dpo/margin": 7.621954917907715, "fcm_dpo/q_t": 0.3873851001262665, "grad_norm": 21.773149490356445, "learning_rate": 4.624313574873786e-07, "logits/chosen": 0.3005369305610657, "logits/rejected": 0.21291136741638184, "logps/chosen": -68.45964813232422, "logps/ref_chosen": -55.434722900390625, "logps/ref_rejected": -77.81967163085938, "logps/rejected": -98.466552734375, "loss": 1.1175, "margin_dpo/margin_mean": 7.6219563484191895, "margin_dpo/margin_std": 13.109130859375, "step": 173 }, { "KL/chosen_KL_mean": -14.391660690307617, "KL/mean": -18.400985717773438, "KL/rejected_KL_mean": -22.41031265258789, "KL/std": 10.439952850341797, "epoch": 0.26303854875283444, "fcm_dpo/beta": 0.06982284039258957, "fcm_dpo/delta": -0.17004986107349396, "fcm_dpo/margin": 8.018655776977539, "fcm_dpo/q_t": 0.3786467909812927, "grad_norm": 22.08015251159668, "learning_rate": 4.61731282057198e-07, "logits/chosen": 0.285661518573761, "logits/rejected": 0.2192431539297104, "logps/chosen": -71.56361389160156, "logps/ref_chosen": -57.17195129394531, "logps/ref_rejected": -85.47578430175781, "logps/rejected": -107.88609313964844, "loss": 1.0428, "margin_dpo/margin_mean": 8.018654823303223, "margin_dpo/margin_std": 11.424968719482422, "step": 174 }, { "KL/chosen_KL_mean": -13.223051071166992, "KL/mean": -17.665573120117188, "KL/rejected_KL_mean": -22.10810089111328, "KL/std": 10.922414779663086, "epoch": 0.26455026455026454, "fcm_dpo/beta": 0.06739327311515808, "fcm_dpo/delta": -0.21115991473197937, "fcm_dpo/margin": 8.885052680969238, "fcm_dpo/q_t": 0.3742350935935974, "grad_norm": 21.33765411376953, "learning_rate": 4.6102528404790965e-07, "logits/chosen": 0.34087687730789185, "logits/rejected": 0.3097060024738312, "logps/chosen": -80.88867950439453, "logps/ref_chosen": -67.6656265258789, "logps/ref_rejected": -84.36766815185547, "logps/rejected": -106.47576904296875, "loss": 1.0334, "margin_dpo/margin_mean": 8.885052680969238, "margin_dpo/margin_std": 12.609004974365234, "step": 175 }, { "KL/chosen_KL_mean": -15.523975372314453, "KL/mean": -18.484966278076172, "KL/rejected_KL_mean": -21.44595718383789, "KL/std": 11.551431655883789, "epoch": 0.2660619803476946, "fcm_dpo/beta": 0.06620974093675613, "fcm_dpo/delta": 0.007484421133995056, "fcm_dpo/margin": 5.921982765197754, "fcm_dpo/q_t": 0.41712304949760437, "grad_norm": 25.927288055419922, "learning_rate": 4.603133832077953e-07, "logits/chosen": 0.28880423307418823, "logits/rejected": 0.26152804493904114, "logps/chosen": -93.38273620605469, "logps/ref_chosen": -77.8587646484375, "logps/ref_rejected": -81.08732604980469, "logps/rejected": -102.53327941894531, "loss": 1.1915, "margin_dpo/margin_mean": 5.921982765197754, "margin_dpo/margin_std": 12.589743614196777, "step": 176 }, { "KL/chosen_KL_mean": -12.250782012939453, "KL/mean": -18.0947265625, "KL/rejected_KL_mean": -23.938678741455078, "KL/std": 11.902139663696289, "epoch": 0.2675736961451247, "fcm_dpo/beta": 0.06285828351974487, "fcm_dpo/delta": -0.36344408988952637, "fcm_dpo/margin": 11.687898635864258, "fcm_dpo/q_t": 0.339341938495636, "grad_norm": 23.54242515563965, "learning_rate": 4.5959559945025183e-07, "logits/chosen": 0.41696828603744507, "logits/rejected": 0.3189684748649597, "logps/chosen": -67.47117614746094, "logps/ref_chosen": -55.22039794921875, "logps/ref_rejected": -92.54973602294922, "logps/rejected": -116.48841857910156, "loss": 0.9092, "margin_dpo/margin_mean": 11.687898635864258, "margin_dpo/margin_std": 12.291168212890625, "step": 177 }, { "KL/chosen_KL_mean": -13.977239608764648, "KL/mean": -17.54144287109375, "KL/rejected_KL_mean": -21.10564422607422, "KL/std": 11.356380462646484, "epoch": 0.2690854119425548, "fcm_dpo/beta": 0.06110500544309616, "fcm_dpo/delta": -0.03895438462495804, "fcm_dpo/margin": 7.128399848937988, "fcm_dpo/q_t": 0.4032604992389679, "grad_norm": 22.13608741760254, "learning_rate": 4.588719528532341e-07, "logits/chosen": 0.26632487773895264, "logits/rejected": 0.2168237566947937, "logps/chosen": -74.78773498535156, "logps/ref_chosen": -60.81049346923828, "logps/ref_rejected": -81.12973022460938, "logps/rejected": -102.23536682128906, "loss": 1.1164, "margin_dpo/margin_mean": 7.128399848937988, "margin_dpo/margin_std": 11.491066932678223, "step": 178 }, { "KL/chosen_KL_mean": -14.837139129638672, "KL/mean": -19.02896499633789, "KL/rejected_KL_mean": -23.22079086303711, "KL/std": 11.86774730682373, "epoch": 0.2705971277399849, "fcm_dpo/beta": 0.06048303842544556, "fcm_dpo/delta": -0.11276848614215851, "fcm_dpo/margin": 8.383651733398438, "fcm_dpo/q_t": 0.390702486038208, "grad_norm": 21.742908477783203, "learning_rate": 4.581424636586928e-07, "logits/chosen": 0.3228716254234314, "logits/rejected": 0.30639001727104187, "logps/chosen": -80.50885009765625, "logps/ref_chosen": -65.67171478271484, "logps/ref_rejected": -75.32586669921875, "logps/rejected": -98.54666137695312, "loss": 1.1012, "margin_dpo/margin_mean": 8.383651733398438, "margin_dpo/margin_std": 14.014520645141602, "step": 179 }, { "KL/chosen_KL_mean": -11.49049186706543, "KL/mean": -15.210617065429688, "KL/rejected_KL_mean": -18.930740356445312, "KL/std": 11.852537155151367, "epoch": 0.272108843537415, "fcm_dpo/beta": 0.059869199991226196, "fcm_dpo/delta": -0.047591034322977066, "fcm_dpo/margin": 7.440252304077148, "fcm_dpo/q_t": 0.4054303765296936, "grad_norm": 19.28363037109375, "learning_rate": 4.5740715227200897e-07, "logits/chosen": 0.13881587982177734, "logits/rejected": 0.11996544897556305, "logps/chosen": -68.17330169677734, "logps/ref_chosen": -56.68280792236328, "logps/ref_rejected": -64.94414520263672, "logps/rejected": -83.87488555908203, "loss": 1.1436, "margin_dpo/margin_mean": 7.440252304077148, "margin_dpo/margin_std": 13.85991096496582, "step": 180 }, { "KL/chosen_KL_mean": -10.01158332824707, "KL/mean": -15.477460861206055, "KL/rejected_KL_mean": -20.94333267211914, "KL/std": 12.780189514160156, "epoch": 0.273620559334845, "fcm_dpo/beta": 0.05729272961616516, "fcm_dpo/delta": -0.2422669529914856, "fcm_dpo/margin": 10.931753158569336, "fcm_dpo/q_t": 0.3609431982040405, "grad_norm": 19.187829971313477, "learning_rate": 4.566660392614228e-07, "logits/chosen": 0.32175952196121216, "logits/rejected": 0.2816872298717499, "logps/chosen": -70.78762817382812, "logps/ref_chosen": -60.77604675292969, "logps/ref_rejected": -83.98361206054688, "logps/rejected": -104.92694854736328, "loss": 0.9572, "margin_dpo/margin_mean": 10.931753158569336, "margin_dpo/margin_std": 11.934650421142578, "step": 181 }, { "KL/chosen_KL_mean": -13.245315551757812, "KL/mean": -19.128177642822266, "KL/rejected_KL_mean": -25.011043548583984, "KL/std": 13.650596618652344, "epoch": 0.2751322751322751, "fcm_dpo/beta": 0.05450388044118881, "fcm_dpo/delta": -0.2583681643009186, "fcm_dpo/margin": 11.765732765197754, "fcm_dpo/q_t": 0.3621079921722412, "grad_norm": 19.192251205444336, "learning_rate": 4.5591914535745817e-07, "logits/chosen": 0.31452324986457825, "logits/rejected": 0.2316857874393463, "logps/chosen": -73.49909973144531, "logps/ref_chosen": -60.2537841796875, "logps/ref_rejected": -89.7706298828125, "logps/rejected": -114.78167724609375, "loss": 1.0017, "margin_dpo/margin_mean": 11.76573371887207, "margin_dpo/margin_std": 15.48458480834961, "step": 182 }, { "KL/chosen_KL_mean": -16.417619705200195, "KL/mean": -18.79208755493164, "KL/rejected_KL_mean": -21.16655731201172, "KL/std": 13.141827583312988, "epoch": 0.2766439909297052, "fcm_dpo/beta": 0.05492541193962097, "fcm_dpo/delta": 0.14310485124588013, "fcm_dpo/margin": 4.748943328857422, "fcm_dpo/q_t": 0.44205474853515625, "grad_norm": 20.398983001708984, "learning_rate": 4.551664914523433e-07, "logits/chosen": 0.29353460669517517, "logits/rejected": 0.2715173363685608, "logps/chosen": -78.17903900146484, "logps/ref_chosen": -61.76142120361328, "logps/ref_rejected": -72.54627990722656, "logps/rejected": -93.71284484863281, "loss": 1.2654, "margin_dpo/margin_mean": 4.748943328857422, "margin_dpo/margin_std": 13.215112686157227, "step": 183 }, { "KL/chosen_KL_mean": -10.467620849609375, "KL/mean": -14.862070083618164, "KL/rejected_KL_mean": -19.256515502929688, "KL/std": 11.553922653198242, "epoch": 0.2781557067271353, "fcm_dpo/beta": 0.054390810430049896, "fcm_dpo/delta": -0.08345025777816772, "fcm_dpo/margin": 8.788891792297363, "fcm_dpo/q_t": 0.39447537064552307, "grad_norm": 18.428091049194336, "learning_rate": 4.544080985994258e-07, "logits/chosen": 0.40536069869995117, "logits/rejected": 0.339550644159317, "logps/chosen": -57.30834197998047, "logps/ref_chosen": -46.840721130371094, "logps/ref_rejected": -69.3609390258789, "logps/rejected": -88.6174545288086, "loss": 1.0607, "margin_dpo/margin_mean": 8.788890838623047, "margin_dpo/margin_std": 11.725053787231445, "step": 184 }, { "KL/chosen_KL_mean": -13.35588264465332, "KL/mean": -18.210655212402344, "KL/rejected_KL_mean": -23.065425872802734, "KL/std": 13.737922668457031, "epoch": 0.2796674225245654, "fcm_dpo/beta": 0.0531897135078907, "fcm_dpo/delta": -0.12439177185297012, "fcm_dpo/margin": 9.709543228149414, "fcm_dpo/q_t": 0.39072197675704956, "grad_norm": 17.2995548248291, "learning_rate": 4.5364398801258394e-07, "logits/chosen": 0.3512234091758728, "logits/rejected": 0.3028402030467987, "logps/chosen": -65.6770248413086, "logps/ref_chosen": -52.32114028930664, "logps/ref_rejected": -68.3885726928711, "logps/rejected": -91.45399475097656, "loss": 1.1187, "margin_dpo/margin_mean": 9.70954418182373, "margin_dpo/margin_std": 16.909622192382812, "step": 185 }, { "KL/chosen_KL_mean": -11.858358383178711, "KL/mean": -17.2672176361084, "KL/rejected_KL_mean": -22.676071166992188, "KL/std": 13.716083526611328, "epoch": 0.2811791383219955, "fcm_dpo/beta": 0.05181612819433212, "fcm_dpo/delta": -0.17022064328193665, "fcm_dpo/margin": 10.817716598510742, "fcm_dpo/q_t": 0.38037991523742676, "grad_norm": 20.882518768310547, "learning_rate": 4.5287418106563354e-07, "logits/chosen": 0.25211092829704285, "logits/rejected": 0.20962807536125183, "logps/chosen": -79.27848815917969, "logps/ref_chosen": -67.42012786865234, "logps/ref_rejected": -82.50968933105469, "logps/rejected": -105.18576049804688, "loss": 1.0654, "margin_dpo/margin_mean": 10.817716598510742, "margin_dpo/margin_std": 16.649192810058594, "step": 186 }, { "KL/chosen_KL_mean": -14.678447723388672, "KL/mean": -19.588550567626953, "KL/rejected_KL_mean": -24.4986572265625, "KL/std": 14.59811019897461, "epoch": 0.28269085411942557, "fcm_dpo/beta": 0.050496138632297516, "fcm_dpo/delta": -0.10132630914449692, "fcm_dpo/margin": 9.820213317871094, "fcm_dpo/q_t": 0.39432644844055176, "grad_norm": 20.352502822875977, "learning_rate": 4.520986992917297e-07, "logits/chosen": 0.3316032886505127, "logits/rejected": 0.2751980423927307, "logps/chosen": -90.20394134521484, "logps/ref_chosen": -75.52549743652344, "logps/ref_rejected": -94.76289367675781, "logps/rejected": -119.26155090332031, "loss": 1.1129, "margin_dpo/margin_mean": 9.820212364196777, "margin_dpo/margin_std": 16.846996307373047, "step": 187 }, { "KL/chosen_KL_mean": -13.222888946533203, "KL/mean": -18.66843032836914, "KL/rejected_KL_mean": -24.113971710205078, "KL/std": 15.047087669372559, "epoch": 0.2842025699168556, "fcm_dpo/beta": 0.049329958856105804, "fcm_dpo/delta": -0.14497430622577667, "fcm_dpo/margin": 10.891085624694824, "fcm_dpo/q_t": 0.3854219317436218, "grad_norm": 18.09861946105957, "learning_rate": 4.5131756438276466e-07, "logits/chosen": 0.36570295691490173, "logits/rejected": 0.3198516368865967, "logps/chosen": -84.7462158203125, "logps/ref_chosen": -71.52333068847656, "logps/ref_rejected": -78.29949951171875, "logps/rejected": -102.41346740722656, "loss": 1.0563, "margin_dpo/margin_mean": 10.89108657836914, "margin_dpo/margin_std": 16.060014724731445, "step": 188 }, { "KL/chosen_KL_mean": -12.530065536499023, "KL/mean": -17.523853302001953, "KL/rejected_KL_mean": -22.51763916015625, "KL/std": 14.935396194458008, "epoch": 0.2857142857142857, "fcm_dpo/beta": 0.04785740375518799, "fcm_dpo/delta": -0.08463504910469055, "fcm_dpo/margin": 9.98757266998291, "fcm_dpo/q_t": 0.3948071599006653, "grad_norm": 20.050233840942383, "learning_rate": 4.5053079818876096e-07, "logits/chosen": 0.35361284017562866, "logits/rejected": 0.36745667457580566, "logps/chosen": -84.70632934570312, "logps/ref_chosen": -72.17626953125, "logps/ref_rejected": -75.26313781738281, "logps/rejected": -97.78077697753906, "loss": 1.102, "margin_dpo/margin_mean": 9.987573623657227, "margin_dpo/margin_std": 15.788932800292969, "step": 189 }, { "KL/chosen_KL_mean": -11.82335090637207, "KL/mean": -18.92856788635254, "KL/rejected_KL_mean": -26.033790588378906, "KL/std": 15.521673202514648, "epoch": 0.2872260015117158, "fcm_dpo/beta": 0.046157702803611755, "fcm_dpo/delta": -0.27415594458580017, "fcm_dpo/margin": 14.210436820983887, "fcm_dpo/q_t": 0.35759294033050537, "grad_norm": 22.220190048217773, "learning_rate": 4.4973842271726024e-07, "logits/chosen": 0.39344653487205505, "logits/rejected": 0.24719932675361633, "logps/chosen": -66.44762420654297, "logps/ref_chosen": -54.624271392822266, "logps/ref_rejected": -101.47068786621094, "logps/rejected": -127.50447845458984, "loss": 0.9607, "margin_dpo/margin_mean": 14.210436820983887, "margin_dpo/margin_std": 16.54659652709961, "step": 190 }, { "KL/chosen_KL_mean": -16.818561553955078, "KL/mean": -22.14589500427246, "KL/rejected_KL_mean": -27.473224639892578, "KL/std": 15.95837688446045, "epoch": 0.2887377173091459, "fcm_dpo/beta": 0.04499708116054535, "fcm_dpo/delta": -0.08340764790773392, "fcm_dpo/margin": 10.6546630859375, "fcm_dpo/q_t": 0.39313390851020813, "grad_norm": 19.53053092956543, "learning_rate": 4.48940460132708e-07, "logits/chosen": 0.4083729386329651, "logits/rejected": 0.3787025213241577, "logps/chosen": -89.7510757446289, "logps/ref_chosen": -72.93251037597656, "logps/ref_rejected": -89.95103454589844, "logps/rejected": -117.42425537109375, "loss": 1.1007, "margin_dpo/margin_mean": 10.6546630859375, "margin_dpo/margin_std": 17.215587615966797, "step": 191 }, { "KL/chosen_KL_mean": -18.174646377563477, "KL/mean": -21.69927215576172, "KL/rejected_KL_mean": -25.223896026611328, "KL/std": 16.775699615478516, "epoch": 0.29024943310657597, "fcm_dpo/beta": 0.045455966144800186, "fcm_dpo/delta": 0.08187517523765564, "fcm_dpo/margin": 7.049251556396484, "fcm_dpo/q_t": 0.4282623529434204, "grad_norm": 16.56797981262207, "learning_rate": 4.481369327558329e-07, "logits/chosen": 0.405287504196167, "logits/rejected": 0.37762707471847534, "logps/chosen": -72.17576599121094, "logps/ref_chosen": -54.001121520996094, "logps/ref_rejected": -63.531551361083984, "logps/rejected": -88.75544738769531, "loss": 1.1946, "margin_dpo/margin_mean": 7.049251556396484, "margin_dpo/margin_std": 14.657774925231934, "step": 192 }, { "KL/chosen_KL_mean": -14.546079635620117, "KL/mean": -21.136138916015625, "KL/rejected_KL_mean": -27.726200103759766, "KL/std": 15.746472358703613, "epoch": 0.29176114890400606, "fcm_dpo/beta": 0.04400428384542465, "fcm_dpo/delta": -0.1927540898323059, "fcm_dpo/margin": 13.180118560791016, "fcm_dpo/q_t": 0.3722004294395447, "grad_norm": 15.471779823303223, "learning_rate": 4.47327863063023e-07, "logits/chosen": 0.3393857181072235, "logits/rejected": 0.3160160183906555, "logps/chosen": -71.29535675048828, "logps/ref_chosen": -56.74927520751953, "logps/ref_rejected": -58.80629348754883, "logps/rejected": -86.5324935913086, "loss": 1.0014, "margin_dpo/margin_mean": 13.180119514465332, "margin_dpo/margin_std": 16.18925666809082, "step": 193 }, { "KL/chosen_KL_mean": -15.568546295166016, "KL/mean": -20.744672775268555, "KL/rejected_KL_mean": -25.920806884765625, "KL/std": 15.98554801940918, "epoch": 0.29327286470143615, "fcm_dpo/beta": 0.043137095868587494, "fcm_dpo/delta": -0.05045357346534729, "fcm_dpo/margin": 10.352262496948242, "fcm_dpo/q_t": 0.40236881375312805, "grad_norm": 18.62397575378418, "learning_rate": 4.4651327368569684e-07, "logits/chosen": 0.3981412351131439, "logits/rejected": 0.3676302433013916, "logps/chosen": -72.21798706054688, "logps/ref_chosen": -56.64944076538086, "logps/ref_rejected": -69.98954772949219, "logps/rejected": -95.91035461425781, "loss": 1.1542, "margin_dpo/margin_mean": 10.352262496948242, "margin_dpo/margin_std": 19.430416107177734, "step": 194 }, { "KL/chosen_KL_mean": -18.541385650634766, "KL/mean": -25.322803497314453, "KL/rejected_KL_mean": -32.10423278808594, "KL/std": 18.18044090270996, "epoch": 0.2947845804988662, "fcm_dpo/beta": 0.04202111437916756, "fcm_dpo/delta": -0.18148893117904663, "fcm_dpo/margin": 13.562843322753906, "fcm_dpo/q_t": 0.37634384632110596, "grad_norm": 18.791791915893555, "learning_rate": 4.4569318740967043e-07, "logits/chosen": 0.30191123485565186, "logits/rejected": 0.30708640813827515, "logps/chosen": -88.95116424560547, "logps/ref_chosen": -70.40977478027344, "logps/ref_rejected": -74.39448547363281, "logps/rejected": -106.49871826171875, "loss": 1.036, "margin_dpo/margin_mean": 13.562843322753906, "margin_dpo/margin_std": 18.799701690673828, "step": 195 }, { "KL/chosen_KL_mean": -17.88593864440918, "KL/mean": -23.226863861083984, "KL/rejected_KL_mean": -28.567790985107422, "KL/std": 18.183853149414062, "epoch": 0.2962962962962963, "fcm_dpo/beta": 0.04170439392328262, "fcm_dpo/delta": -0.04776248335838318, "fcm_dpo/margin": 10.68185043334961, "fcm_dpo/q_t": 0.39945489168167114, "grad_norm": 18.276790618896484, "learning_rate": 4.448676271745197e-07, "logits/chosen": 0.37690073251724243, "logits/rejected": 0.3343625068664551, "logps/chosen": -77.11351013183594, "logps/ref_chosen": -59.227577209472656, "logps/ref_rejected": -83.54757690429688, "logps/rejected": -112.11536407470703, "loss": 1.0993, "margin_dpo/margin_mean": 10.68185043334961, "margin_dpo/margin_std": 16.383975982666016, "step": 196 }, { "KL/chosen_KL_mean": -15.438322067260742, "KL/mean": -22.16175651550293, "KL/rejected_KL_mean": -28.88519287109375, "KL/std": 17.83734130859375, "epoch": 0.29780801209372637, "fcm_dpo/beta": 0.040766216814517975, "fcm_dpo/delta": -0.15669496357440948, "fcm_dpo/margin": 13.446867942810059, "fcm_dpo/q_t": 0.38244131207466125, "grad_norm": 17.88084602355957, "learning_rate": 4.440366160729392e-07, "logits/chosen": 0.4925154447555542, "logits/rejected": 0.4390217363834381, "logps/chosen": -66.96745300292969, "logps/ref_chosen": -51.52912902832031, "logps/ref_rejected": -73.70631408691406, "logps/rejected": -102.59150695800781, "loss": 1.1024, "margin_dpo/margin_mean": 13.446868896484375, "margin_dpo/margin_std": 22.26742935180664, "step": 197 }, { "KL/chosen_KL_mean": -16.668659210205078, "KL/mean": -23.898550033569336, "KL/rejected_KL_mean": -31.12844467163086, "KL/std": 17.481998443603516, "epoch": 0.29931972789115646, "fcm_dpo/beta": 0.039299823343753815, "fcm_dpo/delta": -0.17802850902080536, "fcm_dpo/margin": 14.459779739379883, "fcm_dpo/q_t": 0.37265992164611816, "grad_norm": 17.161579132080078, "learning_rate": 4.432001773500957e-07, "logits/chosen": 0.44338977336883545, "logits/rejected": 0.401350200176239, "logps/chosen": -76.45133972167969, "logps/ref_chosen": -59.78268051147461, "logps/ref_rejected": -72.24533081054688, "logps/rejected": -103.37377166748047, "loss": 1.0073, "margin_dpo/margin_mean": 14.459779739379883, "margin_dpo/margin_std": 17.56760025024414, "step": 198 }, { "KL/chosen_KL_mean": -20.50527000427246, "KL/mean": -26.135860443115234, "KL/rejected_KL_mean": -31.76644515991211, "KL/std": 18.526927947998047, "epoch": 0.30083144368858655, "fcm_dpo/beta": 0.03888273611664772, "fcm_dpo/delta": -0.04109527915716171, "fcm_dpo/margin": 11.26118278503418, "fcm_dpo/q_t": 0.40396648645401, "grad_norm": 16.385759353637695, "learning_rate": 4.4235833440297856e-07, "logits/chosen": 0.41728949546813965, "logits/rejected": 0.3250621557235718, "logps/chosen": -76.89204406738281, "logps/ref_chosen": -56.38677215576172, "logps/ref_rejected": -74.56779479980469, "logps/rejected": -106.33424377441406, "loss": 1.1651, "margin_dpo/margin_mean": 11.261181831359863, "margin_dpo/margin_std": 21.53316879272461, "step": 199 }, { "KL/chosen_KL_mean": -15.848258972167969, "KL/mean": -23.894195556640625, "KL/rejected_KL_mean": -31.940135955810547, "KL/std": 19.302684783935547, "epoch": 0.30234315948601664, "fcm_dpo/beta": 0.03706369549036026, "fcm_dpo/delta": -0.2104116529226303, "fcm_dpo/margin": 16.091880798339844, "fcm_dpo/q_t": 0.37147045135498047, "grad_norm": 15.661738395690918, "learning_rate": 4.415111107797445e-07, "logits/chosen": 0.4774281978607178, "logits/rejected": 0.3997184634208679, "logps/chosen": -73.6725845336914, "logps/ref_chosen": -57.82432556152344, "logps/ref_rejected": -89.28246307373047, "logps/rejected": -121.22259521484375, "loss": 1.0369, "margin_dpo/margin_mean": 16.091880798339844, "margin_dpo/margin_std": 22.954631805419922, "step": 200 }, { "KL/chosen_KL_mean": -22.75146484375, "KL/mean": -29.275501251220703, "KL/rejected_KL_mean": -35.799537658691406, "KL/std": 20.26548957824707, "epoch": 0.30385487528344673, "fcm_dpo/beta": 0.036228105425834656, "fcm_dpo/delta": -0.07731332629919052, "fcm_dpo/margin": 13.04806900024414, "fcm_dpo/q_t": 0.3976749777793884, "grad_norm": 17.741933822631836, "learning_rate": 4.4065853017905953e-07, "logits/chosen": 0.47501319646835327, "logits/rejected": 0.426998496055603, "logps/chosen": -81.751220703125, "logps/ref_chosen": -58.999759674072266, "logps/ref_rejected": -84.67575073242188, "logps/rejected": -120.47528076171875, "loss": 1.1085, "margin_dpo/margin_mean": 13.04806900024414, "margin_dpo/margin_std": 21.63117218017578, "step": 201 }, { "KL/chosen_KL_mean": -18.69136619567871, "KL/mean": -26.704818725585938, "KL/rejected_KL_mean": -34.71826934814453, "KL/std": 19.59726333618164, "epoch": 0.30536659108087677, "fcm_dpo/beta": 0.03546854481101036, "fcm_dpo/delta": -0.1782231628894806, "fcm_dpo/margin": 16.026901245117188, "fcm_dpo/q_t": 0.37232154607772827, "grad_norm": 17.821508407592773, "learning_rate": 4.3980061644943575e-07, "logits/chosen": 0.35442060232162476, "logits/rejected": 0.2809903621673584, "logps/chosen": -66.35201263427734, "logps/ref_chosen": -47.660648345947266, "logps/ref_rejected": -73.63249969482422, "logps/rejected": -108.35076904296875, "loss": 1.0311, "margin_dpo/margin_mean": 16.026901245117188, "margin_dpo/margin_std": 21.388687133789062, "step": 202 }, { "KL/chosen_KL_mean": -23.500404357910156, "KL/mean": -30.568737030029297, "KL/rejected_KL_mean": -37.63706970214844, "KL/std": 19.979999542236328, "epoch": 0.30687830687830686, "fcm_dpo/beta": 0.03459791839122772, "fcm_dpo/delta": -0.09357498586177826, "fcm_dpo/margin": 14.136667251586914, "fcm_dpo/q_t": 0.3933200538158417, "grad_norm": 19.208473205566406, "learning_rate": 4.3893739358856455e-07, "logits/chosen": 0.46573925018310547, "logits/rejected": 0.3855220079421997, "logps/chosen": -85.82594299316406, "logps/ref_chosen": -62.32553482055664, "logps/ref_rejected": -99.37226104736328, "logps/rejected": -137.00933837890625, "loss": 1.0852, "margin_dpo/margin_mean": 14.136667251586914, "margin_dpo/margin_std": 21.907751083374023, "step": 203 }, { "KL/chosen_KL_mean": -21.444040298461914, "KL/mean": -29.211734771728516, "KL/rejected_KL_mean": -36.979427337646484, "KL/std": 21.172447204589844, "epoch": 0.30839002267573695, "fcm_dpo/beta": 0.033351823687553406, "fcm_dpo/delta": -0.12830305099487305, "fcm_dpo/margin": 15.535384178161621, "fcm_dpo/q_t": 0.38586777448654175, "grad_norm": 17.047832489013672, "learning_rate": 4.380688857426449e-07, "logits/chosen": 0.4149578809738159, "logits/rejected": 0.3435972034931183, "logps/chosen": -72.07335662841797, "logps/ref_chosen": -50.62931823730469, "logps/ref_rejected": -66.60475158691406, "logps/rejected": -103.58418273925781, "loss": 1.058, "margin_dpo/margin_mean": 15.535385131835938, "margin_dpo/margin_std": 21.494243621826172, "step": 204 }, { "KL/chosen_KL_mean": -24.892501831054688, "KL/mean": -32.499061584472656, "KL/rejected_KL_mean": -40.10561752319336, "KL/std": 21.863265991210938, "epoch": 0.30990173847316704, "fcm_dpo/beta": 0.03289718180894852, "fcm_dpo/delta": -0.10606454312801361, "fcm_dpo/margin": 15.213113784790039, "fcm_dpo/q_t": 0.39138591289520264, "grad_norm": 20.646154403686523, "learning_rate": 4.3719511720570814e-07, "logits/chosen": 0.4746994972229004, "logits/rejected": 0.40480706095695496, "logps/chosen": -95.2486801147461, "logps/ref_chosen": -70.3561782836914, "logps/ref_rejected": -93.39848327636719, "logps/rejected": -133.50408935546875, "loss": 1.092, "margin_dpo/margin_mean": 15.213114738464355, "margin_dpo/margin_std": 24.427967071533203, "step": 205 }, { "KL/chosen_KL_mean": -27.019397735595703, "KL/mean": -32.2818489074707, "KL/rejected_KL_mean": -37.54429626464844, "KL/std": 21.710651397705078, "epoch": 0.31141345427059713, "fcm_dpo/beta": 0.033220645040273666, "fcm_dpo/delta": 0.05127355828881264, "fcm_dpo/margin": 10.524890899658203, "fcm_dpo/q_t": 0.42648404836654663, "grad_norm": 18.123523712158203, "learning_rate": 4.363161124189387e-07, "logits/chosen": 0.496069073677063, "logits/rejected": 0.47948509454727173, "logps/chosen": -94.66487884521484, "logps/ref_chosen": -67.64547729492188, "logps/ref_rejected": -79.89584350585938, "logps/rejected": -117.44013977050781, "loss": 1.2423, "margin_dpo/margin_mean": 10.52489185333252, "margin_dpo/margin_std": 26.20848274230957, "step": 206 }, { "KL/chosen_KL_mean": -31.685041427612305, "KL/mean": -39.493553161621094, "KL/rejected_KL_mean": -47.30206298828125, "KL/std": 22.86937713623047, "epoch": 0.3129251700680272, "fcm_dpo/beta": 0.032525695860385895, "fcm_dpo/delta": -0.1139114499092102, "fcm_dpo/margin": 15.617015838623047, "fcm_dpo/q_t": 0.39173316955566406, "grad_norm": 15.79115104675293, "learning_rate": 4.3543189596998986e-07, "logits/chosen": 0.42408090829849243, "logits/rejected": 0.35405462980270386, "logps/chosen": -99.34923553466797, "logps/ref_chosen": -67.66419219970703, "logps/ref_rejected": -85.10249328613281, "logps/rejected": -132.40455627441406, "loss": 1.0747, "margin_dpo/margin_mean": 15.61701774597168, "margin_dpo/margin_std": 23.907955169677734, "step": 207 }, { "KL/chosen_KL_mean": -25.132305145263672, "KL/mean": -29.537364959716797, "KL/rejected_KL_mean": -33.942420959472656, "KL/std": 21.290771484375, "epoch": 0.3144368858654573, "fcm_dpo/beta": 0.032919660210609436, "fcm_dpo/delta": 0.11329277604818344, "fcm_dpo/margin": 8.810117721557617, "fcm_dpo/q_t": 0.4377458691596985, "grad_norm": 22.17745018005371, "learning_rate": 4.3454249259229664e-07, "logits/chosen": 0.4305967688560486, "logits/rejected": 0.4044501781463623, "logps/chosen": -82.864013671875, "logps/ref_chosen": -57.731712341308594, "logps/ref_rejected": -74.19276428222656, "logps/rejected": -108.13518524169922, "loss": 1.2647, "margin_dpo/margin_mean": 8.810117721557617, "margin_dpo/margin_std": 23.99683952331543, "step": 208 }, { "KL/chosen_KL_mean": -26.022214889526367, "KL/mean": -35.826934814453125, "KL/rejected_KL_mean": -45.631656646728516, "KL/std": 24.46694564819336, "epoch": 0.31594860166288735, "fcm_dpo/beta": 0.03194945305585861, "fcm_dpo/delta": -0.24169307947158813, "fcm_dpo/margin": 19.609439849853516, "fcm_dpo/q_t": 0.3655078411102295, "grad_norm": 18.581724166870117, "learning_rate": 4.336479271643833e-07, "logits/chosen": 0.382016658782959, "logits/rejected": 0.327199250459671, "logps/chosen": -94.57229614257812, "logps/ref_chosen": -68.55007934570312, "logps/ref_rejected": -87.90541076660156, "logps/rejected": -133.53707885742188, "loss": 1.0394, "margin_dpo/margin_mean": 19.609439849853516, "margin_dpo/margin_std": 28.908449172973633, "step": 209 }, { "KL/chosen_KL_mean": -25.5579891204834, "KL/mean": -34.911468505859375, "KL/rejected_KL_mean": -44.26494598388672, "KL/std": 26.126291275024414, "epoch": 0.31746031746031744, "fcm_dpo/beta": 0.030834760516881943, "fcm_dpo/delta": -0.18722277879714966, "fcm_dpo/margin": 18.706954956054688, "fcm_dpo/q_t": 0.3771812915802002, "grad_norm": 16.534156799316406, "learning_rate": 4.327482247091679e-07, "logits/chosen": 0.5074554681777954, "logits/rejected": 0.40456104278564453, "logps/chosen": -82.82626342773438, "logps/ref_chosen": -57.268272399902344, "logps/ref_rejected": -85.72807312011719, "logps/rejected": -129.99302673339844, "loss": 1.0423, "margin_dpo/margin_mean": 18.706954956054688, "margin_dpo/margin_std": 26.688194274902344, "step": 210 }, { "KL/chosen_KL_mean": -23.136999130249023, "KL/mean": -32.110008239746094, "KL/rejected_KL_mean": -41.0830192565918, "KL/std": 23.694541931152344, "epoch": 0.31897203325774753, "fcm_dpo/beta": 0.029924146831035614, "fcm_dpo/delta": -0.14464184641838074, "fcm_dpo/margin": 17.946014404296875, "fcm_dpo/q_t": 0.38534092903137207, "grad_norm": 20.95856285095215, "learning_rate": 4.3184341039326217e-07, "logits/chosen": 0.5217145681381226, "logits/rejected": 0.4215095043182373, "logps/chosen": -76.7777099609375, "logps/ref_chosen": -53.640708923339844, "logps/ref_rejected": -93.0387954711914, "logps/rejected": -134.12181091308594, "loss": 1.0483, "margin_dpo/margin_mean": 17.946014404296875, "margin_dpo/margin_std": 25.320432662963867, "step": 211 }, { "KL/chosen_KL_mean": -27.168684005737305, "KL/mean": -36.86266326904297, "KL/rejected_KL_mean": -46.556636810302734, "KL/std": 25.865568161010742, "epoch": 0.3204837490551776, "fcm_dpo/beta": 0.02877327799797058, "fcm_dpo/delta": -0.16739435493946075, "fcm_dpo/margin": 19.387954711914062, "fcm_dpo/q_t": 0.3769418001174927, "grad_norm": 14.226759910583496, "learning_rate": 4.309335095262675e-07, "logits/chosen": 0.5224308967590332, "logits/rejected": 0.4461041986942291, "logps/chosen": -84.53543090820312, "logps/ref_chosen": -57.36674499511719, "logps/ref_rejected": -79.89643096923828, "logps/rejected": -126.45306396484375, "loss": 1.0449, "margin_dpo/margin_mean": 19.387954711914062, "margin_dpo/margin_std": 27.3609619140625, "step": 212 }, { "KL/chosen_KL_mean": -20.49267578125, "KL/mean": -31.65011215209961, "KL/rejected_KL_mean": -42.80754852294922, "KL/std": 26.360469818115234, "epoch": 0.3219954648526077, "fcm_dpo/beta": 0.02747097983956337, "fcm_dpo/delta": -0.22853778302669525, "fcm_dpo/margin": 22.314865112304688, "fcm_dpo/q_t": 0.3668210506439209, "grad_norm": 13.588920593261719, "learning_rate": 4.3001854756006724e-07, "logits/chosen": 0.48388439416885376, "logits/rejected": 0.4597597122192383, "logps/chosen": -85.71379089355469, "logps/ref_chosen": -65.22111511230469, "logps/ref_rejected": -80.1810302734375, "logps/rejected": -122.98857116699219, "loss": 1.0092, "margin_dpo/margin_mean": 22.314865112304688, "margin_dpo/margin_std": 29.59410858154297, "step": 213 }, { "KL/chosen_KL_mean": -24.507936477661133, "KL/mean": -35.89101791381836, "KL/rejected_KL_mean": -47.27410125732422, "KL/std": 26.936935424804688, "epoch": 0.3235071806500378, "fcm_dpo/beta": 0.02660961076617241, "fcm_dpo/delta": -0.2185751348733902, "fcm_dpo/margin": 22.76616668701172, "fcm_dpo/q_t": 0.3699638247489929, "grad_norm": 17.4736270904541, "learning_rate": 4.290985500881143e-07, "logits/chosen": 0.36848658323287964, "logits/rejected": 0.34894949197769165, "logps/chosen": -85.80026245117188, "logps/ref_chosen": -61.292327880859375, "logps/ref_rejected": -67.69841003417969, "logps/rejected": -114.97250366210938, "loss": 1.0248, "margin_dpo/margin_mean": 22.76616668701172, "margin_dpo/margin_std": 30.800565719604492, "step": 214 }, { "KL/chosen_KL_mean": -30.14183235168457, "KL/mean": -41.4921875, "KL/rejected_KL_mean": -52.84253692626953, "KL/std": 28.326629638671875, "epoch": 0.3250188964474679, "fcm_dpo/beta": 0.025595759972929955, "fcm_dpo/delta": -0.19203141331672668, "fcm_dpo/margin": 22.700714111328125, "fcm_dpo/q_t": 0.3764479160308838, "grad_norm": 15.755528450012207, "learning_rate": 4.281735428447157e-07, "logits/chosen": 0.36972734332084656, "logits/rejected": 0.2591157853603363, "logps/chosen": -94.01097106933594, "logps/ref_chosen": -63.869136810302734, "logps/ref_rejected": -98.7657241821289, "logps/rejected": -151.60826110839844, "loss": 1.0497, "margin_dpo/margin_mean": 22.700716018676758, "margin_dpo/margin_std": 32.51849365234375, "step": 215 }, { "KL/chosen_KL_mean": -26.490917205810547, "KL/mean": -38.6580810546875, "KL/rejected_KL_mean": -50.82523727416992, "KL/std": 28.62359619140625, "epoch": 0.32653061224489793, "fcm_dpo/beta": 0.024382300674915314, "fcm_dpo/delta": -0.2054787278175354, "fcm_dpo/margin": 24.334320068359375, "fcm_dpo/q_t": 0.36771076917648315, "grad_norm": 21.268993377685547, "learning_rate": 4.2724355170431247e-07, "logits/chosen": 0.5125446915626526, "logits/rejected": 0.42290613055229187, "logps/chosen": -94.31587219238281, "logps/ref_chosen": -67.824951171875, "logps/ref_rejected": -96.40231323242188, "logps/rejected": -147.22755432128906, "loss": 0.9914, "margin_dpo/margin_mean": 24.334320068359375, "margin_dpo/margin_std": 29.2112979888916, "step": 216 }, { "KL/chosen_KL_mean": -30.6090087890625, "KL/mean": -43.44352340698242, "KL/rejected_KL_mean": -56.278038024902344, "KL/std": 30.038503646850586, "epoch": 0.328042328042328, "fcm_dpo/beta": 0.02329612709581852, "fcm_dpo/delta": -0.21151894330978394, "fcm_dpo/margin": 25.669031143188477, "fcm_dpo/q_t": 0.3687037229537964, "grad_norm": 13.630374908447266, "learning_rate": 4.26308602680756e-07, "logits/chosen": 0.46526721119880676, "logits/rejected": 0.3592798709869385, "logps/chosen": -91.11399841308594, "logps/ref_chosen": -60.5049934387207, "logps/ref_rejected": -84.26618194580078, "logps/rejected": -140.54421997070312, "loss": 0.9964, "margin_dpo/margin_mean": 25.669031143188477, "margin_dpo/margin_std": 31.728784561157227, "step": 217 }, { "KL/chosen_KL_mean": -33.458717346191406, "KL/mean": -41.555747985839844, "KL/rejected_KL_mean": -49.65278625488281, "KL/std": 29.070512771606445, "epoch": 0.3295540438397581, "fcm_dpo/beta": 0.02337442710995674, "fcm_dpo/delta": 0.020912915468215942, "fcm_dpo/margin": 16.194072723388672, "fcm_dpo/q_t": 0.4174070358276367, "grad_norm": 16.084028244018555, "learning_rate": 4.253687219265803e-07, "logits/chosen": 0.3525770306587219, "logits/rejected": 0.3516993820667267, "logps/chosen": -104.05303192138672, "logps/ref_chosen": -70.59431457519531, "logps/ref_rejected": -73.89038848876953, "logps/rejected": -123.54317474365234, "loss": 1.2067, "margin_dpo/margin_mean": 16.194074630737305, "margin_dpo/margin_std": 35.17280960083008, "step": 218 }, { "KL/chosen_KL_mean": -32.29864501953125, "KL/mean": -40.812965393066406, "KL/rejected_KL_mean": -49.32728576660156, "KL/std": 30.129287719726562, "epoch": 0.3310657596371882, "fcm_dpo/beta": 0.023134605959057808, "fcm_dpo/delta": 0.0059198178350925446, "fcm_dpo/margin": 17.028644561767578, "fcm_dpo/q_t": 0.4128570556640625, "grad_norm": 16.333072662353516, "learning_rate": 4.2442393573227043e-07, "logits/chosen": 0.44589486718177795, "logits/rejected": 0.40339186787605286, "logps/chosen": -92.78958892822266, "logps/ref_chosen": -60.490943908691406, "logps/ref_rejected": -75.85001373291016, "logps/rejected": -125.17729187011719, "loss": 1.1414, "margin_dpo/margin_mean": 17.028642654418945, "margin_dpo/margin_std": 29.682361602783203, "step": 219 }, { "KL/chosen_KL_mean": -27.624902725219727, "KL/mean": -37.65066909790039, "KL/rejected_KL_mean": -47.67643356323242, "KL/std": 30.960710525512695, "epoch": 0.3325774754346183, "fcm_dpo/beta": 0.02288152649998665, "fcm_dpo/delta": -0.06287384033203125, "fcm_dpo/margin": 20.051528930664062, "fcm_dpo/q_t": 0.40230315923690796, "grad_norm": 13.608770370483398, "learning_rate": 4.234742705255272e-07, "logits/chosen": 0.5344582796096802, "logits/rejected": 0.46789342164993286, "logps/chosen": -72.6383056640625, "logps/ref_chosen": -45.013397216796875, "logps/ref_rejected": -70.49369812011719, "logps/rejected": -118.17013549804688, "loss": 1.1337, "margin_dpo/margin_mean": 20.05152702331543, "margin_dpo/margin_std": 35.65072250366211, "step": 220 }, { "KL/chosen_KL_mean": -27.55103874206543, "KL/mean": -38.84452438354492, "KL/rejected_KL_mean": -50.13801574707031, "KL/std": 30.538475036621094, "epoch": 0.3340891912320484, "fcm_dpo/beta": 0.022582344710826874, "fcm_dpo/delta": -0.11589628458023071, "fcm_dpo/margin": 22.58698081970215, "fcm_dpo/q_t": 0.3906116783618927, "grad_norm": 16.422992706298828, "learning_rate": 4.22519752870528e-07, "logits/chosen": 0.512248158454895, "logits/rejected": 0.4338856339454651, "logps/chosen": -86.64688110351562, "logps/ref_chosen": -59.09584045410156, "logps/ref_rejected": -88.64388275146484, "logps/rejected": -138.78189086914062, "loss": 1.0787, "margin_dpo/margin_mean": 22.58698081970215, "margin_dpo/margin_std": 35.058292388916016, "step": 221 }, { "KL/chosen_KL_mean": -28.761634826660156, "KL/mean": -43.88484573364258, "KL/rejected_KL_mean": -59.00804901123047, "KL/std": 34.609222412109375, "epoch": 0.3356009070294785, "fcm_dpo/beta": 0.02151619829237461, "fcm_dpo/delta": -0.2687666416168213, "fcm_dpo/margin": 30.246416091918945, "fcm_dpo/q_t": 0.3583294153213501, "grad_norm": 15.557254791259766, "learning_rate": 4.2156040946718343e-07, "logits/chosen": 0.5562437772750854, "logits/rejected": 0.46473515033721924, "logps/chosen": -84.75932312011719, "logps/ref_chosen": -55.9976921081543, "logps/ref_rejected": -111.94727325439453, "logps/rejected": -170.955322265625, "loss": 0.9796, "margin_dpo/margin_mean": 30.246417999267578, "margin_dpo/margin_std": 37.03383255004883, "step": 222 }, { "KL/chosen_KL_mean": -34.243709564208984, "KL/mean": -47.062828063964844, "KL/rejected_KL_mean": -59.88194274902344, "KL/std": 34.03950500488281, "epoch": 0.3371126228269085, "fcm_dpo/beta": 0.020738139748573303, "fcm_dpo/delta": -0.13991129398345947, "fcm_dpo/margin": 25.63823699951172, "fcm_dpo/q_t": 0.3834461569786072, "grad_norm": 15.027467727661133, "learning_rate": 4.2059626715039065e-07, "logits/chosen": 0.5505378246307373, "logits/rejected": 0.4921608865261078, "logps/chosen": -94.1351318359375, "logps/ref_chosen": -59.891422271728516, "logps/ref_rejected": -86.28954315185547, "logps/rejected": -146.17147827148438, "loss": 1.0376, "margin_dpo/margin_mean": 25.63823699951172, "margin_dpo/margin_std": 34.131752014160156, "step": 223 }, { "KL/chosen_KL_mean": -39.61954116821289, "KL/mean": -47.04735565185547, "KL/rejected_KL_mean": -54.475162506103516, "KL/std": 33.57848358154297, "epoch": 0.3386243386243386, "fcm_dpo/beta": 0.02092786133289337, "fcm_dpo/delta": 0.0920458734035492, "fcm_dpo/margin": 14.855623245239258, "fcm_dpo/q_t": 0.4327733814716339, "grad_norm": 21.00340461730957, "learning_rate": 4.1962735288928304e-07, "logits/chosen": 0.5788267850875854, "logits/rejected": 0.5559718608856201, "logps/chosen": -103.66417694091797, "logps/ref_chosen": -64.04463195800781, "logps/ref_rejected": -75.05450439453125, "logps/rejected": -129.5296630859375, "loss": 1.2273, "margin_dpo/margin_mean": 14.855621337890625, "margin_dpo/margin_std": 35.59134292602539, "step": 224 }, { "KL/chosen_KL_mean": -37.78761291503906, "KL/mean": -51.74939727783203, "KL/rejected_KL_mean": -65.711181640625, "KL/std": 37.48065185546875, "epoch": 0.3401360544217687, "fcm_dpo/beta": 0.020355746150016785, "fcm_dpo/delta": -0.18072660267353058, "fcm_dpo/margin": 27.923568725585938, "fcm_dpo/q_t": 0.3758217692375183, "grad_norm": 14.574810028076172, "learning_rate": 4.186536937864752e-07, "logits/chosen": 0.5684969425201416, "logits/rejected": 0.4474506974220276, "logps/chosen": -103.88343811035156, "logps/ref_chosen": -66.0958251953125, "logps/ref_rejected": -97.68675231933594, "logps/rejected": -163.39793395996094, "loss": 1.0172, "margin_dpo/margin_mean": 27.923568725585938, "margin_dpo/margin_std": 35.680519104003906, "step": 225 }, { "KL/chosen_KL_mean": -35.753868103027344, "KL/mean": -46.9656982421875, "KL/rejected_KL_mean": -58.17752456665039, "KL/std": 34.91864776611328, "epoch": 0.3416477702191988, "fcm_dpo/beta": 0.020231008529663086, "fcm_dpo/delta": -0.056395478546619415, "fcm_dpo/margin": 22.42365837097168, "fcm_dpo/q_t": 0.40082597732543945, "grad_norm": 14.549175262451172, "learning_rate": 4.176753170773052e-07, "logits/chosen": 0.5831120610237122, "logits/rejected": 0.5354971885681152, "logps/chosen": -87.17073822021484, "logps/ref_chosen": -51.4168701171875, "logps/ref_rejected": -66.30068969726562, "logps/rejected": -124.47821044921875, "loss": 1.1442, "margin_dpo/margin_mean": 22.42365837097168, "margin_dpo/margin_std": 41.22904586791992, "step": 226 }, { "KL/chosen_KL_mean": -38.926422119140625, "KL/mean": -50.359954833984375, "KL/rejected_KL_mean": -61.79349136352539, "KL/std": 38.167694091796875, "epoch": 0.3431594860166289, "fcm_dpo/beta": 0.019856732338666916, "fcm_dpo/delta": -0.05691502243280411, "fcm_dpo/margin": 22.86707305908203, "fcm_dpo/q_t": 0.4031808376312256, "grad_norm": 16.369155883789062, "learning_rate": 4.166922501290729e-07, "logits/chosen": 0.6283693313598633, "logits/rejected": 0.5883187651634216, "logps/chosen": -96.91619873046875, "logps/ref_chosen": -57.989776611328125, "logps/ref_rejected": -75.05464172363281, "logps/rejected": -136.84814453125, "loss": 1.1451, "margin_dpo/margin_mean": 22.867074966430664, "margin_dpo/margin_std": 42.62117004394531, "step": 227 }, { "KL/chosen_KL_mean": -42.122718811035156, "KL/mean": -54.82220458984375, "KL/rejected_KL_mean": -67.52168273925781, "KL/std": 36.266571044921875, "epoch": 0.34467120181405897, "fcm_dpo/beta": 0.019605718553066254, "fcm_dpo/delta": -0.10297306627035141, "fcm_dpo/margin": 25.39896011352539, "fcm_dpo/q_t": 0.39041581749916077, "grad_norm": 15.85113525390625, "learning_rate": 4.1570452044027405e-07, "logits/chosen": 0.5884374976158142, "logits/rejected": 0.5055756568908691, "logps/chosen": -97.68208312988281, "logps/ref_chosen": -55.55936813354492, "logps/ref_rejected": -77.02364349365234, "logps/rejected": -144.54531860351562, "loss": 1.0724, "margin_dpo/margin_mean": 25.39896011352539, "margin_dpo/margin_std": 37.421138763427734, "step": 228 }, { "KL/chosen_KL_mean": -34.708621978759766, "KL/mean": -46.20718765258789, "KL/rejected_KL_mean": -57.70575714111328, "KL/std": 35.467079162597656, "epoch": 0.34618291761148906, "fcm_dpo/beta": 0.019299227744340897, "fcm_dpo/delta": -0.04590361565351486, "fcm_dpo/margin": 22.997133255004883, "fcm_dpo/q_t": 0.4034169912338257, "grad_norm": 22.405696868896484, "learning_rate": 4.147121556398312e-07, "logits/chosen": 0.6827352046966553, "logits/rejected": 0.6061787605285645, "logps/chosen": -85.50328063964844, "logps/ref_chosen": -50.79466247558594, "logps/ref_rejected": -78.4474105834961, "logps/rejected": -136.15316772460938, "loss": 1.1506, "margin_dpo/margin_mean": 22.997133255004883, "margin_dpo/margin_std": 43.55659484863281, "step": 229 }, { "KL/chosen_KL_mean": -38.71097946166992, "KL/mean": -52.039764404296875, "KL/rejected_KL_mean": -65.36854553222656, "KL/std": 39.001365661621094, "epoch": 0.3476946334089191, "fcm_dpo/beta": 0.01911795511841774, "fcm_dpo/delta": -0.1162446141242981, "fcm_dpo/margin": 26.657567977905273, "fcm_dpo/q_t": 0.3882319927215576, "grad_norm": 14.44546890258789, "learning_rate": 4.137151834863213e-07, "logits/chosen": 0.5673795938491821, "logits/rejected": 0.5688091516494751, "logps/chosen": -95.44020080566406, "logps/ref_chosen": -56.729225158691406, "logps/ref_rejected": -62.99180603027344, "logps/rejected": -128.3603515625, "loss": 1.0652, "margin_dpo/margin_mean": 26.65756607055664, "margin_dpo/margin_std": 37.82829284667969, "step": 230 }, { "KL/chosen_KL_mean": -44.53082275390625, "KL/mean": -64.40026092529297, "KL/rejected_KL_mean": -84.26969909667969, "KL/std": 38.41732406616211, "epoch": 0.3492063492063492, "fcm_dpo/beta": 0.017824435606598854, "fcm_dpo/delta": -0.3340993821620941, "fcm_dpo/margin": 39.738887786865234, "fcm_dpo/q_t": 0.3404316008090973, "grad_norm": 14.944037437438965, "learning_rate": 4.1271363186719835e-07, "logits/chosen": 0.49198731780052185, "logits/rejected": 0.4875026047229767, "logps/chosen": -117.12792205810547, "logps/ref_chosen": -72.59709930419922, "logps/ref_rejected": -86.2322998046875, "logps/rejected": -170.5019989013672, "loss": 0.9208, "margin_dpo/margin_mean": 39.73888397216797, "margin_dpo/margin_std": 41.09081268310547, "step": 231 }, { "KL/chosen_KL_mean": -44.682891845703125, "KL/mean": -58.00080108642578, "KL/rejected_KL_mean": -71.31871032714844, "KL/std": 41.824562072753906, "epoch": 0.3507180650037793, "fcm_dpo/beta": 0.017421916127204895, "fcm_dpo/delta": -0.06721210479736328, "fcm_dpo/margin": 26.635814666748047, "fcm_dpo/q_t": 0.4011406898498535, "grad_norm": 15.53917407989502, "learning_rate": 4.1170752879801436e-07, "logits/chosen": 0.539695143699646, "logits/rejected": 0.5095845460891724, "logps/chosen": -112.80142211914062, "logps/ref_chosen": -68.1185302734375, "logps/ref_rejected": -83.79415893554688, "logps/rejected": -155.1128692626953, "loss": 1.1295, "margin_dpo/margin_mean": 26.635812759399414, "margin_dpo/margin_std": 47.687801361083984, "step": 232 }, { "KL/chosen_KL_mean": -51.6092643737793, "KL/mean": -61.849483489990234, "KL/rejected_KL_mean": -72.08970642089844, "KL/std": 39.40514373779297, "epoch": 0.35222978080120937, "fcm_dpo/beta": 0.01731543242931366, "fcm_dpo/delta": 0.04540575295686722, "fcm_dpo/margin": 20.48044204711914, "fcm_dpo/q_t": 0.4230949878692627, "grad_norm": 14.722317695617676, "learning_rate": 4.106969024216348e-07, "logits/chosen": 0.599013090133667, "logits/rejected": 0.5388975739479065, "logps/chosen": -106.67941284179688, "logps/ref_chosen": -55.070152282714844, "logps/ref_rejected": -66.61845397949219, "logps/rejected": -138.70816040039062, "loss": 1.1796, "margin_dpo/margin_mean": 20.48044204711914, "margin_dpo/margin_std": 39.12367248535156, "step": 233 }, { "KL/chosen_KL_mean": -46.10203170776367, "KL/mean": -56.328800201416016, "KL/rejected_KL_mean": -66.5555648803711, "KL/std": 38.45304489135742, "epoch": 0.35374149659863946, "fcm_dpo/beta": 0.017331501469016075, "fcm_dpo/delta": -0.11514575034379959, "fcm_dpo/margin": 20.453533172607422, "fcm_dpo/q_t": 0.42306917905807495, "grad_norm": 20.012310028076172, "learning_rate": 4.09681781007452e-07, "logits/chosen": 0.5130677819252014, "logits/rejected": 0.506496787071228, "logps/chosen": -102.02792358398438, "logps/ref_chosen": -55.92589569091797, "logps/ref_rejected": -51.11608123779297, "logps/rejected": -117.67164611816406, "loss": 1.2025, "margin_dpo/margin_mean": 20.45353126525879, "margin_dpo/margin_std": 41.53302001953125, "step": 234 }, { "KL/chosen_KL_mean": -40.315879821777344, "KL/mean": -57.65726089477539, "KL/rejected_KL_mean": -74.99864196777344, "KL/std": 41.41876220703125, "epoch": 0.35525321239606955, "fcm_dpo/beta": 0.016705987975001335, "fcm_dpo/delta": -0.1900615096092224, "fcm_dpo/margin": 34.68276596069336, "fcm_dpo/q_t": 0.36982935667037964, "grad_norm": 14.829582214355469, "learning_rate": 4.08662192950594e-07, "logits/chosen": 0.6136411428451538, "logits/rejected": 0.5992604494094849, "logps/chosen": -104.85560607910156, "logps/ref_chosen": -64.53972625732422, "logps/ref_rejected": -77.69151306152344, "logps/rejected": -152.69015502929688, "loss": 0.9932, "margin_dpo/margin_mean": 34.682762145996094, "margin_dpo/margin_std": 40.54000473022461, "step": 235 }, { "KL/chosen_KL_mean": -61.0404052734375, "KL/mean": -74.00381469726562, "KL/rejected_KL_mean": -86.96721649169922, "KL/std": 40.741188049316406, "epoch": 0.35676492819349964, "fcm_dpo/beta": 0.016382815316319466, "fcm_dpo/delta": -0.02608785405755043, "fcm_dpo/margin": 25.926815032958984, "fcm_dpo/q_t": 0.4062482714653015, "grad_norm": 14.39378547668457, "learning_rate": 4.076381667711306e-07, "logits/chosen": 0.5750887393951416, "logits/rejected": 0.5613845586776733, "logps/chosen": -132.1951446533203, "logps/ref_chosen": -71.15473937988281, "logps/ref_rejected": -84.88541412353516, "logps/rejected": -171.85263061523438, "loss": 1.1396, "margin_dpo/margin_mean": 25.926815032958984, "margin_dpo/margin_std": 46.30166244506836, "step": 236 }, { "KL/chosen_KL_mean": -51.05704116821289, "KL/mean": -66.93670654296875, "KL/rejected_KL_mean": -82.8163833618164, "KL/std": 40.99628448486328, "epoch": 0.35827664399092973, "fcm_dpo/beta": 0.01615855097770691, "fcm_dpo/delta": -0.1191396713256836, "fcm_dpo/margin": 31.75934410095215, "fcm_dpo/q_t": 0.3859434723854065, "grad_norm": 17.10631561279297, "learning_rate": 4.066097311132753e-07, "logits/chosen": 0.5949573516845703, "logits/rejected": 0.5852821469306946, "logps/chosen": -127.19905090332031, "logps/ref_chosen": -76.14201354980469, "logps/ref_rejected": -80.88479614257812, "logps/rejected": -163.701171875, "loss": 1.0751, "margin_dpo/margin_mean": 31.759342193603516, "margin_dpo/margin_std": 47.10854721069336, "step": 237 }, { "KL/chosen_KL_mean": -45.8564567565918, "KL/mean": -61.5413932800293, "KL/rejected_KL_mean": -77.22633361816406, "KL/std": 42.1279411315918, "epoch": 0.35978835978835977, "fcm_dpo/beta": 0.015698228031396866, "fcm_dpo/delta": -0.09846552461385727, "fcm_dpo/margin": 31.36986541748047, "fcm_dpo/q_t": 0.39032959938049316, "grad_norm": 19.96809959411621, "learning_rate": 4.0557691474458414e-07, "logits/chosen": 0.5656386017799377, "logits/rejected": 0.5574653148651123, "logps/chosen": -114.74130249023438, "logps/ref_chosen": -68.88484954833984, "logps/ref_rejected": -75.8946304321289, "logps/rejected": -153.1209716796875, "loss": 1.0689, "margin_dpo/margin_mean": 31.36986541748047, "margin_dpo/margin_std": 44.51061248779297, "step": 238 }, { "KL/chosen_KL_mean": -56.033790588378906, "KL/mean": -71.40078735351562, "KL/rejected_KL_mean": -86.76777648925781, "KL/std": 45.64823913574219, "epoch": 0.36130007558578986, "fcm_dpo/beta": 0.015608660876750946, "fcm_dpo/delta": -0.08397899568080902, "fcm_dpo/margin": 30.733985900878906, "fcm_dpo/q_t": 0.39459508657455444, "grad_norm": 18.36031723022461, "learning_rate": 4.045397465551513e-07, "logits/chosen": 0.6763525009155273, "logits/rejected": 0.5366228222846985, "logps/chosen": -112.80561828613281, "logps/ref_chosen": -56.771827697753906, "logps/ref_rejected": -116.23050689697266, "logps/rejected": -202.998291015625, "loss": 1.0954, "margin_dpo/margin_mean": 30.733989715576172, "margin_dpo/margin_std": 47.917232513427734, "step": 239 }, { "KL/chosen_KL_mean": -54.066680908203125, "KL/mean": -73.76835632324219, "KL/rejected_KL_mean": -93.47004699707031, "KL/std": 45.91557312011719, "epoch": 0.36281179138321995, "fcm_dpo/beta": 0.014995868317782879, "fcm_dpo/delta": -0.2027919590473175, "fcm_dpo/margin": 39.40336608886719, "fcm_dpo/q_t": 0.36883121728897095, "grad_norm": 13.07312297821045, "learning_rate": 4.0349825555680045e-07, "logits/chosen": 0.5906857848167419, "logits/rejected": 0.49216535687446594, "logps/chosen": -107.42079162597656, "logps/ref_chosen": -53.35411071777344, "logps/ref_rejected": -80.12019348144531, "logps/rejected": -173.59024047851562, "loss": 0.996, "margin_dpo/margin_mean": 39.40336608886719, "margin_dpo/margin_std": 47.73066711425781, "step": 240 }, { "KL/chosen_KL_mean": -50.90334701538086, "KL/mean": -63.605812072753906, "KL/rejected_KL_mean": -76.30827331542969, "KL/std": 42.9580078125, "epoch": 0.36432350718065004, "fcm_dpo/beta": 0.0149228535592556, "fcm_dpo/delta": 0.021481888368725777, "fcm_dpo/margin": 25.404935836791992, "fcm_dpo/q_t": 0.4165813624858856, "grad_norm": 17.74489402770996, "learning_rate": 4.0245247088227377e-07, "logits/chosen": 0.5852094292640686, "logits/rejected": 0.5466130375862122, "logps/chosen": -122.79875946044922, "logps/ref_chosen": -71.89541625976562, "logps/ref_rejected": -83.03492736816406, "logps/rejected": -159.34320068359375, "loss": 1.1587, "margin_dpo/margin_mean": 25.404937744140625, "margin_dpo/margin_std": 47.308311462402344, "step": 241 }, { "KL/chosen_KL_mean": -52.51380920410156, "KL/mean": -69.51727294921875, "KL/rejected_KL_mean": -86.52073669433594, "KL/std": 45.80069351196289, "epoch": 0.36583522297808013, "fcm_dpo/beta": 0.014580689370632172, "fcm_dpo/delta": -0.10227706283330917, "fcm_dpo/margin": 34.006919860839844, "fcm_dpo/q_t": 0.3903062343597412, "grad_norm": 13.116487503051758, "learning_rate": 4.0140242178441665e-07, "logits/chosen": 0.5915254354476929, "logits/rejected": 0.5709429383277893, "logps/chosen": -110.44123840332031, "logps/ref_chosen": -57.927433013916016, "logps/ref_rejected": -67.838623046875, "logps/rejected": -154.35934448242188, "loss": 1.0696, "margin_dpo/margin_mean": 34.00692367553711, "margin_dpo/margin_std": 49.14616394042969, "step": 242 }, { "KL/chosen_KL_mean": -51.1470832824707, "KL/mean": -66.16419982910156, "KL/rejected_KL_mean": -81.18131256103516, "KL/std": 45.306007385253906, "epoch": 0.3673469387755102, "fcm_dpo/beta": 0.014499957673251629, "fcm_dpo/delta": -0.03716906160116196, "fcm_dpo/margin": 30.034236907958984, "fcm_dpo/q_t": 0.40246307849884033, "grad_norm": 17.26784896850586, "learning_rate": 4.003481376353596e-07, "logits/chosen": 0.5729988813400269, "logits/rejected": 0.5832550525665283, "logps/chosen": -125.42375946044922, "logps/ref_chosen": -74.27667236328125, "logps/ref_rejected": -73.24340057373047, "logps/rejected": -154.42471313476562, "loss": 1.104, "margin_dpo/margin_mean": 30.034236907958984, "margin_dpo/margin_std": 46.801856994628906, "step": 243 }, { "KL/chosen_KL_mean": -48.39063262939453, "KL/mean": -68.6911849975586, "KL/rejected_KL_mean": -88.99174499511719, "KL/std": 45.00447082519531, "epoch": 0.3688586545729403, "fcm_dpo/beta": 0.01415896974503994, "fcm_dpo/delta": -0.18511970341205597, "fcm_dpo/margin": 40.60111999511719, "fcm_dpo/q_t": 0.3705122470855713, "grad_norm": 13.084891319274902, "learning_rate": 3.9928964792569654e-07, "logits/chosen": 0.6256475448608398, "logits/rejected": 0.5371542572975159, "logps/chosen": -101.75453186035156, "logps/ref_chosen": -53.36390686035156, "logps/ref_rejected": -71.10276794433594, "logps/rejected": -160.09451293945312, "loss": 0.993, "margin_dpo/margin_mean": 40.60111999511719, "margin_dpo/margin_std": 46.31639099121094, "step": 244 }, { "KL/chosen_KL_mean": -53.580116271972656, "KL/mean": -76.88776397705078, "KL/rejected_KL_mean": -100.19540405273438, "KL/std": 46.22126770019531, "epoch": 0.37037037037037035, "fcm_dpo/beta": 0.01345006562769413, "fcm_dpo/delta": -0.2425159513950348, "fcm_dpo/margin": 46.61528015136719, "fcm_dpo/q_t": 0.35643166303634644, "grad_norm": 17.690349578857422, "learning_rate": 3.982269822636601e-07, "logits/chosen": 0.6387679576873779, "logits/rejected": 0.6115980744361877, "logps/chosen": -124.77522277832031, "logps/ref_chosen": -71.19510650634766, "logps/ref_rejected": -80.76235961914062, "logps/rejected": -180.957763671875, "loss": 0.9451, "margin_dpo/margin_mean": 46.61528015136719, "margin_dpo/margin_std": 46.82879638671875, "step": 245 }, { "KL/chosen_KL_mean": -63.579872131347656, "KL/mean": -81.17933654785156, "KL/rejected_KL_mean": -98.77880859375, "KL/std": 46.690460205078125, "epoch": 0.37188208616780044, "fcm_dpo/beta": 0.013229056261479855, "fcm_dpo/delta": -0.06901253014802933, "fcm_dpo/margin": 35.198936462402344, "fcm_dpo/q_t": 0.3967069089412689, "grad_norm": 14.13879108428955, "learning_rate": 3.971601703742932e-07, "logits/chosen": 0.6410825252532959, "logits/rejected": 0.5778796672821045, "logps/chosen": -135.200927734375, "logps/ref_chosen": -71.62104797363281, "logps/ref_rejected": -94.03392028808594, "logps/rejected": -192.81272888183594, "loss": 1.1142, "margin_dpo/margin_mean": 35.198936462402344, "margin_dpo/margin_std": 58.479515075683594, "step": 246 }, { "KL/chosen_KL_mean": -67.45832824707031, "KL/mean": -78.08027648925781, "KL/rejected_KL_mean": -88.70221710205078, "KL/std": 45.65076446533203, "epoch": 0.37339380196523053, "fcm_dpo/beta": 0.013414295390248299, "fcm_dpo/delta": 0.11734433472156525, "fcm_dpo/margin": 21.243894577026367, "fcm_dpo/q_t": 0.4357885718345642, "grad_norm": 17.8750057220459, "learning_rate": 3.960892420986177e-07, "logits/chosen": 0.6398141980171204, "logits/rejected": 0.6296453475952148, "logps/chosen": -147.48086547851562, "logps/ref_chosen": -80.02254486083984, "logps/ref_rejected": -89.22705841064453, "logps/rejected": -177.9292755126953, "loss": 1.2307, "margin_dpo/margin_mean": 21.243894577026367, "margin_dpo/margin_std": 49.95774841308594, "step": 247 }, { "KL/chosen_KL_mean": -57.298828125, "KL/mean": -76.10064697265625, "KL/rejected_KL_mean": -94.9024658203125, "KL/std": 50.01144790649414, "epoch": 0.3749055177626606, "fcm_dpo/beta": 0.013268327340483665, "fcm_dpo/delta": -0.10401751101016998, "fcm_dpo/margin": 37.6036491394043, "fcm_dpo/q_t": 0.3911561369895935, "grad_norm": 15.191692352294922, "learning_rate": 3.9501422739279953e-07, "logits/chosen": 0.6122466325759888, "logits/rejected": 0.6708812117576599, "logps/chosen": -122.67678833007812, "logps/ref_chosen": -65.37796020507812, "logps/ref_rejected": -61.365787506103516, "logps/rejected": -156.26824951171875, "loss": 1.0769, "margin_dpo/margin_mean": 37.6036491394043, "margin_dpo/margin_std": 56.18958282470703, "step": 248 }, { "KL/chosen_KL_mean": -72.68350219726562, "KL/mean": -80.0612564086914, "KL/rejected_KL_mean": -87.43902587890625, "KL/std": 46.41590118408203, "epoch": 0.3764172335600907, "fcm_dpo/beta": 0.013199391774833202, "fcm_dpo/delta": 0.035816043615341187, "fcm_dpo/margin": 14.755523681640625, "fcm_dpo/q_t": 0.4553859233856201, "grad_norm": 17.512218475341797, "learning_rate": 3.9393515632731094e-07, "logits/chosen": 0.6221470236778259, "logits/rejected": 0.6636273860931396, "logps/chosen": -147.2849578857422, "logps/ref_chosen": -74.60145568847656, "logps/ref_rejected": -63.79338455200195, "logps/rejected": -151.23240661621094, "loss": 1.3242, "margin_dpo/margin_mean": 14.755523681640625, "margin_dpo/margin_std": 53.123321533203125, "step": 249 }, { "KL/chosen_KL_mean": -62.269371032714844, "KL/mean": -81.58802032470703, "KL/rejected_KL_mean": -100.90668487548828, "KL/std": 50.64720153808594, "epoch": 0.3779289493575208, "fcm_dpo/beta": 0.013095343485474586, "fcm_dpo/delta": -0.111575648188591, "fcm_dpo/margin": 38.637306213378906, "fcm_dpo/q_t": 0.3866844177246094, "grad_norm": 13.674488067626953, "learning_rate": 3.9285205908608934e-07, "logits/chosen": 0.7028018236160278, "logits/rejected": 0.6593612432479858, "logps/chosen": -124.20758056640625, "logps/ref_chosen": -61.938209533691406, "logps/ref_rejected": -72.21602630615234, "logps/rejected": -173.12271118164062, "loss": 1.0556, "margin_dpo/margin_mean": 38.63731002807617, "margin_dpo/margin_std": 53.086402893066406, "step": 250 }, { "KL/chosen_KL_mean": -68.74446105957031, "KL/mean": -80.33143615722656, "KL/rejected_KL_mean": -91.91839599609375, "KL/std": 46.451534271240234, "epoch": 0.3794406651549509, "fcm_dpo/beta": 0.013123005628585815, "fcm_dpo/delta": 0.09900989383459091, "fcm_dpo/margin": 23.173925399780273, "fcm_dpo/q_t": 0.4319212734699249, "grad_norm": 21.88115119934082, "learning_rate": 3.9176496596569265e-07, "logits/chosen": 0.6787819862365723, "logits/rejected": 0.6353781223297119, "logps/chosen": -135.60140991210938, "logps/ref_chosen": -66.85694885253906, "logps/ref_rejected": -84.83396911621094, "logps/rejected": -176.7523651123047, "loss": 1.2205, "margin_dpo/margin_mean": 23.173927307128906, "margin_dpo/margin_std": 53.315067291259766, "step": 251 }, { "KL/chosen_KL_mean": -63.414527893066406, "KL/mean": -73.9803466796875, "KL/rejected_KL_mean": -84.54615783691406, "KL/std": 48.33477020263672, "epoch": 0.38095238095238093, "fcm_dpo/beta": 0.013572916388511658, "fcm_dpo/delta": 0.1134091168642044, "fcm_dpo/margin": 21.131629943847656, "fcm_dpo/q_t": 0.4354143738746643, "grad_norm": 22.012300491333008, "learning_rate": 3.9067390737445254e-07, "logits/chosen": 0.6178069114685059, "logits/rejected": 0.5614415407180786, "logps/chosen": -119.63845825195312, "logps/ref_chosen": -56.22393035888672, "logps/ref_rejected": -77.1136245727539, "logps/rejected": -161.6597900390625, "loss": 1.2487, "margin_dpo/margin_mean": 21.131629943847656, "margin_dpo/margin_std": 52.747032165527344, "step": 252 }, { "KL/chosen_KL_mean": -63.61356735229492, "KL/mean": -77.61599731445312, "KL/rejected_KL_mean": -91.61842346191406, "KL/std": 47.1943244934082, "epoch": 0.382464096749811, "fcm_dpo/beta": 0.013471575453877449, "fcm_dpo/delta": 0.0228419192135334, "fcm_dpo/margin": 28.00486183166504, "fcm_dpo/q_t": 0.4155396521091461, "grad_norm": 14.961233139038086, "learning_rate": 3.8957891383162304e-07, "logits/chosen": 0.7191380262374878, "logits/rejected": 0.6768438220024109, "logps/chosen": -115.82357788085938, "logps/ref_chosen": -52.21001434326172, "logps/ref_rejected": -58.75764846801758, "logps/rejected": -150.37606811523438, "loss": 1.1456, "margin_dpo/margin_mean": 28.004863739013672, "margin_dpo/margin_std": 47.672386169433594, "step": 253 }, { "KL/chosen_KL_mean": -68.19625854492188, "KL/mean": -83.46250915527344, "KL/rejected_KL_mean": -98.72874450683594, "KL/std": 49.697425842285156, "epoch": 0.3839758125472411, "fcm_dpo/beta": 0.01349942572414875, "fcm_dpo/delta": -0.013055309653282166, "fcm_dpo/margin": 30.532485961914062, "fcm_dpo/q_t": 0.40936195850372314, "grad_norm": 14.441313743591309, "learning_rate": 3.884800159665276e-07, "logits/chosen": 0.6286872625350952, "logits/rejected": 0.5743802189826965, "logps/chosen": -133.83258056640625, "logps/ref_chosen": -65.63632202148438, "logps/ref_rejected": -82.34425354003906, "logps/rejected": -181.072998046875, "loss": 1.1219, "margin_dpo/margin_mean": 30.532485961914062, "margin_dpo/margin_std": 49.943382263183594, "step": 254 }, { "KL/chosen_KL_mean": -60.290618896484375, "KL/mean": -79.58023071289062, "KL/rejected_KL_mean": -98.86984252929688, "KL/std": 49.87480926513672, "epoch": 0.3854875283446712, "fcm_dpo/beta": 0.013280254788696766, "fcm_dpo/delta": -0.11871804296970367, "fcm_dpo/margin": 38.579219818115234, "fcm_dpo/q_t": 0.38657641410827637, "grad_norm": 18.216447830200195, "learning_rate": 3.873772445177015e-07, "logits/chosen": 0.5999297499656677, "logits/rejected": 0.5706052184104919, "logps/chosen": -128.2017059326172, "logps/ref_chosen": -67.91108703613281, "logps/ref_rejected": -83.89114379882812, "logps/rejected": -182.760986328125, "loss": 1.0558, "margin_dpo/margin_mean": 38.57921600341797, "margin_dpo/margin_std": 54.264957427978516, "step": 255 }, { "KL/chosen_KL_mean": -73.21519470214844, "KL/mean": -90.58259582519531, "KL/rejected_KL_mean": -107.94999694824219, "KL/std": 46.76201629638672, "epoch": 0.3869992441421013, "fcm_dpo/beta": 0.013069174252450466, "fcm_dpo/delta": -0.05695026367902756, "fcm_dpo/margin": 34.73479080200195, "fcm_dpo/q_t": 0.40087705850601196, "grad_norm": 17.64289093017578, "learning_rate": 3.862706303320329e-07, "logits/chosen": 0.6389260292053223, "logits/rejected": 0.570476770401001, "logps/chosen": -136.71517944335938, "logps/ref_chosen": -63.49998474121094, "logps/ref_rejected": -90.77104187011719, "logps/rejected": -198.72103881835938, "loss": 1.1229, "margin_dpo/margin_mean": 34.73479080200195, "margin_dpo/margin_std": 59.88385772705078, "step": 256 }, { "KL/chosen_KL_mean": -69.81033325195312, "KL/mean": -91.40444946289062, "KL/rejected_KL_mean": -112.9985580444336, "KL/std": 52.70113754272461, "epoch": 0.3885109599395314, "fcm_dpo/beta": 0.012741890735924244, "fcm_dpo/delta": -0.15949344635009766, "fcm_dpo/margin": 43.188232421875, "fcm_dpo/q_t": 0.37954503297805786, "grad_norm": 14.254016876220703, "learning_rate": 3.851602043638994e-07, "logits/chosen": 0.6155068874359131, "logits/rejected": 0.5465556383132935, "logps/chosen": -140.41098022460938, "logps/ref_chosen": -70.60064697265625, "logps/ref_rejected": -108.58313751220703, "logps/rejected": -221.58169555664062, "loss": 1.0363, "margin_dpo/margin_mean": 43.188232421875, "margin_dpo/margin_std": 59.28783416748047, "step": 257 }, { "KL/chosen_KL_mean": -70.38436889648438, "KL/mean": -86.61546325683594, "KL/rejected_KL_mean": -102.84657287597656, "KL/std": 45.8340950012207, "epoch": 0.3900226757369615, "fcm_dpo/beta": 0.012654589489102364, "fcm_dpo/delta": -0.01126368809491396, "fcm_dpo/margin": 32.46220397949219, "fcm_dpo/q_t": 0.40449464321136475, "grad_norm": 16.840524673461914, "learning_rate": 3.840459976743023e-07, "logits/chosen": 0.6729997396469116, "logits/rejected": 0.6183003187179565, "logps/chosen": -129.63853454589844, "logps/ref_chosen": -59.25416564941406, "logps/ref_rejected": -85.58709716796875, "logps/rejected": -188.43365478515625, "loss": 1.0877, "margin_dpo/margin_mean": 32.46220397949219, "margin_dpo/margin_std": 43.45000457763672, "step": 258 }, { "KL/chosen_KL_mean": -60.93222427368164, "KL/mean": -85.822998046875, "KL/rejected_KL_mean": -110.71377563476562, "KL/std": 51.75875473022461, "epoch": 0.3915343915343915, "fcm_dpo/beta": 0.012163689360022545, "fcm_dpo/delta": -0.2205704152584076, "fcm_dpo/margin": 49.78155517578125, "fcm_dpo/q_t": 0.36391448974609375, "grad_norm": 14.892366409301758, "learning_rate": 3.8292804142999796e-07, "logits/chosen": 0.5799644589424133, "logits/rejected": 0.47284597158432007, "logps/chosen": -126.36709594726562, "logps/ref_chosen": -65.43487548828125, "logps/ref_rejected": -95.41731262207031, "logps/rejected": -206.13108825683594, "loss": 0.9932, "margin_dpo/margin_mean": 49.78155517578125, "margin_dpo/margin_std": 59.923851013183594, "step": 259 }, { "KL/chosen_KL_mean": -61.398841857910156, "KL/mean": -81.24398803710938, "KL/rejected_KL_mean": -101.08912658691406, "KL/std": 49.07374572753906, "epoch": 0.3930461073318216, "fcm_dpo/beta": 0.011900585144758224, "fcm_dpo/delta": -0.07667770981788635, "fcm_dpo/margin": 39.69029235839844, "fcm_dpo/q_t": 0.3952261209487915, "grad_norm": 15.002617835998535, "learning_rate": 3.818063669026256e-07, "logits/chosen": 0.6235231161117554, "logits/rejected": 0.5295395851135254, "logps/chosen": -110.48841857910156, "logps/ref_chosen": -49.08958435058594, "logps/ref_rejected": -79.01708221435547, "logps/rejected": -180.106201171875, "loss": 1.0902, "margin_dpo/margin_mean": 39.69029235839844, "margin_dpo/margin_std": 60.59455871582031, "step": 260 }, { "KL/chosen_KL_mean": -60.26841354370117, "KL/mean": -77.18064880371094, "KL/rejected_KL_mean": -94.09290313720703, "KL/std": 47.547637939453125, "epoch": 0.3945578231292517, "fcm_dpo/beta": 0.011928768828511238, "fcm_dpo/delta": -0.0037367884069681168, "fcm_dpo/margin": 33.82448959350586, "fcm_dpo/q_t": 0.40984904766082764, "grad_norm": 16.326772689819336, "learning_rate": 3.806810054678331e-07, "logits/chosen": 0.5075196027755737, "logits/rejected": 0.5455510020256042, "logps/chosen": -131.14080810546875, "logps/ref_chosen": -70.87239074707031, "logps/ref_rejected": -65.01522064208984, "logps/rejected": -159.10812377929688, "loss": 1.1194, "margin_dpo/margin_mean": 33.824493408203125, "margin_dpo/margin_std": 53.84626007080078, "step": 261 }, { "KL/chosen_KL_mean": -65.84159851074219, "KL/mean": -81.57791900634766, "KL/rejected_KL_mean": -97.31422424316406, "KL/std": 46.80791091918945, "epoch": 0.3960695389266818, "fcm_dpo/beta": 0.011966339312493801, "fcm_dpo/delta": 0.02416188269853592, "fcm_dpo/margin": 31.47262191772461, "fcm_dpo/q_t": 0.41486579179763794, "grad_norm": 15.958047866821289, "learning_rate": 3.7955198860439887e-07, "logits/chosen": 0.6846475601196289, "logits/rejected": 0.6157461404800415, "logps/chosen": -133.71221923828125, "logps/ref_chosen": -67.8706283569336, "logps/ref_rejected": -88.7205810546875, "logps/rejected": -186.03480529785156, "loss": 1.1325, "margin_dpo/margin_mean": 31.472623825073242, "margin_dpo/margin_std": 50.958343505859375, "step": 262 }, { "KL/chosen_KL_mean": -61.87702941894531, "KL/mean": -77.3545913696289, "KL/rejected_KL_mean": -92.8321533203125, "KL/std": 47.284324645996094, "epoch": 0.3975812547241119, "fcm_dpo/beta": 0.012026961892843246, "fcm_dpo/delta": 0.028671840205788612, "fcm_dpo/margin": 30.955129623413086, "fcm_dpo/q_t": 0.41534680128097534, "grad_norm": 14.962888717651367, "learning_rate": 3.784193478933516e-07, "logits/chosen": 0.5917189121246338, "logits/rejected": 0.47899508476257324, "logps/chosen": -117.07160949707031, "logps/ref_chosen": -55.194583892822266, "logps/ref_rejected": -80.54048156738281, "logps/rejected": -173.37265014648438, "loss": 1.141, "margin_dpo/margin_mean": 30.95513153076172, "margin_dpo/margin_std": 52.228553771972656, "step": 263 }, { "KL/chosen_KL_mean": -64.47840881347656, "KL/mean": -81.5421142578125, "KL/rejected_KL_mean": -98.60582733154297, "KL/std": 51.82183074951172, "epoch": 0.39909297052154197, "fcm_dpo/beta": 0.011982956901192665, "fcm_dpo/delta": -0.009435471147298813, "fcm_dpo/margin": 34.127403259277344, "fcm_dpo/q_t": 0.40906399488449097, "grad_norm": 14.751063346862793, "learning_rate": 3.7728311501708674e-07, "logits/chosen": 0.5228692889213562, "logits/rejected": 0.47892439365386963, "logps/chosen": -147.6490936279297, "logps/ref_chosen": -83.17068481445312, "logps/ref_rejected": -88.33625793457031, "logps/rejected": -186.94207763671875, "loss": 1.13, "margin_dpo/margin_mean": 34.127403259277344, "margin_dpo/margin_std": 57.67744445800781, "step": 264 }, { "KL/chosen_KL_mean": -62.788230895996094, "KL/mean": -81.142822265625, "KL/rejected_KL_mean": -99.49742889404297, "KL/std": 52.152587890625, "epoch": 0.40060468631897206, "fcm_dpo/beta": 0.011970747262239456, "fcm_dpo/delta": -0.04135889559984207, "fcm_dpo/margin": 36.709197998046875, "fcm_dpo/q_t": 0.40274354815483093, "grad_norm": 12.999602317810059, "learning_rate": 3.7614332175848027e-07, "logits/chosen": 0.6839252710342407, "logits/rejected": 0.6158575415611267, "logps/chosen": -114.45108032226562, "logps/ref_chosen": -51.66284942626953, "logps/ref_rejected": -67.1720962524414, "logps/rejected": -166.66952514648438, "loss": 1.1362, "margin_dpo/margin_mean": 36.709197998046875, "margin_dpo/margin_std": 64.98656463623047, "step": 265 }, { "KL/chosen_KL_mean": -59.46649169921875, "KL/mean": -77.75037384033203, "KL/rejected_KL_mean": -96.03425598144531, "KL/std": 52.215267181396484, "epoch": 0.4021164021164021, "fcm_dpo/beta": 0.011831636540591717, "fcm_dpo/delta": -0.03418291360139847, "fcm_dpo/margin": 36.56776428222656, "fcm_dpo/q_t": 0.4037179946899414, "grad_norm": 17.84242057800293, "learning_rate": 3.75e-07, "logits/chosen": 0.5979915261268616, "logits/rejected": 0.5204529762268066, "logps/chosen": -116.91698455810547, "logps/ref_chosen": -57.45049285888672, "logps/ref_rejected": -77.60826110839844, "logps/rejected": -173.64251708984375, "loss": 1.108, "margin_dpo/margin_mean": 36.56776809692383, "margin_dpo/margin_std": 58.076812744140625, "step": 266 }, { "KL/chosen_KL_mean": -54.2578239440918, "KL/mean": -68.69037628173828, "KL/rejected_KL_mean": -83.1229248046875, "KL/std": 48.732933044433594, "epoch": 0.4036281179138322, "fcm_dpo/beta": 0.012029076926410198, "fcm_dpo/delta": 0.052326202392578125, "fcm_dpo/margin": 28.865095138549805, "fcm_dpo/q_t": 0.4219139814376831, "grad_norm": 15.21697998046875, "learning_rate": 3.738531817228131e-07, "logits/chosen": 0.66780024766922, "logits/rejected": 0.6482948064804077, "logps/chosen": -109.29317474365234, "logps/ref_chosen": -55.03535079956055, "logps/ref_rejected": -66.0953369140625, "logps/rejected": -149.21826171875, "loss": 1.1956, "margin_dpo/margin_mean": 28.865097045898438, "margin_dpo/margin_std": 58.339569091796875, "step": 267 }, { "KL/chosen_KL_mean": -54.20028305053711, "KL/mean": -71.07929992675781, "KL/rejected_KL_mean": -87.95832061767578, "KL/std": 51.59467315673828, "epoch": 0.4051398337112623, "fcm_dpo/beta": 0.01196706760674715, "fcm_dpo/delta": -0.004256328567862511, "fcm_dpo/margin": 33.75802993774414, "fcm_dpo/q_t": 0.4097514748573303, "grad_norm": 13.947291374206543, "learning_rate": 3.7270289900589204e-07, "logits/chosen": 0.5228180289268494, "logits/rejected": 0.5083709955215454, "logps/chosen": -119.27202606201172, "logps/ref_chosen": -65.07174682617188, "logps/ref_rejected": -71.42485809326172, "logps/rejected": -159.3831787109375, "loss": 1.1071, "margin_dpo/margin_mean": 33.75802993774414, "margin_dpo/margin_std": 50.174339294433594, "step": 268 }, { "KL/chosen_KL_mean": -60.456459045410156, "KL/mean": -81.71235656738281, "KL/rejected_KL_mean": -102.96824645996094, "KL/std": 54.69612503051758, "epoch": 0.40665154950869237, "fcm_dpo/beta": 0.01182098500430584, "fcm_dpo/delta": -0.10809072107076645, "fcm_dpo/margin": 42.51179122924805, "fcm_dpo/q_t": 0.38763895630836487, "grad_norm": 13.600680351257324, "learning_rate": 3.7154918402511714e-07, "logits/chosen": 0.7140573263168335, "logits/rejected": 0.6650851964950562, "logps/chosen": -127.59266662597656, "logps/ref_chosen": -67.1362075805664, "logps/ref_rejected": -82.55778503417969, "logps/rejected": -185.52603149414062, "loss": 1.0495, "margin_dpo/margin_mean": 42.51179122924805, "margin_dpo/margin_std": 55.717506408691406, "step": 269 }, { "KL/chosen_KL_mean": -62.103599548339844, "KL/mean": -78.8128433227539, "KL/rejected_KL_mean": -95.52208709716797, "KL/std": 53.32160949707031, "epoch": 0.40816326530612246, "fcm_dpo/beta": 0.011670410633087158, "fcm_dpo/delta": 0.01030159741640091, "fcm_dpo/margin": 33.41849899291992, "fcm_dpo/q_t": 0.4117770791053772, "grad_norm": 14.696253776550293, "learning_rate": 3.7039206905237656e-07, "logits/chosen": 0.6845616102218628, "logits/rejected": 0.5967895984649658, "logps/chosen": -128.79229736328125, "logps/ref_chosen": -66.6886978149414, "logps/ref_rejected": -85.16129302978516, "logps/rejected": -180.68338012695312, "loss": 1.1368, "margin_dpo/margin_mean": 33.41849899291992, "margin_dpo/margin_std": 57.01043701171875, "step": 270 }, { "KL/chosen_KL_mean": -63.44163513183594, "KL/mean": -76.80795288085938, "KL/rejected_KL_mean": -90.17426300048828, "KL/std": 55.60846710205078, "epoch": 0.40967498110355255, "fcm_dpo/beta": 0.011818543076515198, "fcm_dpo/delta": 0.08686529844999313, "fcm_dpo/margin": 26.732616424560547, "fcm_dpo/q_t": 0.434378057718277, "grad_norm": 17.31071662902832, "learning_rate": 3.692315864546635e-07, "logits/chosen": 0.6486247777938843, "logits/rejected": 0.58237624168396, "logps/chosen": -135.84918212890625, "logps/ref_chosen": -72.40754699707031, "logps/ref_rejected": -92.06311798095703, "logps/rejected": -182.2373809814453, "loss": 1.234, "margin_dpo/margin_mean": 26.732616424560547, "margin_dpo/margin_std": 65.48910522460938, "step": 271 }, { "KL/chosen_KL_mean": -53.223960876464844, "KL/mean": -79.23175811767578, "KL/rejected_KL_mean": -105.23955535888672, "KL/std": 51.41781997680664, "epoch": 0.41118669690098264, "fcm_dpo/beta": 0.011572392657399178, "fcm_dpo/delta": -0.21442440152168274, "fcm_dpo/margin": 52.015586853027344, "fcm_dpo/q_t": 0.3613682985305786, "grad_norm": 15.099222183227539, "learning_rate": 3.6806776869317067e-07, "logits/chosen": 0.6465578079223633, "logits/rejected": 0.6684205532073975, "logps/chosen": -119.82536315917969, "logps/ref_chosen": -66.60140228271484, "logps/ref_rejected": -67.74340057373047, "logps/rejected": -172.9829559326172, "loss": 0.9487, "margin_dpo/margin_mean": 52.015586853027344, "margin_dpo/margin_std": 49.51817321777344, "step": 272 }, { "KL/chosen_KL_mean": -73.37745666503906, "KL/mean": -92.10945129394531, "KL/rejected_KL_mean": -110.8414306640625, "KL/std": 55.56861114501953, "epoch": 0.4126984126984127, "fcm_dpo/beta": 0.011334793642163277, "fcm_dpo/delta": -0.025967109948396683, "fcm_dpo/margin": 37.463966369628906, "fcm_dpo/q_t": 0.40638309717178345, "grad_norm": 17.103038787841797, "learning_rate": 3.669006483223828e-07, "logits/chosen": 0.6392084360122681, "logits/rejected": 0.5658600926399231, "logps/chosen": -130.73233032226562, "logps/ref_chosen": -57.35487747192383, "logps/ref_rejected": -84.17168426513672, "logps/rejected": -195.01312255859375, "loss": 1.1537, "margin_dpo/margin_mean": 37.46397018432617, "margin_dpo/margin_std": 70.37014770507812, "step": 273 }, { "KL/chosen_KL_mean": -71.62263488769531, "KL/mean": -91.7503890991211, "KL/rejected_KL_mean": -111.87814331054688, "KL/std": 55.358795166015625, "epoch": 0.41421012849584277, "fcm_dpo/beta": 0.011266610585153103, "fcm_dpo/delta": -0.05605369806289673, "fcm_dpo/margin": 40.25550842285156, "fcm_dpo/q_t": 0.3987223207950592, "grad_norm": 13.076324462890625, "learning_rate": 3.657302579891656e-07, "logits/chosen": 0.5529523491859436, "logits/rejected": 0.5354233980178833, "logps/chosen": -131.26412963867188, "logps/ref_chosen": -59.64149475097656, "logps/ref_rejected": -68.29348754882812, "logps/rejected": -180.171630859375, "loss": 1.1094, "margin_dpo/margin_mean": 40.25550842285156, "margin_dpo/margin_std": 65.55795288085938, "step": 274 }, { "KL/chosen_KL_mean": -70.55653381347656, "KL/mean": -90.97258758544922, "KL/rejected_KL_mean": -111.38862609863281, "KL/std": 53.519981384277344, "epoch": 0.41572184429327286, "fcm_dpo/beta": 0.011159954592585564, "fcm_dpo/delta": -0.05834575742483139, "fcm_dpo/margin": 40.83209991455078, "fcm_dpo/q_t": 0.39650213718414307, "grad_norm": 14.841992378234863, "learning_rate": 3.645566304318526e-07, "logits/chosen": 0.6303668022155762, "logits/rejected": 0.545132040977478, "logps/chosen": -123.82318115234375, "logps/ref_chosen": -53.26664352416992, "logps/ref_rejected": -73.84062194824219, "logps/rejected": -185.229248046875, "loss": 1.0729, "margin_dpo/margin_mean": 40.83209991455078, "margin_dpo/margin_std": 55.855072021484375, "step": 275 }, { "KL/chosen_KL_mean": -66.35244750976562, "KL/mean": -86.4322509765625, "KL/rejected_KL_mean": -106.51205444335938, "KL/std": 50.89011764526367, "epoch": 0.41723356009070295, "fcm_dpo/beta": 0.011026125401258469, "fcm_dpo/delta": -0.04477095603942871, "fcm_dpo/margin": 40.15960693359375, "fcm_dpo/q_t": 0.3993651270866394, "grad_norm": 15.211221694946289, "learning_rate": 3.633797984793294e-07, "logits/chosen": 0.5870778560638428, "logits/rejected": 0.5541011095046997, "logps/chosen": -119.37323760986328, "logps/ref_chosen": -53.02079772949219, "logps/ref_rejected": -61.56678771972656, "logps/rejected": -168.07884216308594, "loss": 1.0821, "margin_dpo/margin_mean": 40.15961456298828, "margin_dpo/margin_std": 56.424461364746094, "step": 276 }, { "KL/chosen_KL_mean": -76.11033630371094, "KL/mean": -86.88145446777344, "KL/rejected_KL_mean": -97.652587890625, "KL/std": 54.547332763671875, "epoch": 0.41874527588813304, "fcm_dpo/beta": 0.011240396648645401, "fcm_dpo/delta": 0.16203390061855316, "fcm_dpo/margin": 21.542251586914062, "fcm_dpo/q_t": 0.4464568793773651, "grad_norm": 20.440866470336914, "learning_rate": 3.6219979505011555e-07, "logits/chosen": 0.6785659790039062, "logits/rejected": 0.7067118287086487, "logps/chosen": -147.54331970214844, "logps/ref_chosen": -71.43299102783203, "logps/ref_rejected": -67.65852355957031, "logps/rejected": -165.3111114501953, "loss": 1.2769, "margin_dpo/margin_mean": 21.542253494262695, "margin_dpo/margin_std": 62.680477142333984, "step": 277 }, { "KL/chosen_KL_mean": -78.5838623046875, "KL/mean": -98.24070739746094, "KL/rejected_KL_mean": -117.89755249023438, "KL/std": 58.49961853027344, "epoch": 0.42025699168556313, "fcm_dpo/beta": 0.01136256754398346, "fcm_dpo/delta": -0.04974301904439926, "fcm_dpo/margin": 39.31369400024414, "fcm_dpo/q_t": 0.39944133162498474, "grad_norm": 20.18327522277832, "learning_rate": 3.6101665315144353e-07, "logits/chosen": 0.573406994342804, "logits/rejected": 0.5146926641464233, "logps/chosen": -145.6946258544922, "logps/ref_chosen": -67.11076354980469, "logps/ref_rejected": -88.74851989746094, "logps/rejected": -206.64608764648438, "loss": 1.1067, "margin_dpo/margin_mean": 39.31369400024414, "margin_dpo/margin_std": 61.28262710571289, "step": 278 }, { "KL/chosen_KL_mean": -61.78923797607422, "KL/mean": -88.9472885131836, "KL/rejected_KL_mean": -116.1053466796875, "KL/std": 57.12919998168945, "epoch": 0.4217687074829932, "fcm_dpo/beta": 0.0109342560172081, "fcm_dpo/delta": -0.20577079057693481, "fcm_dpo/margin": 54.31610107421875, "fcm_dpo/q_t": 0.36407333612442017, "grad_norm": 14.203146934509277, "learning_rate": 3.5983040587833563e-07, "logits/chosen": 0.5994449853897095, "logits/rejected": 0.5615173578262329, "logps/chosen": -116.28672790527344, "logps/ref_chosen": -54.49748611450195, "logps/ref_rejected": -70.42373657226562, "logps/rejected": -186.52908325195312, "loss": 0.9612, "margin_dpo/margin_mean": 54.31610107421875, "margin_dpo/margin_std": 54.377685546875, "step": 279 }, { "KL/chosen_KL_mean": -67.08258056640625, "KL/mean": -94.1540298461914, "KL/rejected_KL_mean": -121.2254867553711, "KL/std": 59.60696792602539, "epoch": 0.42328042328042326, "fcm_dpo/beta": 0.010509304702281952, "fcm_dpo/delta": -0.17919118702411652, "fcm_dpo/margin": 54.14289855957031, "fcm_dpo/q_t": 0.3707593083381653, "grad_norm": 11.810566902160645, "learning_rate": 3.586410864126781e-07, "logits/chosen": 0.6714847087860107, "logits/rejected": 0.6250628232955933, "logps/chosen": -127.51538848876953, "logps/ref_chosen": -60.43281173706055, "logps/ref_rejected": -78.39051818847656, "logps/rejected": -199.61599731445312, "loss": 0.9715, "margin_dpo/margin_mean": 54.14289855957031, "margin_dpo/margin_std": 55.1671257019043, "step": 280 }, { "KL/chosen_KL_mean": -70.63385772705078, "KL/mean": -94.31147766113281, "KL/rejected_KL_mean": -117.98908996582031, "KL/std": 58.223365783691406, "epoch": 0.42479213907785335, "fcm_dpo/beta": 0.010295085608959198, "fcm_dpo/delta": -0.09190942347049713, "fcm_dpo/margin": 47.35523986816406, "fcm_dpo/q_t": 0.39007121324539185, "grad_norm": 13.856719017028809, "learning_rate": 3.574487280222929e-07, "logits/chosen": 0.6418502926826477, "logits/rejected": 0.6694193482398987, "logps/chosen": -130.91595458984375, "logps/ref_chosen": -60.2820930480957, "logps/ref_rejected": -62.04009246826172, "logps/rejected": -180.0291748046875, "loss": 1.0588, "margin_dpo/margin_mean": 47.35523986816406, "margin_dpo/margin_std": 64.03118896484375, "step": 281 }, { "KL/chosen_KL_mean": -77.45885467529297, "KL/mean": -99.65185546875, "KL/rejected_KL_mean": -121.8448715209961, "KL/std": 58.28876876831055, "epoch": 0.42630385487528344, "fcm_dpo/beta": 0.010223493911325932, "fcm_dpo/delta": -0.057356491684913635, "fcm_dpo/margin": 44.386009216308594, "fcm_dpo/q_t": 0.3992747664451599, "grad_norm": 16.32402992248535, "learning_rate": 3.562533640600075e-07, "logits/chosen": 0.5796880722045898, "logits/rejected": 0.532652735710144, "logps/chosen": -138.08277893066406, "logps/ref_chosen": -60.623924255371094, "logps/ref_rejected": -68.67400360107422, "logps/rejected": -190.5188751220703, "loss": 1.1023, "margin_dpo/margin_mean": 44.386009216308594, "margin_dpo/margin_std": 67.81919860839844, "step": 282 }, { "KL/chosen_KL_mean": -77.25079345703125, "KL/mean": -98.12049102783203, "KL/rejected_KL_mean": -118.99020385742188, "KL/std": 59.12428283691406, "epoch": 0.42781557067271353, "fcm_dpo/beta": 0.010089964605867863, "fcm_dpo/delta": -0.02210230752825737, "fcm_dpo/margin": 41.739402770996094, "fcm_dpo/q_t": 0.4049845337867737, "grad_norm": 16.064359664916992, "learning_rate": 3.550550279627215e-07, "logits/chosen": 0.6336187720298767, "logits/rejected": 0.5247352123260498, "logps/chosen": -144.89854431152344, "logps/ref_chosen": -67.64775085449219, "logps/ref_rejected": -99.96835327148438, "logps/rejected": -218.95855712890625, "loss": 1.1101, "margin_dpo/margin_mean": 41.739402770996094, "margin_dpo/margin_std": 65.20516204833984, "step": 283 }, { "KL/chosen_KL_mean": -72.93248748779297, "KL/mean": -95.26510620117188, "KL/rejected_KL_mean": -117.59771728515625, "KL/std": 61.540306091308594, "epoch": 0.4293272864701436, "fcm_dpo/beta": 0.009952565655112267, "fcm_dpo/delta": -0.04708627983927727, "fcm_dpo/margin": 44.66523742675781, "fcm_dpo/q_t": 0.3982611894607544, "grad_norm": 12.55010986328125, "learning_rate": 3.5385375325047163e-07, "logits/chosen": 0.6949187517166138, "logits/rejected": 0.6278550028800964, "logps/chosen": -129.89991760253906, "logps/ref_chosen": -56.96742630004883, "logps/ref_rejected": -86.36236572265625, "logps/rejected": -203.9600830078125, "loss": 1.0738, "margin_dpo/margin_mean": 44.66523742675781, "margin_dpo/margin_std": 59.5830078125, "step": 284 }, { "KL/chosen_KL_mean": -85.50302124023438, "KL/mean": -103.44571685791016, "KL/rejected_KL_mean": -121.38841247558594, "KL/std": 56.31591796875, "epoch": 0.4308390022675737, "fcm_dpo/beta": 0.010065186768770218, "fcm_dpo/delta": 0.039796821773052216, "fcm_dpo/margin": 35.885398864746094, "fcm_dpo/q_t": 0.4194878935813904, "grad_norm": 17.57463264465332, "learning_rate": 3.5264957352549375e-07, "logits/chosen": 0.6707309484481812, "logits/rejected": 0.6428732872009277, "logps/chosen": -157.15911865234375, "logps/ref_chosen": -71.65611267089844, "logps/ref_rejected": -81.63829803466797, "logps/rejected": -203.02670288085938, "loss": 1.1526, "margin_dpo/margin_mean": 35.88539505004883, "margin_dpo/margin_std": 62.94104766845703, "step": 285 }, { "KL/chosen_KL_mean": -80.76374816894531, "KL/mean": -109.85708618164062, "KL/rejected_KL_mean": -138.950439453125, "KL/std": 60.531402587890625, "epoch": 0.4323507180650038, "fcm_dpo/beta": 0.00978100299835205, "fcm_dpo/delta": -0.17936107516288757, "fcm_dpo/margin": 58.18669128417969, "fcm_dpo/q_t": 0.3713955581188202, "grad_norm": 13.661906242370605, "learning_rate": 3.514425224712835e-07, "logits/chosen": 0.5708847641944885, "logits/rejected": 0.47632908821105957, "logps/chosen": -141.84326171875, "logps/ref_chosen": -61.07952117919922, "logps/ref_rejected": -91.28128051757812, "logps/rejected": -230.23171997070312, "loss": 0.9846, "margin_dpo/margin_mean": 58.18669509887695, "margin_dpo/margin_std": 64.16620635986328, "step": 286 }, { "KL/chosen_KL_mean": -67.0542221069336, "KL/mean": -94.71749114990234, "KL/rejected_KL_mean": -122.38075256347656, "KL/std": 61.530731201171875, "epoch": 0.43386243386243384, "fcm_dpo/beta": 0.009508990682661533, "fcm_dpo/delta": -0.1330246478319168, "fcm_dpo/margin": 55.32652282714844, "fcm_dpo/q_t": 0.3804447650909424, "grad_norm": 12.890028953552246, "learning_rate": 3.502326338516534e-07, "logits/chosen": 0.7085520029067993, "logits/rejected": 0.6656246185302734, "logps/chosen": -113.09001159667969, "logps/ref_chosen": -46.035789489746094, "logps/ref_rejected": -59.95293426513672, "logps/rejected": -182.33367919921875, "loss": 1.0121, "margin_dpo/margin_mean": 55.32652282714844, "margin_dpo/margin_std": 63.78333282470703, "step": 287 }, { "KL/chosen_KL_mean": -84.3800048828125, "KL/mean": -105.1005630493164, "KL/rejected_KL_mean": -125.82112121582031, "KL/std": 59.49192810058594, "epoch": 0.43537414965986393, "fcm_dpo/beta": 0.00945931114256382, "fcm_dpo/delta": 0.008291497826576233, "fcm_dpo/margin": 41.441131591796875, "fcm_dpo/q_t": 0.4110769033432007, "grad_norm": 17.03542137145996, "learning_rate": 3.490199415097892e-07, "logits/chosen": 0.5387293100357056, "logits/rejected": 0.478929340839386, "logps/chosen": -149.77084350585938, "logps/ref_chosen": -65.3908462524414, "logps/ref_rejected": -88.53607940673828, "logps/rejected": -214.35720825195312, "loss": 1.1204, "margin_dpo/margin_mean": 41.44112777709961, "margin_dpo/margin_std": 64.76750183105469, "step": 288 }, { "KL/chosen_KL_mean": -86.93623352050781, "KL/mean": -105.4552993774414, "KL/rejected_KL_mean": -123.974365234375, "KL/std": 60.03764724731445, "epoch": 0.436885865457294, "fcm_dpo/beta": 0.009551433846354485, "fcm_dpo/delta": 0.047566674649715424, "fcm_dpo/margin": 37.038116455078125, "fcm_dpo/q_t": 0.4222422242164612, "grad_norm": 13.532853126525879, "learning_rate": 3.4780447936730247e-07, "logits/chosen": 0.7148054242134094, "logits/rejected": 0.6745719909667969, "logps/chosen": -141.5298614501953, "logps/ref_chosen": -54.5936279296875, "logps/ref_rejected": -67.20855712890625, "logps/rejected": -191.18292236328125, "loss": 1.171, "margin_dpo/margin_mean": 37.038116455078125, "margin_dpo/margin_std": 70.36173248291016, "step": 289 }, { "KL/chosen_KL_mean": -96.75265502929688, "KL/mean": -120.16667175292969, "KL/rejected_KL_mean": -143.58067321777344, "KL/std": 58.69277572631836, "epoch": 0.4383975812547241, "fcm_dpo/beta": 0.00943966954946518, "fcm_dpo/delta": -0.04456711933016777, "fcm_dpo/margin": 46.82802200317383, "fcm_dpo/q_t": 0.3998667299747467, "grad_norm": 16.931427001953125, "learning_rate": 3.465862814232821e-07, "logits/chosen": 0.7094553709030151, "logits/rejected": 0.6341379284858704, "logps/chosen": -158.13723754882812, "logps/ref_chosen": -61.38457489013672, "logps/ref_rejected": -91.92778015136719, "logps/rejected": -235.50845336914062, "loss": 1.0928, "margin_dpo/margin_mean": 46.82802200317383, "margin_dpo/margin_std": 69.21977233886719, "step": 290 }, { "KL/chosen_KL_mean": -90.10990142822266, "KL/mean": -117.04533386230469, "KL/rejected_KL_mean": -143.98077392578125, "KL/std": 59.01616668701172, "epoch": 0.4399092970521542, "fcm_dpo/beta": 0.009374039247632027, "fcm_dpo/delta": -0.11086567491292953, "fcm_dpo/margin": 53.87086868286133, "fcm_dpo/q_t": 0.3872455060482025, "grad_norm": 14.803458213806152, "learning_rate": 3.4536538175334343e-07, "logits/chosen": 0.8204612731933594, "logits/rejected": 0.741672933101654, "logps/chosen": -140.97293090820312, "logps/ref_chosen": -50.863037109375, "logps/ref_rejected": -82.20868682861328, "logps/rejected": -226.189453125, "loss": 1.0403, "margin_dpo/margin_mean": 53.87086486816406, "margin_dpo/margin_std": 67.89382934570312, "step": 291 }, { "KL/chosen_KL_mean": -89.85836791992188, "KL/mean": -109.6369857788086, "KL/rejected_KL_mean": -129.41561889648438, "KL/std": 62.5750732421875, "epoch": 0.4414210128495843, "fcm_dpo/beta": 0.009300924837589264, "fcm_dpo/delta": 0.03329315781593323, "fcm_dpo/margin": 39.557254791259766, "fcm_dpo/q_t": 0.41779154539108276, "grad_norm": 15.200834274291992, "learning_rate": 3.4414181450867465e-07, "logits/chosen": 0.6663825511932373, "logits/rejected": 0.6153430938720703, "logps/chosen": -154.20724487304688, "logps/ref_chosen": -64.34888458251953, "logps/ref_rejected": -72.86434173583984, "logps/rejected": -202.27996826171875, "loss": 1.1589, "margin_dpo/margin_mean": 39.5572509765625, "margin_dpo/margin_std": 72.9423599243164, "step": 292 }, { "KL/chosen_KL_mean": -89.22817993164062, "KL/mean": -118.22019958496094, "KL/rejected_KL_mean": -147.21221923828125, "KL/std": 63.70530700683594, "epoch": 0.4429327286470144, "fcm_dpo/beta": 0.009160864166915417, "fcm_dpo/delta": -0.13831757009029388, "fcm_dpo/margin": 57.984046936035156, "fcm_dpo/q_t": 0.3815430700778961, "grad_norm": 12.239211082458496, "learning_rate": 3.4291561391508185e-07, "logits/chosen": 0.7890902757644653, "logits/rejected": 0.6947133541107178, "logps/chosen": -144.09764099121094, "logps/ref_chosen": -54.869468688964844, "logps/ref_rejected": -81.858642578125, "logps/rejected": -229.07086181640625, "loss": 1.043, "margin_dpo/margin_mean": 57.984046936035156, "margin_dpo/margin_std": 77.6739501953125, "step": 293 }, { "KL/chosen_KL_mean": -94.31707763671875, "KL/mean": -113.79631042480469, "KL/rejected_KL_mean": -133.27554321289062, "KL/std": 64.919921875, "epoch": 0.4444444444444444, "fcm_dpo/beta": 0.009092319756746292, "fcm_dpo/delta": 0.04737778753042221, "fcm_dpo/margin": 38.958457946777344, "fcm_dpo/q_t": 0.4201389253139496, "grad_norm": 12.569846153259277, "learning_rate": 3.4168681427203153e-07, "logits/chosen": 0.6751217246055603, "logits/rejected": 0.6271165609359741, "logps/chosen": -150.98797607421875, "logps/ref_chosen": -56.670902252197266, "logps/ref_rejected": -70.32819366455078, "logps/rejected": -203.60372924804688, "loss": 1.1334, "margin_dpo/margin_mean": 38.958457946777344, "margin_dpo/margin_std": 60.146690368652344, "step": 294 }, { "KL/chosen_KL_mean": -99.99388122558594, "KL/mean": -118.2969970703125, "KL/rejected_KL_mean": -136.60012817382812, "KL/std": 61.911251068115234, "epoch": 0.4459561602418745, "fcm_dpo/beta": 0.009216606616973877, "fcm_dpo/delta": 0.06484965234994888, "fcm_dpo/margin": 36.60624694824219, "fcm_dpo/q_t": 0.42461496591567993, "grad_norm": 16.885400772094727, "learning_rate": 3.4045544995169125e-07, "logits/chosen": 0.6669223308563232, "logits/rejected": 0.5581985116004944, "logps/chosen": -150.394775390625, "logps/ref_chosen": -50.40088653564453, "logps/ref_rejected": -83.43521881103516, "logps/rejected": -220.03533935546875, "loss": 1.1682, "margin_dpo/margin_mean": 36.60624694824219, "margin_dpo/margin_std": 67.60258483886719, "step": 295 }, { "KL/chosen_KL_mean": -102.53762817382812, "KL/mean": -124.98433685302734, "KL/rejected_KL_mean": -147.4310302734375, "KL/std": 65.74832153320312, "epoch": 0.4474678760393046, "fcm_dpo/beta": 0.009207111783325672, "fcm_dpo/delta": -0.014326311647891998, "fcm_dpo/margin": 44.89341735839844, "fcm_dpo/q_t": 0.40750962495803833, "grad_norm": 13.373530387878418, "learning_rate": 3.392215553979679e-07, "logits/chosen": 0.6259567737579346, "logits/rejected": 0.5756511092185974, "logps/chosen": -171.68797302246094, "logps/ref_chosen": -69.15034484863281, "logps/ref_rejected": -89.60166931152344, "logps/rejected": -237.03271484375, "loss": 1.1283, "margin_dpo/margin_mean": 44.89341735839844, "margin_dpo/margin_std": 75.29107666015625, "step": 296 }, { "KL/chosen_KL_mean": -106.02978515625, "KL/mean": -130.51258850097656, "KL/rejected_KL_mean": -154.99539184570312, "KL/std": 57.90068817138672, "epoch": 0.4489795918367347, "fcm_dpo/beta": 0.009207883849740028, "fcm_dpo/delta": -0.05345804616808891, "fcm_dpo/margin": 48.965606689453125, "fcm_dpo/q_t": 0.39636701345443726, "grad_norm": 13.408093452453613, "learning_rate": 3.3798516512554485e-07, "logits/chosen": 0.6601380705833435, "logits/rejected": 0.6036698818206787, "logps/chosen": -164.04608154296875, "logps/ref_chosen": -58.01630401611328, "logps/ref_rejected": -69.95780944824219, "logps/rejected": -224.95321655273438, "loss": 1.0602, "margin_dpo/margin_mean": 48.965606689453125, "margin_dpo/margin_std": 60.579673767089844, "step": 297 }, { "KL/chosen_KL_mean": -103.89498901367188, "KL/mean": -122.39523315429688, "KL/rejected_KL_mean": -140.89549255371094, "KL/std": 61.32954406738281, "epoch": 0.4504913076341648, "fcm_dpo/beta": 0.009213869459927082, "fcm_dpo/delta": 0.06121082603931427, "fcm_dpo/margin": 37.00048828125, "fcm_dpo/q_t": 0.4239236116409302, "grad_norm": 12.870407104492188, "learning_rate": 3.367463137189156e-07, "logits/chosen": 0.7620357275009155, "logits/rejected": 0.7042220830917358, "logps/chosen": -160.06430053710938, "logps/ref_chosen": -56.1693115234375, "logps/ref_rejected": -68.55052185058594, "logps/rejected": -209.44601440429688, "loss": 1.1781, "margin_dpo/margin_mean": 37.00048828125, "margin_dpo/margin_std": 71.90211486816406, "step": 298 }, { "KL/chosen_KL_mean": -104.1708755493164, "KL/mean": -120.55952453613281, "KL/rejected_KL_mean": -136.9481658935547, "KL/std": 60.09037780761719, "epoch": 0.4520030234315949, "fcm_dpo/beta": 0.009359677322208881, "fcm_dpo/delta": 0.09627296775579453, "fcm_dpo/margin": 32.777286529541016, "fcm_dpo/q_t": 0.4320846498012543, "grad_norm": 18.66891098022461, "learning_rate": 3.355050358314172e-07, "logits/chosen": 0.541263222694397, "logits/rejected": 0.5110787153244019, "logps/chosen": -166.48867797851562, "logps/ref_chosen": -62.31780242919922, "logps/ref_rejected": -72.60028839111328, "logps/rejected": -209.5484619140625, "loss": 1.2256, "margin_dpo/margin_mean": 32.777286529541016, "margin_dpo/margin_std": 76.10530090332031, "step": 299 }, { "KL/chosen_KL_mean": -105.92295837402344, "KL/mean": -126.46493530273438, "KL/rejected_KL_mean": -147.00689697265625, "KL/std": 60.57496643066406, "epoch": 0.45351473922902497, "fcm_dpo/beta": 0.009462611749768257, "fcm_dpo/delta": 0.011636088602244854, "fcm_dpo/margin": 41.083927154541016, "fcm_dpo/q_t": 0.41320398449897766, "grad_norm": 15.427064895629883, "learning_rate": 3.3426136618426043e-07, "logits/chosen": 0.6564747095108032, "logits/rejected": 0.5867733359336853, "logps/chosen": -166.30453491210938, "logps/ref_chosen": -60.38157653808594, "logps/ref_rejected": -75.45442199707031, "logps/rejected": -222.46131896972656, "loss": 1.1443, "margin_dpo/margin_mean": 41.08393096923828, "margin_dpo/margin_std": 71.81248474121094, "step": 300 }, { "KL/chosen_KL_mean": -104.05447387695312, "KL/mean": -121.67876434326172, "KL/rejected_KL_mean": -139.30307006835938, "KL/std": 59.82817077636719, "epoch": 0.455026455026455, "fcm_dpo/beta": 0.009520202875137329, "fcm_dpo/delta": 0.06667570024728775, "fcm_dpo/margin": 35.24860382080078, "fcm_dpo/q_t": 0.42482560873031616, "grad_norm": 13.765838623046875, "learning_rate": 3.3301533956555885e-07, "logits/chosen": 0.713027834892273, "logits/rejected": 0.6836607456207275, "logps/chosen": -156.90536499023438, "logps/ref_chosen": -52.85089111328125, "logps/ref_rejected": -69.97584533691406, "logps/rejected": -209.27891540527344, "loss": 1.182, "margin_dpo/margin_mean": 35.24860382080078, "margin_dpo/margin_std": 69.6445083618164, "step": 301 }, { "KL/chosen_KL_mean": -108.92539978027344, "KL/mean": -123.26554870605469, "KL/rejected_KL_mean": -137.605712890625, "KL/std": 61.54811096191406, "epoch": 0.4565381708238851, "fcm_dpo/beta": 0.009755447506904602, "fcm_dpo/delta": 0.12373100966215134, "fcm_dpo/margin": 28.680299758911133, "fcm_dpo/q_t": 0.43714994192123413, "grad_norm": 18.384824752807617, "learning_rate": 3.317669908293554e-07, "logits/chosen": 0.5518735647201538, "logits/rejected": 0.4927511215209961, "logps/chosen": -175.89190673828125, "logps/ref_chosen": -66.96650695800781, "logps/ref_rejected": -88.09510803222656, "logps/rejected": -225.7008056640625, "loss": 1.2291, "margin_dpo/margin_mean": 28.680301666259766, "margin_dpo/margin_std": 68.15020751953125, "step": 302 }, { "KL/chosen_KL_mean": -97.18345642089844, "KL/mean": -122.90306091308594, "KL/rejected_KL_mean": -148.62265014648438, "KL/std": 62.800289154052734, "epoch": 0.4580498866213152, "fcm_dpo/beta": 0.009679718874394894, "fcm_dpo/delta": -0.10298407077789307, "fcm_dpo/margin": 51.439186096191406, "fcm_dpo/q_t": 0.38958287239074707, "grad_norm": 12.350512504577637, "learning_rate": 3.3051635489464793e-07, "logits/chosen": 0.657049298286438, "logits/rejected": 0.5822213888168335, "logps/chosen": -159.30499267578125, "logps/ref_chosen": -62.12152862548828, "logps/ref_rejected": -90.31204223632812, "logps/rejected": -238.9346923828125, "loss": 1.082, "margin_dpo/margin_mean": 51.439186096191406, "margin_dpo/margin_std": 78.26869201660156, "step": 303 }, { "KL/chosen_KL_mean": -85.36961364746094, "KL/mean": -112.04019165039062, "KL/rejected_KL_mean": -138.7107696533203, "KL/std": 66.2973403930664, "epoch": 0.4595616024187453, "fcm_dpo/beta": 0.009458931162953377, "fcm_dpo/delta": -0.11018989235162735, "fcm_dpo/margin": 53.34115219116211, "fcm_dpo/q_t": 0.383327841758728, "grad_norm": 12.269899368286133, "learning_rate": 3.292634667444117e-07, "logits/chosen": 0.6840123534202576, "logits/rejected": 0.6246076822280884, "logps/chosen": -146.064697265625, "logps/ref_chosen": -60.695091247558594, "logps/ref_rejected": -78.2525405883789, "logps/rejected": -216.96331787109375, "loss": 1.0185, "margin_dpo/margin_mean": 53.341156005859375, "margin_dpo/margin_std": 59.39166259765625, "step": 304 }, { "KL/chosen_KL_mean": -98.3131103515625, "KL/mean": -116.91204833984375, "KL/rejected_KL_mean": -135.51097106933594, "KL/std": 61.239044189453125, "epoch": 0.46107331821617537, "fcm_dpo/beta": 0.009442286565899849, "fcm_dpo/delta": 0.05038090795278549, "fcm_dpo/margin": 37.19786834716797, "fcm_dpo/q_t": 0.4206337034702301, "grad_norm": 14.711054801940918, "learning_rate": 3.280083614246217e-07, "logits/chosen": 0.6261643171310425, "logits/rejected": 0.6597653031349182, "logps/chosen": -171.01223754882812, "logps/ref_chosen": -72.69914245605469, "logps/ref_rejected": -65.65670776367188, "logps/rejected": -201.16769409179688, "loss": 1.1829, "margin_dpo/margin_mean": 37.19786834716797, "margin_dpo/margin_std": 73.96134185791016, "step": 305 }, { "KL/chosen_KL_mean": -92.8389663696289, "KL/mean": -113.00958251953125, "KL/rejected_KL_mean": -133.18020629882812, "KL/std": 58.908294677734375, "epoch": 0.46258503401360546, "fcm_dpo/beta": 0.009471280500292778, "fcm_dpo/delta": 0.017888210713863373, "fcm_dpo/margin": 40.34123229980469, "fcm_dpo/q_t": 0.4124301075935364, "grad_norm": 13.316874504089355, "learning_rate": 3.267510740432719e-07, "logits/chosen": 0.7246155738830566, "logits/rejected": 0.6098471879959106, "logps/chosen": -146.8094940185547, "logps/ref_chosen": -53.97052764892578, "logps/ref_rejected": -71.02423095703125, "logps/rejected": -204.20443725585938, "loss": 1.1139, "margin_dpo/margin_mean": 40.34123229980469, "margin_dpo/margin_std": 57.43824005126953, "step": 306 }, { "KL/chosen_KL_mean": -87.2799301147461, "KL/mean": -96.95132446289062, "KL/rejected_KL_mean": -106.62272644042969, "KL/std": 58.196372985839844, "epoch": 0.46409674981103555, "fcm_dpo/beta": 0.009879468940198421, "fcm_dpo/delta": 0.21212655305862427, "fcm_dpo/margin": 19.342792510986328, "fcm_dpo/q_t": 0.4589805006980896, "grad_norm": 18.865333557128906, "learning_rate": 3.2549163976939285e-07, "logits/chosen": 0.6826301217079163, "logits/rejected": 0.632476806640625, "logps/chosen": -144.6930389404297, "logps/ref_chosen": -57.413108825683594, "logps/ref_rejected": -68.68010711669922, "logps/rejected": -175.30282592773438, "loss": 1.3379, "margin_dpo/margin_mean": 19.342790603637695, "margin_dpo/margin_std": 75.95152282714844, "step": 307 }, { "KL/chosen_KL_mean": -86.84735870361328, "KL/mean": -106.39009094238281, "KL/rejected_KL_mean": -125.93280029296875, "KL/std": 60.022674560546875, "epoch": 0.4656084656084656, "fcm_dpo/beta": 0.009924834594130516, "fcm_dpo/delta": 0.012152183800935745, "fcm_dpo/margin": 39.08544158935547, "fcm_dpo/q_t": 0.41299301385879517, "grad_norm": 11.46269702911377, "learning_rate": 3.2423009383206874e-07, "logits/chosen": 0.647024929523468, "logits/rejected": 0.6352015733718872, "logps/chosen": -153.44615173339844, "logps/ref_chosen": -66.59879302978516, "logps/ref_rejected": -74.337158203125, "logps/rejected": -200.26995849609375, "loss": 1.1458, "margin_dpo/margin_mean": 39.08544158935547, "margin_dpo/margin_std": 68.68582153320312, "step": 308 }, { "KL/chosen_KL_mean": -98.74766540527344, "KL/mean": -118.67848205566406, "KL/rejected_KL_mean": -138.6092987060547, "KL/std": 57.710323333740234, "epoch": 0.4671201814058957, "fcm_dpo/beta": 0.00993155688047409, "fcm_dpo/delta": 0.003720022737979889, "fcm_dpo/margin": 39.86164093017578, "fcm_dpo/q_t": 0.4107176661491394, "grad_norm": 12.478151321411133, "learning_rate": 3.229664715194511e-07, "logits/chosen": 0.6989116668701172, "logits/rejected": 0.6386100053787231, "logps/chosen": -164.14239501953125, "logps/ref_chosen": -65.39474487304688, "logps/ref_rejected": -75.70930480957031, "logps/rejected": -214.318603515625, "loss": 1.1265, "margin_dpo/margin_mean": 39.86164093017578, "margin_dpo/margin_std": 63.787353515625, "step": 309 }, { "KL/chosen_KL_mean": -101.77289581298828, "KL/mean": -113.84951782226562, "KL/rejected_KL_mean": -125.9261474609375, "KL/std": 53.13010025024414, "epoch": 0.46863189720332576, "fcm_dpo/beta": 0.01020658016204834, "fcm_dpo/delta": 0.15757328271865845, "fcm_dpo/margin": 24.153255462646484, "fcm_dpo/q_t": 0.4470118284225464, "grad_norm": 14.50604248046875, "learning_rate": 3.2170080817777257e-07, "logits/chosen": 0.6527221202850342, "logits/rejected": 0.6363627314567566, "logps/chosen": -176.441162109375, "logps/ref_chosen": -74.66827392578125, "logps/ref_rejected": -80.5689697265625, "logps/rejected": -206.4951171875, "loss": 1.2556, "margin_dpo/margin_mean": 24.153257369995117, "margin_dpo/margin_std": 63.91498565673828, "step": 310 }, { "KL/chosen_KL_mean": -81.11943054199219, "KL/mean": -102.99217224121094, "KL/rejected_KL_mean": -124.86492156982422, "KL/std": 66.24354553222656, "epoch": 0.47014361300075586, "fcm_dpo/beta": 0.01017935760319233, "fcm_dpo/delta": -0.048217982053756714, "fcm_dpo/margin": 43.74549102783203, "fcm_dpo/q_t": 0.4035261273384094, "grad_norm": 13.386900901794434, "learning_rate": 3.204331392103574e-07, "logits/chosen": 0.5761426091194153, "logits/rejected": 0.4230411648750305, "logps/chosen": -140.8574676513672, "logps/ref_chosen": -59.738033294677734, "logps/ref_rejected": -93.60757446289062, "logps/rejected": -218.47250366210938, "loss": 1.1018, "margin_dpo/margin_mean": 43.74549102783203, "margin_dpo/margin_std": 69.48627471923828, "step": 311 }, { "KL/chosen_KL_mean": -82.9559097290039, "KL/mean": -107.94522857666016, "KL/rejected_KL_mean": -132.93453979492188, "KL/std": 58.801048278808594, "epoch": 0.47165532879818595, "fcm_dpo/beta": 0.010070530697703362, "fcm_dpo/delta": -0.10864290595054626, "fcm_dpo/margin": 49.97863006591797, "fcm_dpo/q_t": 0.38455379009246826, "grad_norm": 11.932977676391602, "learning_rate": 3.1916350007663176e-07, "logits/chosen": 0.6558113098144531, "logits/rejected": 0.558535099029541, "logps/chosen": -136.7723388671875, "logps/ref_chosen": -53.816436767578125, "logps/ref_rejected": -68.6575698852539, "logps/rejected": -201.59210205078125, "loss": 1.0222, "margin_dpo/margin_mean": 49.97863006591797, "margin_dpo/margin_std": 57.203495025634766, "step": 312 }, { "KL/chosen_KL_mean": -86.02486419677734, "KL/mean": -98.96048736572266, "KL/rejected_KL_mean": -111.8961181640625, "KL/std": 57.0450439453125, "epoch": 0.47316704459561604, "fcm_dpo/beta": 0.010199690237641335, "fcm_dpo/delta": 0.139798104763031, "fcm_dpo/margin": 25.871265411376953, "fcm_dpo/q_t": 0.44218185544013977, "grad_norm": 13.864587783813477, "learning_rate": 3.178919262911314e-07, "logits/chosen": 0.6773887872695923, "logits/rejected": 0.6556931734085083, "logps/chosen": -145.9822235107422, "logps/ref_chosen": -59.957359313964844, "logps/ref_rejected": -69.31729888916016, "logps/rejected": -181.21340942382812, "loss": 1.237, "margin_dpo/margin_mean": 25.871265411376953, "margin_dpo/margin_std": 63.11532974243164, "step": 313 }, { "KL/chosen_KL_mean": -80.63925170898438, "KL/mean": -105.95073699951172, "KL/rejected_KL_mean": -131.26222229003906, "KL/std": 57.55366134643555, "epoch": 0.47467876039304613, "fcm_dpo/beta": 0.010059855878353119, "fcm_dpo/delta": -0.11579519510269165, "fcm_dpo/margin": 50.62295913696289, "fcm_dpo/q_t": 0.3863465487957001, "grad_norm": 12.481680870056152, "learning_rate": 3.166184534225087e-07, "logits/chosen": 0.6313918828964233, "logits/rejected": 0.6649610996246338, "logps/chosen": -150.90740966796875, "logps/ref_chosen": -70.26815795898438, "logps/ref_rejected": -69.23971557617188, "logps/rejected": -200.50193786621094, "loss": 1.0367, "margin_dpo/margin_mean": 50.622955322265625, "margin_dpo/margin_std": 63.79005432128906, "step": 314 }, { "KL/chosen_KL_mean": -86.05230712890625, "KL/mean": -105.64132690429688, "KL/rejected_KL_mean": -125.23036193847656, "KL/std": 58.11652374267578, "epoch": 0.47619047619047616, "fcm_dpo/beta": 0.010089458897709846, "fcm_dpo/delta": 0.004446987062692642, "fcm_dpo/margin": 39.17805099487305, "fcm_dpo/q_t": 0.40984123945236206, "grad_norm": 13.194849014282227, "learning_rate": 3.1534311709253723e-07, "logits/chosen": 0.543517529964447, "logits/rejected": 0.5082773566246033, "logps/chosen": -153.8470001220703, "logps/ref_chosen": -67.79469299316406, "logps/ref_rejected": -74.55148315429688, "logps/rejected": -199.78184509277344, "loss": 1.1191, "margin_dpo/margin_mean": 39.17805480957031, "margin_dpo/margin_std": 60.50445556640625, "step": 315 }, { "KL/chosen_KL_mean": -80.4795150756836, "KL/mean": -107.60636901855469, "KL/rejected_KL_mean": -134.7332000732422, "KL/std": 60.75721740722656, "epoch": 0.47770219198790626, "fcm_dpo/beta": 0.009939778596162796, "fcm_dpo/delta": -0.14816725254058838, "fcm_dpo/margin": 54.25369644165039, "fcm_dpo/q_t": 0.3790704011917114, "grad_norm": 13.798815727233887, "learning_rate": 3.1406595297511564e-07, "logits/chosen": 0.5641090273857117, "logits/rejected": 0.41707244515419006, "logps/chosen": -135.76800537109375, "logps/ref_chosen": -55.288482666015625, "logps/ref_rejected": -96.15723419189453, "logps/rejected": -230.89044189453125, "loss": 1.0166, "margin_dpo/margin_mean": 54.25369644165039, "margin_dpo/margin_std": 61.49580383300781, "step": 316 }, { "KL/chosen_KL_mean": -75.87356567382812, "KL/mean": -102.3594970703125, "KL/rejected_KL_mean": -128.84542846679688, "KL/std": 56.36368179321289, "epoch": 0.47921390778533635, "fcm_dpo/beta": 0.009550904855132103, "fcm_dpo/delta": -0.11237458139657974, "fcm_dpo/margin": 52.97185516357422, "fcm_dpo/q_t": 0.3837541937828064, "grad_norm": 16.33446502685547, "learning_rate": 3.1278699679526975e-07, "logits/chosen": 0.6727269887924194, "logits/rejected": 0.6263071298599243, "logps/chosen": -130.45494079589844, "logps/ref_chosen": -54.58137512207031, "logps/ref_rejected": -72.77232360839844, "logps/rejected": -201.6177520751953, "loss": 1.0265, "margin_dpo/margin_mean": 52.97185516357422, "margin_dpo/margin_std": 62.5837516784668, "step": 317 }, { "KL/chosen_KL_mean": -83.718994140625, "KL/mean": -103.1109848022461, "KL/rejected_KL_mean": -122.50297546386719, "KL/std": 60.44317626953125, "epoch": 0.48072562358276644, "fcm_dpo/beta": 0.009556062519550323, "fcm_dpo/delta": 0.030469154939055443, "fcm_dpo/margin": 38.78398895263672, "fcm_dpo/q_t": 0.42044275999069214, "grad_norm": 12.4547700881958, "learning_rate": 3.1150628432815336e-07, "logits/chosen": 0.684049129486084, "logits/rejected": 0.6094720959663391, "logps/chosen": -136.6072235107422, "logps/ref_chosen": -52.88822937011719, "logps/ref_rejected": -80.63988494873047, "logps/rejected": -203.1428680419922, "loss": 1.1845, "margin_dpo/margin_mean": 38.78398513793945, "margin_dpo/margin_std": 80.1749038696289, "step": 318 }, { "KL/chosen_KL_mean": -80.73106384277344, "KL/mean": -107.32977294921875, "KL/rejected_KL_mean": -133.928466796875, "KL/std": 62.93165969848633, "epoch": 0.48223733938019653, "fcm_dpo/beta": 0.009435629472136497, "fcm_dpo/delta": -0.10740967094898224, "fcm_dpo/margin": 53.197418212890625, "fcm_dpo/q_t": 0.3889296054840088, "grad_norm": 12.824899673461914, "learning_rate": 3.1022385139804707e-07, "logits/chosen": 0.6069538593292236, "logits/rejected": 0.5893139839172363, "logps/chosen": -145.09439086914062, "logps/ref_chosen": -64.36333465576172, "logps/ref_rejected": -79.47296142578125, "logps/rejected": -213.40142822265625, "loss": 1.0517, "margin_dpo/margin_mean": 53.197418212890625, "margin_dpo/margin_std": 72.72633361816406, "step": 319 }, { "KL/chosen_KL_mean": -75.99618530273438, "KL/mean": -97.96165466308594, "KL/rejected_KL_mean": -119.9271240234375, "KL/std": 63.95793914794922, "epoch": 0.4837490551776266, "fcm_dpo/beta": 0.009483279660344124, "fcm_dpo/delta": -0.02047700434923172, "fcm_dpo/margin": 43.93092727661133, "fcm_dpo/q_t": 0.40889182686805725, "grad_norm": 14.631209373474121, "learning_rate": 3.0893973387735683e-07, "logits/chosen": 0.5133136510848999, "logits/rejected": 0.4730033278465271, "logps/chosen": -125.554931640625, "logps/ref_chosen": -49.558746337890625, "logps/ref_rejected": -71.23444366455078, "logps/rejected": -191.16156005859375, "loss": 1.1239, "margin_dpo/margin_mean": 43.93092346191406, "margin_dpo/margin_std": 68.785400390625, "step": 320 }, { "KL/chosen_KL_mean": -81.69332885742188, "KL/mean": -105.12799072265625, "KL/rejected_KL_mean": -128.56268310546875, "KL/std": 57.98059844970703, "epoch": 0.4852607709750567, "fcm_dpo/beta": 0.009233346208930016, "fcm_dpo/delta": -0.03561249375343323, "fcm_dpo/margin": 46.869346618652344, "fcm_dpo/q_t": 0.40108391642570496, "grad_norm": 19.632156372070312, "learning_rate": 3.0765396768561004e-07, "logits/chosen": 0.626788854598999, "logits/rejected": 0.6063965559005737, "logps/chosen": -133.77859497070312, "logps/ref_chosen": -52.08526611328125, "logps/ref_rejected": -55.58674621582031, "logps/rejected": -184.1494140625, "loss": 1.1006, "margin_dpo/margin_mean": 46.869346618652344, "margin_dpo/margin_std": 68.83665466308594, "step": 321 }, { "KL/chosen_KL_mean": -92.54100799560547, "KL/mean": -121.60884094238281, "KL/rejected_KL_mean": -150.67666625976562, "KL/std": 64.49148559570312, "epoch": 0.48677248677248675, "fcm_dpo/beta": 0.009122923016548157, "fcm_dpo/delta": -0.13744433224201202, "fcm_dpo/margin": 58.13566970825195, "fcm_dpo/q_t": 0.37799185514450073, "grad_norm": 12.562987327575684, "learning_rate": 3.063665887884511e-07, "logits/chosen": 0.6769028902053833, "logits/rejected": 0.5882055759429932, "logps/chosen": -139.9451141357422, "logps/ref_chosen": -47.404109954833984, "logps/ref_rejected": -73.4260025024414, "logps/rejected": -224.1026611328125, "loss": 1.0043, "margin_dpo/margin_mean": 58.13567352294922, "margin_dpo/margin_std": 63.69596862792969, "step": 322 }, { "KL/chosen_KL_mean": -94.64340209960938, "KL/mean": -114.46276092529297, "KL/rejected_KL_mean": -134.28213500976562, "KL/std": 62.43727111816406, "epoch": 0.48828420256991684, "fcm_dpo/beta": 0.009102044627070427, "fcm_dpo/delta": 0.040557216852903366, "fcm_dpo/margin": 39.63875198364258, "fcm_dpo/q_t": 0.421923965215683, "grad_norm": 13.756513595581055, "learning_rate": 3.0507763319663517e-07, "logits/chosen": 0.5732111930847168, "logits/rejected": 0.49360191822052, "logps/chosen": -164.64968872070312, "logps/ref_chosen": -70.00630187988281, "logps/ref_rejected": -86.96690368652344, "logps/rejected": -221.24905395507812, "loss": 1.1802, "margin_dpo/margin_mean": 39.63875198364258, "margin_dpo/margin_std": 80.1014633178711, "step": 323 }, { "KL/chosen_KL_mean": -84.78605651855469, "KL/mean": -111.35850524902344, "KL/rejected_KL_mean": -137.93096923828125, "KL/std": 70.51113891601562, "epoch": 0.4897959183673469, "fcm_dpo/beta": 0.008961576968431473, "fcm_dpo/delta": -0.08072944730520248, "fcm_dpo/margin": 53.1449089050293, "fcm_dpo/q_t": 0.39130061864852905, "grad_norm": 17.810380935668945, "learning_rate": 3.0378713696502097e-07, "logits/chosen": 0.6615591645240784, "logits/rejected": 0.6014628410339355, "logps/chosen": -140.67486572265625, "logps/ref_chosen": -55.88882064819336, "logps/ref_rejected": -75.23088073730469, "logps/rejected": -213.16184997558594, "loss": 1.0472, "margin_dpo/margin_mean": 53.1449089050293, "margin_dpo/margin_std": 65.71917724609375, "step": 324 }, { "KL/chosen_KL_mean": -104.56004333496094, "KL/mean": -129.3292236328125, "KL/rejected_KL_mean": -154.098388671875, "KL/std": 67.77610778808594, "epoch": 0.491307634164777, "fcm_dpo/beta": 0.008865425363183022, "fcm_dpo/delta": -0.04157440736889839, "fcm_dpo/margin": 49.538360595703125, "fcm_dpo/q_t": 0.4014624059200287, "grad_norm": 14.201213836669922, "learning_rate": 3.0249513619156206e-07, "logits/chosen": 0.606431245803833, "logits/rejected": 0.5380966067314148, "logps/chosen": -168.70706176757812, "logps/ref_chosen": -64.14701843261719, "logps/ref_rejected": -79.91143798828125, "logps/rejected": -234.00982666015625, "loss": 1.1141, "margin_dpo/margin_mean": 49.53835678100586, "margin_dpo/margin_std": 81.64930725097656, "step": 325 }, { "KL/chosen_KL_mean": -121.84555053710938, "KL/mean": -132.28628540039062, "KL/rejected_KL_mean": -142.72705078125, "KL/std": 65.33851623535156, "epoch": 0.4928193499622071, "fcm_dpo/beta": 0.009159738197922707, "fcm_dpo/delta": 0.21309442818164825, "fcm_dpo/margin": 20.88149070739746, "fcm_dpo/q_t": 0.4580095410346985, "grad_norm": 15.675823211669922, "learning_rate": 3.012016670162977e-07, "logits/chosen": 0.5414941310882568, "logits/rejected": 0.5495314002037048, "logps/chosen": -197.37686157226562, "logps/ref_chosen": -75.53131103515625, "logps/ref_rejected": -76.5898666381836, "logps/rejected": -219.31692504882812, "loss": 1.312, "margin_dpo/margin_mean": 20.881492614746094, "margin_dpo/margin_std": 72.97334289550781, "step": 326 }, { "KL/chosen_KL_mean": -109.90470886230469, "KL/mean": -129.81394958496094, "KL/rejected_KL_mean": -149.72317504882812, "KL/std": 68.6119384765625, "epoch": 0.4943310657596372, "fcm_dpo/beta": 0.009345386177301407, "fcm_dpo/delta": 0.028610462322831154, "fcm_dpo/margin": 39.81847381591797, "fcm_dpo/q_t": 0.4180833101272583, "grad_norm": 16.183422088623047, "learning_rate": 2.99906765620341e-07, "logits/chosen": 0.5542730689048767, "logits/rejected": 0.5263379216194153, "logps/chosen": -179.24188232421875, "logps/ref_chosen": -69.33717346191406, "logps/ref_rejected": -73.37751770019531, "logps/rejected": -223.1007080078125, "loss": 1.1743, "margin_dpo/margin_mean": 39.81846618652344, "margin_dpo/margin_std": 78.34571838378906, "step": 327 }, { "KL/chosen_KL_mean": -96.4388427734375, "KL/mean": -120.63749694824219, "KL/rejected_KL_mean": -144.83615112304688, "KL/std": 68.9698486328125, "epoch": 0.4958427815570673, "fcm_dpo/beta": 0.009272612631320953, "fcm_dpo/delta": -0.05103527009487152, "fcm_dpo/margin": 48.39728927612305, "fcm_dpo/q_t": 0.39925065636634827, "grad_norm": 14.83205795288086, "learning_rate": 2.9861046822486766e-07, "logits/chosen": 0.5593709349632263, "logits/rejected": 0.519730269908905, "logps/chosen": -158.14508056640625, "logps/ref_chosen": -61.70623016357422, "logps/ref_rejected": -83.73808288574219, "logps/rejected": -228.57421875, "loss": 1.0764, "margin_dpo/margin_mean": 48.39728546142578, "margin_dpo/margin_std": 66.56529998779297, "step": 328 }, { "KL/chosen_KL_mean": -107.68646240234375, "KL/mean": -132.2223358154297, "KL/rejected_KL_mean": -156.75819396972656, "KL/std": 71.65765380859375, "epoch": 0.4973544973544973, "fcm_dpo/beta": 0.009159384295344353, "fcm_dpo/delta": -0.05180327221751213, "fcm_dpo/margin": 49.07172393798828, "fcm_dpo/q_t": 0.4012608826160431, "grad_norm": 14.937561988830566, "learning_rate": 2.9731281109010253e-07, "logits/chosen": 0.6999276876449585, "logits/rejected": 0.6358014345169067, "logps/chosen": -172.18487548828125, "logps/ref_chosen": -64.4984130859375, "logps/ref_rejected": -83.6591796875, "logps/rejected": -240.4173583984375, "loss": 1.0909, "margin_dpo/margin_mean": 49.071720123291016, "margin_dpo/margin_std": 73.95594787597656, "step": 329 }, { "KL/chosen_KL_mean": -92.5936508178711, "KL/mean": -118.45382690429688, "KL/rejected_KL_mean": -144.31399536132812, "KL/std": 67.32807922363281, "epoch": 0.4988662131519274, "fcm_dpo/beta": 0.009068884886801243, "fcm_dpo/delta": -0.07242908328771591, "fcm_dpo/margin": 51.72032928466797, "fcm_dpo/q_t": 0.3960675597190857, "grad_norm": 14.681183815002441, "learning_rate": 2.9601383051430505e-07, "logits/chosen": 0.6028814315795898, "logits/rejected": 0.53467857837677, "logps/chosen": -147.39828491210938, "logps/ref_chosen": -54.80464172363281, "logps/ref_rejected": -75.3194351196289, "logps/rejected": -219.6334228515625, "loss": 1.1115, "margin_dpo/margin_mean": 51.720333099365234, "margin_dpo/margin_std": 85.41941833496094, "step": 330 }, { "KL/chosen_KL_mean": -100.29885864257812, "KL/mean": -132.77366638183594, "KL/rejected_KL_mean": -165.24847412109375, "KL/std": 71.77218627929688, "epoch": 0.5003779289493575, "fcm_dpo/beta": 0.008779097348451614, "fcm_dpo/delta": -0.1802905797958374, "fcm_dpo/margin": 64.94962310791016, "fcm_dpo/q_t": 0.37083327770233154, "grad_norm": 12.52709674835205, "learning_rate": 2.947135628327544e-07, "logits/chosen": 0.7146250009536743, "logits/rejected": 0.6876611709594727, "logps/chosen": -159.54144287109375, "logps/ref_chosen": -59.242584228515625, "logps/ref_rejected": -69.87483215332031, "logps/rejected": -235.12330627441406, "loss": 1.003, "margin_dpo/margin_mean": 64.94961547851562, "margin_dpo/margin_std": 78.14910888671875, "step": 331 }, { "KL/chosen_KL_mean": -101.7713623046875, "KL/mean": -129.58682250976562, "KL/rejected_KL_mean": -157.40225219726562, "KL/std": 68.52835083007812, "epoch": 0.5018896447467877, "fcm_dpo/beta": 0.008652852848172188, "fcm_dpo/delta": -0.08633846044540405, "fcm_dpo/margin": 55.63090896606445, "fcm_dpo/q_t": 0.3918297290802002, "grad_norm": 12.344429016113281, "learning_rate": 2.934120444167326e-07, "logits/chosen": 0.5712667107582092, "logits/rejected": 0.5271207690238953, "logps/chosen": -168.88111877441406, "logps/ref_chosen": -67.10975646972656, "logps/ref_rejected": -77.11839294433594, "logps/rejected": -234.52066040039062, "loss": 1.0523, "margin_dpo/margin_mean": 55.63090515136719, "margin_dpo/margin_std": 69.46024322509766, "step": 332 }, { "KL/chosen_KL_mean": -109.85867309570312, "KL/mean": -139.08721923828125, "KL/rejected_KL_mean": -168.31573486328125, "KL/std": 68.70384216308594, "epoch": 0.5034013605442177, "fcm_dpo/beta": 0.008428534492850304, "fcm_dpo/delta": -0.09740786254405975, "fcm_dpo/margin": 58.457061767578125, "fcm_dpo/q_t": 0.38885655999183655, "grad_norm": 12.776857376098633, "learning_rate": 2.921093116725076e-07, "logits/chosen": 0.6594347953796387, "logits/rejected": 0.5772568583488464, "logps/chosen": -168.23980712890625, "logps/ref_chosen": -58.381134033203125, "logps/ref_rejected": -85.02839660644531, "logps/rejected": -253.34413146972656, "loss": 1.0402, "margin_dpo/margin_mean": 58.457061767578125, "margin_dpo/margin_std": 73.11105346679688, "step": 333 }, { "KL/chosen_KL_mean": -104.70248413085938, "KL/mean": -125.80743408203125, "KL/rejected_KL_mean": -146.91238403320312, "KL/std": 70.07809448242188, "epoch": 0.5049130763416477, "fcm_dpo/beta": 0.008443513885140419, "fcm_dpo/delta": 0.045077141374349594, "fcm_dpo/margin": 42.209896087646484, "fcm_dpo/q_t": 0.42087602615356445, "grad_norm": 13.54383659362793, "learning_rate": 2.9080540104031484e-07, "logits/chosen": 0.6578313112258911, "logits/rejected": 0.6086193919181824, "logps/chosen": -171.594482421875, "logps/ref_chosen": -66.89199829101562, "logps/ref_rejected": -91.83695220947266, "logps/rejected": -238.74932861328125, "loss": 1.1727, "margin_dpo/margin_mean": 42.209896087646484, "margin_dpo/margin_std": 81.92105865478516, "step": 334 }, { "KL/chosen_KL_mean": -106.31073760986328, "KL/mean": -129.83917236328125, "KL/rejected_KL_mean": -153.36761474609375, "KL/std": 69.8459701538086, "epoch": 0.5064247921390779, "fcm_dpo/beta": 0.008487050421535969, "fcm_dpo/delta": 0.00023875758051872253, "fcm_dpo/margin": 47.05687713623047, "fcm_dpo/q_t": 0.4113461375236511, "grad_norm": 17.41133689880371, "learning_rate": 2.895003489933375e-07, "logits/chosen": 0.6846311688423157, "logits/rejected": 0.6434398889541626, "logps/chosen": -167.8251953125, "logps/ref_chosen": -61.51445770263672, "logps/ref_rejected": -75.68916320800781, "logps/rejected": -229.05677795410156, "loss": 1.1342, "margin_dpo/margin_mean": 47.05687713623047, "margin_dpo/margin_std": 79.80413818359375, "step": 335 }, { "KL/chosen_KL_mean": -117.79705810546875, "KL/mean": -142.47894287109375, "KL/rejected_KL_mean": -167.1608428955078, "KL/std": 72.92349243164062, "epoch": 0.5079365079365079, "fcm_dpo/beta": 0.00837024487555027, "fcm_dpo/delta": -0.014911554753780365, "fcm_dpo/margin": 49.36377716064453, "fcm_dpo/q_t": 0.4105210602283478, "grad_norm": 13.49311637878418, "learning_rate": 2.8819419203668675e-07, "logits/chosen": 0.557252049446106, "logits/rejected": 0.5326744914054871, "logps/chosen": -186.64712524414062, "logps/ref_chosen": -68.85006713867188, "logps/ref_rejected": -92.99603271484375, "logps/rejected": -260.1568603515625, "loss": 1.1244, "margin_dpo/margin_mean": 49.36377716064453, "margin_dpo/margin_std": 81.3376693725586, "step": 336 }, { "KL/chosen_KL_mean": -122.55076599121094, "KL/mean": -142.1136016845703, "KL/rejected_KL_mean": -161.67642211914062, "KL/std": 71.70664978027344, "epoch": 0.509448223733938, "fcm_dpo/beta": 0.00851651281118393, "fcm_dpo/delta": 0.06903576850891113, "fcm_dpo/margin": 39.125640869140625, "fcm_dpo/q_t": 0.42501017451286316, "grad_norm": 12.491925239562988, "learning_rate": 2.8688696670638053e-07, "logits/chosen": 0.5466808080673218, "logits/rejected": 0.5131638050079346, "logps/chosen": -195.7386016845703, "logps/ref_chosen": -73.18783569335938, "logps/ref_rejected": -86.89118957519531, "logps/rejected": -248.56761169433594, "loss": 1.1792, "margin_dpo/margin_mean": 39.125640869140625, "margin_dpo/margin_std": 76.22183227539062, "step": 337 }, { "KL/chosen_KL_mean": -118.58080291748047, "KL/mean": -139.35684204101562, "KL/rejected_KL_mean": -160.13287353515625, "KL/std": 68.07583618164062, "epoch": 0.5109599395313681, "fcm_dpo/beta": 0.008585257455706596, "fcm_dpo/delta": 0.04488629475235939, "fcm_dpo/margin": 41.552066802978516, "fcm_dpo/q_t": 0.41984856128692627, "grad_norm": 12.422159194946289, "learning_rate": 2.8557870956832133e-07, "logits/chosen": 0.6132569313049316, "logits/rejected": 0.5857648849487305, "logps/chosen": -182.52041625976562, "logps/ref_chosen": -63.939613342285156, "logps/ref_rejected": -75.34243774414062, "logps/rejected": -235.47531127929688, "loss": 1.1626, "margin_dpo/margin_mean": 41.552066802978516, "margin_dpo/margin_std": 76.6851806640625, "step": 338 }, { "KL/chosen_KL_mean": -96.97520446777344, "KL/mean": -120.72698974609375, "KL/rejected_KL_mean": -144.478759765625, "KL/std": 67.07359313964844, "epoch": 0.5124716553287982, "fcm_dpo/beta": 0.00859010498970747, "fcm_dpo/delta": -0.008500054478645325, "fcm_dpo/margin": 47.50355529785156, "fcm_dpo/q_t": 0.4071218967437744, "grad_norm": 13.649033546447754, "learning_rate": 2.842694572172736e-07, "logits/chosen": 0.7820595502853394, "logits/rejected": 0.6895301342010498, "logps/chosen": -142.5243377685547, "logps/ref_chosen": -45.54913330078125, "logps/ref_rejected": -67.0482177734375, "logps/rejected": -211.5269775390625, "loss": 1.1103, "margin_dpo/margin_mean": 47.5035514831543, "margin_dpo/margin_std": 71.7981185913086, "step": 339 }, { "KL/chosen_KL_mean": -109.6189193725586, "KL/mean": -132.94656372070312, "KL/rejected_KL_mean": -156.27420043945312, "KL/std": 73.74958801269531, "epoch": 0.5139833711262283, "fcm_dpo/beta": 0.008629849180579185, "fcm_dpo/delta": -0.002979576587677002, "fcm_dpo/margin": 46.65528106689453, "fcm_dpo/q_t": 0.41186124086380005, "grad_norm": 12.804910659790039, "learning_rate": 2.8295924627584004e-07, "logits/chosen": 0.6743656396865845, "logits/rejected": 0.6509165167808533, "logps/chosen": -163.62457275390625, "logps/ref_chosen": -54.00564956665039, "logps/ref_rejected": -61.314430236816406, "logps/rejected": -217.58863830566406, "loss": 1.1527, "margin_dpo/margin_mean": 46.65528106689453, "margin_dpo/margin_std": 86.06853485107422, "step": 340 }, { "KL/chosen_KL_mean": -108.90359497070312, "KL/mean": -135.95693969726562, "KL/rejected_KL_mean": -163.01025390625, "KL/std": 72.9114990234375, "epoch": 0.5154950869236583, "fcm_dpo/beta": 0.00841558538377285, "fcm_dpo/delta": -0.06184224784374237, "fcm_dpo/margin": 54.10666275024414, "fcm_dpo/q_t": 0.3978581130504608, "grad_norm": 14.011096000671387, "learning_rate": 2.816481133934373e-07, "logits/chosen": 0.6594457626342773, "logits/rejected": 0.6075119972229004, "logps/chosen": -172.29869079589844, "logps/ref_chosen": -63.39509582519531, "logps/ref_rejected": -76.20973205566406, "logps/rejected": -239.22000122070312, "loss": 1.0938, "margin_dpo/margin_mean": 54.10666275024414, "margin_dpo/margin_std": 77.17527770996094, "step": 341 }, { "KL/chosen_KL_mean": -108.45635986328125, "KL/mean": -135.4060821533203, "KL/rejected_KL_mean": -162.35580444335938, "KL/std": 76.76370239257812, "epoch": 0.5170068027210885, "fcm_dpo/beta": 0.008395740762352943, "fcm_dpo/delta": -0.05540557950735092, "fcm_dpo/margin": 53.899452209472656, "fcm_dpo/q_t": 0.3998180329799652, "grad_norm": 12.18912410736084, "learning_rate": 2.8033609524527046e-07, "logits/chosen": 0.7089950442314148, "logits/rejected": 0.6643673181533813, "logps/chosen": -161.50418090820312, "logps/ref_chosen": -53.047813415527344, "logps/ref_rejected": -68.2854232788086, "logps/rejected": -230.6412353515625, "loss": 1.0868, "margin_dpo/margin_mean": 53.899452209472656, "margin_dpo/margin_std": 78.99366760253906, "step": 342 }, { "KL/chosen_KL_mean": -101.50072479248047, "KL/mean": -121.11677551269531, "KL/rejected_KL_mean": -140.73281860351562, "KL/std": 71.01953125, "epoch": 0.5185185185185185, "fcm_dpo/beta": 0.00842662900686264, "fcm_dpo/delta": 0.07119348645210266, "fcm_dpo/margin": 39.232112884521484, "fcm_dpo/q_t": 0.4252588152885437, "grad_norm": 11.982053756713867, "learning_rate": 2.7902322853130753e-07, "logits/chosen": 0.5230769515037537, "logits/rejected": 0.5136237740516663, "logps/chosen": -172.07925415039062, "logps/ref_chosen": -70.57852935791016, "logps/ref_rejected": -84.73873901367188, "logps/rejected": -225.4715576171875, "loss": 1.1768, "margin_dpo/margin_mean": 39.23210906982422, "margin_dpo/margin_std": 73.24249267578125, "step": 343 }, { "KL/chosen_KL_mean": -111.83555603027344, "KL/mean": -140.12460327148438, "KL/rejected_KL_mean": -168.41366577148438, "KL/std": 71.88983154296875, "epoch": 0.5200302343159486, "fcm_dpo/beta": 0.008404587395489216, "fcm_dpo/delta": -0.07948686927556992, "fcm_dpo/margin": 56.57810592651367, "fcm_dpo/q_t": 0.3939194977283478, "grad_norm": 15.187760353088379, "learning_rate": 2.7770954997525274e-07, "logits/chosen": 0.6720702648162842, "logits/rejected": 0.5977091789245605, "logps/chosen": -167.6465606689453, "logps/ref_chosen": -55.811004638671875, "logps/ref_rejected": -84.77637481689453, "logps/rejected": -253.19003295898438, "loss": 1.0683, "margin_dpo/margin_mean": 56.57810974121094, "margin_dpo/margin_std": 79.43333435058594, "step": 344 }, { "KL/chosen_KL_mean": -90.22946166992188, "KL/mean": -114.63905334472656, "KL/rejected_KL_mean": -139.0486602783203, "KL/std": 65.07427978515625, "epoch": 0.5215419501133787, "fcm_dpo/beta": 0.008393699303269386, "fcm_dpo/delta": -0.01029902882874012, "fcm_dpo/margin": 48.81920623779297, "fcm_dpo/q_t": 0.40778762102127075, "grad_norm": 13.599929809570312, "learning_rate": 2.7639509632351927e-07, "logits/chosen": 0.691046416759491, "logits/rejected": 0.6441917419433594, "logps/chosen": -148.01556396484375, "logps/ref_chosen": -57.78609848022461, "logps/ref_rejected": -78.91847229003906, "logps/rejected": -217.96713256835938, "loss": 1.1166, "margin_dpo/margin_mean": 48.81920623779297, "margin_dpo/margin_std": 77.43244171142578, "step": 345 }, { "KL/chosen_KL_mean": -98.14485168457031, "KL/mean": -127.07212829589844, "KL/rejected_KL_mean": -155.99942016601562, "KL/std": 75.85902404785156, "epoch": 0.5230536659108088, "fcm_dpo/beta": 0.008301769383251667, "fcm_dpo/delta": -0.08439122140407562, "fcm_dpo/margin": 57.854549407958984, "fcm_dpo/q_t": 0.3912527859210968, "grad_norm": 14.651946067810059, "learning_rate": 2.7507990434420123e-07, "logits/chosen": 0.674179196357727, "logits/rejected": 0.5830097198486328, "logps/chosen": -154.4299774169922, "logps/ref_chosen": -56.285125732421875, "logps/ref_rejected": -91.15303039550781, "logps/rejected": -247.15243530273438, "loss": 1.0674, "margin_dpo/margin_mean": 57.854549407958984, "margin_dpo/margin_std": 78.946533203125, "step": 346 }, { "KL/chosen_KL_mean": -107.67420959472656, "KL/mean": -130.18055725097656, "KL/rejected_KL_mean": -152.6868896484375, "KL/std": 72.40330505371094, "epoch": 0.5245653817082389, "fcm_dpo/beta": 0.008253240026533604, "fcm_dpo/delta": 0.02960244007408619, "fcm_dpo/margin": 45.0127067565918, "fcm_dpo/q_t": 0.4166538715362549, "grad_norm": 15.322473526000977, "learning_rate": 2.737640108260456e-07, "logits/chosen": 0.8045604228973389, "logits/rejected": 0.7491985559463501, "logps/chosen": -161.1737518310547, "logps/ref_chosen": -53.499542236328125, "logps/ref_rejected": -72.52565002441406, "logps/rejected": -225.21255493164062, "loss": 1.1389, "margin_dpo/margin_mean": 45.0127067565918, "margin_dpo/margin_std": 75.2705078125, "step": 347 }, { "KL/chosen_KL_mean": -96.49987030029297, "KL/mean": -123.30168914794922, "KL/rejected_KL_mean": -150.10350036621094, "KL/std": 69.12874603271484, "epoch": 0.5260770975056689, "fcm_dpo/beta": 0.008198895491659641, "fcm_dpo/delta": -0.04165411740541458, "fcm_dpo/margin": 53.603633880615234, "fcm_dpo/q_t": 0.4039544463157654, "grad_norm": 13.738903999328613, "learning_rate": 2.724474525774229e-07, "logits/chosen": 0.7492630481719971, "logits/rejected": 0.7167302370071411, "logps/chosen": -147.28671264648438, "logps/ref_chosen": -50.78684997558594, "logps/ref_rejected": -68.63732147216797, "logps/rejected": -218.74081420898438, "loss": 1.1037, "margin_dpo/margin_mean": 53.6036376953125, "margin_dpo/margin_std": 84.41233825683594, "step": 348 }, { "KL/chosen_KL_mean": -95.79183959960938, "KL/mean": -123.4986572265625, "KL/rejected_KL_mean": -151.2054443359375, "KL/std": 74.21245574951172, "epoch": 0.527588813303099, "fcm_dpo/beta": 0.00817472767084837, "fcm_dpo/delta": -0.0556333102285862, "fcm_dpo/margin": 55.413612365722656, "fcm_dpo/q_t": 0.399311363697052, "grad_norm": 13.029300689697266, "learning_rate": 2.711302664252973e-07, "logits/chosen": 0.6797126531600952, "logits/rejected": 0.5815830230712891, "logps/chosen": -149.11685180664062, "logps/ref_chosen": -53.325008392333984, "logps/ref_rejected": -83.21236419677734, "logps/rejected": -234.41781616210938, "loss": 1.0839, "margin_dpo/margin_mean": 55.413612365722656, "margin_dpo/margin_std": 79.81897735595703, "step": 349 }, { "KL/chosen_KL_mean": -102.28199005126953, "KL/mean": -136.58529663085938, "KL/rejected_KL_mean": -170.88861083984375, "KL/std": 79.92459869384766, "epoch": 0.5291005291005291, "fcm_dpo/beta": 0.007922561839222908, "fcm_dpo/delta": -0.1521030217409134, "fcm_dpo/margin": 68.60662841796875, "fcm_dpo/q_t": 0.376028448343277, "grad_norm": 16.23141098022461, "learning_rate": 2.698124892141971e-07, "logits/chosen": 0.6232413053512573, "logits/rejected": 0.5423790216445923, "logps/chosen": -163.9077606201172, "logps/ref_chosen": -61.625770568847656, "logps/ref_rejected": -87.63627624511719, "logps/rejected": -258.52490234375, "loss": 1.0016, "margin_dpo/margin_mean": 68.60662841796875, "margin_dpo/margin_std": 77.74540710449219, "step": 350 }, { "KL/chosen_KL_mean": -97.53581237792969, "KL/mean": -122.48731231689453, "KL/rejected_KL_mean": -147.43881225585938, "KL/std": 68.1795883178711, "epoch": 0.5306122448979592, "fcm_dpo/beta": 0.007859094068408012, "fcm_dpo/delta": 0.007957909256219864, "fcm_dpo/margin": 49.903011322021484, "fcm_dpo/q_t": 0.4095124900341034, "grad_norm": 13.02753734588623, "learning_rate": 2.6849415780518357e-07, "logits/chosen": 0.5934625267982483, "logits/rejected": 0.515455424785614, "logps/chosen": -153.79214477539062, "logps/ref_chosen": -56.2563362121582, "logps/ref_rejected": -79.11589813232422, "logps/rejected": -226.55471801757812, "loss": 1.1386, "margin_dpo/margin_mean": 49.90301513671875, "margin_dpo/margin_std": 84.64741516113281, "step": 351 }, { "KL/chosen_KL_mean": -97.65863037109375, "KL/mean": -126.34355163574219, "KL/rejected_KL_mean": -155.02847290039062, "KL/std": 74.74703979492188, "epoch": 0.5321239606953893, "fcm_dpo/beta": 0.007809111848473549, "fcm_dpo/delta": -0.0504891499876976, "fcm_dpo/margin": 57.369850158691406, "fcm_dpo/q_t": 0.3992422819137573, "grad_norm": 12.689568519592285, "learning_rate": 2.6717530907482024e-07, "logits/chosen": 0.6787396669387817, "logits/rejected": 0.6228320598602295, "logps/chosen": -160.7105712890625, "logps/ref_chosen": -63.05195236206055, "logps/ref_rejected": -85.52035522460938, "logps/rejected": -240.548828125, "loss": 1.0786, "margin_dpo/margin_mean": 57.369850158691406, "margin_dpo/margin_std": 80.87751007080078, "step": 352 }, { "KL/chosen_KL_mean": -95.69207763671875, "KL/mean": -122.7952880859375, "KL/rejected_KL_mean": -149.8984832763672, "KL/std": 70.47689819335938, "epoch": 0.5336356764928194, "fcm_dpo/beta": 0.007794972509145737, "fcm_dpo/delta": -0.023562796413898468, "fcm_dpo/margin": 54.2064094543457, "fcm_dpo/q_t": 0.40339455008506775, "grad_norm": 12.208295822143555, "learning_rate": 2.658559799141411e-07, "logits/chosen": 0.6846418976783752, "logits/rejected": 0.6894683837890625, "logps/chosen": -164.70126342773438, "logps/ref_chosen": -69.00918579101562, "logps/ref_rejected": -72.65840148925781, "logps/rejected": -222.556884765625, "loss": 1.0904, "margin_dpo/margin_mean": 54.20640563964844, "margin_dpo/margin_std": 75.77108764648438, "step": 353 }, { "KL/chosen_KL_mean": -96.5929183959961, "KL/mean": -127.48405456542969, "KL/rejected_KL_mean": -158.37518310546875, "KL/std": 68.72752380371094, "epoch": 0.5351473922902494, "fcm_dpo/beta": 0.007651031948626041, "fcm_dpo/delta": -0.07695139944553375, "fcm_dpo/margin": 61.78227615356445, "fcm_dpo/q_t": 0.39242392778396606, "grad_norm": 13.139801025390625, "learning_rate": 2.6453620722761895e-07, "logits/chosen": 0.7695657014846802, "logits/rejected": 0.6296084523200989, "logps/chosen": -136.38125610351562, "logps/ref_chosen": -39.78833770751953, "logps/ref_rejected": -69.56885528564453, "logps/rejected": -227.94406127929688, "loss": 1.0635, "margin_dpo/margin_mean": 61.78227996826172, "margin_dpo/margin_std": 82.7366943359375, "step": 354 }, { "KL/chosen_KL_mean": -103.11363220214844, "KL/mean": -134.59609985351562, "KL/rejected_KL_mean": -166.0785675048828, "KL/std": 77.29907989501953, "epoch": 0.5366591080876795, "fcm_dpo/beta": 0.007575302384793758, "fcm_dpo/delta": -0.0807543694972992, "fcm_dpo/margin": 62.96492004394531, "fcm_dpo/q_t": 0.3925732970237732, "grad_norm": 15.88890552520752, "learning_rate": 2.632160279321328e-07, "logits/chosen": 0.7179785966873169, "logits/rejected": 0.5774426460266113, "logps/chosen": -149.3690185546875, "logps/ref_chosen": -46.25537872314453, "logps/ref_rejected": -78.20236206054688, "logps/rejected": -244.28091430664062, "loss": 1.0732, "margin_dpo/margin_mean": 62.96492385864258, "margin_dpo/margin_std": 89.31265258789062, "step": 355 }, { "KL/chosen_KL_mean": -97.25872802734375, "KL/mean": -123.20895385742188, "KL/rejected_KL_mean": -149.1591796875, "KL/std": 73.4035873413086, "epoch": 0.5381708238851096, "fcm_dpo/beta": 0.00750363152474165, "fcm_dpo/delta": 0.010530386120080948, "fcm_dpo/margin": 51.900447845458984, "fcm_dpo/q_t": 0.41315823793411255, "grad_norm": 12.564486503601074, "learning_rate": 2.618954789559356e-07, "logits/chosen": 0.660346269607544, "logits/rejected": 0.5757850408554077, "logps/chosen": -145.16488647460938, "logps/ref_chosen": -47.906158447265625, "logps/ref_rejected": -74.29397583007812, "logps/rejected": -223.45315551757812, "loss": 1.1549, "margin_dpo/margin_mean": 51.90044403076172, "margin_dpo/margin_std": 94.7741470336914, "step": 356 }, { "KL/chosen_KL_mean": -115.6062240600586, "KL/mean": -137.5577392578125, "KL/rejected_KL_mean": -159.50926208496094, "KL/std": 72.64031982421875, "epoch": 0.5396825396825397, "fcm_dpo/beta": 0.007546389475464821, "fcm_dpo/delta": 0.06960046291351318, "fcm_dpo/margin": 43.90301513671875, "fcm_dpo/q_t": 0.4237821698188782, "grad_norm": 13.361268997192383, "learning_rate": 2.6057459723762076e-07, "logits/chosen": 0.6856166124343872, "logits/rejected": 0.6604490280151367, "logps/chosen": -178.24122619628906, "logps/ref_chosen": -62.63500213623047, "logps/ref_rejected": -65.11399841308594, "logps/rejected": -224.62326049804688, "loss": 1.1781, "margin_dpo/margin_mean": 43.90301513671875, "margin_dpo/margin_std": 79.18133544921875, "step": 357 }, { "KL/chosen_KL_mean": -111.6898193359375, "KL/mean": -142.54754638671875, "KL/rejected_KL_mean": -173.40528869628906, "KL/std": 76.13407897949219, "epoch": 0.5411942554799698, "fcm_dpo/beta": 0.007627150975167751, "fcm_dpo/delta": -0.07491104304790497, "fcm_dpo/margin": 61.71546173095703, "fcm_dpo/q_t": 0.3949502408504486, "grad_norm": 16.380239486694336, "learning_rate": 2.5925341972508954e-07, "logits/chosen": 0.589666485786438, "logits/rejected": 0.6041041612625122, "logps/chosen": -178.89944458007812, "logps/ref_chosen": -67.20960998535156, "logps/ref_rejected": -69.34715270996094, "logps/rejected": -242.75244140625, "loss": 1.0715, "margin_dpo/margin_mean": 61.71546173095703, "margin_dpo/margin_std": 84.65841674804688, "step": 358 }, { "KL/chosen_KL_mean": -126.52949523925781, "KL/mean": -141.81393432617188, "KL/rejected_KL_mean": -157.098388671875, "KL/std": 74.38471984863281, "epoch": 0.5427059712773998, "fcm_dpo/beta": 0.007773904595524073, "fcm_dpo/delta": 0.16444545984268188, "fcm_dpo/margin": 30.568878173828125, "fcm_dpo/q_t": 0.44677430391311646, "grad_norm": 15.262069702148438, "learning_rate": 2.579319833745169e-07, "logits/chosen": 0.628871500492096, "logits/rejected": 0.5945923328399658, "logps/chosen": -189.05528259277344, "logps/ref_chosen": -62.52578353881836, "logps/ref_rejected": -76.63114929199219, "logps/rejected": -233.72952270507812, "loss": 1.2533, "margin_dpo/margin_mean": 30.56888198852539, "margin_dpo/margin_std": 77.31507110595703, "step": 359 }, { "KL/chosen_KL_mean": -121.8115234375, "KL/mean": -148.08889770507812, "KL/rejected_KL_mean": -174.36624145507812, "KL/std": 79.44476318359375, "epoch": 0.54421768707483, "fcm_dpo/beta": 0.007763609290122986, "fcm_dpo/delta": -0.008455626666545868, "fcm_dpo/margin": 52.55471420288086, "fcm_dpo/q_t": 0.4087502360343933, "grad_norm": 12.670330047607422, "learning_rate": 2.5661032514931834e-07, "logits/chosen": 0.5793747901916504, "logits/rejected": 0.48438405990600586, "logps/chosen": -185.29925537109375, "logps/ref_chosen": -63.48772048950195, "logps/ref_rejected": -90.6891098022461, "logps/rejected": -265.05535888671875, "loss": 1.108, "margin_dpo/margin_mean": 52.554718017578125, "margin_dpo/margin_std": 79.45843505859375, "step": 360 }, { "KL/chosen_KL_mean": -116.59442901611328, "KL/mean": -147.58245849609375, "KL/rejected_KL_mean": -178.57049560546875, "KL/std": 77.82331848144531, "epoch": 0.54572940287226, "fcm_dpo/beta": 0.007699761539697647, "fcm_dpo/delta": -0.08099059760570526, "fcm_dpo/margin": 61.97608184814453, "fcm_dpo/q_t": 0.3913147449493408, "grad_norm": 12.27685546875, "learning_rate": 2.552884820191154e-07, "logits/chosen": 0.7169207334518433, "logits/rejected": 0.6682182550430298, "logps/chosen": -174.51156616210938, "logps/ref_chosen": -57.917144775390625, "logps/ref_rejected": -72.39089965820312, "logps/rejected": -250.96139526367188, "loss": 1.0528, "margin_dpo/margin_mean": 61.97608184814453, "margin_dpo/margin_std": 79.43408203125, "step": 361 }, { "KL/chosen_KL_mean": -118.05259704589844, "KL/mean": -151.82545471191406, "KL/rejected_KL_mean": -185.5983123779297, "KL/std": 84.501220703125, "epoch": 0.54724111866969, "fcm_dpo/beta": 0.0075636012479662895, "fcm_dpo/delta": -0.11687410622835159, "fcm_dpo/margin": 67.54571533203125, "fcm_dpo/q_t": 0.3894726037979126, "grad_norm": 13.32232666015625, "learning_rate": 2.53966490958702e-07, "logits/chosen": 0.7364634275436401, "logits/rejected": 0.6219512224197388, "logps/chosen": -181.49606323242188, "logps/ref_chosen": -63.4434700012207, "logps/ref_rejected": -103.45516967773438, "logps/rejected": -289.053466796875, "loss": 1.0648, "margin_dpo/margin_mean": 67.54571533203125, "margin_dpo/margin_std": 99.06629180908203, "step": 362 }, { "KL/chosen_KL_mean": -123.294921875, "KL/mean": -155.24533081054688, "KL/rejected_KL_mean": -187.1957550048828, "KL/std": 75.24211883544922, "epoch": 0.5487528344671202, "fcm_dpo/beta": 0.007382534444332123, "fcm_dpo/delta": -0.07541361451148987, "fcm_dpo/margin": 63.90083312988281, "fcm_dpo/q_t": 0.39400172233581543, "grad_norm": 14.595932006835938, "learning_rate": 2.526443889470099e-07, "logits/chosen": 0.7039732933044434, "logits/rejected": 0.5658940076828003, "logps/chosen": -171.94674682617188, "logps/ref_chosen": -48.65182876586914, "logps/ref_rejected": -88.65904235839844, "logps/rejected": -275.85479736328125, "loss": 1.0634, "margin_dpo/margin_mean": 63.90083694458008, "margin_dpo/margin_std": 87.35243225097656, "step": 363 }, { "KL/chosen_KL_mean": -109.17286682128906, "KL/mean": -144.913330078125, "KL/rejected_KL_mean": -180.65379333496094, "KL/std": 83.70720672607422, "epoch": 0.5502645502645502, "fcm_dpo/beta": 0.007225670851767063, "fcm_dpo/delta": -0.12290657311677933, "fcm_dpo/margin": 71.4809341430664, "fcm_dpo/q_t": 0.3865968585014343, "grad_norm": 11.887857437133789, "learning_rate": 2.513222129660744e-07, "logits/chosen": 0.603469729423523, "logits/rejected": 0.5128209590911865, "logps/chosen": -167.0439453125, "logps/ref_chosen": -57.87107467651367, "logps/ref_rejected": -80.95503234863281, "logps/rejected": -261.60882568359375, "loss": 1.0645, "margin_dpo/margin_mean": 71.4809341430664, "margin_dpo/margin_std": 104.8623046875, "step": 364 }, { "KL/chosen_KL_mean": -102.30352020263672, "KL/mean": -135.4107208251953, "KL/rejected_KL_mean": -168.51791381835938, "KL/std": 87.6494369506836, "epoch": 0.5517762660619804, "fcm_dpo/beta": 0.007087586913257837, "fcm_dpo/delta": -0.07303847372531891, "fcm_dpo/margin": 66.21439361572266, "fcm_dpo/q_t": 0.39243191480636597, "grad_norm": 11.069833755493164, "learning_rate": 2.5e-07, "logits/chosen": 0.651065468788147, "logits/rejected": 0.6449978351593018, "logps/chosen": -167.24569702148438, "logps/ref_chosen": -64.94217681884766, "logps/ref_rejected": -74.8599853515625, "logps/rejected": -243.37789916992188, "loss": 1.039, "margin_dpo/margin_mean": 66.21439361572266, "margin_dpo/margin_std": 77.24095153808594, "step": 365 }, { "KL/chosen_KL_mean": -101.74160766601562, "KL/mean": -129.18167114257812, "KL/rejected_KL_mean": -156.62171936035156, "KL/std": 77.08753967285156, "epoch": 0.5532879818594104, "fcm_dpo/beta": 0.007117936387658119, "fcm_dpo/delta": 0.009379461407661438, "fcm_dpo/margin": 54.8801155090332, "fcm_dpo/q_t": 0.41335535049438477, "grad_norm": 13.69518756866455, "learning_rate": 2.486777870339255e-07, "logits/chosen": 0.6271668672561646, "logits/rejected": 0.6168335676193237, "logps/chosen": -156.9075927734375, "logps/ref_chosen": -55.16598129272461, "logps/ref_rejected": -65.26121520996094, "logps/rejected": -221.8829345703125, "loss": 1.1506, "margin_dpo/margin_mean": 54.88011169433594, "margin_dpo/margin_std": 99.16736602783203, "step": 366 }, { "KL/chosen_KL_mean": -116.36273193359375, "KL/mean": -143.30245971679688, "KL/rejected_KL_mean": -170.2421875, "KL/std": 78.35556030273438, "epoch": 0.5547996976568406, "fcm_dpo/beta": 0.007099668495357037, "fcm_dpo/delta": 0.01813933067023754, "fcm_dpo/margin": 53.87945556640625, "fcm_dpo/q_t": 0.41123294830322266, "grad_norm": 12.775952339172363, "learning_rate": 2.4735561105299014e-07, "logits/chosen": 0.663223147392273, "logits/rejected": 0.5544114112854004, "logps/chosen": -172.37319946289062, "logps/ref_chosen": -56.01046371459961, "logps/ref_rejected": -77.31010437011719, "logps/rejected": -247.5522918701172, "loss": 1.1225, "margin_dpo/margin_mean": 53.879451751708984, "margin_dpo/margin_std": 83.09695434570312, "step": 367 }, { "KL/chosen_KL_mean": -128.00091552734375, "KL/mean": -153.79397583007812, "KL/rejected_KL_mean": -179.5870361328125, "KL/std": 77.07600402832031, "epoch": 0.5563114134542706, "fcm_dpo/beta": 0.007140608504414558, "fcm_dpo/delta": 0.03284718468785286, "fcm_dpo/margin": 51.58610534667969, "fcm_dpo/q_t": 0.41517072916030884, "grad_norm": 13.476099014282227, "learning_rate": 2.46033509041298e-07, "logits/chosen": 0.5109145045280457, "logits/rejected": 0.5121912360191345, "logps/chosen": -202.8302001953125, "logps/ref_chosen": -74.82927703857422, "logps/ref_rejected": -76.11680603027344, "logps/rejected": -255.70384216308594, "loss": 1.1322, "margin_dpo/margin_mean": 51.58610534667969, "margin_dpo/margin_std": 81.55989837646484, "step": 368 }, { "KL/chosen_KL_mean": -119.14222717285156, "KL/mean": -142.11587524414062, "KL/rejected_KL_mean": -165.08953857421875, "KL/std": 77.98490905761719, "epoch": 0.5578231292517006, "fcm_dpo/beta": 0.007270177826285362, "fcm_dpo/delta": 0.0675322413444519, "fcm_dpo/margin": 45.947303771972656, "fcm_dpo/q_t": 0.42452847957611084, "grad_norm": 12.494619369506836, "learning_rate": 2.447115179808846e-07, "logits/chosen": 0.6886367797851562, "logits/rejected": 0.6335114240646362, "logps/chosen": -177.46844482421875, "logps/ref_chosen": -58.32621765136719, "logps/ref_rejected": -80.92183685302734, "logps/rejected": -246.01138305664062, "loss": 1.1678, "margin_dpo/margin_mean": 45.947303771972656, "margin_dpo/margin_std": 83.54072570800781, "step": 369 }, { "KL/chosen_KL_mean": -114.87617492675781, "KL/mean": -147.30535888671875, "KL/rejected_KL_mean": -179.7345733642578, "KL/std": 81.04230499267578, "epoch": 0.5593348450491308, "fcm_dpo/beta": 0.007172431331127882, "fcm_dpo/delta": -0.06864120811223984, "fcm_dpo/margin": 64.8583984375, "fcm_dpo/q_t": 0.393879234790802, "grad_norm": 12.52617073059082, "learning_rate": 2.4338967485068164e-07, "logits/chosen": 0.7644656896591187, "logits/rejected": 0.6927477717399597, "logps/chosen": -167.7598876953125, "logps/ref_chosen": -52.88372039794922, "logps/ref_rejected": -79.43692016601562, "logps/rejected": -259.1714782714844, "loss": 1.0801, "margin_dpo/margin_mean": 64.85840606689453, "margin_dpo/margin_std": 94.23211669921875, "step": 370 }, { "KL/chosen_KL_mean": -115.05244445800781, "KL/mean": -142.79544067382812, "KL/rejected_KL_mean": -170.53843688964844, "KL/std": 77.92510986328125, "epoch": 0.5608465608465608, "fcm_dpo/beta": 0.007202905602753162, "fcm_dpo/delta": -0.000254802405834198, "fcm_dpo/margin": 55.48599624633789, "fcm_dpo/q_t": 0.4096566438674927, "grad_norm": 15.842639923095703, "learning_rate": 2.420680166254831e-07, "logits/chosen": 0.8232964277267456, "logits/rejected": 0.788852870464325, "logps/chosen": -164.2766571044922, "logps/ref_chosen": -49.224212646484375, "logps/ref_rejected": -63.348472595214844, "logps/rejected": -233.88690185546875, "loss": 1.1171, "margin_dpo/margin_mean": 55.486000061035156, "margin_dpo/margin_std": 85.15362548828125, "step": 371 }, { "KL/chosen_KL_mean": -123.14059448242188, "KL/mean": -139.93470764160156, "KL/rejected_KL_mean": -156.72882080078125, "KL/std": 77.38941192626953, "epoch": 0.562358276643991, "fcm_dpo/beta": 0.007378925569355488, "fcm_dpo/delta": 0.1541157066822052, "fcm_dpo/margin": 33.58821487426758, "fcm_dpo/q_t": 0.4454403519630432, "grad_norm": 16.654294967651367, "learning_rate": 2.4074658027491044e-07, "logits/chosen": 0.6877930760383606, "logits/rejected": 0.5879219770431519, "logps/chosen": -175.41015625, "logps/ref_chosen": -52.269554138183594, "logps/ref_rejected": -72.99522399902344, "logps/rejected": -229.72402954101562, "loss": 1.2892, "margin_dpo/margin_mean": 33.588218688964844, "margin_dpo/margin_std": 101.15355682373047, "step": 372 }, { "KL/chosen_KL_mean": -136.29205322265625, "KL/mean": -160.035888671875, "KL/rejected_KL_mean": -183.77972412109375, "KL/std": 79.77000427246094, "epoch": 0.563869992441421, "fcm_dpo/beta": 0.007445330731570721, "fcm_dpo/delta": 0.048104479908943176, "fcm_dpo/margin": 47.48767852783203, "fcm_dpo/q_t": 0.4214223027229309, "grad_norm": 13.432877540588379, "learning_rate": 2.394254027623792e-07, "logits/chosen": 0.6861114501953125, "logits/rejected": 0.6134101152420044, "logps/chosen": -197.40504455566406, "logps/ref_chosen": -61.112998962402344, "logps/ref_rejected": -76.24851989746094, "logps/rejected": -260.0282287597656, "loss": 1.1955, "margin_dpo/margin_mean": 47.48767852783203, "margin_dpo/margin_std": 100.75148010253906, "step": 373 }, { "KL/chosen_KL_mean": -116.14297485351562, "KL/mean": -153.75967407226562, "KL/rejected_KL_mean": -191.37637329101562, "KL/std": 82.49403381347656, "epoch": 0.5653817082388511, "fcm_dpo/beta": 0.007278250530362129, "fcm_dpo/delta": -0.1563607156276703, "fcm_dpo/margin": 75.23340606689453, "fcm_dpo/q_t": 0.37690961360931396, "grad_norm": 13.282299995422363, "learning_rate": 2.381045210440644e-07, "logits/chosen": 0.6100819110870361, "logits/rejected": 0.6109589338302612, "logps/chosen": -188.8121795654297, "logps/ref_chosen": -72.66920471191406, "logps/ref_rejected": -76.83158874511719, "logps/rejected": -268.20794677734375, "loss": 1.0268, "margin_dpo/margin_mean": 75.23340606689453, "margin_dpo/margin_std": 97.93537902832031, "step": 374 }, { "KL/chosen_KL_mean": -113.97125244140625, "KL/mean": -142.1385040283203, "KL/rejected_KL_mean": -170.3057403564453, "KL/std": 88.80377197265625, "epoch": 0.5668934240362812, "fcm_dpo/beta": 0.007202597334980965, "fcm_dpo/delta": -0.006175890564918518, "fcm_dpo/margin": 56.33448791503906, "fcm_dpo/q_t": 0.40949082374572754, "grad_norm": 15.551325798034668, "learning_rate": 2.3678397206786715e-07, "logits/chosen": 0.7072441577911377, "logits/rejected": 0.6454247236251831, "logps/chosen": -171.65455627441406, "logps/ref_chosen": -57.68330383300781, "logps/ref_rejected": -79.34097290039062, "logps/rejected": -249.64671325683594, "loss": 1.133, "margin_dpo/margin_mean": 56.33448028564453, "margin_dpo/margin_std": 96.07463073730469, "step": 375 }, { "KL/chosen_KL_mean": -122.6119384765625, "KL/mean": -156.71728515625, "KL/rejected_KL_mean": -190.8226318359375, "KL/std": 88.48606872558594, "epoch": 0.5684051398337112, "fcm_dpo/beta": 0.007107567973434925, "fcm_dpo/delta": -0.0894772857427597, "fcm_dpo/margin": 68.2106704711914, "fcm_dpo/q_t": 0.3942412734031677, "grad_norm": 13.224767684936523, "learning_rate": 2.3546379277238103e-07, "logits/chosen": 0.7721911668777466, "logits/rejected": 0.6945620775222778, "logps/chosen": -174.2860107421875, "logps/ref_chosen": -51.674072265625, "logps/ref_rejected": -75.69713592529297, "logps/rejected": -266.519775390625, "loss": 1.0832, "margin_dpo/margin_mean": 68.2106704711914, "margin_dpo/margin_std": 104.27572631835938, "step": 376 }, { "KL/chosen_KL_mean": -124.97073364257812, "KL/mean": -148.96510314941406, "KL/rejected_KL_mean": -172.95948791503906, "KL/std": 75.81988525390625, "epoch": 0.5699168556311414, "fcm_dpo/beta": 0.007183787412941456, "fcm_dpo/delta": 0.05659983307123184, "fcm_dpo/margin": 47.98875045776367, "fcm_dpo/q_t": 0.42107149958610535, "grad_norm": 13.312004089355469, "learning_rate": 2.3414402008585886e-07, "logits/chosen": 0.6740518808364868, "logits/rejected": 0.6495604515075684, "logps/chosen": -171.14926147460938, "logps/ref_chosen": -46.17853546142578, "logps/ref_rejected": -57.756500244140625, "logps/rejected": -230.71597290039062, "loss": 1.1638, "margin_dpo/margin_mean": 47.988746643066406, "margin_dpo/margin_std": 86.49845886230469, "step": 377 }, { "KL/chosen_KL_mean": -125.39057922363281, "KL/mean": -147.92562866210938, "KL/rejected_KL_mean": -170.46072387695312, "KL/std": 82.615234375, "epoch": 0.5714285714285714, "fcm_dpo/beta": 0.0072835953906178474, "fcm_dpo/delta": 0.07346655428409576, "fcm_dpo/margin": 45.07012176513672, "fcm_dpo/q_t": 0.42542755603790283, "grad_norm": 13.203572273254395, "learning_rate": 2.3282469092517977e-07, "logits/chosen": 0.7316404581069946, "logits/rejected": 0.6816238164901733, "logps/chosen": -184.6094512939453, "logps/ref_chosen": -59.21887969970703, "logps/ref_rejected": -71.24818420410156, "logps/rejected": -241.70889282226562, "loss": 1.1769, "margin_dpo/margin_mean": 45.07012176513672, "margin_dpo/margin_std": 84.25788116455078, "step": 378 }, { "KL/chosen_KL_mean": -121.52671813964844, "KL/mean": -154.25457763671875, "KL/rejected_KL_mean": -186.98245239257812, "KL/std": 85.5152587890625, "epoch": 0.5729402872260015, "fcm_dpo/beta": 0.007202798500657082, "fcm_dpo/delta": -0.07496453821659088, "fcm_dpo/margin": 65.45574951171875, "fcm_dpo/q_t": 0.3955724537372589, "grad_norm": 14.618745803833008, "learning_rate": 2.3150584219481643e-07, "logits/chosen": 0.671619176864624, "logits/rejected": 0.5937076807022095, "logps/chosen": -197.84329223632812, "logps/ref_chosen": -76.31658935546875, "logps/ref_rejected": -104.26200103759766, "logps/rejected": -291.24444580078125, "loss": 1.074, "margin_dpo/margin_mean": 65.45574951171875, "margin_dpo/margin_std": 94.66348266601562, "step": 379 }, { "KL/chosen_KL_mean": -105.07964324951172, "KL/mean": -143.19851684570312, "KL/rejected_KL_mean": -181.31741333007812, "KL/std": 78.93547058105469, "epoch": 0.5744520030234316, "fcm_dpo/beta": 0.007026575040072203, "fcm_dpo/delta": -0.14325784146785736, "fcm_dpo/margin": 76.23777770996094, "fcm_dpo/q_t": 0.37770792841911316, "grad_norm": 11.684196472167969, "learning_rate": 2.3018751078580283e-07, "logits/chosen": 0.7029041051864624, "logits/rejected": 0.6619795560836792, "logps/chosen": -166.36280822753906, "logps/ref_chosen": -61.283164978027344, "logps/ref_rejected": -72.38892364501953, "logps/rejected": -253.7063446044922, "loss": 1.023, "margin_dpo/margin_mean": 76.23777770996094, "margin_dpo/margin_std": 94.23179626464844, "step": 380 }, { "KL/chosen_KL_mean": -129.57518005371094, "KL/mean": -144.18365478515625, "KL/rejected_KL_mean": -158.7921142578125, "KL/std": 81.25999450683594, "epoch": 0.5759637188208617, "fcm_dpo/beta": 0.007202141918241978, "fcm_dpo/delta": 0.1924881935119629, "fcm_dpo/margin": 29.21695327758789, "fcm_dpo/q_t": 0.45364683866500854, "grad_norm": 12.961562156677246, "learning_rate": 2.288697335747027e-07, "logits/chosen": 0.6437864303588867, "logits/rejected": 0.6247032880783081, "logps/chosen": -187.78916931152344, "logps/ref_chosen": -58.2139892578125, "logps/ref_rejected": -60.78669357299805, "logps/rejected": -219.5788116455078, "loss": 1.3017, "margin_dpo/margin_mean": 29.21695327758789, "margin_dpo/margin_std": 97.07288360595703, "step": 381 }, { "KL/chosen_KL_mean": -129.46324157714844, "KL/mean": -156.5772705078125, "KL/rejected_KL_mean": -183.69131469726562, "KL/std": 81.28971862792969, "epoch": 0.5774754346182918, "fcm_dpo/beta": 0.007276223041117191, "fcm_dpo/delta": 0.005313534289598465, "fcm_dpo/margin": 54.228084564208984, "fcm_dpo/q_t": 0.41020467877388, "grad_norm": 13.063127517700195, "learning_rate": 2.2755254742257706e-07, "logits/chosen": 0.6734954118728638, "logits/rejected": 0.6170482635498047, "logps/chosen": -191.28855895996094, "logps/ref_chosen": -61.82532501220703, "logps/ref_rejected": -83.0452880859375, "logps/rejected": -266.7366027832031, "loss": 1.1109, "margin_dpo/margin_mean": 54.228084564208984, "margin_dpo/margin_std": 79.92002868652344, "step": 382 }, { "KL/chosen_KL_mean": -130.00021362304688, "KL/mean": -154.45675659179688, "KL/rejected_KL_mean": -178.91326904296875, "KL/std": 81.607421875, "epoch": 0.5789871504157218, "fcm_dpo/beta": 0.0072705792263150215, "fcm_dpo/delta": 0.0458693653345108, "fcm_dpo/margin": 48.91305923461914, "fcm_dpo/q_t": 0.42183059453964233, "grad_norm": 15.296006202697754, "learning_rate": 2.2623598917395436e-07, "logits/chosen": 0.5694981813430786, "logits/rejected": 0.6031885147094727, "logps/chosen": -210.5634765625, "logps/ref_chosen": -80.56326293945312, "logps/ref_rejected": -74.62922668457031, "logps/rejected": -253.54249572753906, "loss": 1.1927, "margin_dpo/margin_mean": 48.91305923461914, "margin_dpo/margin_std": 102.63812255859375, "step": 383 }, { "KL/chosen_KL_mean": -124.91947937011719, "KL/mean": -152.39324951171875, "KL/rejected_KL_mean": -179.86700439453125, "KL/std": 80.42024993896484, "epoch": 0.5804988662131519, "fcm_dpo/beta": 0.0073145655915141106, "fcm_dpo/delta": -0.00199760589748621, "fcm_dpo/margin": 54.94753646850586, "fcm_dpo/q_t": 0.40669381618499756, "grad_norm": 15.579803466796875, "learning_rate": 2.2492009565579875e-07, "logits/chosen": 0.7384841442108154, "logits/rejected": 0.6944015622138977, "logps/chosen": -190.39462280273438, "logps/ref_chosen": -65.47514343261719, "logps/ref_rejected": -79.67378234863281, "logps/rejected": -259.5408020019531, "loss": 1.1154, "margin_dpo/margin_mean": 54.94753646850586, "margin_dpo/margin_std": 85.66532135009766, "step": 384 }, { "KL/chosen_KL_mean": -123.13201904296875, "KL/mean": -159.01617431640625, "KL/rejected_KL_mean": -194.9003143310547, "KL/std": 82.08739471435547, "epoch": 0.582010582010582, "fcm_dpo/beta": 0.0072093624621629715, "fcm_dpo/delta": -0.12364254891872406, "fcm_dpo/margin": 71.76829528808594, "fcm_dpo/q_t": 0.383453369140625, "grad_norm": 12.84723949432373, "learning_rate": 2.2360490367648084e-07, "logits/chosen": 0.6259031295776367, "logits/rejected": 0.5812070965766907, "logps/chosen": -189.18853759765625, "logps/ref_chosen": -66.0565185546875, "logps/ref_rejected": -86.68023681640625, "logps/rejected": -281.58056640625, "loss": 1.0275, "margin_dpo/margin_mean": 71.76829528808594, "margin_dpo/margin_std": 87.22917175292969, "step": 385 }, { "KL/chosen_KL_mean": -142.86434936523438, "KL/mean": -166.49163818359375, "KL/rejected_KL_mean": -190.11892700195312, "KL/std": 85.83015441894531, "epoch": 0.5835222978080121, "fcm_dpo/beta": 0.007186328526586294, "fcm_dpo/delta": 0.06257650256156921, "fcm_dpo/margin": 47.25459289550781, "fcm_dpo/q_t": 0.42166996002197266, "grad_norm": 13.215150833129883, "learning_rate": 2.2229045002474724e-07, "logits/chosen": 0.6145610809326172, "logits/rejected": 0.5554145574569702, "logps/chosen": -218.48800659179688, "logps/ref_chosen": -75.6236572265625, "logps/ref_rejected": -92.62330627441406, "logps/rejected": -282.74224853515625, "loss": 1.1694, "margin_dpo/margin_mean": 47.25458908081055, "margin_dpo/margin_std": 87.60594940185547, "step": 386 }, { "KL/chosen_KL_mean": -125.47482299804688, "KL/mean": -159.42498779296875, "KL/rejected_KL_mean": -193.3751678466797, "KL/std": 79.78838348388672, "epoch": 0.5850340136054422, "fcm_dpo/beta": 0.007123415358364582, "fcm_dpo/delta": -0.08804966509342194, "fcm_dpo/margin": 67.90034484863281, "fcm_dpo/q_t": 0.39012467861175537, "grad_norm": 13.807373046875, "learning_rate": 2.209767714686924e-07, "logits/chosen": 0.7212274670600891, "logits/rejected": 0.6039391756057739, "logps/chosen": -172.696533203125, "logps/ref_chosen": -47.22170639038086, "logps/ref_rejected": -87.338134765625, "logps/rejected": -280.71331787109375, "loss": 1.0358, "margin_dpo/margin_mean": 67.90033721923828, "margin_dpo/margin_std": 80.91610717773438, "step": 387 }, { "KL/chosen_KL_mean": -128.02667236328125, "KL/mean": -149.42562866210938, "KL/rejected_KL_mean": -170.8245849609375, "KL/std": 84.12289428710938, "epoch": 0.5865457294028723, "fcm_dpo/beta": 0.007163255475461483, "fcm_dpo/delta": 0.09631352126598358, "fcm_dpo/margin": 42.79789352416992, "fcm_dpo/q_t": 0.4336731433868408, "grad_norm": 13.67048168182373, "learning_rate": 2.1966390475472954e-07, "logits/chosen": 0.6444123983383179, "logits/rejected": 0.639690101146698, "logps/chosen": -202.60614013671875, "logps/ref_chosen": -74.5794677734375, "logps/ref_rejected": -79.92558288574219, "logps/rejected": -250.7501678466797, "loss": 1.2238, "margin_dpo/margin_mean": 42.797889709472656, "margin_dpo/margin_std": 99.72714233398438, "step": 388 }, { "KL/chosen_KL_mean": -121.66131591796875, "KL/mean": -155.45494079589844, "KL/rejected_KL_mean": -189.24856567382812, "KL/std": 81.51905059814453, "epoch": 0.5880574452003023, "fcm_dpo/beta": 0.007154286373406649, "fcm_dpo/delta": -0.0876961201429367, "fcm_dpo/margin": 67.58724212646484, "fcm_dpo/q_t": 0.3913712501525879, "grad_norm": 24.15449333190918, "learning_rate": 2.1835188660656265e-07, "logits/chosen": 0.7004282474517822, "logits/rejected": 0.6616201400756836, "logps/chosen": -183.28567504882812, "logps/ref_chosen": -61.624366760253906, "logps/ref_rejected": -76.50978088378906, "logps/rejected": -265.75836181640625, "loss": 1.0559, "margin_dpo/margin_mean": 67.58724212646484, "margin_dpo/margin_std": 90.30506896972656, "step": 389 }, { "KL/chosen_KL_mean": -109.01703643798828, "KL/mean": -135.18896484375, "KL/rejected_KL_mean": -161.36090087890625, "KL/std": 78.40666198730469, "epoch": 0.5895691609977324, "fcm_dpo/beta": 0.0071219149976968765, "fcm_dpo/delta": 0.028195127844810486, "fcm_dpo/margin": 52.34385681152344, "fcm_dpo/q_t": 0.41536635160446167, "grad_norm": 11.698328971862793, "learning_rate": 2.170407537241599e-07, "logits/chosen": 0.7919474840164185, "logits/rejected": 0.7127720713615417, "logps/chosen": -154.88890075683594, "logps/ref_chosen": -45.871864318847656, "logps/ref_rejected": -61.305999755859375, "logps/rejected": -222.66690063476562, "loss": 1.1296, "margin_dpo/margin_mean": 52.34385681152344, "margin_dpo/margin_std": 82.52626037597656, "step": 390 }, { "KL/chosen_KL_mean": -122.2738037109375, "KL/mean": -153.88844299316406, "KL/rejected_KL_mean": -185.50308227539062, "KL/std": 80.52210235595703, "epoch": 0.5910808767951625, "fcm_dpo/beta": 0.007075528148561716, "fcm_dpo/delta": -0.049968231469392776, "fcm_dpo/margin": 63.22930145263672, "fcm_dpo/q_t": 0.39903032779693604, "grad_norm": 12.636250495910645, "learning_rate": 2.1573054278272636e-07, "logits/chosen": 0.6821379065513611, "logits/rejected": 0.6111657619476318, "logps/chosen": -180.4608154296875, "logps/ref_chosen": -58.18701171875, "logps/ref_rejected": -83.63442993164062, "logps/rejected": -269.13751220703125, "loss": 1.1019, "margin_dpo/margin_mean": 63.22930145263672, "margin_dpo/margin_std": 98.30987548828125, "step": 391 }, { "KL/chosen_KL_mean": -109.65264892578125, "KL/mean": -144.0840606689453, "KL/rejected_KL_mean": -178.51547241210938, "KL/std": 86.02153015136719, "epoch": 0.5925925925925926, "fcm_dpo/beta": 0.007030046544969082, "fcm_dpo/delta": -0.0886000394821167, "fcm_dpo/margin": 68.86282348632812, "fcm_dpo/q_t": 0.3925461173057556, "grad_norm": 11.746094703674316, "learning_rate": 2.1442129043167873e-07, "logits/chosen": 0.6962657570838928, "logits/rejected": 0.6370902061462402, "logps/chosen": -179.39718627929688, "logps/ref_chosen": -69.7445297241211, "logps/ref_rejected": -94.05877685546875, "logps/rejected": -272.5742492675781, "loss": 1.0767, "margin_dpo/margin_mean": 68.86282348632812, "margin_dpo/margin_std": 101.075439453125, "step": 392 }, { "KL/chosen_KL_mean": -123.72262573242188, "KL/mean": -159.5321044921875, "KL/rejected_KL_mean": -195.34158325195312, "KL/std": 83.92768859863281, "epoch": 0.5941043083900227, "fcm_dpo/beta": 0.006823638454079628, "fcm_dpo/delta": -0.09428704530000687, "fcm_dpo/margin": 71.61897277832031, "fcm_dpo/q_t": 0.3880813419818878, "grad_norm": 11.2382230758667, "learning_rate": 2.131130332936195e-07, "logits/chosen": 0.6831046342849731, "logits/rejected": 0.6424210071563721, "logps/chosen": -176.05752563476562, "logps/ref_chosen": -52.33489990234375, "logps/ref_rejected": -74.33809661865234, "logps/rejected": -269.6796875, "loss": 1.0332, "margin_dpo/margin_mean": 71.61896514892578, "margin_dpo/margin_std": 83.35065460205078, "step": 393 }, { "KL/chosen_KL_mean": -119.21342468261719, "KL/mean": -149.92340087890625, "KL/rejected_KL_mean": -180.6333770751953, "KL/std": 77.53596496582031, "epoch": 0.5956160241874527, "fcm_dpo/beta": 0.006829938385635614, "fcm_dpo/delta": -0.020546775311231613, "fcm_dpo/margin": 61.419960021972656, "fcm_dpo/q_t": 0.40248793363571167, "grad_norm": 11.583659172058105, "learning_rate": 2.1180580796331323e-07, "logits/chosen": 0.7224094867706299, "logits/rejected": 0.6902562379837036, "logps/chosen": -179.88955688476562, "logps/ref_chosen": -60.6761360168457, "logps/ref_rejected": -71.36074829101562, "logps/rejected": -251.994140625, "loss": 1.0789, "margin_dpo/margin_mean": 61.419960021972656, "margin_dpo/margin_std": 77.19164276123047, "step": 394 }, { "KL/chosen_KL_mean": -122.50703430175781, "KL/mean": -150.58975219726562, "KL/rejected_KL_mean": -178.67247009277344, "KL/std": 80.42475891113281, "epoch": 0.5971277399848829, "fcm_dpo/beta": 0.0068355584517121315, "fcm_dpo/delta": 0.016474729403853416, "fcm_dpo/margin": 56.165428161621094, "fcm_dpo/q_t": 0.4139012396335602, "grad_norm": 13.971101760864258, "learning_rate": 2.104996510066625e-07, "logits/chosen": 0.7060272693634033, "logits/rejected": 0.6042532920837402, "logps/chosen": -173.11135864257812, "logps/ref_chosen": -50.60432434082031, "logps/ref_rejected": -77.08731079101562, "logps/rejected": -255.75978088378906, "loss": 1.1185, "margin_dpo/margin_mean": 56.165428161621094, "margin_dpo/margin_std": 85.66535949707031, "step": 395 }, { "KL/chosen_KL_mean": -116.39704132080078, "KL/mean": -146.19480895996094, "KL/rejected_KL_mean": -175.99258422851562, "KL/std": 89.64775085449219, "epoch": 0.5986394557823129, "fcm_dpo/beta": 0.0067558130249381065, "fcm_dpo/delta": -0.0043720826506614685, "fcm_dpo/margin": 59.59553146362305, "fcm_dpo/q_t": 0.4080699682235718, "grad_norm": 11.438410758972168, "learning_rate": 2.0919459895968517e-07, "logits/chosen": 0.7179821729660034, "logits/rejected": 0.6186529994010925, "logps/chosen": -167.75665283203125, "logps/ref_chosen": -51.35961151123047, "logps/ref_rejected": -79.89360046386719, "logps/rejected": -255.88616943359375, "loss": 1.0956, "margin_dpo/margin_mean": 59.59553146362305, "margin_dpo/margin_std": 77.72926330566406, "step": 396 }, { "KL/chosen_KL_mean": -133.34695434570312, "KL/mean": -150.32850646972656, "KL/rejected_KL_mean": -167.31004333496094, "KL/std": 81.386474609375, "epoch": 0.600151171579743, "fcm_dpo/beta": 0.006991279777139425, "fcm_dpo/delta": 0.1663690060377121, "fcm_dpo/margin": 33.96311950683594, "fcm_dpo/q_t": 0.4465146064758301, "grad_norm": 12.702045440673828, "learning_rate": 2.078906883274924e-07, "logits/chosen": 0.6114163994789124, "logits/rejected": 0.5590552687644958, "logps/chosen": -199.8031768798828, "logps/ref_chosen": -66.45622253417969, "logps/ref_rejected": -85.74736785888672, "logps/rejected": -253.0574188232422, "loss": 1.2715, "margin_dpo/margin_mean": 33.96311950683594, "margin_dpo/margin_std": 96.10679626464844, "step": 397 }, { "KL/chosen_KL_mean": -113.849853515625, "KL/mean": -151.95046997070312, "KL/rejected_KL_mean": -190.05108642578125, "KL/std": 85.21319580078125, "epoch": 0.6016628873771731, "fcm_dpo/beta": 0.0068751135841012, "fcm_dpo/delta": -0.13184233009815216, "fcm_dpo/margin": 76.20124053955078, "fcm_dpo/q_t": 0.38280850648880005, "grad_norm": 10.512614250183105, "learning_rate": 2.065879555832674e-07, "logits/chosen": 0.6477320194244385, "logits/rejected": 0.5816516876220703, "logps/chosen": -163.09410095214844, "logps/ref_chosen": -49.244239807128906, "logps/ref_rejected": -75.18949127197266, "logps/rejected": -265.2405700683594, "loss": 1.0135, "margin_dpo/margin_mean": 76.20124053955078, "margin_dpo/margin_std": 89.14450073242188, "step": 398 }, { "KL/chosen_KL_mean": -132.7601776123047, "KL/mean": -176.23660278320312, "KL/rejected_KL_mean": -219.71304321289062, "KL/std": 94.06497192382812, "epoch": 0.6031746031746031, "fcm_dpo/beta": 0.006639732047915459, "fcm_dpo/delta": -0.1895592212677002, "fcm_dpo/margin": 86.952880859375, "fcm_dpo/q_t": 0.3730096220970154, "grad_norm": 12.22734260559082, "learning_rate": 2.052864371672457e-07, "logits/chosen": 0.6234591007232666, "logits/rejected": 0.4737226963043213, "logps/chosen": -201.06695556640625, "logps/ref_chosen": -68.30679321289062, "logps/ref_rejected": -113.2708511352539, "logps/rejected": -332.98388671875, "loss": 0.9958, "margin_dpo/margin_mean": 86.95288848876953, "margin_dpo/margin_std": 105.41049194335938, "step": 399 }, { "KL/chosen_KL_mean": -147.748779296875, "KL/mean": -172.35479736328125, "KL/rejected_KL_mean": -196.96080017089844, "KL/std": 86.04580688476562, "epoch": 0.6046863189720333, "fcm_dpo/beta": 0.006632579956203699, "fcm_dpo/delta": 0.07553811371326447, "fcm_dpo/margin": 49.212013244628906, "fcm_dpo/q_t": 0.42484426498413086, "grad_norm": 17.143085479736328, "learning_rate": 2.0398616948569493e-07, "logits/chosen": 0.6802634000778198, "logits/rejected": 0.6172671318054199, "logps/chosen": -219.37527465820312, "logps/ref_chosen": -71.62649536132812, "logps/ref_rejected": -90.98765563964844, "logps/rejected": -287.9484558105469, "loss": 1.167, "margin_dpo/margin_mean": 49.212013244628906, "margin_dpo/margin_std": 84.21356201171875, "step": 400 }, { "KL/chosen_KL_mean": -112.90580749511719, "KL/mean": -147.96231079101562, "KL/rejected_KL_mean": -183.01881408691406, "KL/std": 92.42097473144531, "epoch": 0.6061980347694633, "fcm_dpo/beta": 0.006614885292947292, "fcm_dpo/delta": -0.0674346387386322, "fcm_dpo/margin": 70.11300659179688, "fcm_dpo/q_t": 0.3951302170753479, "grad_norm": 9.65842342376709, "learning_rate": 2.0268718890989752e-07, "logits/chosen": 0.7155077457427979, "logits/rejected": 0.6205871105194092, "logps/chosen": -166.63076782226562, "logps/ref_chosen": -53.72495651245117, "logps/ref_rejected": -75.06304931640625, "logps/rejected": -258.08184814453125, "loss": 1.0542, "margin_dpo/margin_mean": 70.11299896240234, "margin_dpo/margin_std": 88.1304702758789, "step": 401 }, { "KL/chosen_KL_mean": -123.09654235839844, "KL/mean": -150.04502868652344, "KL/rejected_KL_mean": -176.9935302734375, "KL/std": 79.02203369140625, "epoch": 0.6077097505668935, "fcm_dpo/beta": 0.006622655317187309, "fcm_dpo/delta": 0.04423164948821068, "fcm_dpo/margin": 53.89698791503906, "fcm_dpo/q_t": 0.41926831007003784, "grad_norm": 13.637972831726074, "learning_rate": 2.013895317751323e-07, "logits/chosen": 0.6671075820922852, "logits/rejected": 0.646820068359375, "logps/chosen": -184.970458984375, "logps/ref_chosen": -61.873931884765625, "logps/ref_rejected": -66.15198516845703, "logps/rejected": -243.1455078125, "loss": 1.1591, "margin_dpo/margin_mean": 53.89698028564453, "margin_dpo/margin_std": 95.33618927001953, "step": 402 }, { "KL/chosen_KL_mean": -129.99464416503906, "KL/mean": -165.91558837890625, "KL/rejected_KL_mean": -201.83653259277344, "KL/std": 91.60020446777344, "epoch": 0.6092214663643235, "fcm_dpo/beta": 0.006587062031030655, "fcm_dpo/delta": -0.07708509266376495, "fcm_dpo/margin": 71.84190368652344, "fcm_dpo/q_t": 0.39418938755989075, "grad_norm": 10.517227172851562, "learning_rate": 2.0009323437965898e-07, "logits/chosen": 0.8143857717514038, "logits/rejected": 0.722244143486023, "logps/chosen": -181.31614685058594, "logps/ref_chosen": -51.321502685546875, "logps/ref_rejected": -86.54010772705078, "logps/rejected": -288.37664794921875, "loss": 1.0617, "margin_dpo/margin_mean": 71.84190368652344, "margin_dpo/margin_std": 96.05760192871094, "step": 403 }, { "KL/chosen_KL_mean": -122.32210540771484, "KL/mean": -158.37295532226562, "KL/rejected_KL_mean": -194.42381286621094, "KL/std": 92.72088623046875, "epoch": 0.6107331821617535, "fcm_dpo/beta": 0.006459000986069441, "fcm_dpo/delta": -0.07044952362775803, "fcm_dpo/margin": 72.10169982910156, "fcm_dpo/q_t": 0.3962360620498657, "grad_norm": 14.719085693359375, "learning_rate": 1.9879833298370237e-07, "logits/chosen": 0.6476539969444275, "logits/rejected": 0.5492456555366516, "logps/chosen": -184.58499145507812, "logps/ref_chosen": -62.26288604736328, "logps/ref_rejected": -95.19029998779297, "logps/rejected": -289.6141052246094, "loss": 1.0753, "margin_dpo/margin_mean": 72.10169982910156, "margin_dpo/margin_std": 100.69046020507812, "step": 404 }, { "KL/chosen_KL_mean": -125.48126220703125, "KL/mean": -154.83218383789062, "KL/rejected_KL_mean": -184.18310546875, "KL/std": 83.02670288085938, "epoch": 0.6122448979591837, "fcm_dpo/beta": 0.006452477071434259, "fcm_dpo/delta": 0.02102312445640564, "fcm_dpo/margin": 58.70184326171875, "fcm_dpo/q_t": 0.41416507959365845, "grad_norm": 11.768880844116211, "learning_rate": 1.975048638084379e-07, "logits/chosen": 0.7292817831039429, "logits/rejected": 0.6814388036727905, "logps/chosen": -176.06561279296875, "logps/ref_chosen": -50.5843391418457, "logps/ref_rejected": -65.43156433105469, "logps/rejected": -249.6146697998047, "loss": 1.1197, "margin_dpo/margin_mean": 58.70184326171875, "margin_dpo/margin_std": 84.7135009765625, "step": 405 }, { "KL/chosen_KL_mean": -120.27396392822266, "KL/mean": -158.66598510742188, "KL/rejected_KL_mean": -197.05801391601562, "KL/std": 89.59160614013672, "epoch": 0.6137566137566137, "fcm_dpo/beta": 0.006429762579500675, "fcm_dpo/delta": -0.09845130890607834, "fcm_dpo/margin": 76.78402709960938, "fcm_dpo/q_t": 0.3877302408218384, "grad_norm": 12.265458106994629, "learning_rate": 1.9621286303497914e-07, "logits/chosen": 0.7168718576431274, "logits/rejected": 0.5531203746795654, "logps/chosen": -169.26956176757812, "logps/ref_chosen": -48.99560546875, "logps/ref_rejected": -92.47774505615234, "logps/rejected": -289.5357666015625, "loss": 1.0502, "margin_dpo/margin_mean": 76.78402709960938, "margin_dpo/margin_std": 100.21034240722656, "step": 406 }, { "KL/chosen_KL_mean": -151.61819458007812, "KL/mean": -181.9169921875, "KL/rejected_KL_mean": -212.21580505371094, "KL/std": 101.87554168701172, "epoch": 0.6152683295540439, "fcm_dpo/beta": 0.006411860231310129, "fcm_dpo/delta": 0.011711766943335533, "fcm_dpo/margin": 60.59760665893555, "fcm_dpo/q_t": 0.41196078062057495, "grad_norm": 13.501534461975098, "learning_rate": 1.9492236680336483e-07, "logits/chosen": 0.558795690536499, "logits/rejected": 0.49235716462135315, "logps/chosen": -241.0187530517578, "logps/ref_chosen": -89.40056610107422, "logps/ref_rejected": -99.28775024414062, "logps/rejected": -311.5035400390625, "loss": 1.1392, "margin_dpo/margin_mean": 60.59761047363281, "margin_dpo/margin_std": 103.35305786132812, "step": 407 }, { "KL/chosen_KL_mean": -113.6641845703125, "KL/mean": -156.4038848876953, "KL/rejected_KL_mean": -199.1436004638672, "KL/std": 84.85147094726562, "epoch": 0.6167800453514739, "fcm_dpo/beta": 0.006298558786511421, "fcm_dpo/delta": -0.1461481899023056, "fcm_dpo/margin": 85.47942352294922, "fcm_dpo/q_t": 0.37603700160980225, "grad_norm": 10.623373985290527, "learning_rate": 1.9363341121154895e-07, "logits/chosen": 0.6735584735870361, "logits/rejected": 0.5957342386245728, "logps/chosen": -168.36810302734375, "logps/ref_chosen": -54.70391845703125, "logps/ref_rejected": -73.98648834228516, "logps/rejected": -273.1300964355469, "loss": 0.9966, "margin_dpo/margin_mean": 85.47941589355469, "margin_dpo/margin_std": 89.57831573486328, "step": 408 }, { "KL/chosen_KL_mean": -140.25930786132812, "KL/mean": -159.85809326171875, "KL/rejected_KL_mean": -179.4569091796875, "KL/std": 73.4033203125, "epoch": 0.618291761148904, "fcm_dpo/beta": 0.006353562697768211, "fcm_dpo/delta": 0.15491655468940735, "fcm_dpo/margin": 39.19761657714844, "fcm_dpo/q_t": 0.44314515590667725, "grad_norm": 12.572002410888672, "learning_rate": 1.9234603231438994e-07, "logits/chosen": 0.6764421463012695, "logits/rejected": 0.6854976415634155, "logps/chosen": -202.37753295898438, "logps/ref_chosen": -62.11822509765625, "logps/ref_rejected": -61.933509826660156, "logps/rejected": -241.39041137695312, "loss": 1.2212, "margin_dpo/margin_mean": 39.19761657714844, "margin_dpo/margin_std": 83.82356262207031, "step": 409 }, { "KL/chosen_KL_mean": -131.65499877929688, "KL/mean": -166.28590393066406, "KL/rejected_KL_mean": -200.9167938232422, "KL/std": 82.87063598632812, "epoch": 0.6198034769463341, "fcm_dpo/beta": 0.006320708431303501, "fcm_dpo/delta": -0.041026365011930466, "fcm_dpo/margin": 69.26181030273438, "fcm_dpo/q_t": 0.3986797332763672, "grad_norm": 11.368917465209961, "learning_rate": 1.9106026612264315e-07, "logits/chosen": 0.6804258823394775, "logits/rejected": 0.6558456420898438, "logps/chosen": -193.45767211914062, "logps/ref_chosen": -61.80266189575195, "logps/ref_rejected": -76.60002136230469, "logps/rejected": -277.5168151855469, "loss": 1.0623, "margin_dpo/margin_mean": 69.2618179321289, "margin_dpo/margin_std": 81.38176727294922, "step": 410 }, { "KL/chosen_KL_mean": -138.29071044921875, "KL/mean": -171.49337768554688, "KL/rejected_KL_mean": -204.696044921875, "KL/std": 89.67030334472656, "epoch": 0.6213151927437641, "fcm_dpo/beta": 0.0063368408009409904, "fcm_dpo/delta": -0.021709755063056946, "fcm_dpo/margin": 66.40533447265625, "fcm_dpo/q_t": 0.4044159948825836, "grad_norm": 11.313436508178711, "learning_rate": 1.8977614860195296e-07, "logits/chosen": 0.7408417463302612, "logits/rejected": 0.6807792782783508, "logps/chosen": -192.736083984375, "logps/ref_chosen": -54.44539260864258, "logps/ref_rejected": -74.5650863647461, "logps/rejected": -279.2611083984375, "loss": 1.0935, "margin_dpo/margin_mean": 66.40534210205078, "margin_dpo/margin_std": 95.49729919433594, "step": 411 }, { "KL/chosen_KL_mean": -141.05941772460938, "KL/mean": -171.43373107910156, "KL/rejected_KL_mean": -201.80804443359375, "KL/std": 77.73248291015625, "epoch": 0.6228269085411943, "fcm_dpo/beta": 0.006320852320641279, "fcm_dpo/delta": 0.016446245834231377, "fcm_dpo/margin": 60.74862289428711, "fcm_dpo/q_t": 0.41138458251953125, "grad_norm": 11.807281494140625, "learning_rate": 1.8849371567184662e-07, "logits/chosen": 0.690960168838501, "logits/rejected": 0.6279210448265076, "logps/chosen": -196.3074951171875, "logps/ref_chosen": -55.248085021972656, "logps/ref_rejected": -68.96623229980469, "logps/rejected": -270.7742614746094, "loss": 1.1047, "margin_dpo/margin_mean": 60.748619079589844, "margin_dpo/margin_std": 82.77706909179688, "step": 412 }, { "KL/chosen_KL_mean": -156.57864379882812, "KL/mean": -185.267822265625, "KL/rejected_KL_mean": -213.95700073242188, "KL/std": 90.07394409179688, "epoch": 0.6243386243386243, "fcm_dpo/beta": 0.006390274967998266, "fcm_dpo/delta": 0.03452521562576294, "fcm_dpo/margin": 57.37835693359375, "fcm_dpo/q_t": 0.41872984170913696, "grad_norm": 14.247014045715332, "learning_rate": 1.872130032047302e-07, "logits/chosen": 0.531366229057312, "logits/rejected": 0.49875760078430176, "logps/chosen": -225.29937744140625, "logps/ref_chosen": -68.72074890136719, "logps/ref_rejected": -78.76539611816406, "logps/rejected": -292.722412109375, "loss": 1.1705, "margin_dpo/margin_mean": 57.37835693359375, "margin_dpo/margin_std": 110.43424224853516, "step": 413 }, { "KL/chosen_KL_mean": -138.19000244140625, "KL/mean": -172.3620147705078, "KL/rejected_KL_mean": -206.5340576171875, "KL/std": 96.0737075805664, "epoch": 0.6258503401360545, "fcm_dpo/beta": 0.006364773027598858, "fcm_dpo/delta": -0.03656965494155884, "fcm_dpo/margin": 68.34403991699219, "fcm_dpo/q_t": 0.3993681073188782, "grad_norm": 10.928391456604004, "learning_rate": 1.8593404702488436e-07, "logits/chosen": 0.7208206653594971, "logits/rejected": 0.6589357852935791, "logps/chosen": -192.32821655273438, "logps/ref_chosen": -54.138214111328125, "logps/ref_rejected": -74.65741729736328, "logps/rejected": -281.19146728515625, "loss": 1.0749, "margin_dpo/margin_mean": 68.34403991699219, "margin_dpo/margin_std": 88.88926696777344, "step": 414 }, { "KL/chosen_KL_mean": -136.01686096191406, "KL/mean": -164.8328857421875, "KL/rejected_KL_mean": -193.64889526367188, "KL/std": 89.09803771972656, "epoch": 0.6273620559334845, "fcm_dpo/beta": 0.006370305083692074, "fcm_dpo/delta": 0.03412775322794914, "fcm_dpo/margin": 57.63203811645508, "fcm_dpo/q_t": 0.4166450798511505, "grad_norm": 11.460617065429688, "learning_rate": 1.846568829074628e-07, "logits/chosen": 0.7241143584251404, "logits/rejected": 0.7067319750785828, "logps/chosen": -191.9354248046875, "logps/ref_chosen": -55.91856002807617, "logps/ref_rejected": -61.747703552246094, "logps/rejected": -255.3966064453125, "loss": 1.1407, "margin_dpo/margin_mean": 57.632041931152344, "margin_dpo/margin_std": 96.32302856445312, "step": 415 }, { "KL/chosen_KL_mean": -146.26498413085938, "KL/mean": -170.3350830078125, "KL/rejected_KL_mean": -194.40518188476562, "KL/std": 93.53575134277344, "epoch": 0.6288737717309146, "fcm_dpo/beta": 0.006337391212582588, "fcm_dpo/delta": -0.08305186033248901, "fcm_dpo/margin": 48.14020919799805, "fcm_dpo/q_t": 0.43126845359802246, "grad_norm": 14.00360107421875, "learning_rate": 1.8338154657749128e-07, "logits/chosen": 0.6633551716804504, "logits/rejected": 0.6139250993728638, "logps/chosen": -200.98806762695312, "logps/ref_chosen": -54.72308349609375, "logps/ref_rejected": -69.17388916015625, "logps/rejected": -263.5790710449219, "loss": 1.1968, "margin_dpo/margin_mean": 48.14020919799805, "margin_dpo/margin_std": 90.29344177246094, "step": 416 }, { "KL/chosen_KL_mean": -151.7848358154297, "KL/mean": -183.6634521484375, "KL/rejected_KL_mean": -215.5420684814453, "KL/std": 87.87062072753906, "epoch": 0.6303854875283447, "fcm_dpo/beta": 0.006316634826362133, "fcm_dpo/delta": -0.003457318991422653, "fcm_dpo/margin": 63.75724411010742, "fcm_dpo/q_t": 0.4064716696739197, "grad_norm": 12.411998748779297, "learning_rate": 1.8210807370886849e-07, "logits/chosen": 0.8052752017974854, "logits/rejected": 0.7394411563873291, "logps/chosen": -208.5760955810547, "logps/ref_chosen": -56.791259765625, "logps/ref_rejected": -68.7791748046875, "logps/rejected": -284.32122802734375, "loss": 1.1314, "margin_dpo/margin_mean": 63.75724792480469, "margin_dpo/margin_std": 106.73062133789062, "step": 417 }, { "KL/chosen_KL_mean": -160.93739318847656, "KL/mean": -187.81124877929688, "KL/rejected_KL_mean": -214.68505859375, "KL/std": 98.37525939941406, "epoch": 0.6318972033257747, "fcm_dpo/beta": 0.006278286688029766, "fcm_dpo/delta": 0.06338182091712952, "fcm_dpo/margin": 53.74766540527344, "fcm_dpo/q_t": 0.42482703924179077, "grad_norm": 13.669818878173828, "learning_rate": 1.8083649992336825e-07, "logits/chosen": 0.7154676914215088, "logits/rejected": 0.7190610766410828, "logps/chosen": -230.04537963867188, "logps/ref_chosen": -69.10798645019531, "logps/ref_rejected": -75.09132385253906, "logps/rejected": -289.7763977050781, "loss": 1.1789, "margin_dpo/margin_mean": 53.74766540527344, "margin_dpo/margin_std": 99.94847106933594, "step": 418 }, { "KL/chosen_KL_mean": -130.17111206054688, "KL/mean": -168.118408203125, "KL/rejected_KL_mean": -206.0657196044922, "KL/std": 90.11360168457031, "epoch": 0.6334089191232048, "fcm_dpo/beta": 0.006277072709053755, "fcm_dpo/delta": -0.08035100251436234, "fcm_dpo/margin": 75.89460754394531, "fcm_dpo/q_t": 0.39397668838500977, "grad_norm": 11.692562103271484, "learning_rate": 1.7956686078964255e-07, "logits/chosen": 0.6163479089736938, "logits/rejected": 0.5674476027488708, "logps/chosen": -188.34288024902344, "logps/ref_chosen": -58.1717643737793, "logps/ref_rejected": -71.67066955566406, "logps/rejected": -277.73638916015625, "loss": 1.0612, "margin_dpo/margin_mean": 75.89460754394531, "margin_dpo/margin_std": 103.62324523925781, "step": 419 }, { "KL/chosen_KL_mean": -160.61224365234375, "KL/mean": -180.34158325195312, "KL/rejected_KL_mean": -200.0709228515625, "KL/std": 93.16465759277344, "epoch": 0.6349206349206349, "fcm_dpo/beta": 0.00638390239328146, "fcm_dpo/delta": 0.15220247209072113, "fcm_dpo/margin": 39.45867156982422, "fcm_dpo/q_t": 0.4451296329498291, "grad_norm": 12.861660957336426, "learning_rate": 1.782991918222275e-07, "logits/chosen": 0.7003469467163086, "logits/rejected": 0.655767560005188, "logps/chosen": -217.66575622558594, "logps/ref_chosen": -57.05351257324219, "logps/ref_rejected": -62.670982360839844, "logps/rejected": -262.7419128417969, "loss": 1.2613, "margin_dpo/margin_mean": 39.45867156982422, "margin_dpo/margin_std": 106.66671752929688, "step": 420 }, { "KL/chosen_KL_mean": -147.0243377685547, "KL/mean": -176.02650451660156, "KL/rejected_KL_mean": -205.02865600585938, "KL/std": 92.48200988769531, "epoch": 0.636432350718065, "fcm_dpo/beta": 0.006444813683629036, "fcm_dpo/delta": 0.026773784309625626, "fcm_dpo/margin": 58.004310607910156, "fcm_dpo/q_t": 0.41775447130203247, "grad_norm": 14.737412452697754, "learning_rate": 1.7703352848054887e-07, "logits/chosen": 0.6828116178512573, "logits/rejected": 0.6210600137710571, "logps/chosen": -204.34759521484375, "logps/ref_chosen": -57.32324981689453, "logps/ref_rejected": -75.33782958984375, "logps/rejected": -280.3664855957031, "loss": 1.1888, "margin_dpo/margin_mean": 58.004310607910156, "margin_dpo/margin_std": 120.49269104003906, "step": 421 }, { "KL/chosen_KL_mean": -130.69232177734375, "KL/mean": -168.12051391601562, "KL/rejected_KL_mean": -205.54869079589844, "KL/std": 91.55455780029297, "epoch": 0.6379440665154951, "fcm_dpo/beta": 0.0064270165748894215, "fcm_dpo/delta": -0.08514492213726044, "fcm_dpo/margin": 74.85635375976562, "fcm_dpo/q_t": 0.39063405990600586, "grad_norm": 13.418085098266602, "learning_rate": 1.7576990616793137e-07, "logits/chosen": 0.6663640737533569, "logits/rejected": 0.6602794528007507, "logps/chosen": -197.74990844726562, "logps/ref_chosen": -67.05757141113281, "logps/ref_rejected": -72.12803649902344, "logps/rejected": -277.6767272949219, "loss": 1.0445, "margin_dpo/margin_mean": 74.85636138916016, "margin_dpo/margin_std": 92.13258361816406, "step": 422 }, { "KL/chosen_KL_mean": -132.84735107421875, "KL/mean": -173.277099609375, "KL/rejected_KL_mean": -213.7068328857422, "KL/std": 93.94609832763672, "epoch": 0.6394557823129252, "fcm_dpo/beta": 0.006261817179620266, "fcm_dpo/delta": -0.1122402548789978, "fcm_dpo/margin": 80.85948944091797, "fcm_dpo/q_t": 0.3863791525363922, "grad_norm": 11.530719757080078, "learning_rate": 1.745083602306071e-07, "logits/chosen": 0.7256453633308411, "logits/rejected": 0.657960057258606, "logps/chosen": -186.90902709960938, "logps/ref_chosen": -54.06167221069336, "logps/ref_rejected": -76.64092254638672, "logps/rejected": -290.3477783203125, "loss": 1.0317, "margin_dpo/margin_mean": 80.85948181152344, "margin_dpo/margin_std": 100.18473815917969, "step": 423 }, { "KL/chosen_KL_mean": -141.92503356933594, "KL/mean": -180.5269012451172, "KL/rejected_KL_mean": -219.12875366210938, "KL/std": 88.34001159667969, "epoch": 0.6409674981103552, "fcm_dpo/beta": 0.00612823199480772, "fcm_dpo/delta": -0.07791505008935928, "fcm_dpo/margin": 77.20375061035156, "fcm_dpo/q_t": 0.3932613730430603, "grad_norm": 16.036226272583008, "learning_rate": 1.7324892595672804e-07, "logits/chosen": 0.6484436392784119, "logits/rejected": 0.6037266254425049, "logps/chosen": -195.53390502929688, "logps/ref_chosen": -53.60887145996094, "logps/ref_rejected": -79.2139892578125, "logps/rejected": -298.3427734375, "loss": 1.0579, "margin_dpo/margin_mean": 77.20375061035156, "margin_dpo/margin_std": 101.07964324951172, "step": 424 }, { "KL/chosen_KL_mean": -141.54779052734375, "KL/mean": -171.28265380859375, "KL/rejected_KL_mean": -201.0175323486328, "KL/std": 85.41214752197266, "epoch": 0.6424792139077853, "fcm_dpo/beta": 0.006170147098600864, "fcm_dpo/delta": 0.03432762995362282, "fcm_dpo/margin": 59.469749450683594, "fcm_dpo/q_t": 0.416858047246933, "grad_norm": 12.468502044677734, "learning_rate": 1.7199163857537824e-07, "logits/chosen": 0.7201306819915771, "logits/rejected": 0.6925822496414185, "logps/chosen": -199.96246337890625, "logps/ref_chosen": -58.41468048095703, "logps/ref_rejected": -66.59054565429688, "logps/rejected": -267.60809326171875, "loss": 1.1375, "margin_dpo/margin_mean": 59.469749450683594, "margin_dpo/margin_std": 96.95804595947266, "step": 425 }, { "KL/chosen_KL_mean": -170.64788818359375, "KL/mean": -188.68301391601562, "KL/rejected_KL_mean": -206.7181396484375, "KL/std": 91.15631866455078, "epoch": 0.6439909297052154, "fcm_dpo/beta": 0.006349620409309864, "fcm_dpo/delta": 0.17486482858657837, "fcm_dpo/margin": 36.070255279541016, "fcm_dpo/q_t": 0.4480932652950287, "grad_norm": 16.96051788330078, "learning_rate": 1.7073653325558828e-07, "logits/chosen": 0.6682687997817993, "logits/rejected": 0.6715260744094849, "logps/chosen": -242.35610961914062, "logps/ref_chosen": -71.70822143554688, "logps/ref_rejected": -73.57725524902344, "logps/rejected": -280.29541015625, "loss": 1.2922, "margin_dpo/margin_mean": 36.07025146484375, "margin_dpo/margin_std": 114.49520874023438, "step": 426 }, { "KL/chosen_KL_mean": -161.13392639160156, "KL/mean": -192.15673828125, "KL/rejected_KL_mean": -223.17955017089844, "KL/std": 99.90629577636719, "epoch": 0.6455026455026455, "fcm_dpo/beta": 0.0063988445326685905, "fcm_dpo/delta": 0.0030871136114001274, "fcm_dpo/margin": 62.045616149902344, "fcm_dpo/q_t": 0.4128350615501404, "grad_norm": 13.89861011505127, "learning_rate": 1.6948364510535218e-07, "logits/chosen": 0.7114277482032776, "logits/rejected": 0.6444870233535767, "logps/chosen": -219.77670288085938, "logps/ref_chosen": -58.64276885986328, "logps/ref_rejected": -86.25437927246094, "logps/rejected": -309.4339294433594, "loss": 1.1435, "margin_dpo/margin_mean": 62.045623779296875, "margin_dpo/margin_std": 111.41607666015625, "step": 427 }, { "KL/chosen_KL_mean": -155.09945678710938, "KL/mean": -188.89183044433594, "KL/rejected_KL_mean": -222.6842041015625, "KL/std": 100.60108947753906, "epoch": 0.6470143613000756, "fcm_dpo/beta": 0.006401218008249998, "fcm_dpo/delta": -0.03433738648891449, "fcm_dpo/margin": 67.5847396850586, "fcm_dpo/q_t": 0.4036714434623718, "grad_norm": 12.783305168151855, "learning_rate": 1.6823300917064458e-07, "logits/chosen": 0.6407492160797119, "logits/rejected": 0.5998574495315552, "logps/chosen": -221.6955108642578, "logps/ref_chosen": -66.5960464477539, "logps/ref_rejected": -82.3941650390625, "logps/rejected": -305.078369140625, "loss": 1.1121, "margin_dpo/margin_mean": 67.5847396850586, "margin_dpo/margin_std": 107.8612060546875, "step": 428 }, { "KL/chosen_KL_mean": -157.85076904296875, "KL/mean": -183.78836059570312, "KL/rejected_KL_mean": -209.72593688964844, "KL/std": 85.93083190917969, "epoch": 0.6485260770975056, "fcm_dpo/beta": 0.00645467359572649, "fcm_dpo/delta": 0.06671077758073807, "fcm_dpo/margin": 51.87518310546875, "fcm_dpo/q_t": 0.42263728380203247, "grad_norm": 14.143959999084473, "learning_rate": 1.669846604344412e-07, "logits/chosen": 0.6520216464996338, "logits/rejected": 0.6711635589599609, "logps/chosen": -214.8604736328125, "logps/ref_chosen": -57.00970458984375, "logps/ref_rejected": -59.86549377441406, "logps/rejected": -269.5914306640625, "loss": 1.1735, "margin_dpo/margin_mean": 51.87518310546875, "margin_dpo/margin_std": 95.87495422363281, "step": 429 }, { "KL/chosen_KL_mean": -140.60528564453125, "KL/mean": -183.13192749023438, "KL/rejected_KL_mean": -225.65859985351562, "KL/std": 87.80361938476562, "epoch": 0.6500377928949358, "fcm_dpo/beta": 0.006307076197117567, "fcm_dpo/delta": -0.14426840841770172, "fcm_dpo/margin": 85.05331420898438, "fcm_dpo/q_t": 0.3759458661079407, "grad_norm": 13.13545036315918, "learning_rate": 1.6573863381573954e-07, "logits/chosen": 0.5855617523193359, "logits/rejected": 0.5838299989700317, "logps/chosen": -200.16848754882812, "logps/ref_chosen": -59.563194274902344, "logps/ref_rejected": -70.52289581298828, "logps/rejected": -296.1814880371094, "loss": 1.015, "margin_dpo/margin_mean": 85.05331420898438, "margin_dpo/margin_std": 100.49716186523438, "step": 430 }, { "KL/chosen_KL_mean": -140.24456787109375, "KL/mean": -170.9658203125, "KL/rejected_KL_mean": -201.68710327148438, "KL/std": 89.18963623046875, "epoch": 0.6515495086923658, "fcm_dpo/beta": 0.006269997917115688, "fcm_dpo/delta": 0.015329739078879356, "fcm_dpo/margin": 61.44251251220703, "fcm_dpo/q_t": 0.41269272565841675, "grad_norm": 12.90273380279541, "learning_rate": 1.6449496416858282e-07, "logits/chosen": 0.6572583913803101, "logits/rejected": 0.6016641855239868, "logps/chosen": -190.44488525390625, "logps/ref_chosen": -50.20032501220703, "logps/ref_rejected": -77.81680297851562, "logps/rejected": -279.50390625, "loss": 1.1252, "margin_dpo/margin_mean": 61.4425163269043, "margin_dpo/margin_std": 97.04786682128906, "step": 431 }, { "KL/chosen_KL_mean": -147.359619140625, "KL/mean": -179.90625, "KL/rejected_KL_mean": -212.45291137695312, "KL/std": 88.27741241455078, "epoch": 0.6530612244897959, "fcm_dpo/beta": 0.006269288249313831, "fcm_dpo/delta": -0.008461249992251396, "fcm_dpo/margin": 65.09325408935547, "fcm_dpo/q_t": 0.40682950615882874, "grad_norm": 13.348393440246582, "learning_rate": 1.632536862810844e-07, "logits/chosen": 0.6946985721588135, "logits/rejected": 0.6414108276367188, "logps/chosen": -209.02236938476562, "logps/ref_chosen": -61.662757873535156, "logps/ref_rejected": -83.94496154785156, "logps/rejected": -296.3978576660156, "loss": 1.1148, "margin_dpo/margin_mean": 65.09325408935547, "margin_dpo/margin_std": 101.81596374511719, "step": 432 }, { "KL/chosen_KL_mean": -145.7971954345703, "KL/mean": -183.37362670898438, "KL/rejected_KL_mean": -220.95008850097656, "KL/std": 83.93692016601562, "epoch": 0.654572940287226, "fcm_dpo/beta": 0.006223015021532774, "fcm_dpo/delta": -0.07095064222812653, "fcm_dpo/margin": 75.15287017822266, "fcm_dpo/q_t": 0.3921838700771332, "grad_norm": 12.827315330505371, "learning_rate": 1.6201483487445515e-07, "logits/chosen": 0.7911877632141113, "logits/rejected": 0.7912458181381226, "logps/chosen": -209.5263671875, "logps/ref_chosen": -63.72917938232422, "logps/ref_rejected": -65.8391342163086, "logps/rejected": -286.7892150878906, "loss": 1.0583, "margin_dpo/margin_mean": 75.15287017822266, "margin_dpo/margin_std": 96.3397216796875, "step": 433 }, { "KL/chosen_KL_mean": -124.45191192626953, "KL/mean": -166.2886962890625, "KL/rejected_KL_mean": -208.12547302246094, "KL/std": 93.03741455078125, "epoch": 0.656084656084656, "fcm_dpo/beta": 0.00603717565536499, "fcm_dpo/delta": -0.1126817911863327, "fcm_dpo/margin": 83.67355346679688, "fcm_dpo/q_t": 0.3845771253108978, "grad_norm": 11.618966102600098, "learning_rate": 1.6077844460203204e-07, "logits/chosen": 0.8206672072410583, "logits/rejected": 0.7543665170669556, "logps/chosen": -172.42523193359375, "logps/ref_chosen": -47.97331619262695, "logps/ref_rejected": -72.51132202148438, "logps/rejected": -280.63677978515625, "loss": 1.0545, "margin_dpo/margin_mean": 83.67355346679688, "margin_dpo/margin_std": 112.55625915527344, "step": 434 }, { "KL/chosen_KL_mean": -150.77972412109375, "KL/mean": -181.7442169189453, "KL/rejected_KL_mean": -212.7086944580078, "KL/std": 88.39662170410156, "epoch": 0.6575963718820862, "fcm_dpo/beta": 0.0060953618958592415, "fcm_dpo/delta": 0.02286442741751671, "fcm_dpo/margin": 61.92898941040039, "fcm_dpo/q_t": 0.41460686922073364, "grad_norm": 13.56441879272461, "learning_rate": 1.5954455004830878e-07, "logits/chosen": 0.7735519409179688, "logits/rejected": 0.7344120144844055, "logps/chosen": -207.8399658203125, "logps/ref_chosen": -57.06024932861328, "logps/ref_rejected": -71.69146728515625, "logps/rejected": -284.400146484375, "loss": 1.1378, "margin_dpo/margin_mean": 61.92898941040039, "margin_dpo/margin_std": 102.48014068603516, "step": 435 }, { "KL/chosen_KL_mean": -149.90362548828125, "KL/mean": -175.8869171142578, "KL/rejected_KL_mean": -201.87020874023438, "KL/std": 87.14476776123047, "epoch": 0.6591080876795162, "fcm_dpo/beta": 0.006144003942608833, "fcm_dpo/delta": 0.08344481885433197, "fcm_dpo/margin": 51.96659469604492, "fcm_dpo/q_t": 0.4281252324581146, "grad_norm": 14.366814613342285, "learning_rate": 1.5831318572796847e-07, "logits/chosen": 0.6878930926322937, "logits/rejected": 0.6309799551963806, "logps/chosen": -206.06167602539062, "logps/ref_chosen": -56.158050537109375, "logps/ref_rejected": -67.63787841796875, "logps/rejected": -269.5080871582031, "loss": 1.1935, "margin_dpo/margin_mean": 51.96659469604492, "margin_dpo/margin_std": 106.5383529663086, "step": 436 }, { "KL/chosen_KL_mean": -156.62506103515625, "KL/mean": -187.46817016601562, "KL/rejected_KL_mean": -218.31126403808594, "KL/std": 94.74977111816406, "epoch": 0.6606198034769464, "fcm_dpo/beta": 0.006142673082649708, "fcm_dpo/delta": 0.020575210452079773, "fcm_dpo/margin": 61.68621063232422, "fcm_dpo/q_t": 0.41576099395751953, "grad_norm": 15.018247604370117, "learning_rate": 1.5708438608491815e-07, "logits/chosen": 0.6854249238967896, "logits/rejected": 0.5601704716682434, "logps/chosen": -213.61083984375, "logps/ref_chosen": -56.98578643798828, "logps/ref_rejected": -85.61524963378906, "logps/rejected": -303.926513671875, "loss": 1.1762, "margin_dpo/margin_mean": 61.68621063232422, "margin_dpo/margin_std": 119.7267074584961, "step": 437 }, { "KL/chosen_KL_mean": -134.8008270263672, "KL/mean": -176.47152709960938, "KL/rejected_KL_mean": -218.1422119140625, "KL/std": 98.8070297241211, "epoch": 0.6621315192743764, "fcm_dpo/beta": 0.006119017489254475, "fcm_dpo/delta": -0.11572910845279694, "fcm_dpo/margin": 83.3414077758789, "fcm_dpo/q_t": 0.3877474069595337, "grad_norm": 12.523723602294922, "learning_rate": 1.558581854913253e-07, "logits/chosen": 0.7682843804359436, "logits/rejected": 0.7010859847068787, "logps/chosen": -176.07861328125, "logps/ref_chosen": -41.27777862548828, "logps/ref_rejected": -65.33840942382812, "logps/rejected": -283.48065185546875, "loss": 1.0372, "margin_dpo/margin_mean": 83.3414077758789, "margin_dpo/margin_std": 108.39314270019531, "step": 438 }, { "KL/chosen_KL_mean": -153.8817138671875, "KL/mean": -191.04409790039062, "KL/rejected_KL_mean": -228.2064666748047, "KL/std": 100.76587677001953, "epoch": 0.6636432350718064, "fcm_dpo/beta": 0.006057861726731062, "fcm_dpo/delta": -0.05337338522076607, "fcm_dpo/margin": 74.32474517822266, "fcm_dpo/q_t": 0.3977718949317932, "grad_norm": 12.527008056640625, "learning_rate": 1.5463461824665658e-07, "logits/chosen": 0.6251957416534424, "logits/rejected": 0.5903706550598145, "logps/chosen": -235.2993621826172, "logps/ref_chosen": -81.41764831542969, "logps/ref_rejected": -94.72309875488281, "logps/rejected": -322.9295654296875, "loss": 1.0785, "margin_dpo/margin_mean": 74.32474517822266, "margin_dpo/margin_std": 101.76370239257812, "step": 439 }, { "KL/chosen_KL_mean": -132.0679931640625, "KL/mean": -169.42111206054688, "KL/rejected_KL_mean": -206.7742462158203, "KL/std": 89.31532287597656, "epoch": 0.6651549508692366, "fcm_dpo/beta": 0.005974017083644867, "fcm_dpo/delta": -0.04854672774672508, "fcm_dpo/margin": 74.70626831054688, "fcm_dpo/q_t": 0.39923959970474243, "grad_norm": 20.631868362426758, "learning_rate": 1.534137185767178e-07, "logits/chosen": 0.6927276849746704, "logits/rejected": 0.5932461619377136, "logps/chosen": -174.60617065429688, "logps/ref_chosen": -42.538185119628906, "logps/ref_rejected": -69.78813934326172, "logps/rejected": -276.5623779296875, "loss": 1.0803, "margin_dpo/margin_mean": 74.70626831054688, "margin_dpo/margin_std": 103.83875274658203, "step": 440 }, { "KL/chosen_KL_mean": -133.71435546875, "KL/mean": -175.30303955078125, "KL/rejected_KL_mean": -216.8917236328125, "KL/std": 96.92250061035156, "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.005834102630615234, "fcm_dpo/delta": -0.09035193175077438, "fcm_dpo/margin": 83.17737579345703, "fcm_dpo/q_t": 0.3867751955986023, "grad_norm": 13.054012298583984, "learning_rate": 1.521955206326976e-07, "logits/chosen": 0.6799192428588867, "logits/rejected": 0.5854415893554688, "logps/chosen": -191.30758666992188, "logps/ref_chosen": -57.593223571777344, "logps/ref_rejected": -84.82878875732422, "logps/rejected": -301.72052001953125, "loss": 1.0199, "margin_dpo/margin_mean": 83.1773681640625, "margin_dpo/margin_std": 86.29299926757812, "step": 441 }, { "KL/chosen_KL_mean": -161.66006469726562, "KL/mean": -201.89044189453125, "KL/rejected_KL_mean": -242.12081909179688, "KL/std": 97.12081146240234, "epoch": 0.6681783824640968, "fcm_dpo/beta": 0.005757060833275318, "fcm_dpo/delta": -0.06665948033332825, "fcm_dpo/margin": 80.46076202392578, "fcm_dpo/q_t": 0.3939310908317566, "grad_norm": 13.893742561340332, "learning_rate": 1.5098005849021078e-07, "logits/chosen": 0.7006994485855103, "logits/rejected": 0.6529697179794312, "logps/chosen": -229.12127685546875, "logps/ref_chosen": -67.46121978759766, "logps/ref_rejected": -89.0693588256836, "logps/rejected": -331.190185546875, "loss": 1.0552, "margin_dpo/margin_mean": 80.46076965332031, "margin_dpo/margin_std": 100.84950256347656, "step": 442 }, { "KL/chosen_KL_mean": -133.28994750976562, "KL/mean": -182.92572021484375, "KL/rejected_KL_mean": -232.5614776611328, "KL/std": 102.47624206542969, "epoch": 0.6696900982615268, "fcm_dpo/beta": 0.005609867163002491, "fcm_dpo/delta": -0.16628825664520264, "fcm_dpo/margin": 99.27154541015625, "fcm_dpo/q_t": 0.3736429810523987, "grad_norm": 14.203608512878418, "learning_rate": 1.4976736614834662e-07, "logits/chosen": 0.7221025228500366, "logits/rejected": 0.6557480096817017, "logps/chosen": -188.0860595703125, "logps/ref_chosen": -54.79610061645508, "logps/ref_rejected": -77.80781555175781, "logps/rejected": -310.3692932128906, "loss": 0.9957, "margin_dpo/margin_mean": 99.27154541015625, "margin_dpo/margin_std": 114.22992706298828, "step": 443 }, { "KL/chosen_KL_mean": -164.04290771484375, "KL/mean": -186.19549560546875, "KL/rejected_KL_mean": -208.34808349609375, "KL/std": 101.94633483886719, "epoch": 0.671201814058957, "fcm_dpo/beta": 0.005667264573276043, "fcm_dpo/delta": 0.15300863981246948, "fcm_dpo/margin": 44.3051643371582, "fcm_dpo/q_t": 0.4450858235359192, "grad_norm": 16.502891540527344, "learning_rate": 1.4855747752871654e-07, "logits/chosen": 0.6830138564109802, "logits/rejected": 0.5934052467346191, "logps/chosen": -222.79196166992188, "logps/ref_chosen": -58.749061584472656, "logps/ref_rejected": -86.87396240234375, "logps/rejected": -295.2220458984375, "loss": 1.2501, "margin_dpo/margin_mean": 44.30516052246094, "margin_dpo/margin_std": 112.28422546386719, "step": 444 }, { "KL/chosen_KL_mean": -155.0587921142578, "KL/mean": -196.34661865234375, "KL/rejected_KL_mean": -237.63446044921875, "KL/std": 93.90274047851562, "epoch": 0.672713529856387, "fcm_dpo/beta": 0.0056798215955495834, "fcm_dpo/delta": -0.0723859965801239, "fcm_dpo/margin": 82.57566833496094, "fcm_dpo/q_t": 0.3913373351097107, "grad_norm": 12.268744468688965, "learning_rate": 1.473504264745062e-07, "logits/chosen": 0.6730467081069946, "logits/rejected": 0.6653146743774414, "logps/chosen": -215.97622680664062, "logps/ref_chosen": -60.91743850708008, "logps/ref_rejected": -71.5637435913086, "logps/rejected": -309.19818115234375, "loss": 1.0459, "margin_dpo/margin_mean": 82.57566833496094, "margin_dpo/margin_std": 99.127197265625, "step": 445 }, { "KL/chosen_KL_mean": -141.07745361328125, "KL/mean": -183.24526977539062, "KL/rejected_KL_mean": -225.4130859375, "KL/std": 95.9248275756836, "epoch": 0.674225245653817, "fcm_dpo/beta": 0.005640149116516113, "fcm_dpo/delta": -0.08037254214286804, "fcm_dpo/margin": 84.33563232421875, "fcm_dpo/q_t": 0.3905682861804962, "grad_norm": 11.658855438232422, "learning_rate": 1.461462467495284e-07, "logits/chosen": 0.7361161708831787, "logits/rejected": 0.6562565565109253, "logps/chosen": -189.876708984375, "logps/ref_chosen": -48.79924774169922, "logps/ref_rejected": -71.8719482421875, "logps/rejected": -297.2850341796875, "loss": 1.0373, "margin_dpo/margin_mean": 84.33563232421875, "margin_dpo/margin_std": 91.992919921875, "step": 446 }, { "KL/chosen_KL_mean": -137.0552978515625, "KL/mean": -186.58033752441406, "KL/rejected_KL_mean": -236.10537719726562, "KL/std": 95.55780029296875, "epoch": 0.6757369614512472, "fcm_dpo/beta": 0.005425992421805859, "fcm_dpo/delta": -0.14611226320266724, "fcm_dpo/margin": 99.0500717163086, "fcm_dpo/q_t": 0.37507176399230957, "grad_norm": 12.766183853149414, "learning_rate": 1.4494497203727843e-07, "logits/chosen": 0.6704369783401489, "logits/rejected": 0.565066397190094, "logps/chosen": -190.73802185058594, "logps/ref_chosen": -53.682716369628906, "logps/ref_rejected": -88.17315673828125, "logps/rejected": -324.2785339355469, "loss": 1.0021, "margin_dpo/margin_mean": 99.05007934570312, "margin_dpo/margin_std": 108.26829528808594, "step": 447 }, { "KL/chosen_KL_mean": -149.04644775390625, "KL/mean": -187.95932006835938, "KL/rejected_KL_mean": -226.8721923828125, "KL/std": 98.18172454833984, "epoch": 0.6772486772486772, "fcm_dpo/beta": 0.005390051286667585, "fcm_dpo/delta": -0.020336728543043137, "fcm_dpo/margin": 77.82575988769531, "fcm_dpo/q_t": 0.40321606397628784, "grad_norm": 9.62354564666748, "learning_rate": 1.4374663593999256e-07, "logits/chosen": 0.7249032258987427, "logits/rejected": 0.6728522777557373, "logps/chosen": -202.79769897460938, "logps/ref_chosen": -53.75125503540039, "logps/ref_rejected": -77.17623901367188, "logps/rejected": -304.0484313964844, "loss": 1.085, "margin_dpo/margin_mean": 77.82575988769531, "margin_dpo/margin_std": 105.24398803710938, "step": 448 }, { "KL/chosen_KL_mean": -178.3193359375, "KL/mean": -197.23353576660156, "KL/rejected_KL_mean": -216.14773559570312, "KL/std": 100.89761352539062, "epoch": 0.6787603930461074, "fcm_dpo/beta": 0.005525457672774792, "fcm_dpo/delta": 0.19548821449279785, "fcm_dpo/margin": 37.82838821411133, "fcm_dpo/q_t": 0.452498197555542, "grad_norm": 18.4531307220459, "learning_rate": 1.4255127197770707e-07, "logits/chosen": 0.5460859537124634, "logits/rejected": 0.545451283454895, "logps/chosen": -254.14671325683594, "logps/ref_chosen": -75.82737731933594, "logps/ref_rejected": -82.20687866210938, "logps/rejected": -298.3546142578125, "loss": 1.2628, "margin_dpo/margin_mean": 37.82838821411133, "margin_dpo/margin_std": 99.50639343261719, "step": 449 }, { "KL/chosen_KL_mean": -146.93153381347656, "KL/mean": -179.05615234375, "KL/rejected_KL_mean": -211.1807861328125, "KL/std": 101.90434265136719, "epoch": 0.6802721088435374, "fcm_dpo/beta": 0.005613807588815689, "fcm_dpo/delta": 0.04069120064377785, "fcm_dpo/margin": 64.24925231933594, "fcm_dpo/q_t": 0.4215589761734009, "grad_norm": 12.490607261657715, "learning_rate": 1.4135891358732205e-07, "logits/chosen": 0.8319960832595825, "logits/rejected": 0.7168232798576355, "logps/chosen": -194.0472412109375, "logps/ref_chosen": -47.11572265625, "logps/ref_rejected": -78.7546615600586, "logps/rejected": -289.9354553222656, "loss": 1.16, "margin_dpo/margin_mean": 64.2492446899414, "margin_dpo/margin_std": 117.5666275024414, "step": 450 }, { "KL/chosen_KL_mean": -148.942138671875, "KL/mean": -177.62315368652344, "KL/rejected_KL_mean": -206.30418395996094, "KL/std": 96.27726745605469, "epoch": 0.6817838246409675, "fcm_dpo/beta": 0.005715922452509403, "fcm_dpo/delta": 0.07441489398479462, "fcm_dpo/margin": 57.362030029296875, "fcm_dpo/q_t": 0.42512810230255127, "grad_norm": 13.263283729553223, "learning_rate": 1.4016959412166437e-07, "logits/chosen": 0.6437816619873047, "logits/rejected": 0.5958889722824097, "logps/chosen": -212.29258728027344, "logps/ref_chosen": -63.350440979003906, "logps/ref_rejected": -76.28530883789062, "logps/rejected": -282.5894775390625, "loss": 1.1622, "margin_dpo/margin_mean": 57.362030029296875, "margin_dpo/margin_std": 100.08808898925781, "step": 451 }, { "KL/chosen_KL_mean": -147.36083984375, "KL/mean": -180.61264038085938, "KL/rejected_KL_mean": -213.86444091796875, "KL/std": 92.54578399658203, "epoch": 0.6832955404383976, "fcm_dpo/beta": 0.005742376670241356, "fcm_dpo/delta": 0.018834762275218964, "fcm_dpo/margin": 66.50358581542969, "fcm_dpo/q_t": 0.4136967360973358, "grad_norm": 14.197269439697266, "learning_rate": 1.3898334684855645e-07, "logits/chosen": 0.6498645544052124, "logits/rejected": 0.5683034658432007, "logps/chosen": -202.94668579101562, "logps/ref_chosen": -55.58583450317383, "logps/ref_rejected": -77.68738555908203, "logps/rejected": -291.55181884765625, "loss": 1.1394, "margin_dpo/margin_mean": 66.50358581542969, "margin_dpo/margin_std": 112.99519348144531, "step": 452 }, { "KL/chosen_KL_mean": -147.13145446777344, "KL/mean": -179.03643798828125, "KL/rejected_KL_mean": -210.94140625, "KL/std": 96.53543090820312, "epoch": 0.6848072562358276, "fcm_dpo/beta": 0.005770553834736347, "fcm_dpo/delta": 0.033000145107507706, "fcm_dpo/margin": 63.80995178222656, "fcm_dpo/q_t": 0.41716498136520386, "grad_norm": 14.38304615020752, "learning_rate": 1.3780020494988445e-07, "logits/chosen": 0.637764573097229, "logits/rejected": 0.6154556274414062, "logps/chosen": -208.90966796875, "logps/ref_chosen": -61.778202056884766, "logps/ref_rejected": -71.51403045654297, "logps/rejected": -282.4554443359375, "loss": 1.1463, "margin_dpo/margin_mean": 63.80995178222656, "margin_dpo/margin_std": 109.52471160888672, "step": 453 }, { "KL/chosen_KL_mean": -136.54115295410156, "KL/mean": -172.95950317382812, "KL/rejected_KL_mean": -209.37783813476562, "KL/std": 96.98989868164062, "epoch": 0.6863189720332578, "fcm_dpo/beta": 0.005766098387539387, "fcm_dpo/delta": -0.020890843123197556, "fcm_dpo/margin": 72.83668518066406, "fcm_dpo/q_t": 0.4038616418838501, "grad_norm": 12.730914115905762, "learning_rate": 1.366202015206706e-07, "logits/chosen": 0.6938184499740601, "logits/rejected": 0.6552758812904358, "logps/chosen": -188.13632202148438, "logps/ref_chosen": -51.59515380859375, "logps/ref_rejected": -63.96732711791992, "logps/rejected": -273.34515380859375, "loss": 1.1007, "margin_dpo/margin_mean": 72.836669921875, "margin_dpo/margin_std": 109.37074279785156, "step": 454 }, { "KL/chosen_KL_mean": -156.71600341796875, "KL/mean": -192.44114685058594, "KL/rejected_KL_mean": -228.16629028320312, "KL/std": 96.54014587402344, "epoch": 0.6878306878306878, "fcm_dpo/beta": 0.0057487934827804565, "fcm_dpo/delta": -0.011250181123614311, "fcm_dpo/margin": 71.4502944946289, "fcm_dpo/q_t": 0.4079058766365051, "grad_norm": 13.477120399475098, "learning_rate": 1.354433695681474e-07, "logits/chosen": 0.5650009512901306, "logits/rejected": 0.5374402403831482, "logps/chosen": -227.36770629882812, "logps/ref_chosen": -70.65170288085938, "logps/ref_rejected": -77.44276428222656, "logps/rejected": -305.60906982421875, "loss": 1.0988, "margin_dpo/margin_mean": 71.45030212402344, "margin_dpo/margin_std": 103.77528381347656, "step": 455 }, { "KL/chosen_KL_mean": -153.38815307617188, "KL/mean": -186.91957092285156, "KL/rejected_KL_mean": -220.4510040283203, "KL/std": 96.58705139160156, "epoch": 0.6893424036281179, "fcm_dpo/beta": 0.005753940902650356, "fcm_dpo/delta": 0.014651095494627953, "fcm_dpo/margin": 67.0628433227539, "fcm_dpo/q_t": 0.41437584161758423, "grad_norm": 15.417052268981934, "learning_rate": 1.3426974201083439e-07, "logits/chosen": 0.6084275245666504, "logits/rejected": 0.5453929901123047, "logps/chosen": -209.78643798828125, "logps/ref_chosen": -56.398284912109375, "logps/ref_rejected": -82.61642456054688, "logps/rejected": -303.06744384765625, "loss": 1.1289, "margin_dpo/margin_mean": 67.0628433227539, "margin_dpo/margin_std": 110.33543395996094, "step": 456 }, { "KL/chosen_KL_mean": -151.1236572265625, "KL/mean": -186.5624237060547, "KL/rejected_KL_mean": -222.00119018554688, "KL/std": 98.57406616210938, "epoch": 0.690854119425548, "fcm_dpo/beta": 0.005745270289480686, "fcm_dpo/delta": -0.007712500169873238, "fcm_dpo/margin": 70.87751770019531, "fcm_dpo/q_t": 0.4060364365577698, "grad_norm": 12.802936553955078, "learning_rate": 1.3309935167761717e-07, "logits/chosen": 0.7779738306999207, "logits/rejected": 0.6983845233917236, "logps/chosen": -195.84423828125, "logps/ref_chosen": -44.72057342529297, "logps/ref_rejected": -68.1158676147461, "logps/rejected": -290.1170654296875, "loss": 1.09, "margin_dpo/margin_mean": 70.87751770019531, "margin_dpo/margin_std": 94.67393493652344, "step": 457 }, { "KL/chosen_KL_mean": -144.48300170898438, "KL/mean": -183.05062866210938, "KL/rejected_KL_mean": -221.61825561523438, "KL/std": 101.5442886352539, "epoch": 0.6923658352229781, "fcm_dpo/beta": 0.005730716977268457, "fcm_dpo/delta": -0.04396446794271469, "fcm_dpo/margin": 77.13526153564453, "fcm_dpo/q_t": 0.3995509445667267, "grad_norm": 12.134772300720215, "learning_rate": 1.3193223130682936e-07, "logits/chosen": 0.6931065917015076, "logits/rejected": 0.5775718688964844, "logps/chosen": -194.48867797851562, "logps/ref_chosen": -50.00569152832031, "logps/ref_rejected": -87.50015258789062, "logps/rejected": -309.118408203125, "loss": 1.0893, "margin_dpo/margin_mean": 77.13526916503906, "margin_dpo/margin_std": 113.13229370117188, "step": 458 }, { "KL/chosen_KL_mean": -140.49830627441406, "KL/mean": -185.78858947753906, "KL/rejected_KL_mean": -231.07887268066406, "KL/std": 115.9324951171875, "epoch": 0.6938775510204082, "fcm_dpo/beta": 0.0056596603244543076, "fcm_dpo/delta": -0.11930438876152039, "fcm_dpo/margin": 90.58056640625, "fcm_dpo/q_t": 0.38196009397506714, "grad_norm": 12.148503303527832, "learning_rate": 1.3076841354533658e-07, "logits/chosen": 0.7065992951393127, "logits/rejected": 0.6742749214172363, "logps/chosen": -205.87625122070312, "logps/ref_chosen": -65.37794494628906, "logps/ref_rejected": -88.19244384765625, "logps/rejected": -319.27130126953125, "loss": 1.0332, "margin_dpo/margin_mean": 90.58056640625, "margin_dpo/margin_std": 108.33981323242188, "step": 459 }, { "KL/chosen_KL_mean": -152.5089111328125, "KL/mean": -196.7735595703125, "KL/rejected_KL_mean": -241.03823852539062, "KL/std": 101.97077941894531, "epoch": 0.6953892668178382, "fcm_dpo/beta": 0.0054806992411613464, "fcm_dpo/delta": -0.0901375487446785, "fcm_dpo/margin": 88.52931213378906, "fcm_dpo/q_t": 0.3906601667404175, "grad_norm": 11.680279731750488, "learning_rate": 1.2960793094762345e-07, "logits/chosen": 0.7258821129798889, "logits/rejected": 0.6057487726211548, "logps/chosen": -217.07057189941406, "logps/ref_chosen": -64.5616683959961, "logps/ref_rejected": -88.67890167236328, "logps/rejected": -329.7171325683594, "loss": 1.0422, "margin_dpo/margin_mean": 88.52931213378906, "margin_dpo/margin_std": 110.12898254394531, "step": 460 }, { "KL/chosen_KL_mean": -127.51901245117188, "KL/mean": -172.11732482910156, "KL/rejected_KL_mean": -216.71563720703125, "KL/std": 94.23216247558594, "epoch": 0.6969009826152683, "fcm_dpo/beta": 0.005357278976589441, "fcm_dpo/delta": -0.08428196609020233, "fcm_dpo/margin": 89.1966323852539, "fcm_dpo/q_t": 0.38994884490966797, "grad_norm": 11.4501371383667, "learning_rate": 1.2845081597488286e-07, "logits/chosen": 0.7885807752609253, "logits/rejected": 0.7069375514984131, "logps/chosen": -176.99691772460938, "logps/ref_chosen": -49.4779167175293, "logps/ref_rejected": -72.65262603759766, "logps/rejected": -289.3682861328125, "loss": 1.0401, "margin_dpo/margin_mean": 89.19664001464844, "margin_dpo/margin_std": 101.12211608886719, "step": 461 }, { "KL/chosen_KL_mean": -140.79742431640625, "KL/mean": -184.84768676757812, "KL/rejected_KL_mean": -228.89794921875, "KL/std": 92.286865234375, "epoch": 0.6984126984126984, "fcm_dpo/beta": 0.00531365629285574, "fcm_dpo/delta": -0.07177212089300156, "fcm_dpo/margin": 88.10053253173828, "fcm_dpo/q_t": 0.390781968832016, "grad_norm": 12.073556900024414, "learning_rate": 1.27297100994108e-07, "logits/chosen": 0.6879956722259521, "logits/rejected": 0.6383634209632874, "logps/chosen": -201.29254150390625, "logps/ref_chosen": -60.4951171875, "logps/ref_rejected": -74.82136535644531, "logps/rejected": -303.71929931640625, "loss": 1.0442, "margin_dpo/margin_mean": 88.10052490234375, "margin_dpo/margin_std": 103.91917419433594, "step": 462 }, { "KL/chosen_KL_mean": -168.54791259765625, "KL/mean": -195.15968322753906, "KL/rejected_KL_mean": -221.77145385742188, "KL/std": 93.11166381835938, "epoch": 0.6999244142101285, "fcm_dpo/beta": 0.005378012545406818, "fcm_dpo/delta": 0.11724771559238434, "fcm_dpo/margin": 53.223541259765625, "fcm_dpo/q_t": 0.4336676001548767, "grad_norm": 17.376508712768555, "learning_rate": 1.2614681827718695e-07, "logits/chosen": 0.6730971336364746, "logits/rejected": 0.6703752279281616, "logps/chosen": -236.2330322265625, "logps/ref_chosen": -67.68511962890625, "logps/ref_rejected": -71.32196044921875, "logps/rejected": -293.0934143066406, "loss": 1.1962, "margin_dpo/margin_mean": 53.223541259765625, "margin_dpo/margin_std": 102.88514709472656, "step": 463 }, { "KL/chosen_KL_mean": -152.6397705078125, "KL/mean": -194.53070068359375, "KL/rejected_KL_mean": -236.421630859375, "KL/std": 104.98362731933594, "epoch": 0.7014361300075586, "fcm_dpo/beta": 0.005408740136772394, "fcm_dpo/delta": -0.05576523393392563, "fcm_dpo/margin": 83.78184509277344, "fcm_dpo/q_t": 0.3968367576599121, "grad_norm": 11.669844627380371, "learning_rate": 1.2500000000000005e-07, "logits/chosen": 0.6964342594146729, "logits/rejected": 0.6669223308563232, "logps/chosen": -211.805419921875, "logps/ref_chosen": -59.16564178466797, "logps/ref_rejected": -69.56146240234375, "logps/rejected": -305.98309326171875, "loss": 1.0841, "margin_dpo/margin_mean": 83.78184509277344, "margin_dpo/margin_std": 118.7095718383789, "step": 464 }, { "KL/chosen_KL_mean": -160.89891052246094, "KL/mean": -196.22000122070312, "KL/rejected_KL_mean": -231.54110717773438, "KL/std": 99.98161315917969, "epoch": 0.7029478458049887, "fcm_dpo/beta": 0.00537415686994791, "fcm_dpo/delta": 0.021056313067674637, "fcm_dpo/margin": 70.64216613769531, "fcm_dpo/q_t": 0.41352707147598267, "grad_norm": 12.136297225952148, "learning_rate": 1.238566782415197e-07, "logits/chosen": 0.752052903175354, "logits/rejected": 0.6917625665664673, "logps/chosen": -219.41258239746094, "logps/ref_chosen": -58.513671875, "logps/ref_rejected": -84.31745910644531, "logps/rejected": -315.85858154296875, "loss": 1.1299, "margin_dpo/margin_mean": 70.64217376708984, "margin_dpo/margin_std": 112.321044921875, "step": 465 }, { "KL/chosen_KL_mean": -174.26275634765625, "KL/mean": -194.21337890625, "KL/rejected_KL_mean": -214.16400146484375, "KL/std": 107.25249481201172, "epoch": 0.7044595616024187, "fcm_dpo/beta": 0.005527378059923649, "fcm_dpo/delta": 0.18388685584068298, "fcm_dpo/margin": 39.90126037597656, "fcm_dpo/q_t": 0.4488188624382019, "grad_norm": 18.92510414123535, "learning_rate": 1.2271688498291334e-07, "logits/chosen": 0.682281494140625, "logits/rejected": 0.6877008676528931, "logps/chosen": -247.528564453125, "logps/ref_chosen": -73.26580810546875, "logps/ref_rejected": -74.83621215820312, "logps/rejected": -289.0002136230469, "loss": 1.2685, "margin_dpo/margin_mean": 39.90126037597656, "margin_dpo/margin_std": 110.06634521484375, "step": 466 }, { "KL/chosen_KL_mean": -153.88006591796875, "KL/mean": -187.2439422607422, "KL/rejected_KL_mean": -220.60781860351562, "KL/std": 103.33351135253906, "epoch": 0.7059712773998488, "fcm_dpo/beta": 0.005615277215838432, "fcm_dpo/delta": 0.026298683136701584, "fcm_dpo/margin": 66.72773742675781, "fcm_dpo/q_t": 0.4163486957550049, "grad_norm": 11.329875946044922, "learning_rate": 1.2158065210664848e-07, "logits/chosen": 0.7394921779632568, "logits/rejected": 0.5978328585624695, "logps/chosen": -201.4595489501953, "logps/ref_chosen": -47.57947540283203, "logps/ref_rejected": -78.68522644042969, "logps/rejected": -299.29302978515625, "loss": 1.1286, "margin_dpo/margin_mean": 66.72773742675781, "margin_dpo/margin_std": 107.79798889160156, "step": 467 }, { "KL/chosen_KL_mean": -146.56996154785156, "KL/mean": -190.59567260742188, "KL/rejected_KL_mean": -234.62139892578125, "KL/std": 101.60633850097656, "epoch": 0.7074829931972789, "fcm_dpo/beta": 0.005549922585487366, "fcm_dpo/delta": -0.09322632849216461, "fcm_dpo/margin": 88.05142974853516, "fcm_dpo/q_t": 0.38789719343185425, "grad_norm": 16.00092887878418, "learning_rate": 1.204480113956011e-07, "logits/chosen": 0.6605847477912903, "logits/rejected": 0.6465221643447876, "logps/chosen": -210.49774169921875, "logps/ref_chosen": -63.92778778076172, "logps/ref_rejected": -76.51626586914062, "logps/rejected": -311.1376647949219, "loss": 1.0514, "margin_dpo/margin_mean": 88.05143737792969, "margin_dpo/margin_std": 115.202880859375, "step": 468 }, { "KL/chosen_KL_mean": -149.39480590820312, "KL/mean": -189.18585205078125, "KL/rejected_KL_mean": -228.97686767578125, "KL/std": 100.38656616210938, "epoch": 0.708994708994709, "fcm_dpo/beta": 0.005461008753627539, "fcm_dpo/delta": -0.037061676383018494, "fcm_dpo/margin": 79.58206939697266, "fcm_dpo/q_t": 0.3996996283531189, "grad_norm": 13.444028854370117, "learning_rate": 1.1931899453216697e-07, "logits/chosen": 0.7412175536155701, "logits/rejected": 0.7253165245056152, "logps/chosen": -208.4530029296875, "logps/ref_chosen": -59.05818176269531, "logps/ref_rejected": -75.67672729492188, "logps/rejected": -304.6535949707031, "loss": 1.0622, "margin_dpo/margin_mean": 79.58206939697266, "margin_dpo/margin_std": 93.98836517333984, "step": 469 }, { "KL/chosen_KL_mean": -142.4019317626953, "KL/mean": -181.78958129882812, "KL/rejected_KL_mean": -221.17721557617188, "KL/std": 96.90911865234375, "epoch": 0.7105064247921391, "fcm_dpo/beta": 0.005472735036164522, "fcm_dpo/delta": -0.03261677548289299, "fcm_dpo/margin": 78.77527618408203, "fcm_dpo/q_t": 0.4012529253959656, "grad_norm": 12.413467407226562, "learning_rate": 1.1819363309737438e-07, "logits/chosen": 0.7318305969238281, "logits/rejected": 0.665635883808136, "logps/chosen": -190.26937866210938, "logps/ref_chosen": -47.86743927001953, "logps/ref_rejected": -65.96859741210938, "logps/rejected": -287.14581298828125, "loss": 1.0847, "margin_dpo/margin_mean": 78.77527618408203, "margin_dpo/margin_std": 108.5361328125, "step": 470 }, { "KL/chosen_KL_mean": -138.75448608398438, "KL/mean": -181.73556518554688, "KL/rejected_KL_mean": -224.7166290283203, "KL/std": 96.12451171875, "epoch": 0.7120181405895691, "fcm_dpo/beta": 0.0053847962990403175, "fcm_dpo/delta": -0.06609291583299637, "fcm_dpo/margin": 85.96214294433594, "fcm_dpo/q_t": 0.3922724723815918, "grad_norm": 13.146032333374023, "learning_rate": 1.1707195857000215e-07, "logits/chosen": 0.7092275619506836, "logits/rejected": 0.6556600332260132, "logps/chosen": -196.53233337402344, "logps/ref_chosen": -57.777854919433594, "logps/ref_rejected": -73.81172180175781, "logps/rejected": -298.5283508300781, "loss": 1.0531, "margin_dpo/margin_mean": 85.96214294433594, "margin_dpo/margin_std": 105.41901397705078, "step": 471 }, { "KL/chosen_KL_mean": -139.62513732910156, "KL/mean": -173.63473510742188, "KL/rejected_KL_mean": -207.6443328857422, "KL/std": 100.2338638305664, "epoch": 0.7135298563869993, "fcm_dpo/beta": 0.00541897676885128, "fcm_dpo/delta": 0.03227098286151886, "fcm_dpo/margin": 68.01919555664062, "fcm_dpo/q_t": 0.41764307022094727, "grad_norm": 13.090045928955078, "learning_rate": 1.1595400232569768e-07, "logits/chosen": 0.7192651629447937, "logits/rejected": 0.6732028722763062, "logps/chosen": -195.5338134765625, "logps/ref_chosen": -55.908668518066406, "logps/ref_rejected": -74.70294189453125, "logps/rejected": -282.3472900390625, "loss": 1.159, "margin_dpo/margin_mean": 68.01919555664062, "margin_dpo/margin_std": 125.8399658203125, "step": 472 }, { "KL/chosen_KL_mean": -143.20843505859375, "KL/mean": -185.14447021484375, "KL/rejected_KL_mean": -227.0805206298828, "KL/std": 115.70045471191406, "epoch": 0.7150415721844293, "fcm_dpo/beta": 0.005365893244743347, "fcm_dpo/delta": -0.052427202463150024, "fcm_dpo/margin": 83.87207794189453, "fcm_dpo/q_t": 0.4017806649208069, "grad_norm": 14.348088264465332, "learning_rate": 1.1483979563610069e-07, "logits/chosen": 0.8000465631484985, "logits/rejected": 0.6861811876296997, "logps/chosen": -197.36932373046875, "logps/ref_chosen": -54.16088104248047, "logps/ref_rejected": -92.76789855957031, "logps/rejected": -319.8484191894531, "loss": 1.1088, "margin_dpo/margin_mean": 83.87207794189453, "margin_dpo/margin_std": 137.078125, "step": 473 }, { "KL/chosen_KL_mean": -145.8909912109375, "KL/mean": -181.9802703857422, "KL/rejected_KL_mean": -218.06954956054688, "KL/std": 103.30743408203125, "epoch": 0.7165532879818595, "fcm_dpo/beta": 0.0053678592666983604, "fcm_dpo/delta": 0.013039542362093925, "fcm_dpo/margin": 72.17857360839844, "fcm_dpo/q_t": 0.4138396084308624, "grad_norm": 18.02506446838379, "learning_rate": 1.1372936966796709e-07, "logits/chosen": 0.7723579406738281, "logits/rejected": 0.699033260345459, "logps/chosen": -192.57669067382812, "logps/ref_chosen": -46.685707092285156, "logps/ref_rejected": -71.44731903076172, "logps/rejected": -289.5168762207031, "loss": 1.1372, "margin_dpo/margin_mean": 72.17857360839844, "margin_dpo/margin_std": 123.31417846679688, "step": 474 }, { "KL/chosen_KL_mean": -139.6383514404297, "KL/mean": -190.87100219726562, "KL/rejected_KL_mean": -242.1036376953125, "KL/std": 104.505615234375, "epoch": 0.7180650037792895, "fcm_dpo/beta": 0.005235554650425911, "fcm_dpo/delta": -0.14490701258182526, "fcm_dpo/margin": 102.46532440185547, "fcm_dpo/q_t": 0.3785374164581299, "grad_norm": 9.661964416503906, "learning_rate": 1.126227554822985e-07, "logits/chosen": 0.7241721153259277, "logits/rejected": 0.6713173389434814, "logps/chosen": -198.1256561279297, "logps/ref_chosen": -58.4873046875, "logps/ref_rejected": -87.00187683105469, "logps/rejected": -329.10552978515625, "loss": 1.0044, "margin_dpo/margin_mean": 102.46531677246094, "margin_dpo/margin_std": 115.88256072998047, "step": 475 }, { "KL/chosen_KL_mean": -170.72518920898438, "KL/mean": -202.345703125, "KL/rejected_KL_mean": -233.9662322998047, "KL/std": 108.97540283203125, "epoch": 0.7195767195767195, "fcm_dpo/beta": 0.005290007218718529, "fcm_dpo/delta": 0.06701076030731201, "fcm_dpo/margin": 63.24106979370117, "fcm_dpo/q_t": 0.424893856048584, "grad_norm": 14.140975952148438, "learning_rate": 1.1151998403347243e-07, "logits/chosen": 0.6015628576278687, "logits/rejected": 0.6035597324371338, "logps/chosen": -246.1068115234375, "logps/ref_chosen": -75.38162231445312, "logps/ref_rejected": -76.99822235107422, "logps/rejected": -310.9644775390625, "loss": 1.1701, "margin_dpo/margin_mean": 63.24106979370117, "margin_dpo/margin_std": 116.38935852050781, "step": 476 }, { "KL/chosen_KL_mean": -169.3343048095703, "KL/mean": -205.39453125, "KL/rejected_KL_mean": -241.45474243164062, "KL/std": 109.87100219726562, "epoch": 0.7210884353741497, "fcm_dpo/beta": 0.005296539049595594, "fcm_dpo/delta": 0.018730882555246353, "fcm_dpo/margin": 72.12043762207031, "fcm_dpo/q_t": 0.41515541076660156, "grad_norm": 13.771434783935547, "learning_rate": 1.1042108616837692e-07, "logits/chosen": 0.7227048277854919, "logits/rejected": 0.6724196672439575, "logps/chosen": -230.40768432617188, "logps/ref_chosen": -61.073387145996094, "logps/ref_rejected": -81.34375, "logps/rejected": -322.7984924316406, "loss": 1.1689, "margin_dpo/margin_mean": 72.12043762207031, "margin_dpo/margin_std": 140.45468139648438, "step": 477 }, { "KL/chosen_KL_mean": -149.87953186035156, "KL/mean": -177.552490234375, "KL/rejected_KL_mean": -205.22547912597656, "KL/std": 93.2808837890625, "epoch": 0.7226001511715797, "fcm_dpo/beta": 0.005373564548790455, "fcm_dpo/delta": 0.10587666183710098, "fcm_dpo/margin": 55.34593963623047, "fcm_dpo/q_t": 0.43196284770965576, "grad_norm": 15.536688804626465, "learning_rate": 1.0932609262554746e-07, "logits/chosen": 0.655882716178894, "logits/rejected": 0.6669564247131348, "logps/chosen": -207.04684448242188, "logps/ref_chosen": -57.16731643676758, "logps/ref_rejected": -53.30917739868164, "logps/rejected": -258.53466796875, "loss": 1.2018, "margin_dpo/margin_mean": 55.34593963623047, "margin_dpo/margin_std": 114.13899230957031, "step": 478 }, { "KL/chosen_KL_mean": -154.94833374023438, "KL/mean": -179.95150756835938, "KL/rejected_KL_mean": -204.95465087890625, "KL/std": 93.98583984375, "epoch": 0.7241118669690099, "fcm_dpo/beta": 0.005500415340065956, "fcm_dpo/delta": 0.12866447865962982, "fcm_dpo/margin": 50.00631332397461, "fcm_dpo/q_t": 0.4381939172744751, "grad_norm": 15.030978202819824, "learning_rate": 1.0823503403430734e-07, "logits/chosen": 0.6659060716629028, "logits/rejected": 0.627450704574585, "logps/chosen": -213.86166381835938, "logps/ref_chosen": -58.91331481933594, "logps/ref_rejected": -63.7403450012207, "logps/rejected": -268.69500732421875, "loss": 1.2205, "margin_dpo/margin_mean": 50.006309509277344, "margin_dpo/margin_std": 111.1731185913086, "step": 479 }, { "KL/chosen_KL_mean": -159.55169677734375, "KL/mean": -198.3495330810547, "KL/rejected_KL_mean": -237.14736938476562, "KL/std": 99.51255798339844, "epoch": 0.7256235827664399, "fcm_dpo/beta": 0.005575342103838921, "fcm_dpo/delta": -0.035166725516319275, "fcm_dpo/margin": 77.59567260742188, "fcm_dpo/q_t": 0.40224170684814453, "grad_norm": 15.844165802001953, "learning_rate": 1.0714794091391072e-07, "logits/chosen": 0.6764963269233704, "logits/rejected": 0.6669754981994629, "logps/chosen": -222.352294921875, "logps/ref_chosen": -62.80061340332031, "logps/ref_rejected": -67.58859252929688, "logps/rejected": -304.7359619140625, "loss": 1.1007, "margin_dpo/margin_mean": 77.59567260742188, "margin_dpo/margin_std": 114.90257263183594, "step": 480 }, { "KL/chosen_KL_mean": -156.21351623535156, "KL/mean": -191.5591583251953, "KL/rejected_KL_mean": -226.90480041503906, "KL/std": 102.61424255371094, "epoch": 0.72713529856387, "fcm_dpo/beta": 0.005516592413187027, "fcm_dpo/delta": 0.010359540581703186, "fcm_dpo/margin": 70.6912841796875, "fcm_dpo/q_t": 0.41247764229774475, "grad_norm": 14.178512573242188, "learning_rate": 1.0606484367268906e-07, "logits/chosen": 0.6627799272537231, "logits/rejected": 0.6644883751869202, "logps/chosen": -221.5, "logps/ref_chosen": -65.28649139404297, "logps/ref_rejected": -70.78668212890625, "logps/rejected": -297.69146728515625, "loss": 1.1258, "margin_dpo/margin_mean": 70.69129180908203, "margin_dpo/margin_std": 114.8702392578125, "step": 481 }, { "KL/chosen_KL_mean": -175.76876831054688, "KL/mean": -212.71697998046875, "KL/rejected_KL_mean": -249.66519165039062, "KL/std": 110.07460021972656, "epoch": 0.7286470143613001, "fcm_dpo/beta": 0.005537115037441254, "fcm_dpo/delta": -0.009665248915553093, "fcm_dpo/margin": 73.89642333984375, "fcm_dpo/q_t": 0.41188326478004456, "grad_norm": 16.9229679107666, "learning_rate": 1.0498577260720048e-07, "logits/chosen": 0.6482515335083008, "logits/rejected": 0.4943665862083435, "logps/chosen": -236.67495727539062, "logps/ref_chosen": -60.906185150146484, "logps/ref_rejected": -103.44656372070312, "logps/rejected": -353.11175537109375, "loss": 1.1497, "margin_dpo/margin_mean": 73.89642333984375, "margin_dpo/margin_std": 137.73287963867188, "step": 482 }, { "KL/chosen_KL_mean": -147.17604064941406, "KL/mean": -191.10537719726562, "KL/rejected_KL_mean": -235.03472900390625, "KL/std": 99.93904113769531, "epoch": 0.7301587301587301, "fcm_dpo/beta": 0.005463513545691967, "fcm_dpo/delta": -0.08396557718515396, "fcm_dpo/margin": 87.85865783691406, "fcm_dpo/q_t": 0.39078426361083984, "grad_norm": 12.708328247070312, "learning_rate": 1.0391075790138232e-07, "logits/chosen": 0.7652244567871094, "logits/rejected": 0.6555371284484863, "logps/chosen": -200.36805725097656, "logps/ref_chosen": -53.192012786865234, "logps/ref_rejected": -81.83927154541016, "logps/rejected": -316.8739929199219, "loss": 1.0521, "margin_dpo/margin_mean": 87.85865783691406, "margin_dpo/margin_std": 113.67207336425781, "step": 483 }, { "KL/chosen_KL_mean": -154.01959228515625, "KL/mean": -181.98313903808594, "KL/rejected_KL_mean": -209.94671630859375, "KL/std": 97.48323059082031, "epoch": 0.7316704459561603, "fcm_dpo/beta": 0.005525670014321804, "fcm_dpo/delta": 0.09322261810302734, "fcm_dpo/margin": 55.9271240234375, "fcm_dpo/q_t": 0.42791566252708435, "grad_norm": 17.04726219177246, "learning_rate": 1.0283982962570681e-07, "logits/chosen": 0.772758960723877, "logits/rejected": 0.7413250207901001, "logps/chosen": -211.78904724121094, "logps/ref_chosen": -57.76945877075195, "logps/ref_rejected": -71.6829833984375, "logps/rejected": -281.62969970703125, "loss": 1.1562, "margin_dpo/margin_mean": 55.9271240234375, "margin_dpo/margin_std": 86.11868286132812, "step": 484 }, { "KL/chosen_KL_mean": -156.07125854492188, "KL/mean": -188.55908203125, "KL/rejected_KL_mean": -221.0469207763672, "KL/std": 99.03124237060547, "epoch": 0.7331821617535903, "fcm_dpo/beta": 0.005521639250218868, "fcm_dpo/delta": 0.04180263727903366, "fcm_dpo/margin": 64.97565460205078, "fcm_dpo/q_t": 0.41826534271240234, "grad_norm": 12.852514266967773, "learning_rate": 1.0177301773633992e-07, "logits/chosen": 0.733189582824707, "logits/rejected": 0.7117427587509155, "logps/chosen": -212.7071075439453, "logps/ref_chosen": -56.63584899902344, "logps/ref_rejected": -70.85614013671875, "logps/rejected": -291.9030456542969, "loss": 1.1325, "margin_dpo/margin_mean": 64.97565460205078, "margin_dpo/margin_std": 96.84291076660156, "step": 485 }, { "KL/chosen_KL_mean": -181.23599243164062, "KL/mean": -211.62835693359375, "KL/rejected_KL_mean": -242.020751953125, "KL/std": 114.32650756835938, "epoch": 0.7346938775510204, "fcm_dpo/beta": 0.00562657555565238, "fcm_dpo/delta": 0.060053564608097076, "fcm_dpo/margin": 60.78474426269531, "fcm_dpo/q_t": 0.42665895819664, "grad_norm": 11.995468139648438, "learning_rate": 1.007103520743035e-07, "logits/chosen": 0.7314083576202393, "logits/rejected": 0.6161884069442749, "logps/chosen": -237.5830078125, "logps/ref_chosen": -56.347023010253906, "logps/ref_rejected": -85.97221374511719, "logps/rejected": -327.9929504394531, "loss": 1.2017, "margin_dpo/margin_mean": 60.78474426269531, "margin_dpo/margin_std": 133.8424072265625, "step": 486 }, { "KL/chosen_KL_mean": -162.27162170410156, "KL/mean": -199.17288208007812, "KL/rejected_KL_mean": -236.07412719726562, "KL/std": 104.9486083984375, "epoch": 0.7362055933484505, "fcm_dpo/beta": 0.005627620965242386, "fcm_dpo/delta": -0.016016241163015366, "fcm_dpo/margin": 73.8025131225586, "fcm_dpo/q_t": 0.4103911519050598, "grad_norm": 14.136743545532227, "learning_rate": 9.965186236464046e-08, "logits/chosen": 0.8066281676292419, "logits/rejected": 0.7472410202026367, "logps/chosen": -222.8888397216797, "logps/ref_chosen": -60.617218017578125, "logps/ref_rejected": -82.50975036621094, "logps/rejected": -318.5838623046875, "loss": 1.1205, "margin_dpo/margin_mean": 73.80250549316406, "margin_dpo/margin_std": 125.0999755859375, "step": 487 }, { "KL/chosen_KL_mean": -152.14353942871094, "KL/mean": -192.02452087402344, "KL/rejected_KL_mean": -231.905517578125, "KL/std": 99.62962341308594, "epoch": 0.7377173091458806, "fcm_dpo/beta": 0.005611489526927471, "fcm_dpo/delta": -0.05003175511956215, "fcm_dpo/margin": 79.76195526123047, "fcm_dpo/q_t": 0.39986562728881836, "grad_norm": 16.106365203857422, "learning_rate": 9.859757821558337e-08, "logits/chosen": 0.7334238290786743, "logits/rejected": 0.6768772006034851, "logps/chosen": -215.25259399414062, "logps/ref_chosen": -63.10905075073242, "logps/ref_rejected": -82.49348449707031, "logps/rejected": -314.39898681640625, "loss": 1.0765, "margin_dpo/margin_mean": 79.76195526123047, "margin_dpo/margin_std": 109.84991455078125, "step": 488 }, { "KL/chosen_KL_mean": -178.01058959960938, "KL/mean": -202.5729522705078, "KL/rejected_KL_mean": -227.13528442382812, "KL/std": 110.450927734375, "epoch": 0.7392290249433107, "fcm_dpo/beta": 0.005673965439200401, "fcm_dpo/delta": 0.12477852404117584, "fcm_dpo/margin": 49.12468719482422, "fcm_dpo/q_t": 0.4373658299446106, "grad_norm": 14.4613618850708, "learning_rate": 9.754752911772615e-08, "logits/chosen": 0.6765942573547363, "logits/rejected": 0.6304404735565186, "logps/chosen": -242.9995574951172, "logps/ref_chosen": -64.98896026611328, "logps/ref_rejected": -84.39607238769531, "logps/rejected": -311.5313720703125, "loss": 1.2361, "margin_dpo/margin_mean": 49.12468719482422, "margin_dpo/margin_std": 119.76895141601562, "step": 489 }, { "KL/chosen_KL_mean": -160.54306030273438, "KL/mean": -189.7637481689453, "KL/rejected_KL_mean": -218.9844512939453, "KL/std": 106.82121276855469, "epoch": 0.7407407407407407, "fcm_dpo/beta": 0.005743181798607111, "fcm_dpo/delta": 0.06649347394704819, "fcm_dpo/margin": 58.441383361816406, "fcm_dpo/q_t": 0.4260715842247009, "grad_norm": 12.848529815673828, "learning_rate": 9.650174444319956e-08, "logits/chosen": 0.755155086517334, "logits/rejected": 0.7366271018981934, "logps/chosen": -222.45181274414062, "logps/ref_chosen": -61.90874481201172, "logps/ref_rejected": -70.58566284179688, "logps/rejected": -289.57012939453125, "loss": 1.217, "margin_dpo/margin_mean": 58.441383361816406, "margin_dpo/margin_std": 134.05650329589844, "step": 490 }, { "KL/chosen_KL_mean": -153.87290954589844, "KL/mean": -185.72323608398438, "KL/rejected_KL_mean": -217.57354736328125, "KL/std": 97.37249755859375, "epoch": 0.7422524565381708, "fcm_dpo/beta": 0.0057761832140386105, "fcm_dpo/delta": 0.03252832964062691, "fcm_dpo/margin": 63.70063400268555, "fcm_dpo/q_t": 0.4171232581138611, "grad_norm": 13.742348670959473, "learning_rate": 9.546025344484868e-08, "logits/chosen": 0.6527000069618225, "logits/rejected": 0.5926094055175781, "logps/chosen": -209.34860229492188, "logps/ref_chosen": -55.47570037841797, "logps/ref_rejected": -78.70318603515625, "logps/rejected": -296.2767333984375, "loss": 1.1383, "margin_dpo/margin_mean": 63.70063781738281, "margin_dpo/margin_std": 103.06165313720703, "step": 491 }, { "KL/chosen_KL_mean": -182.7808837890625, "KL/mean": -209.8721923828125, "KL/rejected_KL_mean": -236.9635009765625, "KL/std": 109.27008819580078, "epoch": 0.7437641723356009, "fcm_dpo/beta": 0.005924704950302839, "fcm_dpo/delta": 0.080745629966259, "fcm_dpo/margin": 54.182613372802734, "fcm_dpo/q_t": 0.4289284348487854, "grad_norm": 16.213777542114258, "learning_rate": 9.442308525541589e-08, "logits/chosen": 0.7147728800773621, "logits/rejected": 0.6443384885787964, "logps/chosen": -250.06727600097656, "logps/ref_chosen": -67.28638458251953, "logps/ref_rejected": -82.78628540039062, "logps/rejected": -319.74981689453125, "loss": 1.2201, "margin_dpo/margin_mean": 54.182613372802734, "margin_dpo/margin_std": 122.61883544921875, "step": 492 }, { "KL/chosen_KL_mean": -149.8064727783203, "KL/mean": -190.75125122070312, "KL/rejected_KL_mean": -231.696044921875, "KL/std": 106.93426513671875, "epoch": 0.745275888133031, "fcm_dpo/beta": 0.005893126130104065, "fcm_dpo/delta": -0.08745664358139038, "fcm_dpo/margin": 81.88956451416016, "fcm_dpo/q_t": 0.3927948474884033, "grad_norm": 14.668304443359375, "learning_rate": 9.339026888672468e-08, "logits/chosen": 0.7042903304100037, "logits/rejected": 0.6280442476272583, "logps/chosen": -205.73397827148438, "logps/ref_chosen": -55.92750549316406, "logps/ref_rejected": -79.12149810791016, "logps/rejected": -310.8175354003906, "loss": 1.0729, "margin_dpo/margin_mean": 81.88956451416016, "margin_dpo/margin_std": 115.11085510253906, "step": 493 }, { "KL/chosen_KL_mean": -155.3618621826172, "KL/mean": -190.2035369873047, "KL/rejected_KL_mean": -225.04522705078125, "KL/std": 106.14227294921875, "epoch": 0.7467876039304611, "fcm_dpo/beta": 0.005800245329737663, "fcm_dpo/delta": -0.004481932148337364, "fcm_dpo/margin": 69.683349609375, "fcm_dpo/q_t": 0.41056621074676514, "grad_norm": 15.236351013183594, "learning_rate": 9.236183322886945e-08, "logits/chosen": 0.623336672782898, "logits/rejected": 0.572545051574707, "logps/chosen": -223.31597900390625, "logps/ref_chosen": -67.95410919189453, "logps/ref_rejected": -90.50865173339844, "logps/rejected": -315.55389404296875, "loss": 1.154, "margin_dpo/margin_mean": 69.683349609375, "margin_dpo/margin_std": 131.35519409179688, "step": 494 }, { "KL/chosen_KL_mean": -153.0796661376953, "KL/mean": -181.11660766601562, "KL/rejected_KL_mean": -209.15353393554688, "KL/std": 104.95536804199219, "epoch": 0.7482993197278912, "fcm_dpo/beta": 0.005901433527469635, "fcm_dpo/delta": 0.0706893801689148, "fcm_dpo/margin": 56.073883056640625, "fcm_dpo/q_t": 0.43131160736083984, "grad_norm": 17.017921447753906, "learning_rate": 9.133780704940594e-08, "logits/chosen": 0.7580794095993042, "logits/rejected": 0.6966645121574402, "logps/chosen": -205.7051239013672, "logps/ref_chosen": -52.62546157836914, "logps/ref_rejected": -72.06781005859375, "logps/rejected": -281.2213439941406, "loss": 1.2098, "margin_dpo/margin_mean": 56.073883056640625, "margin_dpo/margin_std": 128.44505310058594, "step": 495 }, { "KL/chosen_KL_mean": -167.5390625, "KL/mean": -206.25115966796875, "KL/rejected_KL_mean": -244.96324157714844, "KL/std": 114.25956726074219, "epoch": 0.7498110355253212, "fcm_dpo/beta": 0.005806931294500828, "fcm_dpo/delta": -0.05337368696928024, "fcm_dpo/margin": 77.4241943359375, "fcm_dpo/q_t": 0.4092605710029602, "grad_norm": 13.090649604797363, "learning_rate": 9.031821899254797e-08, "logits/chosen": 0.7403356432914734, "logits/rejected": 0.6223543882369995, "logps/chosen": -225.13638305664062, "logps/ref_chosen": -57.597320556640625, "logps/ref_rejected": -94.36127471923828, "logps/rejected": -339.32452392578125, "loss": 1.1346, "margin_dpo/margin_mean": 77.4241943359375, "margin_dpo/margin_std": 145.4561004638672, "step": 496 }, { "KL/chosen_KL_mean": -168.29544067382812, "KL/mean": -207.5030975341797, "KL/rejected_KL_mean": -246.71075439453125, "KL/std": 105.35246276855469, "epoch": 0.7513227513227513, "fcm_dpo/beta": 0.005777341779321432, "fcm_dpo/delta": -0.05565295368432999, "fcm_dpo/margin": 78.41531372070312, "fcm_dpo/q_t": 0.3961220383644104, "grad_norm": 13.204294204711914, "learning_rate": 8.930309757836516e-08, "logits/chosen": 0.7045537829399109, "logits/rejected": 0.6701431274414062, "logps/chosen": -241.08538818359375, "logps/ref_chosen": -72.78994750976562, "logps/ref_rejected": -89.48483276367188, "logps/rejected": -336.1955871582031, "loss": 1.0859, "margin_dpo/margin_mean": 78.41531372070312, "margin_dpo/margin_std": 114.64138793945312, "step": 497 }, { "KL/chosen_KL_mean": -151.4757537841797, "KL/mean": -191.3174591064453, "KL/rejected_KL_mean": -231.15916442871094, "KL/std": 99.70770263671875, "epoch": 0.7528344671201814, "fcm_dpo/beta": 0.005714482627809048, "fcm_dpo/delta": -0.05804353579878807, "fcm_dpo/margin": 79.68342590332031, "fcm_dpo/q_t": 0.39497846364974976, "grad_norm": 14.724855422973633, "learning_rate": 8.829247120198563e-08, "logits/chosen": 0.6854042410850525, "logits/rejected": 0.6628051996231079, "logps/chosen": -219.84146118164062, "logps/ref_chosen": -68.36572265625, "logps/ref_rejected": -71.28846740722656, "logps/rejected": -302.4476318359375, "loss": 1.068, "margin_dpo/margin_mean": 79.68342590332031, "margin_dpo/margin_std": 106.51513671875, "step": 498 }, { "KL/chosen_KL_mean": -151.89300537109375, "KL/mean": -192.9249267578125, "KL/rejected_KL_mean": -233.95684814453125, "KL/std": 109.68285369873047, "epoch": 0.7543461829176115, "fcm_dpo/beta": 0.00564592145383358, "fcm_dpo/delta": -0.06639964133501053, "fcm_dpo/margin": 82.06385040283203, "fcm_dpo/q_t": 0.399397611618042, "grad_norm": 18.094348907470703, "learning_rate": 8.728636813280163e-08, "logits/chosen": 0.7410115003585815, "logits/rejected": 0.672810435295105, "logps/chosen": -213.80181884765625, "logps/ref_chosen": -61.90882873535156, "logps/ref_rejected": -91.9411392211914, "logps/rejected": -325.8979797363281, "loss": 1.1248, "margin_dpo/margin_mean": 82.0638427734375, "margin_dpo/margin_std": 142.8561553955078, "step": 499 }, { "KL/chosen_KL_mean": -158.30862426757812, "KL/mean": -192.1842041015625, "KL/rejected_KL_mean": -226.05978393554688, "KL/std": 97.90144348144531, "epoch": 0.7558578987150416, "fcm_dpo/beta": 0.005628244951367378, "fcm_dpo/delta": 0.019335078075528145, "fcm_dpo/margin": 67.75115966796875, "fcm_dpo/q_t": 0.4112476706504822, "grad_norm": 15.212784767150879, "learning_rate": 8.628481651367875e-08, "logits/chosen": 0.6629618406295776, "logits/rejected": 0.6684551239013672, "logps/chosen": -228.53445434570312, "logps/ref_chosen": -70.225830078125, "logps/ref_rejected": -71.72203063964844, "logps/rejected": -297.78179931640625, "loss": 1.1625, "margin_dpo/margin_mean": 67.75115966796875, "margin_dpo/margin_std": 127.54959869384766, "step": 500 }, { "KL/chosen_KL_mean": -159.2582550048828, "KL/mean": -189.79246520996094, "KL/rejected_KL_mean": -220.32667541503906, "KL/std": 104.34779357910156, "epoch": 0.7573696145124716, "fcm_dpo/beta": 0.005706685595214367, "fcm_dpo/delta": 0.05312148109078407, "fcm_dpo/margin": 61.06843185424805, "fcm_dpo/q_t": 0.4191363751888275, "grad_norm": 11.982666015625, "learning_rate": 8.528784436016878e-08, "logits/chosen": 0.7013833522796631, "logits/rejected": 0.7047612071037292, "logps/chosen": -223.8570556640625, "logps/ref_chosen": -64.59880828857422, "logps/ref_rejected": -70.59329223632812, "logps/rejected": -290.91998291015625, "loss": 1.1253, "margin_dpo/margin_mean": 61.06842803955078, "margin_dpo/margin_std": 85.38192749023438, "step": 501 }, { "KL/chosen_KL_mean": -159.037109375, "KL/mean": -192.675537109375, "KL/rejected_KL_mean": -226.31393432617188, "KL/std": 110.60833740234375, "epoch": 0.7588813303099018, "fcm_dpo/beta": 0.005739234387874603, "fcm_dpo/delta": 0.014180617406964302, "fcm_dpo/margin": 67.27684020996094, "fcm_dpo/q_t": 0.4120279550552368, "grad_norm": 17.2750301361084, "learning_rate": 8.4295479559726e-08, "logits/chosen": 0.7115000486373901, "logits/rejected": 0.6628080606460571, "logps/chosen": -224.5037384033203, "logps/ref_chosen": -65.46662902832031, "logps/ref_rejected": -90.22233581542969, "logps/rejected": -316.5362548828125, "loss": 1.1198, "margin_dpo/margin_mean": 67.27684020996094, "margin_dpo/margin_std": 103.1072769165039, "step": 502 }, { "KL/chosen_KL_mean": -142.4674072265625, "KL/mean": -176.8408203125, "KL/rejected_KL_mean": -211.2142333984375, "KL/std": 96.57347106933594, "epoch": 0.7603930461073318, "fcm_dpo/beta": 0.0057237460277974606, "fcm_dpo/delta": 0.0067562200129032135, "fcm_dpo/margin": 68.746826171875, "fcm_dpo/q_t": 0.4091821312904358, "grad_norm": 12.451175689697266, "learning_rate": 8.330774987092712e-08, "logits/chosen": 0.6891748905181885, "logits/rejected": 0.6937826871871948, "logps/chosen": -194.3021697998047, "logps/ref_chosen": -51.83476257324219, "logps/ref_rejected": -57.62522506713867, "logps/rejected": -268.8394775390625, "loss": 1.1277, "margin_dpo/margin_mean": 68.746826171875, "margin_dpo/margin_std": 110.9075698852539, "step": 503 }, { "KL/chosen_KL_mean": -147.31979370117188, "KL/mean": -192.40731811523438, "KL/rejected_KL_mean": -237.49481201171875, "KL/std": 96.30535888671875, "epoch": 0.7619047619047619, "fcm_dpo/beta": 0.005670842714607716, "fcm_dpo/delta": -0.11748392879962921, "fcm_dpo/margin": 90.17503356933594, "fcm_dpo/q_t": 0.38196149468421936, "grad_norm": 13.738269805908203, "learning_rate": 8.232468292269479e-08, "logits/chosen": 0.7036927342414856, "logits/rejected": 0.6851463317871094, "logps/chosen": -215.97097778320312, "logps/ref_chosen": -68.65119934082031, "logps/ref_rejected": -77.91394805908203, "logps/rejected": -315.40875244140625, "loss": 1.0149, "margin_dpo/margin_mean": 90.17504119873047, "margin_dpo/margin_std": 97.8731689453125, "step": 504 }, { "KL/chosen_KL_mean": -155.90220642089844, "KL/mean": -186.30569458007812, "KL/rejected_KL_mean": -216.70916748046875, "KL/std": 111.57499694824219, "epoch": 0.763416477702192, "fcm_dpo/beta": 0.005609945394098759, "fcm_dpo/delta": 0.060340505093336105, "fcm_dpo/margin": 60.806968688964844, "fcm_dpo/q_t": 0.42532825469970703, "grad_norm": 13.912813186645508, "learning_rate": 8.134630621352483e-08, "logits/chosen": 0.7275354266166687, "logits/rejected": 0.6875624656677246, "logps/chosen": -215.90106201171875, "logps/ref_chosen": -59.99884796142578, "logps/ref_rejected": -76.88048553466797, "logps/rejected": -293.58966064453125, "loss": 1.1904, "margin_dpo/margin_mean": 60.806968688964844, "margin_dpo/margin_std": 123.60678100585938, "step": 505 }, { "KL/chosen_KL_mean": -154.84149169921875, "KL/mean": -187.8001708984375, "KL/rejected_KL_mean": -220.75881958007812, "KL/std": 99.20973205566406, "epoch": 0.764928193499622, "fcm_dpo/beta": 0.005677978508174419, "fcm_dpo/delta": 0.026687901467084885, "fcm_dpo/margin": 65.91732025146484, "fcm_dpo/q_t": 0.4146183431148529, "grad_norm": 14.703615188598633, "learning_rate": 8.037264711071698e-08, "logits/chosen": 0.6911933422088623, "logits/rejected": 0.6756948232650757, "logps/chosen": -224.9127960205078, "logps/ref_chosen": -70.07130432128906, "logps/ref_rejected": -82.03775024414062, "logps/rejected": -302.79656982421875, "loss": 1.1628, "margin_dpo/margin_mean": 65.91732025146484, "margin_dpo/margin_std": 123.7665023803711, "step": 506 }, { "KL/chosen_KL_mean": -166.98780822753906, "KL/mean": -203.90573120117188, "KL/rejected_KL_mean": -240.82363891601562, "KL/std": 111.98727416992188, "epoch": 0.7664399092970522, "fcm_dpo/beta": 0.0056532504968345165, "fcm_dpo/delta": -0.01878700777888298, "fcm_dpo/margin": 73.83582305908203, "fcm_dpo/q_t": 0.41168636083602905, "grad_norm": 14.66965389251709, "learning_rate": 7.940373284960933e-08, "logits/chosen": 0.6715726852416992, "logits/rejected": 0.6238397359848022, "logps/chosen": -238.99484252929688, "logps/ref_chosen": -72.00703430175781, "logps/ref_rejected": -93.94987487792969, "logps/rejected": -334.77349853515625, "loss": 1.1429, "margin_dpo/margin_mean": 73.83582305908203, "margin_dpo/margin_std": 132.9780731201172, "step": 507 }, { "KL/chosen_KL_mean": -153.32525634765625, "KL/mean": -196.36053466796875, "KL/rejected_KL_mean": -239.39578247070312, "KL/std": 116.38261413574219, "epoch": 0.7679516250944822, "fcm_dpo/beta": 0.005641060881316662, "fcm_dpo/delta": -0.09029045701026917, "fcm_dpo/margin": 86.07052612304688, "fcm_dpo/q_t": 0.3928912281990051, "grad_norm": 16.511768341064453, "learning_rate": 7.843959053281663e-08, "logits/chosen": 0.660283625125885, "logits/rejected": 0.5391038656234741, "logps/chosen": -213.54518127441406, "logps/ref_chosen": -60.21992492675781, "logps/ref_rejected": -95.9200668334961, "logps/rejected": -335.31585693359375, "loss": 1.0785, "margin_dpo/margin_mean": 86.07052612304688, "margin_dpo/margin_std": 125.32095336914062, "step": 508 }, { "KL/chosen_KL_mean": -163.14154052734375, "KL/mean": -195.06103515625, "KL/rejected_KL_mean": -226.98052978515625, "KL/std": 103.42306518554688, "epoch": 0.7694633408919124, "fcm_dpo/beta": 0.0055953748524188995, "fcm_dpo/delta": 0.04433656483888626, "fcm_dpo/margin": 63.8390007019043, "fcm_dpo/q_t": 0.41883113980293274, "grad_norm": 16.704587936401367, "learning_rate": 7.748024712947204e-08, "logits/chosen": 0.6394075155258179, "logits/rejected": 0.6166356205940247, "logps/chosen": -229.4116973876953, "logps/ref_chosen": -66.27017211914062, "logps/ref_rejected": -71.73065185546875, "logps/rejected": -298.711181640625, "loss": 1.1671, "margin_dpo/margin_mean": 63.83899688720703, "margin_dpo/margin_std": 119.53250122070312, "step": 509 }, { "KL/chosen_KL_mean": -161.60655212402344, "KL/mean": -203.13580322265625, "KL/rejected_KL_mean": -244.6650390625, "KL/std": 113.58114624023438, "epoch": 0.7709750566893424, "fcm_dpo/beta": 0.005545272957533598, "fcm_dpo/delta": -0.06424017250537872, "fcm_dpo/margin": 83.0584716796875, "fcm_dpo/q_t": 0.4023531377315521, "grad_norm": 14.228266716003418, "learning_rate": 7.652572947447272e-08, "logits/chosen": 0.7977691888809204, "logits/rejected": 0.7011072039604187, "logps/chosen": -215.15142822265625, "logps/ref_chosen": -53.54487609863281, "logps/ref_rejected": -91.36648559570312, "logps/rejected": -336.0315246582031, "loss": 1.1203, "margin_dpo/margin_mean": 83.0584716796875, "margin_dpo/margin_std": 144.74627685546875, "step": 510 }, { "KL/chosen_KL_mean": -152.7357177734375, "KL/mean": -199.5233154296875, "KL/rejected_KL_mean": -246.31092834472656, "KL/std": 102.5706787109375, "epoch": 0.7724867724867724, "fcm_dpo/beta": 0.005463335197418928, "fcm_dpo/delta": -0.11711655557155609, "fcm_dpo/margin": 93.5752182006836, "fcm_dpo/q_t": 0.38272392749786377, "grad_norm": 17.340545654296875, "learning_rate": 7.557606426772961e-08, "logits/chosen": 0.7229855060577393, "logits/rejected": 0.6641731262207031, "logps/chosen": -208.58009338378906, "logps/ref_chosen": -55.844383239746094, "logps/ref_rejected": -86.49819946289062, "logps/rejected": -332.80914306640625, "loss": 1.0249, "margin_dpo/margin_mean": 93.57521057128906, "margin_dpo/margin_std": 109.5538330078125, "step": 511 }, { "KL/chosen_KL_mean": -158.43212890625, "KL/mean": -187.49783325195312, "KL/rejected_KL_mean": -216.56353759765625, "KL/std": 95.57821655273438, "epoch": 0.7739984882842026, "fcm_dpo/beta": 0.005476620979607105, "fcm_dpo/delta": 0.08439172059297562, "fcm_dpo/margin": 58.131404876708984, "fcm_dpo/q_t": 0.4264548420906067, "grad_norm": 19.954294204711914, "learning_rate": 7.463127807341966e-08, "logits/chosen": 0.5845282673835754, "logits/rejected": 0.5833408236503601, "logps/chosen": -220.08517456054688, "logps/ref_chosen": -61.653038024902344, "logps/ref_rejected": -72.83148193359375, "logps/rejected": -289.39501953125, "loss": 1.1921, "margin_dpo/margin_mean": 58.131404876708984, "margin_dpo/margin_std": 116.42222595214844, "step": 512 }, { "KL/chosen_KL_mean": -141.0041046142578, "KL/mean": -178.875244140625, "KL/rejected_KL_mean": -216.7464141845703, "KL/std": 98.06171417236328, "epoch": 0.7755102040816326, "fcm_dpo/beta": 0.005491101182997227, "fcm_dpo/delta": -0.016725288704037666, "fcm_dpo/margin": 75.74229431152344, "fcm_dpo/q_t": 0.40563303232192993, "grad_norm": 11.389507293701172, "learning_rate": 7.369139731924401e-08, "logits/chosen": 0.8551111221313477, "logits/rejected": 0.7970899343490601, "logps/chosen": -191.85665893554688, "logps/ref_chosen": -50.85256576538086, "logps/ref_rejected": -69.21754455566406, "logps/rejected": -285.9639587402344, "loss": 1.0901, "margin_dpo/margin_mean": 75.74230194091797, "margin_dpo/margin_std": 104.52362060546875, "step": 513 }, { "KL/chosen_KL_mean": -157.754150390625, "KL/mean": -200.53826904296875, "KL/rejected_KL_mean": -243.3223876953125, "KL/std": 106.9458999633789, "epoch": 0.7770219198790628, "fcm_dpo/beta": 0.005463584326207638, "fcm_dpo/delta": -0.0708877444267273, "fcm_dpo/margin": 85.56825256347656, "fcm_dpo/q_t": 0.3941824734210968, "grad_norm": 15.855218887329102, "learning_rate": 7.275644829568747e-08, "logits/chosen": 0.7480206489562988, "logits/rejected": 0.7150290608406067, "logps/chosen": -227.13906860351562, "logps/ref_chosen": -69.38493347167969, "logps/ref_rejected": -83.32447814941406, "logps/rejected": -326.6468505859375, "loss": 1.0743, "margin_dpo/margin_mean": 85.56825256347656, "margin_dpo/margin_std": 121.04784393310547, "step": 514 }, { "KL/chosen_KL_mean": -165.98629760742188, "KL/mean": -199.73580932617188, "KL/rejected_KL_mean": -233.4853515625, "KL/std": 100.55045318603516, "epoch": 0.7785336356764928, "fcm_dpo/beta": 0.005442567635327578, "fcm_dpo/delta": 0.033889930695295334, "fcm_dpo/margin": 67.4990234375, "fcm_dpo/q_t": 0.41551756858825684, "grad_norm": 14.63206672668457, "learning_rate": 7.182645715528435e-08, "logits/chosen": 0.7209761142730713, "logits/rejected": 0.6388729810714722, "logps/chosen": -219.67333984375, "logps/ref_chosen": -53.687034606933594, "logps/ref_rejected": -83.59614562988281, "logps/rejected": -317.08148193359375, "loss": 1.1453, "margin_dpo/margin_mean": 67.4990234375, "margin_dpo/margin_std": 115.27774047851562, "step": 515 }, { "KL/chosen_KL_mean": -143.6598358154297, "KL/mean": -175.19981384277344, "KL/rejected_KL_mean": -206.7397918701172, "KL/std": 99.82717895507812, "epoch": 0.780045351473923, "fcm_dpo/beta": 0.005517736077308655, "fcm_dpo/delta": 0.05346924811601639, "fcm_dpo/margin": 63.079952239990234, "fcm_dpo/q_t": 0.41942089796066284, "grad_norm": 15.73136043548584, "learning_rate": 7.090144991188568e-08, "logits/chosen": 0.710132896900177, "logits/rejected": 0.6714646816253662, "logps/chosen": -200.56155395507812, "logps/ref_chosen": -56.9017219543457, "logps/ref_rejected": -67.83477783203125, "logps/rejected": -274.5745849609375, "loss": 1.1707, "margin_dpo/margin_mean": 63.079952239990234, "margin_dpo/margin_std": 118.81394958496094, "step": 516 }, { "KL/chosen_KL_mean": -171.2911376953125, "KL/mean": -196.329345703125, "KL/rejected_KL_mean": -221.36752319335938, "KL/std": 105.70317077636719, "epoch": 0.781557067271353, "fcm_dpo/beta": 0.005636849440634251, "fcm_dpo/delta": 0.1204327642917633, "fcm_dpo/margin": 50.076385498046875, "fcm_dpo/q_t": 0.43720489740371704, "grad_norm": 15.20931625366211, "learning_rate": 6.998145243993284e-08, "logits/chosen": 0.7693038582801819, "logits/rejected": 0.7707263827323914, "logps/chosen": -233.0662841796875, "logps/ref_chosen": -61.775142669677734, "logps/ref_rejected": -62.88270950317383, "logps/rejected": -284.250244140625, "loss": 1.2151, "margin_dpo/margin_mean": 50.076385498046875, "margin_dpo/margin_std": 110.11534118652344, "step": 517 }, { "KL/chosen_KL_mean": -145.9417724609375, "KL/mean": -181.2034912109375, "KL/rejected_KL_mean": -216.46519470214844, "KL/std": 102.95506286621094, "epoch": 0.783068783068783, "fcm_dpo/beta": 0.005652305670082569, "fcm_dpo/delta": 0.0014371161814779043, "fcm_dpo/margin": 70.5234146118164, "fcm_dpo/q_t": 0.4121934175491333, "grad_norm": 12.902327537536621, "learning_rate": 6.906649047373245e-08, "logits/chosen": 0.6915863752365112, "logits/rejected": 0.6487727165222168, "logps/chosen": -207.96701049804688, "logps/ref_chosen": -62.02523422241211, "logps/ref_rejected": -79.06085205078125, "logps/rejected": -295.52606201171875, "loss": 1.1253, "margin_dpo/margin_mean": 70.52342224121094, "margin_dpo/margin_std": 116.34848022460938, "step": 518 }, { "KL/chosen_KL_mean": -176.3029022216797, "KL/mean": -196.67068481445312, "KL/rejected_KL_mean": -217.03846740722656, "KL/std": 105.22140502929688, "epoch": 0.7845804988662132, "fcm_dpo/beta": 0.005807607434689999, "fcm_dpo/delta": 0.16692955791950226, "fcm_dpo/margin": 40.735565185546875, "fcm_dpo/q_t": 0.4469537138938904, "grad_norm": 22.4091854095459, "learning_rate": 6.815658960673781e-08, "logits/chosen": 0.7130922675132751, "logits/rejected": 0.6659531593322754, "logps/chosen": -237.90927124023438, "logps/ref_chosen": -61.60636901855469, "logps/ref_rejected": -74.50727844238281, "logps/rejected": -291.5457458496094, "loss": 1.3159, "margin_dpo/margin_mean": 40.735565185546875, "margin_dpo/margin_std": 138.74868774414062, "step": 519 }, { "KL/chosen_KL_mean": -162.3382110595703, "KL/mean": -193.408447265625, "KL/rejected_KL_mean": -224.47869873046875, "KL/std": 105.3907470703125, "epoch": 0.7860922146636432, "fcm_dpo/beta": 0.005840703845024109, "fcm_dpo/delta": 0.03788800537586212, "fcm_dpo/margin": 62.14047622680664, "fcm_dpo/q_t": 0.4192347824573517, "grad_norm": 15.117642402648926, "learning_rate": 6.725177529083209e-08, "logits/chosen": 0.7774897813796997, "logits/rejected": 0.720142662525177, "logps/chosen": -225.21163940429688, "logps/ref_chosen": -62.87343215942383, "logps/ref_rejected": -76.505615234375, "logps/rejected": -300.98431396484375, "loss": 1.1514, "margin_dpo/margin_mean": 62.140480041503906, "margin_dpo/margin_std": 108.19155883789062, "step": 520 }, { "KL/chosen_KL_mean": -160.02740478515625, "KL/mean": -204.9112091064453, "KL/rejected_KL_mean": -249.79498291015625, "KL/std": 104.34115600585938, "epoch": 0.7876039304610734, "fcm_dpo/beta": 0.005790054798126221, "fcm_dpo/delta": -0.12616199254989624, "fcm_dpo/margin": 89.76756286621094, "fcm_dpo/q_t": 0.3843619227409363, "grad_norm": 12.815384864807129, "learning_rate": 6.63520728356167e-08, "logits/chosen": 0.5885381698608398, "logits/rejected": 0.5134579539299011, "logps/chosen": -224.23410034179688, "logps/ref_chosen": -64.20668029785156, "logps/ref_rejected": -92.28083038330078, "logps/rejected": -342.0758056640625, "loss": 1.0372, "margin_dpo/margin_mean": 89.76756286621094, "margin_dpo/margin_std": 118.39604187011719, "step": 521 }, { "KL/chosen_KL_mean": -160.57598876953125, "KL/mean": -187.0106658935547, "KL/rejected_KL_mean": -213.4453582763672, "KL/std": 104.2094955444336, "epoch": 0.7891156462585034, "fcm_dpo/beta": 0.005809691734611988, "fcm_dpo/delta": 0.09589952230453491, "fcm_dpo/margin": 52.869354248046875, "fcm_dpo/q_t": 0.43191057443618774, "grad_norm": 15.976807594299316, "learning_rate": 6.545750740770336e-08, "logits/chosen": 0.7014885544776917, "logits/rejected": 0.6930610537528992, "logps/chosen": -218.94570922851562, "logps/ref_chosen": -58.369720458984375, "logps/ref_rejected": -68.79248046875, "logps/rejected": -282.23785400390625, "loss": 1.2483, "margin_dpo/margin_mean": 52.869354248046875, "margin_dpo/margin_std": 136.038330078125, "step": 522 }, { "KL/chosen_KL_mean": -165.72467041015625, "KL/mean": -197.77651977539062, "KL/rejected_KL_mean": -229.82833862304688, "KL/std": 105.67681121826172, "epoch": 0.7906273620559335, "fcm_dpo/beta": 0.005870272405445576, "fcm_dpo/delta": 0.024627383798360825, "fcm_dpo/margin": 64.10369873046875, "fcm_dpo/q_t": 0.4130624532699585, "grad_norm": 17.706695556640625, "learning_rate": 6.456810403001012e-08, "logits/chosen": 0.6971518397331238, "logits/rejected": 0.5737546682357788, "logps/chosen": -231.43789672851562, "logps/ref_chosen": -65.71324157714844, "logps/ref_rejected": -91.98896789550781, "logps/rejected": -321.81732177734375, "loss": 1.1611, "margin_dpo/margin_mean": 64.10369873046875, "margin_dpo/margin_std": 120.08074188232422, "step": 523 }, { "KL/chosen_KL_mean": -143.75306701660156, "KL/mean": -177.9951171875, "KL/rejected_KL_mean": -212.2371826171875, "KL/std": 101.63809204101562, "epoch": 0.7921390778533636, "fcm_dpo/beta": 0.00589970126748085, "fcm_dpo/delta": -0.00446426123380661, "fcm_dpo/margin": 68.48410034179688, "fcm_dpo/q_t": 0.40784668922424316, "grad_norm": 15.014260292053223, "learning_rate": 6.368388758106134e-08, "logits/chosen": 0.6233705878257751, "logits/rejected": 0.5993573665618896, "logps/chosen": -220.10430908203125, "logps/ref_chosen": -76.35124969482422, "logps/ref_rejected": -89.96072387695312, "logps/rejected": -302.1979064941406, "loss": 1.1119, "margin_dpo/margin_mean": 68.48410034179688, "margin_dpo/margin_std": 104.5627212524414, "step": 524 }, { "KL/chosen_KL_mean": -163.55609130859375, "KL/mean": -190.743408203125, "KL/rejected_KL_mean": -217.9307098388672, "KL/std": 100.99457550048828, "epoch": 0.7936507936507936, "fcm_dpo/beta": 0.00594139751046896, "fcm_dpo/delta": 0.07956387102603912, "fcm_dpo/margin": 54.37462615966797, "fcm_dpo/q_t": 0.42932847142219543, "grad_norm": 19.195297241210938, "learning_rate": 6.280488279429185e-08, "logits/chosen": 0.5288229584693909, "logits/rejected": 0.5252171754837036, "logps/chosen": -239.0518798828125, "logps/ref_chosen": -75.49578857421875, "logps/ref_rejected": -84.04852294921875, "logps/rejected": -301.979248046875, "loss": 1.209, "margin_dpo/margin_mean": 54.37462615966797, "margin_dpo/margin_std": 121.72251892089844, "step": 525 }, { "KL/chosen_KL_mean": -170.69439697265625, "KL/mean": -195.64962768554688, "KL/rejected_KL_mean": -220.6048583984375, "KL/std": 103.0473403930664, "epoch": 0.7951625094482238, "fcm_dpo/beta": 0.006015341263264418, "fcm_dpo/delta": 0.10238677263259888, "fcm_dpo/margin": 49.910484313964844, "fcm_dpo/q_t": 0.4324687719345093, "grad_norm": 15.048983573913574, "learning_rate": 6.193111425735515e-08, "logits/chosen": 0.7008933424949646, "logits/rejected": 0.6313944458961487, "logps/chosen": -231.98681640625, "logps/ref_chosen": -61.29241943359375, "logps/ref_rejected": -82.47763061523438, "logps/rejected": -303.0824890136719, "loss": 1.211, "margin_dpo/margin_mean": 49.910484313964844, "margin_dpo/margin_std": 105.97373962402344, "step": 526 }, { "KL/chosen_KL_mean": -178.74819946289062, "KL/mean": -200.55226135253906, "KL/rejected_KL_mean": -222.35635375976562, "KL/std": 101.82447814941406, "epoch": 0.7966742252456538, "fcm_dpo/beta": 0.006244257092475891, "fcm_dpo/delta": 0.13016949594020844, "fcm_dpo/margin": 43.60816192626953, "fcm_dpo/q_t": 0.4404991865158081, "grad_norm": 17.9735050201416, "learning_rate": 6.106260641143546e-08, "logits/chosen": 0.7643716335296631, "logits/rejected": 0.6801888346672058, "logps/chosen": -240.2208251953125, "logps/ref_chosen": -61.472625732421875, "logps/ref_rejected": -90.52831268310547, "logps/rejected": -312.8846435546875, "loss": 1.2555, "margin_dpo/margin_mean": 43.60816192626953, "margin_dpo/margin_std": 115.52593994140625, "step": 527 }, { "KL/chosen_KL_mean": -157.19871520996094, "KL/mean": -182.855224609375, "KL/rejected_KL_mean": -208.51173400878906, "KL/std": 98.76786804199219, "epoch": 0.7981859410430839, "fcm_dpo/beta": 0.006307326257228851, "fcm_dpo/delta": 0.07893209159374237, "fcm_dpo/margin": 51.313026428222656, "fcm_dpo/q_t": 0.42957398295402527, "grad_norm": 18.043712615966797, "learning_rate": 6.019938355056422e-08, "logits/chosen": 0.6514958143234253, "logits/rejected": 0.5767842531204224, "logps/chosen": -215.99073791503906, "logps/ref_chosen": -58.792015075683594, "logps/ref_rejected": -71.82516479492188, "logps/rejected": -280.3369140625, "loss": 1.2402, "margin_dpo/margin_mean": 51.313026428222656, "margin_dpo/margin_std": 126.47222900390625, "step": 528 }, { "KL/chosen_KL_mean": -150.249267578125, "KL/mean": -200.34048461914062, "KL/rejected_KL_mean": -250.43173217773438, "KL/std": 97.72628021240234, "epoch": 0.799697656840514, "fcm_dpo/beta": 0.006136591546237469, "fcm_dpo/delta": -0.2290957272052765, "fcm_dpo/margin": 100.18245697021484, "fcm_dpo/q_t": 0.36025696992874146, "grad_norm": 17.034433364868164, "learning_rate": 5.934146982094049e-08, "logits/chosen": 0.6221810579299927, "logits/rejected": 0.5700123310089111, "logps/chosen": -205.3202362060547, "logps/ref_chosen": -55.070960998535156, "logps/ref_rejected": -75.44007873535156, "logps/rejected": -325.8717956542969, "loss": 0.9612, "margin_dpo/margin_mean": 100.18244934082031, "margin_dpo/margin_std": 106.74053192138672, "step": 529 }, { "KL/chosen_KL_mean": -152.41957092285156, "KL/mean": -184.40634155273438, "KL/rejected_KL_mean": -216.39312744140625, "KL/std": 104.05105590820312, "epoch": 0.8012093726379441, "fcm_dpo/beta": 0.006076472345739603, "fcm_dpo/delta": 0.011720176786184311, "fcm_dpo/margin": 63.97355270385742, "fcm_dpo/q_t": 0.41505563259124756, "grad_norm": 19.411815643310547, "learning_rate": 5.848888922025552e-08, "logits/chosen": 0.742800235748291, "logits/rejected": 0.6958895921707153, "logps/chosen": -209.16339111328125, "logps/ref_chosen": -56.743812561035156, "logps/ref_rejected": -76.6692123413086, "logps/rejected": -293.06231689453125, "loss": 1.1371, "margin_dpo/margin_mean": 63.97355651855469, "margin_dpo/margin_std": 108.51207733154297, "step": 530 }, { "KL/chosen_KL_mean": -151.02877807617188, "KL/mean": -184.73265075683594, "KL/rejected_KL_mean": -218.4365234375, "KL/std": 102.89830017089844, "epoch": 0.8027210884353742, "fcm_dpo/beta": 0.00608763936907053, "fcm_dpo/delta": -0.010867763310670853, "fcm_dpo/margin": 67.40776062011719, "fcm_dpo/q_t": 0.40898647904396057, "grad_norm": 15.34810733795166, "learning_rate": 5.7641665597021435e-08, "logits/chosen": 0.6947374939918518, "logits/rejected": 0.6132571697235107, "logps/chosen": -202.14523315429688, "logps/ref_chosen": -51.116455078125, "logps/ref_rejected": -79.52884674072266, "logps/rejected": -297.9653625488281, "loss": 1.1247, "margin_dpo/margin_mean": 67.40776824951172, "margin_dpo/margin_std": 111.83642578125, "step": 531 }, { "KL/chosen_KL_mean": -175.77328491210938, "KL/mean": -211.45449829101562, "KL/rejected_KL_mean": -247.13571166992188, "KL/std": 101.09457397460938, "epoch": 0.8042328042328042, "fcm_dpo/beta": 0.006043744273483753, "fcm_dpo/delta": -0.03269847482442856, "fcm_dpo/margin": 71.36245727539062, "fcm_dpo/q_t": 0.40316763520240784, "grad_norm": 15.20186710357666, "learning_rate": 5.679982264990424e-08, "logits/chosen": 0.6399841904640198, "logits/rejected": 0.5887705087661743, "logps/chosen": -234.05322265625, "logps/ref_chosen": -58.279945373535156, "logps/ref_rejected": -78.05426788330078, "logps/rejected": -325.19000244140625, "loss": 1.1102, "margin_dpo/margin_mean": 71.36245727539062, "margin_dpo/margin_std": 113.26620483398438, "step": 532 }, { "KL/chosen_KL_mean": -125.95477294921875, "KL/mean": -164.83905029296875, "KL/rejected_KL_mean": -203.7233123779297, "KL/std": 106.76790618896484, "epoch": 0.8057445200302343, "fcm_dpo/beta": 0.005985047202557325, "fcm_dpo/delta": -0.06859785318374634, "fcm_dpo/margin": 77.76854705810547, "fcm_dpo/q_t": 0.39635133743286133, "grad_norm": 14.38045883178711, "learning_rate": 5.596338392706076e-08, "logits/chosen": 0.8273634314537048, "logits/rejected": 0.7599341869354248, "logps/chosen": -182.37278747558594, "logps/ref_chosen": -56.41801071166992, "logps/ref_rejected": -73.89324951171875, "logps/rejected": -277.6165771484375, "loss": 1.0808, "margin_dpo/margin_mean": 77.76854705810547, "margin_dpo/margin_std": 113.24247741699219, "step": 533 }, { "KL/chosen_KL_mean": -157.62551879882812, "KL/mean": -189.51998901367188, "KL/rejected_KL_mean": -221.4144287109375, "KL/std": 99.86007690429688, "epoch": 0.8072562358276644, "fcm_dpo/beta": 0.005947187077254057, "fcm_dpo/delta": 0.02126619592308998, "fcm_dpo/margin": 63.78891372680664, "fcm_dpo/q_t": 0.41670671105384827, "grad_norm": 14.632553100585938, "learning_rate": 5.513237282548033e-08, "logits/chosen": 0.6824707984924316, "logits/rejected": 0.6461671590805054, "logps/chosen": -218.37420654296875, "logps/ref_chosen": -60.748687744140625, "logps/ref_rejected": -73.8623046875, "logps/rejected": -295.2767333984375, "loss": 1.1674, "margin_dpo/margin_mean": 63.78891372680664, "margin_dpo/margin_std": 123.30252075195312, "step": 534 }, { "KL/chosen_KL_mean": -166.86672973632812, "KL/mean": -197.24063110351562, "KL/rejected_KL_mean": -227.61453247070312, "KL/std": 106.24088287353516, "epoch": 0.8087679516250945, "fcm_dpo/beta": 0.006018165964633226, "fcm_dpo/delta": 0.035570912063121796, "fcm_dpo/margin": 60.74782943725586, "fcm_dpo/q_t": 0.41900062561035156, "grad_norm": 15.982062339782715, "learning_rate": 5.430681259032957e-08, "logits/chosen": 0.6182019710540771, "logits/rejected": 0.557220995426178, "logps/chosen": -228.504150390625, "logps/ref_chosen": -61.637413024902344, "logps/ref_rejected": -80.93138885498047, "logps/rejected": -308.5459289550781, "loss": 1.1686, "margin_dpo/margin_mean": 60.74782943725586, "margin_dpo/margin_std": 115.64102172851562, "step": 535 }, { "KL/chosen_KL_mean": -145.83583068847656, "KL/mean": -192.2183837890625, "KL/rejected_KL_mean": -238.6009521484375, "KL/std": 106.7518539428711, "epoch": 0.8102796674225246, "fcm_dpo/beta": 0.005869024433195591, "fcm_dpo/delta": -0.1532009094953537, "fcm_dpo/margin": 92.76512908935547, "fcm_dpo/q_t": 0.3770672678947449, "grad_norm": 12.833633422851562, "learning_rate": 5.3486726314303175e-08, "logits/chosen": 0.7487941384315491, "logits/rejected": 0.662542462348938, "logps/chosen": -197.7248077392578, "logps/ref_chosen": -51.88897705078125, "logps/ref_rejected": -73.34864044189453, "logps/rejected": -311.9495849609375, "loss": 0.9974, "margin_dpo/margin_mean": 92.76513671875, "margin_dpo/margin_std": 103.5233154296875, "step": 536 }, { "KL/chosen_KL_mean": -165.7203369140625, "KL/mean": -204.83944702148438, "KL/rejected_KL_mean": -243.9585418701172, "KL/std": 113.77987670898438, "epoch": 0.8117913832199547, "fcm_dpo/beta": 0.00576906418427825, "fcm_dpo/delta": -0.05421897768974304, "fcm_dpo/margin": 78.23818969726562, "fcm_dpo/q_t": 0.4007411599159241, "grad_norm": 14.385953903198242, "learning_rate": 5.267213693697695e-08, "logits/chosen": 0.7646770477294922, "logits/rejected": 0.6726725101470947, "logps/chosen": -219.96896362304688, "logps/ref_chosen": -54.248619079589844, "logps/ref_rejected": -94.94343566894531, "logps/rejected": -338.9019775390625, "loss": 1.1018, "margin_dpo/margin_mean": 78.23818969726562, "margin_dpo/margin_std": 123.42788696289062, "step": 537 }, { "KL/chosen_KL_mean": -165.12217712402344, "KL/mean": -204.04302978515625, "KL/rejected_KL_mean": -242.96392822265625, "KL/std": 109.24994659423828, "epoch": 0.8133030990173847, "fcm_dpo/beta": 0.0057380604557693005, "fcm_dpo/delta": -0.04881321266293526, "fcm_dpo/margin": 77.84176635742188, "fcm_dpo/q_t": 0.39897143840789795, "grad_norm": 14.58580493927002, "learning_rate": 5.1863067244167144e-08, "logits/chosen": 0.7028927206993103, "logits/rejected": 0.6771712303161621, "logps/chosen": -235.21571350097656, "logps/ref_chosen": -70.09353637695312, "logps/ref_rejected": -79.49833679199219, "logps/rejected": -322.4622802734375, "loss": 1.0787, "margin_dpo/margin_mean": 77.84175872802734, "margin_dpo/margin_std": 109.56571960449219, "step": 538 }, { "KL/chosen_KL_mean": -172.00189208984375, "KL/mean": -201.82867431640625, "KL/rejected_KL_mean": -231.6554718017578, "KL/std": 103.12753295898438, "epoch": 0.8148148148148148, "fcm_dpo/beta": 0.005756544414907694, "fcm_dpo/delta": 0.05865339934825897, "fcm_dpo/margin": 59.653564453125, "fcm_dpo/q_t": 0.4250602722167969, "grad_norm": 15.257790565490723, "learning_rate": 5.105953986729195e-08, "logits/chosen": 0.6645068526268005, "logits/rejected": 0.5794214606285095, "logps/chosen": -233.93359375, "logps/ref_chosen": -61.93169403076172, "logps/ref_rejected": -84.08946228027344, "logps/rejected": -315.74493408203125, "loss": 1.1636, "margin_dpo/margin_mean": 59.653564453125, "margin_dpo/margin_std": 110.09439086914062, "step": 539 }, { "KL/chosen_KL_mean": -157.47537231445312, "KL/mean": -203.33316040039062, "KL/rejected_KL_mean": -249.19097900390625, "KL/std": 116.36079406738281, "epoch": 0.8163265306122449, "fcm_dpo/beta": 0.005698896944522858, "fcm_dpo/delta": -0.12926650047302246, "fcm_dpo/margin": 91.71561431884766, "fcm_dpo/q_t": 0.3814542591571808, "grad_norm": 14.18432331085205, "learning_rate": 5.026157728273966e-08, "logits/chosen": 0.7328705787658691, "logits/rejected": 0.6384468078613281, "logps/chosen": -220.1796112060547, "logps/ref_chosen": -62.704254150390625, "logps/ref_rejected": -95.63597106933594, "logps/rejected": -344.82696533203125, "loss": 1.0221, "margin_dpo/margin_mean": 91.71561431884766, "margin_dpo/margin_std": 109.49424743652344, "step": 540 }, { "KL/chosen_KL_mean": -153.7994384765625, "KL/mean": -191.71060180664062, "KL/rejected_KL_mean": -229.62179565429688, "KL/std": 102.88682556152344, "epoch": 0.817838246409675, "fcm_dpo/beta": 0.005589553155004978, "fcm_dpo/delta": -0.025420255959033966, "fcm_dpo/margin": 75.82234191894531, "fcm_dpo/q_t": 0.4019586443901062, "grad_norm": 13.555953025817871, "learning_rate": 4.9469201811239035e-08, "logits/chosen": 0.7742887139320374, "logits/rejected": 0.8011547923088074, "logps/chosen": -216.2802734375, "logps/ref_chosen": -62.48084259033203, "logps/ref_rejected": -57.55541229248047, "logps/rejected": -287.17718505859375, "loss": 1.0834, "margin_dpo/margin_mean": 75.82233428955078, "margin_dpo/margin_std": 100.54383850097656, "step": 541 }, { "KL/chosen_KL_mean": -133.67481994628906, "KL/mean": -176.20965576171875, "KL/rejected_KL_mean": -218.7445068359375, "KL/std": 99.16780090332031, "epoch": 0.8193499622071051, "fcm_dpo/beta": 0.0055374205112457275, "fcm_dpo/delta": -0.07474702596664429, "fcm_dpo/margin": 85.06968688964844, "fcm_dpo/q_t": 0.39279815554618835, "grad_norm": 14.162556648254395, "learning_rate": 4.868243561723534e-08, "logits/chosen": 0.7516461610794067, "logits/rejected": 0.7043957114219666, "logps/chosen": -183.1297149658203, "logps/ref_chosen": -49.454891204833984, "logps/ref_rejected": -65.33275604248047, "logps/rejected": -284.0772705078125, "loss": 1.0731, "margin_dpo/margin_mean": 85.06968688964844, "margin_dpo/margin_std": 120.65644073486328, "step": 542 }, { "KL/chosen_KL_mean": -144.69039916992188, "KL/mean": -186.76333618164062, "KL/rejected_KL_mean": -228.83628845214844, "KL/std": 102.19773864746094, "epoch": 0.8208616780045351, "fcm_dpo/beta": 0.005488088354468346, "fcm_dpo/delta": -0.06477857381105423, "fcm_dpo/margin": 84.1458969116211, "fcm_dpo/q_t": 0.3943963348865509, "grad_norm": 11.421889305114746, "learning_rate": 4.790130070827028e-08, "logits/chosen": 0.7144954204559326, "logits/rejected": 0.6286009550094604, "logps/chosen": -195.791259765625, "logps/ref_chosen": -51.100860595703125, "logps/ref_rejected": -76.06130981445312, "logps/rejected": -304.8975830078125, "loss": 1.0675, "margin_dpo/margin_mean": 84.14590454101562, "margin_dpo/margin_std": 111.68165588378906, "step": 543 }, { "KL/chosen_KL_mean": -155.3533172607422, "KL/mean": -202.8394012451172, "KL/rejected_KL_mean": -250.32546997070312, "KL/std": 111.51242065429688, "epoch": 0.8223733938019653, "fcm_dpo/beta": 0.005360930226743221, "fcm_dpo/delta": -0.11500416696071625, "fcm_dpo/margin": 94.97218322753906, "fcm_dpo/q_t": 0.38585346937179565, "grad_norm": 14.714296340942383, "learning_rate": 4.7125818934366454e-08, "logits/chosen": 0.6963378190994263, "logits/rejected": 0.6182718276977539, "logps/chosen": -215.63055419921875, "logps/ref_chosen": -60.2772331237793, "logps/ref_rejected": -88.40553283691406, "logps/rejected": -338.73101806640625, "loss": 1.0491, "margin_dpo/margin_mean": 94.97218322753906, "margin_dpo/margin_std": 128.6180877685547, "step": 544 }, { "KL/chosen_KL_mean": -170.68997192382812, "KL/mean": -197.08816528320312, "KL/rejected_KL_mean": -223.48635864257812, "KL/std": 103.76513671875, "epoch": 0.8238851095993953, "fcm_dpo/beta": 0.005427246913313866, "fcm_dpo/delta": 0.11671829223632812, "fcm_dpo/margin": 52.796348571777344, "fcm_dpo/q_t": 0.43493160605430603, "grad_norm": 14.415922164916992, "learning_rate": 4.635601198741607e-08, "logits/chosen": 0.6525213718414307, "logits/rejected": 0.5933715105056763, "logps/chosen": -232.30523681640625, "logps/ref_chosen": -61.61524963378906, "logps/ref_rejected": -78.71266174316406, "logps/rejected": -302.1990051269531, "loss": 1.2087, "margin_dpo/margin_mean": 52.796348571777344, "margin_dpo/margin_std": 112.15919494628906, "step": 545 }, { "KL/chosen_KL_mean": -160.33554077148438, "KL/mean": -190.79754638671875, "KL/rejected_KL_mean": -221.25955200195312, "KL/std": 99.437255859375, "epoch": 0.8253968253968254, "fcm_dpo/beta": 0.005504027009010315, "fcm_dpo/delta": 0.0669669657945633, "fcm_dpo/margin": 60.92401123046875, "fcm_dpo/q_t": 0.4233905076980591, "grad_norm": 17.242345809936523, "learning_rate": 4.559190140057428e-08, "logits/chosen": 0.7818703651428223, "logits/rejected": 0.7760703563690186, "logps/chosen": -219.6488037109375, "logps/ref_chosen": -59.313262939453125, "logps/ref_rejected": -64.73631286621094, "logps/rejected": -285.995849609375, "loss": 1.1879, "margin_dpo/margin_mean": 60.92401123046875, "margin_dpo/margin_std": 123.54962158203125, "step": 546 }, { "KL/chosen_KL_mean": -139.52874755859375, "KL/mean": -184.1389617919922, "KL/rejected_KL_mean": -228.74916076660156, "KL/std": 104.88829040527344, "epoch": 0.8269085411942555, "fcm_dpo/beta": 0.005440082401037216, "fcm_dpo/delta": -0.09014324843883514, "fcm_dpo/margin": 89.22044372558594, "fcm_dpo/q_t": 0.38920462131500244, "grad_norm": 14.522475242614746, "learning_rate": 4.483350854765672e-08, "logits/chosen": 0.6299358606338501, "logits/rejected": 0.5683769583702087, "logps/chosen": -194.5054931640625, "logps/ref_chosen": -54.97674560546875, "logps/ref_rejected": -75.35922241210938, "logps/rejected": -304.1083984375, "loss": 1.0546, "margin_dpo/margin_mean": 89.2204360961914, "margin_dpo/margin_std": 118.66877746582031, "step": 547 }, { "KL/chosen_KL_mean": -165.9710693359375, "KL/mean": -192.84939575195312, "KL/rejected_KL_mean": -219.7277374267578, "KL/std": 104.1871109008789, "epoch": 0.8284202569916855, "fcm_dpo/beta": 0.0055316779762506485, "fcm_dpo/delta": 0.10533760488033295, "fcm_dpo/margin": 53.756656646728516, "fcm_dpo/q_t": 0.43300092220306396, "grad_norm": 16.411643981933594, "learning_rate": 4.4080854642541826e-08, "logits/chosen": 0.6215442419052124, "logits/rejected": 0.5646119117736816, "logps/chosen": -229.18174743652344, "logps/ref_chosen": -63.21067428588867, "logps/ref_rejected": -81.23347473144531, "logps/rejected": -300.9612121582031, "loss": 1.2022, "margin_dpo/margin_mean": 53.756656646728516, "margin_dpo/margin_std": 111.9897232055664, "step": 548 }, { "KL/chosen_KL_mean": -162.5404815673828, "KL/mean": -196.65106201171875, "KL/rejected_KL_mean": -230.76162719726562, "KL/std": 113.05363464355469, "epoch": 0.8299319727891157, "fcm_dpo/beta": 0.005542172584682703, "fcm_dpo/delta": 0.02252175845205784, "fcm_dpo/margin": 68.22113800048828, "fcm_dpo/q_t": 0.41690272092819214, "grad_norm": 16.461366653442383, "learning_rate": 4.333396073857723e-08, "logits/chosen": 0.7623779773712158, "logits/rejected": 0.6952296495437622, "logps/chosen": -226.81399536132812, "logps/ref_chosen": -64.27351379394531, "logps/ref_rejected": -92.31663513183594, "logps/rejected": -323.0782470703125, "loss": 1.1704, "margin_dpo/margin_mean": 68.22114562988281, "margin_dpo/margin_std": 133.75875854492188, "step": 549 }, { "KL/chosen_KL_mean": -168.4662322998047, "KL/mean": -191.34432983398438, "KL/rejected_KL_mean": -214.22242736816406, "KL/std": 100.89875793457031, "epoch": 0.8314436885865457, "fcm_dpo/beta": 0.0056635793298482895, "fcm_dpo/delta": 0.1448059380054474, "fcm_dpo/margin": 45.75617980957031, "fcm_dpo/q_t": 0.44192641973495483, "grad_norm": 15.279605865478516, "learning_rate": 4.259284772799099e-08, "logits/chosen": 0.7094156742095947, "logits/rejected": 0.6828978061676025, "logps/chosen": -224.69667053222656, "logps/ref_chosen": -56.230438232421875, "logps/ref_rejected": -62.59788513183594, "logps/rejected": -276.8203125, "loss": 1.2405, "margin_dpo/margin_mean": 45.75617599487305, "margin_dpo/margin_std": 109.83035278320312, "step": 550 }, { "KL/chosen_KL_mean": -172.52247619628906, "KL/mean": -201.90504455566406, "KL/rejected_KL_mean": -231.28762817382812, "KL/std": 108.69143676757812, "epoch": 0.8329554043839759, "fcm_dpo/beta": 0.005799220874905586, "fcm_dpo/delta": 0.06108397990465164, "fcm_dpo/margin": 58.76515197753906, "fcm_dpo/q_t": 0.4258086681365967, "grad_norm": 15.20448112487793, "learning_rate": 4.1857536341307176e-08, "logits/chosen": 0.7232894897460938, "logits/rejected": 0.6899063587188721, "logps/chosen": -240.26968383789062, "logps/ref_chosen": -67.74720764160156, "logps/ref_rejected": -87.04285430908203, "logps/rejected": -318.3304748535156, "loss": 1.1593, "margin_dpo/margin_mean": 58.76515197753906, "margin_dpo/margin_std": 106.95538330078125, "step": 551 }, { "KL/chosen_KL_mean": -158.82626342773438, "KL/mean": -192.85626220703125, "KL/rejected_KL_mean": -226.88626098632812, "KL/std": 108.2763900756836, "epoch": 0.8344671201814059, "fcm_dpo/beta": 0.00583791546523571, "fcm_dpo/delta": 0.002204202115535736, "fcm_dpo/margin": 68.06000518798828, "fcm_dpo/q_t": 0.4077278971672058, "grad_norm": 16.67183494567871, "learning_rate": 4.112804714676593e-08, "logits/chosen": 0.6853049993515015, "logits/rejected": 0.631269097328186, "logps/chosen": -221.75253295898438, "logps/ref_chosen": -62.92625427246094, "logps/ref_rejected": -82.98365783691406, "logps/rejected": -309.86993408203125, "loss": 1.1141, "margin_dpo/margin_mean": 68.06000518798828, "margin_dpo/margin_std": 102.41045379638672, "step": 552 }, { "KL/chosen_KL_mean": -172.24127197265625, "KL/mean": -205.81686401367188, "KL/rejected_KL_mean": -239.3924560546875, "KL/std": 105.17045593261719, "epoch": 0.8359788359788359, "fcm_dpo/beta": 0.005816556513309479, "fcm_dpo/delta": 0.009793659672141075, "fcm_dpo/margin": 67.15115356445312, "fcm_dpo/q_t": 0.4150969982147217, "grad_norm": 15.538284301757812, "learning_rate": 4.0404400549748144e-08, "logits/chosen": 0.6806514859199524, "logits/rejected": 0.5783441066741943, "logps/chosen": -228.27976989746094, "logps/ref_chosen": -56.038490295410156, "logps/ref_rejected": -84.48454284667969, "logps/rejected": -323.8769836425781, "loss": 1.1563, "margin_dpo/margin_mean": 67.15115356445312, "margin_dpo/margin_std": 126.45013427734375, "step": 553 }, { "KL/chosen_KL_mean": -155.46331787109375, "KL/mean": -195.21212768554688, "KL/rejected_KL_mean": -234.9609375, "KL/std": 101.96823120117188, "epoch": 0.8374905517762661, "fcm_dpo/beta": 0.005765959620475769, "fcm_dpo/delta": -0.06119035556912422, "fcm_dpo/margin": 79.49761199951172, "fcm_dpo/q_t": 0.39454004168510437, "grad_norm": 14.239871978759766, "learning_rate": 3.968661679220467e-08, "logits/chosen": 0.6696711778640747, "logits/rejected": 0.6575784683227539, "logps/chosen": -219.99392700195312, "logps/ref_chosen": -64.53059387207031, "logps/ref_rejected": -71.2155990600586, "logps/rejected": -306.1765441894531, "loss": 1.0772, "margin_dpo/margin_mean": 79.49761199951172, "margin_dpo/margin_std": 111.55782318115234, "step": 554 }, { "KL/chosen_KL_mean": -170.21527099609375, "KL/mean": -201.37120056152344, "KL/rejected_KL_mean": -232.527099609375, "KL/std": 102.17310333251953, "epoch": 0.8390022675736961, "fcm_dpo/beta": 0.00582374632358551, "fcm_dpo/delta": 0.03709454834461212, "fcm_dpo/margin": 62.311851501464844, "fcm_dpo/q_t": 0.4159049093723297, "grad_norm": 14.529513359069824, "learning_rate": 3.89747159520904e-08, "logits/chosen": 0.68208909034729, "logits/rejected": 0.6605509519577026, "logps/chosen": -236.8671875, "logps/ref_chosen": -66.65191650390625, "logps/ref_rejected": -68.6667251586914, "logps/rejected": -301.19384765625, "loss": 1.1831, "margin_dpo/margin_mean": 62.31185531616211, "margin_dpo/margin_std": 122.26577758789062, "step": 555 }, { "KL/chosen_KL_mean": -168.04849243164062, "KL/mean": -197.58963012695312, "KL/rejected_KL_mean": -227.13076782226562, "KL/std": 106.78985595703125, "epoch": 0.8405139833711263, "fcm_dpo/beta": 0.0058286068961024284, "fcm_dpo/delta": 0.057645734399557114, "fcm_dpo/margin": 59.082279205322266, "fcm_dpo/q_t": 0.42424604296684265, "grad_norm": 15.20958423614502, "learning_rate": 3.826871794280192e-08, "logits/chosen": 0.7351358532905579, "logits/rejected": 0.6876686811447144, "logps/chosen": -220.880859375, "logps/ref_chosen": -52.832366943359375, "logps/ref_rejected": -64.49044036865234, "logps/rejected": -291.6212158203125, "loss": 1.1902, "margin_dpo/margin_mean": 59.082279205322266, "margin_dpo/margin_std": 121.4478759765625, "step": 556 }, { "KL/chosen_KL_mean": -166.92498779296875, "KL/mean": -209.52706909179688, "KL/rejected_KL_mean": -252.129150390625, "KL/std": 104.20233154296875, "epoch": 0.8420256991685563, "fcm_dpo/beta": 0.005754388403147459, "fcm_dpo/delta": -0.09558665752410889, "fcm_dpo/margin": 85.20416259765625, "fcm_dpo/q_t": 0.388553261756897, "grad_norm": 11.867389678955078, "learning_rate": 3.756864251262143e-08, "logits/chosen": 0.8205718398094177, "logits/rejected": 0.7426509857177734, "logps/chosen": -221.96096801757812, "logps/ref_chosen": -55.03598403930664, "logps/ref_rejected": -75.80644989013672, "logps/rejected": -327.93560791015625, "loss": 1.0353, "margin_dpo/margin_mean": 85.20416259765625, "margin_dpo/margin_std": 100.88729095458984, "step": 557 }, { "KL/chosen_KL_mean": -154.03515625, "KL/mean": -201.06326293945312, "KL/rejected_KL_mean": -248.09136962890625, "KL/std": 111.77014923095703, "epoch": 0.8435374149659864, "fcm_dpo/beta": 0.005608959123492241, "fcm_dpo/delta": -0.13560570776462555, "fcm_dpo/margin": 94.05621337890625, "fcm_dpo/q_t": 0.38192081451416016, "grad_norm": 11.089186668395996, "learning_rate": 3.687450924416341e-08, "logits/chosen": 0.7711827754974365, "logits/rejected": 0.7159498929977417, "logps/chosen": -217.26150512695312, "logps/ref_chosen": -63.226348876953125, "logps/ref_rejected": -91.46881866455078, "logps/rejected": -339.5601806640625, "loss": 1.0224, "margin_dpo/margin_mean": 94.05621337890625, "margin_dpo/margin_std": 115.32173156738281, "step": 558 }, { "KL/chosen_KL_mean": -156.6865234375, "KL/mean": -198.32675170898438, "KL/rejected_KL_mean": -239.96702575683594, "KL/std": 109.83236694335938, "epoch": 0.8450491307634165, "fcm_dpo/beta": 0.005508118309080601, "fcm_dpo/delta": -0.06268289685249329, "fcm_dpo/margin": 83.280517578125, "fcm_dpo/q_t": 0.4005870819091797, "grad_norm": 11.729891777038574, "learning_rate": 3.6186337553827743e-08, "logits/chosen": 0.6707146167755127, "logits/rejected": 0.6091177463531494, "logps/chosen": -218.20816040039062, "logps/ref_chosen": -61.521644592285156, "logps/ref_rejected": -82.83859252929688, "logps/rejected": -322.80560302734375, "loss": 1.0981, "margin_dpo/margin_mean": 83.280517578125, "margin_dpo/margin_std": 130.9115753173828, "step": 559 }, { "KL/chosen_KL_mean": -171.60235595703125, "KL/mean": -209.61856079101562, "KL/rejected_KL_mean": -247.63478088378906, "KL/std": 106.89784240722656, "epoch": 0.8465608465608465, "fcm_dpo/beta": 0.00552861113101244, "fcm_dpo/delta": -0.02149122953414917, "fcm_dpo/margin": 76.03242492675781, "fcm_dpo/q_t": 0.4038241505622864, "grad_norm": 16.61585235595703, "learning_rate": 3.550414669125573e-08, "logits/chosen": 0.705360472202301, "logits/rejected": 0.6663854122161865, "logps/chosen": -232.24359130859375, "logps/ref_chosen": -60.64122009277344, "logps/ref_rejected": -78.75474548339844, "logps/rejected": -326.3895263671875, "loss": 1.0842, "margin_dpo/margin_mean": 76.03242492675781, "margin_dpo/margin_std": 100.37289428710938, "step": 560 }, { "KL/chosen_KL_mean": -156.08416748046875, "KL/mean": -191.4915771484375, "KL/rejected_KL_mean": -226.89898681640625, "KL/std": 104.07613372802734, "epoch": 0.8480725623582767, "fcm_dpo/beta": 0.005500419996678829, "fcm_dpo/delta": 0.010878851637244225, "fcm_dpo/margin": 70.81483459472656, "fcm_dpo/q_t": 0.41301605105400085, "grad_norm": 12.828431129455566, "learning_rate": 3.482795573879241e-08, "logits/chosen": 0.6696001291275024, "logits/rejected": 0.6357676982879639, "logps/chosen": -218.582763671875, "logps/ref_chosen": -62.49859619140625, "logps/ref_rejected": -78.72064208984375, "logps/rejected": -305.61962890625, "loss": 1.1241, "margin_dpo/margin_mean": 70.81483459472656, "margin_dpo/margin_std": 112.45053100585938, "step": 561 }, { "KL/chosen_KL_mean": -167.10365295410156, "KL/mean": -212.86196899414062, "KL/rejected_KL_mean": -258.62030029296875, "KL/std": 120.87919616699219, "epoch": 0.8495842781557067, "fcm_dpo/beta": 0.005394667387008667, "fcm_dpo/delta": -0.10014957934617996, "fcm_dpo/margin": 91.51663970947266, "fcm_dpo/q_t": 0.3899437189102173, "grad_norm": 16.289958953857422, "learning_rate": 3.415778361095226e-08, "logits/chosen": 0.6985616683959961, "logits/rejected": 0.6643628478050232, "logps/chosen": -241.88539123535156, "logps/ref_chosen": -74.78173828125, "logps/ref_rejected": -92.63499450683594, "logps/rejected": -351.25531005859375, "loss": 1.0459, "margin_dpo/margin_mean": 91.51663970947266, "margin_dpo/margin_std": 117.03136444091797, "step": 562 }, { "KL/chosen_KL_mean": -139.36734008789062, "KL/mean": -178.2583770751953, "KL/rejected_KL_mean": -217.1494140625, "KL/std": 93.23200225830078, "epoch": 0.8510959939531368, "fcm_dpo/beta": 0.0053891572169959545, "fcm_dpo/delta": -0.02001415193080902, "fcm_dpo/margin": 77.78207397460938, "fcm_dpo/q_t": 0.4036242961883545, "grad_norm": 18.518709182739258, "learning_rate": 3.349364905389032e-08, "logits/chosen": 0.7896828651428223, "logits/rejected": 0.7366276979446411, "logps/chosen": -189.5658416748047, "logps/ref_chosen": -50.19850158691406, "logps/ref_rejected": -66.76687622070312, "logps/rejected": -283.9162902832031, "loss": 1.1081, "margin_dpo/margin_mean": 77.78207397460938, "margin_dpo/margin_std": 119.42278289794922, "step": 563 }, { "KL/chosen_KL_mean": -142.17361450195312, "KL/mean": -191.52587890625, "KL/rejected_KL_mean": -240.87814331054688, "KL/std": 101.90835571289062, "epoch": 0.8526077097505669, "fcm_dpo/beta": 0.0052757966332137585, "fcm_dpo/delta": -0.1275056004524231, "fcm_dpo/margin": 98.70454406738281, "fcm_dpo/q_t": 0.3802381753921509, "grad_norm": 12.03218936920166, "learning_rate": 3.283557064487785e-08, "logits/chosen": 0.679800271987915, "logits/rejected": 0.6504189968109131, "logps/chosen": -197.91445922851562, "logps/ref_chosen": -55.7408447265625, "logps/ref_rejected": -74.82323455810547, "logps/rejected": -315.7013854980469, "loss": 1.0277, "margin_dpo/margin_mean": 98.70453643798828, "margin_dpo/margin_std": 121.65130615234375, "step": 564 }, { "KL/chosen_KL_mean": -172.18453979492188, "KL/mean": -207.91888427734375, "KL/rejected_KL_mean": -243.65325927734375, "KL/std": 105.38109588623047, "epoch": 0.854119425547997, "fcm_dpo/beta": 0.005283844657242298, "fcm_dpo/delta": 0.02286720462143421, "fcm_dpo/margin": 71.46871948242188, "fcm_dpo/q_t": 0.41272348165512085, "grad_norm": 14.575935363769531, "learning_rate": 3.218356679178252e-08, "logits/chosen": 0.7471562623977661, "logits/rejected": 0.6933939456939697, "logps/chosen": -230.52191162109375, "logps/ref_chosen": -58.33738327026367, "logps/ref_rejected": -78.31776428222656, "logps/rejected": -321.97100830078125, "loss": 1.1176, "margin_dpo/margin_mean": 71.46871185302734, "margin_dpo/margin_std": 105.04327392578125, "step": 565 }, { "KL/chosen_KL_mean": -162.09071350097656, "KL/mean": -194.7822265625, "KL/rejected_KL_mean": -227.47377014160156, "KL/std": 108.80207824707031, "epoch": 0.8556311413454271, "fcm_dpo/beta": 0.0053364988416433334, "fcm_dpo/delta": 0.05221934616565704, "fcm_dpo/margin": 65.383056640625, "fcm_dpo/q_t": 0.4226508140563965, "grad_norm": 14.741025924682617, "learning_rate": 3.1537655732553764e-08, "logits/chosen": 0.7219111919403076, "logits/rejected": 0.7103610634803772, "logps/chosen": -233.314453125, "logps/ref_chosen": -71.22373962402344, "logps/ref_rejected": -71.11601257324219, "logps/rejected": -298.58978271484375, "loss": 1.1933, "margin_dpo/margin_mean": 65.383056640625, "margin_dpo/margin_std": 137.35252380371094, "step": 566 }, { "KL/chosen_KL_mean": -155.47418212890625, "KL/mean": -196.4601593017578, "KL/rejected_KL_mean": -237.4461212158203, "KL/std": 102.49577331542969, "epoch": 0.8571428571428571, "fcm_dpo/beta": 0.0052682156674563885, "fcm_dpo/delta": -0.034140560775995255, "fcm_dpo/margin": 81.97193908691406, "fcm_dpo/q_t": 0.4000481367111206, "grad_norm": 11.802513122558594, "learning_rate": 3.089785553471233e-08, "logits/chosen": 0.7189158201217651, "logits/rejected": 0.6299164295196533, "logps/chosen": -208.14346313476562, "logps/ref_chosen": -52.669273376464844, "logps/ref_rejected": -74.34785461425781, "logps/rejected": -311.7939758300781, "loss": 1.0831, "margin_dpo/margin_mean": 81.97193908691406, "margin_dpo/margin_std": 110.23538208007812, "step": 567 }, { "KL/chosen_KL_mean": -142.93621826171875, "KL/mean": -193.83016967773438, "KL/rejected_KL_mean": -244.72412109375, "KL/std": 109.18955993652344, "epoch": 0.8586545729402872, "fcm_dpo/beta": 0.005216483026742935, "fcm_dpo/delta": -0.1383085399866104, "fcm_dpo/margin": 101.78787231445312, "fcm_dpo/q_t": 0.3798676133155823, "grad_norm": 15.167051315307617, "learning_rate": 3.026418409484513e-08, "logits/chosen": 0.7353081107139587, "logits/rejected": 0.6538002490997314, "logps/chosen": -195.11422729492188, "logps/ref_chosen": -52.178001403808594, "logps/ref_rejected": -85.8277587890625, "logps/rejected": -330.5518798828125, "loss": 1.005, "margin_dpo/margin_mean": 101.78787231445312, "margin_dpo/margin_std": 109.98161315917969, "step": 568 }, { "KL/chosen_KL_mean": -168.43011474609375, "KL/mean": -193.41647338867188, "KL/rejected_KL_mean": -218.40283203125, "KL/std": 110.10757446289062, "epoch": 0.8601662887377173, "fcm_dpo/beta": 0.0052167270332574844, "fcm_dpo/delta": 0.14295698702335358, "fcm_dpo/margin": 49.97270202636719, "fcm_dpo/q_t": 0.4411056339740753, "grad_norm": 14.588390350341797, "learning_rate": 2.963665913810451e-08, "logits/chosen": 0.6284128427505493, "logits/rejected": 0.5956965684890747, "logps/chosen": -231.07937622070312, "logps/ref_chosen": -62.649261474609375, "logps/ref_rejected": -75.4298324584961, "logps/rejected": -293.8326721191406, "loss": 1.2268, "margin_dpo/margin_mean": 49.97270202636719, "margin_dpo/margin_std": 109.44187927246094, "step": 569 }, { "KL/chosen_KL_mean": -147.50064086914062, "KL/mean": -195.2564697265625, "KL/rejected_KL_mean": -243.01231384277344, "KL/std": 101.05863189697266, "epoch": 0.8616780045351474, "fcm_dpo/beta": 0.005238240119069815, "fcm_dpo/delta": -0.1055883839726448, "fcm_dpo/margin": 95.51168060302734, "fcm_dpo/q_t": 0.38603097200393677, "grad_norm": 13.221871376037598, "learning_rate": 2.9015298217712453e-08, "logits/chosen": 0.70135498046875, "logits/rejected": 0.6157346963882446, "logps/chosen": -197.54244995117188, "logps/ref_chosen": -50.04179382324219, "logps/ref_rejected": -78.27146911621094, "logps/rejected": -321.2837829589844, "loss": 1.0312, "margin_dpo/margin_mean": 95.51167297363281, "margin_dpo/margin_std": 113.90409851074219, "step": 570 }, { "KL/chosen_KL_mean": -161.2565155029297, "KL/mean": -188.04335021972656, "KL/rejected_KL_mean": -214.83018493652344, "KL/std": 99.07657623291016, "epoch": 0.8631897203325775, "fcm_dpo/beta": 0.0052953860722482204, "fcm_dpo/delta": 0.11900515854358673, "fcm_dpo/margin": 53.57366943359375, "fcm_dpo/q_t": 0.43476077914237976, "grad_norm": 13.77877140045166, "learning_rate": 2.840011871446962e-08, "logits/chosen": 0.7282629609107971, "logits/rejected": 0.7028162479400635, "logps/chosen": -214.913330078125, "logps/ref_chosen": -53.65681457519531, "logps/ref_rejected": -66.13298034667969, "logps/rejected": -280.9631652832031, "loss": 1.2133, "margin_dpo/margin_mean": 53.573665618896484, "margin_dpo/margin_std": 114.42227172851562, "step": 571 }, { "KL/chosen_KL_mean": -161.99856567382812, "KL/mean": -193.23593139648438, "KL/rejected_KL_mean": -224.47332763671875, "KL/std": 105.49080657958984, "epoch": 0.8647014361300076, "fcm_dpo/beta": 0.005384175106883049, "fcm_dpo/delta": 0.06513378024101257, "fcm_dpo/margin": 62.474761962890625, "fcm_dpo/q_t": 0.42261409759521484, "grad_norm": 14.15194034576416, "learning_rate": 2.7791137836269158e-08, "logits/chosen": 0.6913356781005859, "logits/rejected": 0.733739972114563, "logps/chosen": -236.8164825439453, "logps/ref_chosen": -74.81792449951172, "logps/ref_rejected": -65.88681030273438, "logps/rejected": -290.3601379394531, "loss": 1.1585, "margin_dpo/margin_mean": 62.47476577758789, "margin_dpo/margin_std": 106.80880737304688, "step": 572 }, { "KL/chosen_KL_mean": -176.1700897216797, "KL/mean": -211.75979614257812, "KL/rejected_KL_mean": -247.34950256347656, "KL/std": 114.86283111572266, "epoch": 0.8662131519274376, "fcm_dpo/beta": 0.005393402185291052, "fcm_dpo/delta": 0.016733139753341675, "fcm_dpo/margin": 71.179443359375, "fcm_dpo/q_t": 0.41635024547576904, "grad_norm": 15.528578758239746, "learning_rate": 2.718837261761528e-08, "logits/chosen": 0.7083475589752197, "logits/rejected": 0.6636322736740112, "logps/chosen": -244.89572143554688, "logps/ref_chosen": -68.72564697265625, "logps/ref_rejected": -88.16201782226562, "logps/rejected": -335.51153564453125, "loss": 1.1631, "margin_dpo/margin_mean": 71.179443359375, "margin_dpo/margin_std": 137.14370727539062, "step": 573 }, { "KL/chosen_KL_mean": -156.84823608398438, "KL/mean": -202.28956604003906, "KL/rejected_KL_mean": -247.73086547851562, "KL/std": 105.29086303710938, "epoch": 0.8677248677248677, "fcm_dpo/beta": 0.005354342050850391, "fcm_dpo/delta": -0.09120815247297287, "fcm_dpo/margin": 90.88264465332031, "fcm_dpo/q_t": 0.3886836767196655, "grad_norm": 11.730754852294922, "learning_rate": 2.659183991914696e-08, "logits/chosen": 0.7415429949760437, "logits/rejected": 0.6785948276519775, "logps/chosen": -213.16165161132812, "logps/ref_chosen": -56.31340026855469, "logps/ref_rejected": -83.91553497314453, "logps/rejected": -331.64642333984375, "loss": 1.0255, "margin_dpo/margin_mean": 90.88263702392578, "margin_dpo/margin_std": 99.63211059570312, "step": 574 }, { "KL/chosen_KL_mean": -161.31875610351562, "KL/mean": -193.75637817382812, "KL/rejected_KL_mean": -226.19400024414062, "KL/std": 107.47755432128906, "epoch": 0.8692365835222978, "fcm_dpo/beta": 0.005301470868289471, "fcm_dpo/delta": 0.05722144991159439, "fcm_dpo/margin": 64.87521362304688, "fcm_dpo/q_t": 0.4250301122665405, "grad_norm": 13.5238676071167, "learning_rate": 2.600155642716606e-08, "logits/chosen": 0.7329334616661072, "logits/rejected": 0.6541694402694702, "logps/chosen": -225.90289306640625, "logps/ref_chosen": -64.5841293334961, "logps/ref_rejected": -93.47034454345703, "logps/rejected": -319.6643371582031, "loss": 1.1879, "margin_dpo/margin_mean": 64.87521362304688, "margin_dpo/margin_std": 128.93638610839844, "step": 575 }, { "KL/chosen_KL_mean": -143.23312377929688, "KL/mean": -190.56591796875, "KL/rejected_KL_mean": -237.89871215820312, "KL/std": 107.76126861572266, "epoch": 0.8707482993197279, "fcm_dpo/beta": 0.005239981692284346, "fcm_dpo/delta": -0.10286588966846466, "fcm_dpo/margin": 94.66560363769531, "fcm_dpo/q_t": 0.38725581765174866, "grad_norm": 12.78931999206543, "learning_rate": 2.5417538653170754e-08, "logits/chosen": 0.716622531414032, "logits/rejected": 0.6150184869766235, "logps/chosen": -196.51364135742188, "logps/ref_chosen": -53.28052520751953, "logps/ref_rejected": -84.2000503540039, "logps/rejected": -322.0987548828125, "loss": 1.0446, "margin_dpo/margin_mean": 94.66560363769531, "margin_dpo/margin_std": 119.16376495361328, "step": 576 }, { "KL/chosen_KL_mean": -160.7834930419922, "KL/mean": -191.5000762939453, "KL/rejected_KL_mean": -222.2166748046875, "KL/std": 105.21707916259766, "epoch": 0.872260015117158, "fcm_dpo/beta": 0.005305111408233643, "fcm_dpo/delta": 0.07664323598146439, "fcm_dpo/margin": 61.43318176269531, "fcm_dpo/q_t": 0.42474716901779175, "grad_norm": 13.350493431091309, "learning_rate": 2.4839802933393607e-08, "logits/chosen": 0.6843061447143555, "logits/rejected": 0.6764715909957886, "logps/chosen": -223.10818481445312, "logps/ref_chosen": -62.32468795776367, "logps/ref_rejected": -67.300537109375, "logps/rejected": -289.5172119140625, "loss": 1.1767, "margin_dpo/margin_mean": 61.43318176269531, "margin_dpo/margin_std": 112.92008972167969, "step": 577 }, { "KL/chosen_KL_mean": -151.0160369873047, "KL/mean": -181.88204956054688, "KL/rejected_KL_mean": -212.748046875, "KL/std": 109.35968017578125, "epoch": 0.873771730914588, "fcm_dpo/beta": 0.005385764874517918, "fcm_dpo/delta": 0.06986413896083832, "fcm_dpo/margin": 61.73200225830078, "fcm_dpo/q_t": 0.42697471380233765, "grad_norm": 12.926276206970215, "learning_rate": 2.4268365428344733e-08, "logits/chosen": 0.7761894464492798, "logits/rejected": 0.7603508234024048, "logps/chosen": -207.67161560058594, "logps/ref_chosen": -56.65557861328125, "logps/ref_rejected": -68.21835327148438, "logps/rejected": -280.9664001464844, "loss": 1.1874, "margin_dpo/margin_mean": 61.73200988769531, "margin_dpo/margin_std": 125.3292236328125, "step": 578 }, { "KL/chosen_KL_mean": -157.9138946533203, "KL/mean": -202.1259765625, "KL/rejected_KL_mean": -246.3380584716797, "KL/std": 101.47502136230469, "epoch": 0.8752834467120182, "fcm_dpo/beta": 0.005336810369044542, "fcm_dpo/delta": -0.07564752548933029, "fcm_dpo/margin": 88.4241714477539, "fcm_dpo/q_t": 0.3900591731071472, "grad_norm": 12.923798561096191, "learning_rate": 2.3703242122359357e-08, "logits/chosen": 0.6561988592147827, "logits/rejected": 0.6358931064605713, "logps/chosen": -214.72354125976562, "logps/ref_chosen": -56.809661865234375, "logps/ref_rejected": -68.09613037109375, "logps/rejected": -314.4342041015625, "loss": 1.0415, "margin_dpo/margin_mean": 88.4241714477539, "margin_dpo/margin_std": 104.05563354492188, "step": 579 }, { "KL/chosen_KL_mean": -158.97811889648438, "KL/mean": -197.81932067871094, "KL/rejected_KL_mean": -236.66053771972656, "KL/std": 118.46055603027344, "epoch": 0.8767951625094482, "fcm_dpo/beta": 0.005337144713848829, "fcm_dpo/delta": -0.015590623021125793, "fcm_dpo/margin": 77.68240356445312, "fcm_dpo/q_t": 0.4092617630958557, "grad_norm": 14.614388465881348, "learning_rate": 2.3144448823151392e-08, "logits/chosen": 0.6761208772659302, "logits/rejected": 0.6266993880271912, "logps/chosen": -216.67822265625, "logps/ref_chosen": -57.70011520385742, "logps/ref_rejected": -77.90664672851562, "logps/rejected": -314.56719970703125, "loss": 1.1279, "margin_dpo/margin_mean": 77.68241119384766, "margin_dpo/margin_std": 131.71633911132812, "step": 580 }, { "KL/chosen_KL_mean": -174.16661071777344, "KL/mean": -209.36288452148438, "KL/rejected_KL_mean": -244.55918884277344, "KL/std": 105.65506744384766, "epoch": 0.8783068783068783, "fcm_dpo/beta": 0.005308878142386675, "fcm_dpo/delta": 0.02710052765905857, "fcm_dpo/margin": 70.39259338378906, "fcm_dpo/q_t": 0.4171040952205658, "grad_norm": 13.96402359008789, "learning_rate": 2.259200116137039e-08, "logits/chosen": 0.7344849109649658, "logits/rejected": 0.6715294122695923, "logps/chosen": -233.49896240234375, "logps/ref_chosen": -59.332359313964844, "logps/ref_rejected": -83.64482116699219, "logps/rejected": -328.2040100097656, "loss": 1.151, "margin_dpo/margin_mean": 70.39259338378906, "margin_dpo/margin_std": 125.0219955444336, "step": 581 }, { "KL/chosen_KL_mean": -159.0695037841797, "KL/mean": -192.51058959960938, "KL/rejected_KL_mean": -225.95169067382812, "KL/std": 99.99513244628906, "epoch": 0.8798185941043084, "fcm_dpo/beta": 0.0053823720663785934, "fcm_dpo/delta": 0.04131370037794113, "fcm_dpo/margin": 66.88218688964844, "fcm_dpo/q_t": 0.41894370317459106, "grad_norm": 11.404556274414062, "learning_rate": 2.204591459016525e-08, "logits/chosen": 0.7167221903800964, "logits/rejected": 0.7457035779953003, "logps/chosen": -223.23236083984375, "logps/ref_chosen": -64.16285705566406, "logps/ref_rejected": -58.632896423339844, "logps/rejected": -284.5845947265625, "loss": 1.1493, "margin_dpo/margin_mean": 66.88218688964844, "margin_dpo/margin_std": 114.8166275024414, "step": 582 }, { "KL/chosen_KL_mean": -156.87049865722656, "KL/mean": -200.28271484375, "KL/rejected_KL_mean": -243.6949462890625, "KL/std": 113.34555053710938, "epoch": 0.8813303099017384, "fcm_dpo/beta": 0.005323405377566814, "fcm_dpo/delta": -0.06531860679388046, "fcm_dpo/margin": 86.82443237304688, "fcm_dpo/q_t": 0.3956996500492096, "grad_norm": 14.661458015441895, "learning_rate": 2.1506204384751064e-08, "logits/chosen": 0.791956901550293, "logits/rejected": 0.6918524503707886, "logps/chosen": -208.74288940429688, "logps/ref_chosen": -51.87239456176758, "logps/ref_rejected": -83.86331176757812, "logps/rejected": -327.5582580566406, "loss": 1.0855, "margin_dpo/margin_mean": 86.82443237304688, "margin_dpo/margin_std": 128.30801391601562, "step": 583 }, { "KL/chosen_KL_mean": -145.236083984375, "KL/mean": -181.6009979248047, "KL/rejected_KL_mean": -217.96591186523438, "KL/std": 104.1064453125, "epoch": 0.8828420256991686, "fcm_dpo/beta": 0.005305691156536341, "fcm_dpo/delta": 0.014509003609418869, "fcm_dpo/margin": 72.72981262207031, "fcm_dpo/q_t": 0.41576987504959106, "grad_norm": 14.06952953338623, "learning_rate": 2.09728856419826e-08, "logits/chosen": 0.8287737965583801, "logits/rejected": 0.7238653898239136, "logps/chosen": -191.80746459960938, "logps/ref_chosen": -46.571388244628906, "logps/ref_rejected": -80.67969512939453, "logps/rejected": -298.6455993652344, "loss": 1.1556, "margin_dpo/margin_mean": 72.72981262207031, "margin_dpo/margin_std": 134.8087158203125, "step": 584 }, { "KL/chosen_KL_mean": -169.44070434570312, "KL/mean": -196.7550506591797, "KL/rejected_KL_mean": -224.0694122314453, "KL/std": 114.88101196289062, "epoch": 0.8843537414965986, "fcm_dpo/beta": 0.005436314269900322, "fcm_dpo/delta": 0.10527031868696213, "fcm_dpo/margin": 54.628692626953125, "fcm_dpo/q_t": 0.4324593245983124, "grad_norm": 13.1787691116333, "learning_rate": 2.044597327993153e-08, "logits/chosen": 0.6715638637542725, "logits/rejected": 0.6183586716651917, "logps/chosen": -227.56524658203125, "logps/ref_chosen": -58.124534606933594, "logps/ref_rejected": -79.00538635253906, "logps/rejected": -303.0747985839844, "loss": 1.2024, "margin_dpo/margin_mean": 54.628700256347656, "margin_dpo/margin_std": 113.292724609375, "step": 585 }, { "KL/chosen_KL_mean": -155.8141632080078, "KL/mean": -190.7159423828125, "KL/rejected_KL_mean": -225.61773681640625, "KL/std": 95.33213806152344, "epoch": 0.8858654572940288, "fcm_dpo/beta": 0.005444392561912537, "fcm_dpo/delta": 0.020684288814663887, "fcm_dpo/margin": 69.80355834960938, "fcm_dpo/q_t": 0.41235822439193726, "grad_norm": 13.345340728759766, "learning_rate": 1.9925482037469187e-08, "logits/chosen": 0.7436306476593018, "logits/rejected": 0.7003189325332642, "logps/chosen": -209.91580200195312, "logps/ref_chosen": -54.10163879394531, "logps/ref_rejected": -63.72113037109375, "logps/rejected": -289.3388671875, "loss": 1.097, "margin_dpo/margin_mean": 69.8035659790039, "margin_dpo/margin_std": 88.85975646972656, "step": 586 }, { "KL/chosen_KL_mean": -164.1700439453125, "KL/mean": -198.61328125, "KL/rejected_KL_mean": -233.05653381347656, "KL/std": 112.59260559082031, "epoch": 0.8873771730914588, "fcm_dpo/beta": 0.005474920384585857, "fcm_dpo/delta": 0.023741472512483597, "fcm_dpo/margin": 68.88650512695312, "fcm_dpo/q_t": 0.40968871116638184, "grad_norm": 15.52365779876709, "learning_rate": 1.9411426473854687e-08, "logits/chosen": 0.7410461902618408, "logits/rejected": 0.7377340793609619, "logps/chosen": -227.58721923828125, "logps/ref_chosen": -63.41719436645508, "logps/ref_rejected": -63.47003936767578, "logps/rejected": -296.5265808105469, "loss": 1.1936, "margin_dpo/margin_mean": 68.88650512695312, "margin_dpo/margin_std": 151.58599853515625, "step": 587 }, { "KL/chosen_KL_mean": -166.9088592529297, "KL/mean": -206.42413330078125, "KL/rejected_KL_mean": -245.9394073486328, "KL/std": 111.02951049804688, "epoch": 0.8888888888888888, "fcm_dpo/beta": 0.005442941095679998, "fcm_dpo/delta": -0.03191109001636505, "fcm_dpo/margin": 79.03054809570312, "fcm_dpo/q_t": 0.4048862159252167, "grad_norm": 17.76612091064453, "learning_rate": 1.890382096832699e-08, "logits/chosen": 0.7640427947044373, "logits/rejected": 0.7215005159378052, "logps/chosen": -229.10989379882812, "logps/ref_chosen": -62.20103454589844, "logps/ref_rejected": -82.10249328613281, "logps/rejected": -328.0419006347656, "loss": 1.1259, "margin_dpo/margin_mean": 79.03054809570312, "margin_dpo/margin_std": 135.61080932617188, "step": 588 }, { "KL/chosen_KL_mean": -156.30934143066406, "KL/mean": -196.30322265625, "KL/rejected_KL_mean": -236.297119140625, "KL/std": 106.37506103515625, "epoch": 0.890400604686319, "fcm_dpo/beta": 0.005437628366053104, "fcm_dpo/delta": -0.036562640219926834, "fcm_dpo/margin": 79.98777770996094, "fcm_dpo/q_t": 0.3991260528564453, "grad_norm": 12.22568416595459, "learning_rate": 1.840267971970344e-08, "logits/chosen": 0.7192363739013672, "logits/rejected": 0.6883012652397156, "logps/chosen": -213.02294921875, "logps/ref_chosen": -56.71361541748047, "logps/ref_rejected": -76.7366943359375, "logps/rejected": -313.0338134765625, "loss": 1.0678, "margin_dpo/margin_mean": 79.98777770996094, "margin_dpo/margin_std": 100.07125854492188, "step": 589 }, { "KL/chosen_KL_mean": -170.75979614257812, "KL/mean": -207.8297882080078, "KL/rejected_KL_mean": -244.89979553222656, "KL/std": 99.0168685913086, "epoch": 0.891912320483749, "fcm_dpo/beta": 0.005432046018540859, "fcm_dpo/delta": -0.003115180879831314, "fcm_dpo/margin": 74.14000701904297, "fcm_dpo/q_t": 0.4076809287071228, "grad_norm": 16.128381729125977, "learning_rate": 1.7908016745981856e-08, "logits/chosen": 0.6945627927780151, "logits/rejected": 0.6577792763710022, "logps/chosen": -237.2736053466797, "logps/ref_chosen": -66.5138168334961, "logps/ref_rejected": -85.70820617675781, "logps/rejected": -330.6080017089844, "loss": 1.1049, "margin_dpo/margin_mean": 74.13999938964844, "margin_dpo/margin_std": 106.85514831542969, "step": 590 }, { "KL/chosen_KL_mean": -145.10528564453125, "KL/mean": -192.39013671875, "KL/rejected_KL_mean": -239.6749725341797, "KL/std": 115.98556518554688, "epoch": 0.8934240362811792, "fcm_dpo/beta": 0.0052824802696704865, "fcm_dpo/delta": -0.10702680796384811, "fcm_dpo/margin": 94.5697021484375, "fcm_dpo/q_t": 0.3904598355293274, "grad_norm": 14.471207618713379, "learning_rate": 1.7419845883949098e-08, "logits/chosen": 0.8123652338981628, "logits/rejected": 0.7523353099822998, "logps/chosen": -205.80245971679688, "logps/ref_chosen": -60.697181701660156, "logps/ref_rejected": -86.12278747558594, "logps/rejected": -325.7977600097656, "loss": 1.0776, "margin_dpo/margin_mean": 94.5697021484375, "margin_dpo/margin_std": 141.56353759765625, "step": 591 }, { "KL/chosen_KL_mean": -158.5380401611328, "KL/mean": -194.74591064453125, "KL/rejected_KL_mean": -230.95382690429688, "KL/std": 99.0007095336914, "epoch": 0.8949357520786092, "fcm_dpo/beta": 0.0052817342802882195, "fcm_dpo/delta": 0.017707686871290207, "fcm_dpo/margin": 72.41576385498047, "fcm_dpo/q_t": 0.4149664640426636, "grad_norm": 14.247783660888672, "learning_rate": 1.6938180788793556e-08, "logits/chosen": 0.7698496580123901, "logits/rejected": 0.6575839519500732, "logps/chosen": -209.77536010742188, "logps/ref_chosen": -51.237327575683594, "logps/ref_rejected": -81.60242462158203, "logps/rejected": -312.5562438964844, "loss": 1.1246, "margin_dpo/margin_mean": 72.41576385498047, "margin_dpo/margin_std": 112.3682632446289, "step": 592 }, { "KL/chosen_KL_mean": -131.92184448242188, "KL/mean": -170.8389892578125, "KL/rejected_KL_mean": -209.75611877441406, "KL/std": 101.70758819580078, "epoch": 0.8964474678760394, "fcm_dpo/beta": 0.005294942297041416, "fcm_dpo/delta": -0.012750206515192986, "fcm_dpo/margin": 77.83428955078125, "fcm_dpo/q_t": 0.4051951766014099, "grad_norm": 15.31247615814209, "learning_rate": 1.6463034933723336e-08, "logits/chosen": 0.7372517585754395, "logits/rejected": 0.6430565118789673, "logps/chosen": -174.0018310546875, "logps/ref_chosen": -42.08000183105469, "logps/ref_rejected": -68.47499084472656, "logps/rejected": -278.2311096191406, "loss": 1.1067, "margin_dpo/margin_mean": 77.83428955078125, "margin_dpo/margin_std": 117.07554626464844, "step": 593 }, { "KL/chosen_KL_mean": -166.61257934570312, "KL/mean": -200.24847412109375, "KL/rejected_KL_mean": -233.8843994140625, "KL/std": 100.83384704589844, "epoch": 0.8979591836734694, "fcm_dpo/beta": 0.00533599779009819, "fcm_dpo/delta": 0.04256129637360573, "fcm_dpo/margin": 67.27182006835938, "fcm_dpo/q_t": 0.4163721799850464, "grad_norm": 12.89805793762207, "learning_rate": 1.5994421609589385e-08, "logits/chosen": 0.6495029926300049, "logits/rejected": 0.6361119747161865, "logps/chosen": -230.271240234375, "logps/ref_chosen": -63.658668518066406, "logps/ref_rejected": -70.35597229003906, "logps/rejected": -304.2403564453125, "loss": 1.131, "margin_dpo/margin_mean": 67.27182006835938, "margin_dpo/margin_std": 102.46228790283203, "step": 594 }, { "KL/chosen_KL_mean": -154.23065185546875, "KL/mean": -198.49163818359375, "KL/rejected_KL_mean": -242.75262451171875, "KL/std": 107.8331298828125, "epoch": 0.8994708994708994, "fcm_dpo/beta": 0.005307753570377827, "fcm_dpo/delta": -0.07329151034355164, "fcm_dpo/margin": 88.52198791503906, "fcm_dpo/q_t": 0.3952780067920685, "grad_norm": 11.26756763458252, "learning_rate": 1.553235392451377e-08, "logits/chosen": 0.8032486438751221, "logits/rejected": 0.7115143537521362, "logps/chosen": -210.44940185546875, "logps/ref_chosen": -56.21875762939453, "logps/ref_rejected": -83.95773315429688, "logps/rejected": -326.71038818359375, "loss": 1.0858, "margin_dpo/margin_mean": 88.52198791503906, "margin_dpo/margin_std": 133.67189025878906, "step": 595 }, { "KL/chosen_KL_mean": -178.541015625, "KL/mean": -195.6376953125, "KL/rejected_KL_mean": -212.734375, "KL/std": 99.39712524414062, "epoch": 0.9009826152683296, "fcm_dpo/beta": 0.005426476243883371, "fcm_dpo/delta": 0.21902009844779968, "fcm_dpo/margin": 34.19335174560547, "fcm_dpo/q_t": 0.45916134119033813, "grad_norm": 16.031299591064453, "learning_rate": 1.507684480352292e-08, "logits/chosen": 0.63373863697052, "logits/rejected": 0.6557974219322205, "logps/chosen": -247.0218963623047, "logps/ref_chosen": -68.48088073730469, "logps/ref_rejected": -61.732967376708984, "logps/rejected": -274.46734619140625, "loss": 1.2901, "margin_dpo/margin_mean": 34.19335174560547, "margin_dpo/margin_std": 105.52054595947266, "step": 596 }, { "KL/chosen_KL_mean": -136.6261444091797, "KL/mean": -170.66302490234375, "KL/rejected_KL_mean": -204.69989013671875, "KL/std": 87.61666870117188, "epoch": 0.9024943310657596, "fcm_dpo/beta": 0.005537579767405987, "fcm_dpo/delta": 0.0237730722874403, "fcm_dpo/margin": 68.0737533569336, "fcm_dpo/q_t": 0.4140661358833313, "grad_norm": 11.68276309967041, "learning_rate": 1.4627906988186111e-08, "logits/chosen": 0.7150381803512573, "logits/rejected": 0.6985746622085571, "logps/chosen": -185.483642578125, "logps/ref_chosen": -48.85750961303711, "logps/ref_rejected": -55.068084716796875, "logps/rejected": -259.7679748535156, "loss": 1.1229, "margin_dpo/margin_mean": 68.0737533569336, "margin_dpo/margin_std": 103.95882415771484, "step": 597 }, { "KL/chosen_KL_mean": -178.44558715820312, "KL/mean": -200.84381103515625, "KL/rejected_KL_mean": -223.24200439453125, "KL/std": 104.78968811035156, "epoch": 0.9040060468631897, "fcm_dpo/beta": 0.005672060884535313, "fcm_dpo/delta": 0.14913338422775269, "fcm_dpo/margin": 44.79642105102539, "fcm_dpo/q_t": 0.44390666484832764, "grad_norm": 13.62793254852295, "learning_rate": 1.4185553036259095e-08, "logits/chosen": 0.7172830104827881, "logits/rejected": 0.6382390260696411, "logps/chosen": -237.33274841308594, "logps/ref_chosen": -58.88715362548828, "logps/ref_rejected": -81.43145751953125, "logps/rejected": -304.6734619140625, "loss": 1.2437, "margin_dpo/margin_mean": 44.796417236328125, "margin_dpo/margin_std": 112.09017181396484, "step": 598 }, { "KL/chosen_KL_mean": -180.0388946533203, "KL/mean": -206.260009765625, "KL/rejected_KL_mean": -232.48114013671875, "KL/std": 100.97914123535156, "epoch": 0.9055177626606198, "fcm_dpo/beta": 0.005800665356218815, "fcm_dpo/delta": 0.09832191467285156, "fcm_dpo/margin": 52.442222595214844, "fcm_dpo/q_t": 0.43262118101119995, "grad_norm": 16.923696517944336, "learning_rate": 1.3749795321332885e-08, "logits/chosen": 0.7779073715209961, "logits/rejected": 0.7366725206375122, "logps/chosen": -237.64608764648438, "logps/ref_chosen": -57.60719299316406, "logps/ref_rejected": -71.80469512939453, "logps/rejected": -304.28582763671875, "loss": 1.2083, "margin_dpo/margin_mean": 52.44222640991211, "margin_dpo/margin_std": 114.17462158203125, "step": 599 }, { "KL/chosen_KL_mean": -163.935546875, "KL/mean": -195.71664428710938, "KL/rejected_KL_mean": -227.49774169921875, "KL/std": 111.24919128417969, "epoch": 0.9070294784580499, "fcm_dpo/beta": 0.005879203788936138, "fcm_dpo/delta": 0.02595154196023941, "fcm_dpo/margin": 63.562198638916016, "fcm_dpo/q_t": 0.4192517399787903, "grad_norm": 17.71397590637207, "learning_rate": 1.3320646032487393e-08, "logits/chosen": 0.766040563583374, "logits/rejected": 0.7092505097389221, "logps/chosen": -222.3778533935547, "logps/ref_chosen": -58.44231414794922, "logps/ref_rejected": -83.64639282226562, "logps/rejected": -311.1441345214844, "loss": 1.1623, "margin_dpo/margin_mean": 63.56219482421875, "margin_dpo/margin_std": 118.50975799560547, "step": 600 }, { "KL/chosen_KL_mean": -147.6629638671875, "KL/mean": -190.49981689453125, "KL/rejected_KL_mean": -233.33670043945312, "KL/std": 115.73957061767578, "epoch": 0.90854119425548, "fcm_dpo/beta": 0.005733265075832605, "fcm_dpo/delta": -0.09673769026994705, "fcm_dpo/margin": 85.6737289428711, "fcm_dpo/q_t": 0.3916672468185425, "grad_norm": 13.079845428466797, "learning_rate": 1.2898117173950868e-08, "logits/chosen": 0.7403521537780762, "logits/rejected": 0.6603990793228149, "logps/chosen": -203.2572784423828, "logps/ref_chosen": -55.59432601928711, "logps/ref_rejected": -83.68630981445312, "logps/rejected": -317.02301025390625, "loss": 1.0838, "margin_dpo/margin_mean": 85.67372131347656, "margin_dpo/margin_std": 131.4403076171875, "step": 601 }, { "KL/chosen_KL_mean": -136.67974853515625, "KL/mean": -176.0723876953125, "KL/rejected_KL_mean": -215.46502685546875, "KL/std": 99.88746643066406, "epoch": 0.91005291005291, "fcm_dpo/beta": 0.005706445779651403, "fcm_dpo/delta": -0.05199021100997925, "fcm_dpo/margin": 78.78524780273438, "fcm_dpo/q_t": 0.39741697907447815, "grad_norm": 13.740655899047852, "learning_rate": 1.2482220564763667e-08, "logits/chosen": 0.7226306200027466, "logits/rejected": 0.6935929656028748, "logps/chosen": -193.02894592285156, "logps/ref_chosen": -56.349185943603516, "logps/ref_rejected": -71.9959716796875, "logps/rejected": -287.46099853515625, "loss": 1.0727, "margin_dpo/margin_mean": 78.7852554321289, "margin_dpo/margin_std": 106.68989562988281, "step": 602 }, { "KL/chosen_KL_mean": -149.92149353027344, "KL/mean": -188.60498046875, "KL/rejected_KL_mean": -227.28848266601562, "KL/std": 99.36843872070312, "epoch": 0.9115646258503401, "fcm_dpo/beta": 0.005627095699310303, "fcm_dpo/delta": -0.03714841976761818, "fcm_dpo/margin": 77.36700439453125, "fcm_dpo/q_t": 0.40159088373184204, "grad_norm": 14.270939826965332, "learning_rate": 1.2072967838448051e-08, "logits/chosen": 0.6830695867538452, "logits/rejected": 0.6326911449432373, "logps/chosen": -203.08987426757812, "logps/ref_chosen": -53.16838836669922, "logps/ref_rejected": -73.8604736328125, "logps/rejected": -301.14898681640625, "loss": 1.0973, "margin_dpo/margin_mean": 77.36700439453125, "margin_dpo/margin_std": 116.27694702148438, "step": 603 }, { "KL/chosen_KL_mean": -147.6639862060547, "KL/mean": -180.3508758544922, "KL/rejected_KL_mean": -213.0377655029297, "KL/std": 95.46358489990234, "epoch": 0.9130763416477702, "fcm_dpo/beta": 0.005667116958647966, "fcm_dpo/delta": 0.030490772798657417, "fcm_dpo/margin": 65.37378692626953, "fcm_dpo/q_t": 0.41794219613075256, "grad_norm": 16.183061599731445, "learning_rate": 1.1670370442682459e-08, "logits/chosen": 0.6764880418777466, "logits/rejected": 0.6834244132041931, "logps/chosen": -220.31341552734375, "logps/ref_chosen": -72.64942169189453, "logps/ref_rejected": -69.8792724609375, "logps/rejected": -282.91705322265625, "loss": 1.1664, "margin_dpo/margin_mean": 65.37378692626953, "margin_dpo/margin_std": 125.50735473632812, "step": 604 }, { "KL/chosen_KL_mean": -166.05267333984375, "KL/mean": -202.0325469970703, "KL/rejected_KL_mean": -238.01242065429688, "KL/std": 99.51739501953125, "epoch": 0.9145880574452003, "fcm_dpo/beta": 0.0056418003514409065, "fcm_dpo/delta": -0.006424885243177414, "fcm_dpo/margin": 71.95973205566406, "fcm_dpo/q_t": 0.40800243616104126, "grad_norm": 15.322138786315918, "learning_rate": 1.1274439638981532e-08, "logits/chosen": 0.7631226778030396, "logits/rejected": 0.7103903293609619, "logps/chosen": -227.66552734375, "logps/ref_chosen": -61.61284637451172, "logps/ref_rejected": -79.34398651123047, "logps/rejected": -317.3564147949219, "loss": 1.1273, "margin_dpo/margin_mean": 71.95973205566406, "margin_dpo/margin_std": 118.6252212524414, "step": 605 }, { "KL/chosen_KL_mean": -146.68096923828125, "KL/mean": -189.8187255859375, "KL/rejected_KL_mean": -232.95648193359375, "KL/std": 103.99552917480469, "epoch": 0.9160997732426304, "fcm_dpo/beta": 0.005604826845228672, "fcm_dpo/delta": -0.08777381479740143, "fcm_dpo/margin": 86.2755126953125, "fcm_dpo/q_t": 0.39236563444137573, "grad_norm": 15.880330085754395, "learning_rate": 1.0885186502381016e-08, "logits/chosen": 0.6865954995155334, "logits/rejected": 0.6223019361495972, "logps/chosen": -201.14520263671875, "logps/ref_chosen": -54.46424102783203, "logps/ref_rejected": -79.62708282470703, "logps/rejected": -312.58355712890625, "loss": 1.0655, "margin_dpo/margin_mean": 86.2755126953125, "margin_dpo/margin_std": 120.58283996582031, "step": 606 }, { "KL/chosen_KL_mean": -161.45761108398438, "KL/mean": -198.7611083984375, "KL/rejected_KL_mean": -236.0645751953125, "KL/std": 100.57814025878906, "epoch": 0.9176114890400605, "fcm_dpo/beta": 0.005495757795870304, "fcm_dpo/delta": -0.01238175481557846, "fcm_dpo/margin": 74.60693359375, "fcm_dpo/q_t": 0.4067836403846741, "grad_norm": 12.885636329650879, "learning_rate": 1.0502621921127774e-08, "logits/chosen": 0.6752293109893799, "logits/rejected": 0.6519873142242432, "logps/chosen": -224.3184814453125, "logps/ref_chosen": -62.86086654663086, "logps/ref_rejected": -72.5501937866211, "logps/rejected": -308.61474609375, "loss": 1.119, "margin_dpo/margin_mean": 74.60693359375, "margin_dpo/margin_std": 114.7344970703125, "step": 607 }, { "KL/chosen_KL_mean": -164.59835815429688, "KL/mean": -205.1026153564453, "KL/rejected_KL_mean": -245.6068572998047, "KL/std": 110.28776550292969, "epoch": 0.9191232048374905, "fcm_dpo/beta": 0.005514230579137802, "fcm_dpo/delta": -0.04886316880583763, "fcm_dpo/margin": 81.00849914550781, "fcm_dpo/q_t": 0.3991330862045288, "grad_norm": 13.691490173339844, "learning_rate": 1.0126756596375685e-08, "logits/chosen": 0.7040418982505798, "logits/rejected": 0.6250287294387817, "logps/chosen": -227.77908325195312, "logps/ref_chosen": -63.18071746826172, "logps/ref_rejected": -99.15888214111328, "logps/rejected": -344.7657470703125, "loss": 1.078, "margin_dpo/margin_mean": 81.00849914550781, "margin_dpo/margin_std": 112.37265014648438, "step": 608 }, { "KL/chosen_KL_mean": -146.50006103515625, "KL/mean": -189.08883666992188, "KL/rejected_KL_mean": -231.67758178710938, "KL/std": 99.16598510742188, "epoch": 0.9206349206349206, "fcm_dpo/beta": 0.0054123280569911, "fcm_dpo/delta": -0.06470134854316711, "fcm_dpo/margin": 85.17752075195312, "fcm_dpo/q_t": 0.39238014817237854, "grad_norm": 12.95341682434082, "learning_rate": 9.757601041885694e-09, "logits/chosen": 0.785415530204773, "logits/rejected": 0.7473229169845581, "logps/chosen": -195.123291015625, "logps/ref_chosen": -48.62322235107422, "logps/ref_rejected": -68.28271484375, "logps/rejected": -299.9602966308594, "loss": 1.0493, "margin_dpo/margin_mean": 85.17752075195312, "margin_dpo/margin_std": 98.3773193359375, "step": 609 }, { "KL/chosen_KL_mean": -160.15623474121094, "KL/mean": -201.11212158203125, "KL/rejected_KL_mean": -242.06800842285156, "KL/std": 109.5565414428711, "epoch": 0.9221466364323507, "fcm_dpo/beta": 0.005365372635424137, "fcm_dpo/delta": -0.041763849556446075, "fcm_dpo/margin": 81.91178131103516, "fcm_dpo/q_t": 0.4016649127006531, "grad_norm": 12.831808090209961, "learning_rate": 9.395165583732379e-09, "logits/chosen": 0.7413656711578369, "logits/rejected": 0.7341820001602173, "logps/chosen": -232.82135009765625, "logps/ref_chosen": -72.66513061523438, "logps/ref_rejected": -87.15310668945312, "logps/rejected": -329.22113037109375, "loss": 1.0972, "margin_dpo/margin_mean": 81.91178894042969, "margin_dpo/margin_std": 123.76081085205078, "step": 610 }, { "KL/chosen_KL_mean": -147.50057983398438, "KL/mean": -180.25575256347656, "KL/rejected_KL_mean": -213.01092529296875, "KL/std": 96.39732360839844, "epoch": 0.9236583522297808, "fcm_dpo/beta": 0.005401009693741798, "fcm_dpo/delta": 0.047877293080091476, "fcm_dpo/margin": 65.51034545898438, "fcm_dpo/q_t": 0.4180787205696106, "grad_norm": 13.849523544311523, "learning_rate": 9.03946036001449e-09, "logits/chosen": 0.7488099932670593, "logits/rejected": 0.7005026340484619, "logps/chosen": -195.8091583251953, "logps/ref_chosen": -48.30857849121094, "logps/ref_rejected": -70.6141128540039, "logps/rejected": -283.62506103515625, "loss": 1.1319, "margin_dpo/margin_mean": 65.51034545898438, "margin_dpo/margin_std": 98.97489929199219, "step": 611 }, { "KL/chosen_KL_mean": -162.2694549560547, "KL/mean": -210.58135986328125, "KL/rejected_KL_mean": -258.89324951171875, "KL/std": 104.07249450683594, "epoch": 0.9251700680272109, "fcm_dpo/beta": 0.005330703221261501, "fcm_dpo/delta": -0.1212921291589737, "fcm_dpo/margin": 96.62378692626953, "fcm_dpo/q_t": 0.3814992904663086, "grad_norm": 12.464214324951172, "learning_rate": 8.690495320571839e-09, "logits/chosen": 0.6381398439407349, "logits/rejected": 0.5688310861587524, "logps/chosen": -223.50100708007812, "logps/ref_chosen": -61.23155975341797, "logps/ref_rejected": -94.37979888916016, "logps/rejected": -353.2730407714844, "loss": 1.02, "margin_dpo/margin_mean": 96.62378692626953, "margin_dpo/margin_std": 113.25186157226562, "step": 612 }, { "KL/chosen_KL_mean": -140.88748168945312, "KL/mean": -183.08731079101562, "KL/rejected_KL_mean": -225.2871551513672, "KL/std": 110.63365173339844, "epoch": 0.926681783824641, "fcm_dpo/beta": 0.005276457406580448, "fcm_dpo/delta": -0.04751266539096832, "fcm_dpo/margin": 84.39967346191406, "fcm_dpo/q_t": 0.39712822437286377, "grad_norm": 11.47330093383789, "learning_rate": 8.348280226706722e-09, "logits/chosen": 0.6416479349136353, "logits/rejected": 0.6371424198150635, "logps/chosen": -194.87057495117188, "logps/ref_chosen": -53.98310852050781, "logps/ref_rejected": -58.32208251953125, "logps/rejected": -283.6092529296875, "loss": 1.0729, "margin_dpo/margin_mean": 84.39967346191406, "margin_dpo/margin_std": 112.11402893066406, "step": 613 }, { "KL/chosen_KL_mean": -160.6935272216797, "KL/mean": -200.44207763671875, "KL/rejected_KL_mean": -240.19061279296875, "KL/std": 94.97077941894531, "epoch": 0.9281934996220711, "fcm_dpo/beta": 0.005260440520942211, "fcm_dpo/delta": -0.019519072026014328, "fcm_dpo/margin": 79.4970703125, "fcm_dpo/q_t": 0.4030989706516266, "grad_norm": 15.34008502960205, "learning_rate": 8.012824650910937e-09, "logits/chosen": 0.7439731359481812, "logits/rejected": 0.7354336977005005, "logps/chosen": -220.9365692138672, "logps/ref_chosen": -60.24303436279297, "logps/ref_rejected": -72.26258850097656, "logps/rejected": -312.45318603515625, "loss": 1.0915, "margin_dpo/margin_mean": 79.4970703125, "margin_dpo/margin_std": 109.16985321044922, "step": 614 }, { "KL/chosen_KL_mean": -169.2880859375, "KL/mean": -212.89573669433594, "KL/rejected_KL_mean": -256.50341796875, "KL/std": 123.34716033935547, "epoch": 0.9297052154195011, "fcm_dpo/beta": 0.005149473436176777, "fcm_dpo/delta": -0.052740830928087234, "fcm_dpo/margin": 87.21531677246094, "fcm_dpo/q_t": 0.399341881275177, "grad_norm": 13.794776916503906, "learning_rate": 7.684137976598088e-09, "logits/chosen": 0.6767693161964417, "logits/rejected": 0.618577778339386, "logps/chosen": -241.38275146484375, "logps/ref_chosen": -72.09467315673828, "logps/ref_rejected": -104.02980041503906, "logps/rejected": -360.533203125, "loss": 1.1001, "margin_dpo/margin_mean": 87.21531677246094, "margin_dpo/margin_std": 134.67889404296875, "step": 615 }, { "KL/chosen_KL_mean": -161.08840942382812, "KL/mean": -194.94741821289062, "KL/rejected_KL_mean": -228.80642700195312, "KL/std": 113.65570068359375, "epoch": 0.9312169312169312, "fcm_dpo/beta": 0.0052018859423696995, "fcm_dpo/delta": 0.04951094835996628, "fcm_dpo/margin": 67.71803283691406, "fcm_dpo/q_t": 0.41964447498321533, "grad_norm": 12.714033126831055, "learning_rate": 7.36222939784098e-09, "logits/chosen": 0.7352438569068909, "logits/rejected": 0.6616415977478027, "logps/chosen": -219.61912536621094, "logps/ref_chosen": -58.530723571777344, "logps/ref_rejected": -75.48025512695312, "logps/rejected": -304.28668212890625, "loss": 1.1425, "margin_dpo/margin_mean": 67.71803283691406, "margin_dpo/margin_std": 109.96153259277344, "step": 616 }, { "KL/chosen_KL_mean": -169.83023071289062, "KL/mean": -201.72604370117188, "KL/rejected_KL_mean": -233.621826171875, "KL/std": 104.35264587402344, "epoch": 0.9327286470143613, "fcm_dpo/beta": 0.005237380973994732, "fcm_dpo/delta": 0.06759772449731827, "fcm_dpo/margin": 63.79159927368164, "fcm_dpo/q_t": 0.4250500798225403, "grad_norm": 16.894447326660156, "learning_rate": 7.047107919114586e-09, "logits/chosen": 0.7362958788871765, "logits/rejected": 0.6854845285415649, "logps/chosen": -227.4389190673828, "logps/ref_chosen": -57.608673095703125, "logps/ref_rejected": -81.22109985351562, "logps/rejected": -314.8429260253906, "loss": 1.1671, "margin_dpo/margin_mean": 63.791595458984375, "margin_dpo/margin_std": 112.23297882080078, "step": 617 }, { "KL/chosen_KL_mean": -156.01065063476562, "KL/mean": -194.25592041015625, "KL/rejected_KL_mean": -232.50120544433594, "KL/std": 114.61581420898438, "epoch": 0.9342403628117913, "fcm_dpo/beta": 0.005274903029203415, "fcm_dpo/delta": -0.0038125645369291306, "fcm_dpo/margin": 76.49055480957031, "fcm_dpo/q_t": 0.410194456577301, "grad_norm": 15.224896430969238, "learning_rate": 6.738782355044048e-09, "logits/chosen": 0.7007787227630615, "logits/rejected": 0.6046236753463745, "logps/chosen": -212.7065887451172, "logps/ref_chosen": -56.69594192504883, "logps/ref_rejected": -85.92362976074219, "logps/rejected": -318.4248352050781, "loss": 1.1073, "margin_dpo/margin_mean": 76.49055480957031, "margin_dpo/margin_std": 115.4001235961914, "step": 618 }, { "KL/chosen_KL_mean": -149.36265563964844, "KL/mean": -190.2163543701172, "KL/rejected_KL_mean": -231.070068359375, "KL/std": 108.53770446777344, "epoch": 0.9357520786092215, "fcm_dpo/beta": 0.005254857242107391, "fcm_dpo/delta": -0.03081546165049076, "fcm_dpo/margin": 81.70741271972656, "fcm_dpo/q_t": 0.4019641876220703, "grad_norm": 13.710566520690918, "learning_rate": 6.437261330158206e-09, "logits/chosen": 0.8129085302352905, "logits/rejected": 0.7362833619117737, "logps/chosen": -203.42108154296875, "logps/ref_chosen": -54.05841827392578, "logps/ref_rejected": -83.55493927001953, "logps/rejected": -314.625, "loss": 1.0929, "margin_dpo/margin_mean": 81.70740509033203, "margin_dpo/margin_std": 119.22251892089844, "step": 619 }, { "KL/chosen_KL_mean": -154.7310333251953, "KL/mean": -186.06826782226562, "KL/rejected_KL_mean": -217.40548706054688, "KL/std": 98.9871826171875, "epoch": 0.9372637944066515, "fcm_dpo/beta": 0.005364430136978626, "fcm_dpo/delta": 0.06292795389890671, "fcm_dpo/margin": 62.67444610595703, "fcm_dpo/q_t": 0.4228057861328125, "grad_norm": 13.484994888305664, "learning_rate": 6.142553278648238e-09, "logits/chosen": 0.7374277114868164, "logits/rejected": 0.7379218339920044, "logps/chosen": -218.1007537841797, "logps/ref_chosen": -63.36971664428711, "logps/ref_rejected": -65.68269348144531, "logps/rejected": -283.0881652832031, "loss": 1.1642, "margin_dpo/margin_mean": 62.67444610595703, "margin_dpo/margin_std": 104.49517059326172, "step": 620 }, { "KL/chosen_KL_mean": -151.9742431640625, "KL/mean": -188.44854736328125, "KL/rejected_KL_mean": -224.9228515625, "KL/std": 106.9564208984375, "epoch": 0.9387755102040817, "fcm_dpo/beta": 0.0053487373515963554, "fcm_dpo/delta": 0.009966127574443817, "fcm_dpo/margin": 72.94859313964844, "fcm_dpo/q_t": 0.4138706624507904, "grad_norm": 14.086136817932129, "learning_rate": 5.854666444131934e-09, "logits/chosen": 0.7628237009048462, "logits/rejected": 0.6525827646255493, "logps/chosen": -204.2954559326172, "logps/ref_chosen": -52.321224212646484, "logps/ref_rejected": -88.09001159667969, "logps/rejected": -313.0128479003906, "loss": 1.141, "margin_dpo/margin_mean": 72.94860076904297, "margin_dpo/margin_std": 126.93898010253906, "step": 621 }, { "KL/chosen_KL_mean": -167.28756713867188, "KL/mean": -204.14352416992188, "KL/rejected_KL_mean": -240.99948120117188, "KL/std": 111.68397521972656, "epoch": 0.9402872260015117, "fcm_dpo/beta": 0.005355824716389179, "fcm_dpo/delta": 0.005186108872294426, "fcm_dpo/margin": 73.71192932128906, "fcm_dpo/q_t": 0.41089174151420593, "grad_norm": 16.46088409423828, "learning_rate": 5.573608879422875e-09, "logits/chosen": 0.6860193610191345, "logits/rejected": 0.6397637128829956, "logps/chosen": -227.15301513671875, "logps/ref_chosen": -59.86545944213867, "logps/ref_rejected": -81.86668395996094, "logps/rejected": -322.86614990234375, "loss": 1.121, "margin_dpo/margin_mean": 73.71192932128906, "margin_dpo/margin_std": 115.60111999511719, "step": 622 }, { "KL/chosen_KL_mean": -176.53656005859375, "KL/mean": -214.87255859375, "KL/rejected_KL_mean": -253.20855712890625, "KL/std": 117.26280212402344, "epoch": 0.9417989417989417, "fcm_dpo/beta": 0.005307169631123543, "fcm_dpo/delta": -0.007740970700979233, "fcm_dpo/margin": 76.6719970703125, "fcm_dpo/q_t": 0.40785908699035645, "grad_norm": 14.635030746459961, "learning_rate": 5.299388446305342e-09, "logits/chosen": 0.7361361980438232, "logits/rejected": 0.6836817264556885, "logps/chosen": -243.90501403808594, "logps/ref_chosen": -67.36846160888672, "logps/ref_rejected": -82.02733612060547, "logps/rejected": -335.23590087890625, "loss": 1.1183, "margin_dpo/margin_mean": 76.67198944091797, "margin_dpo/margin_std": 121.35376739501953, "step": 623 }, { "KL/chosen_KL_mean": -149.57037353515625, "KL/mean": -190.38003540039062, "KL/rejected_KL_mean": -231.18972778320312, "KL/std": 110.87216186523438, "epoch": 0.9433106575963719, "fcm_dpo/beta": 0.0052911280654370785, "fcm_dpo/delta": -0.03351764753460884, "fcm_dpo/margin": 81.61933898925781, "fcm_dpo/q_t": 0.4038599133491516, "grad_norm": 15.522960662841797, "learning_rate": 5.03201281531429e-09, "logits/chosen": 0.7098270654678345, "logits/rejected": 0.6187886595726013, "logps/chosen": -200.596923828125, "logps/ref_chosen": -51.02655029296875, "logps/ref_rejected": -76.49203491210938, "logps/rejected": -307.6817626953125, "loss": 1.0977, "margin_dpo/margin_mean": 81.61933898925781, "margin_dpo/margin_std": 122.6544189453125, "step": 624 }, { "KL/chosen_KL_mean": -150.91305541992188, "KL/mean": -183.52633666992188, "KL/rejected_KL_mean": -216.1396484375, "KL/std": 105.79248046875, "epoch": 0.9448223733938019, "fcm_dpo/beta": 0.005336887203156948, "fcm_dpo/delta": 0.053777653723955154, "fcm_dpo/margin": 65.22659301757812, "fcm_dpo/q_t": 0.42401954531669617, "grad_norm": 12.910602569580078, "learning_rate": 4.7714894655209174e-09, "logits/chosen": 0.8031026124954224, "logits/rejected": 0.7135155200958252, "logps/chosen": -205.12066650390625, "logps/ref_chosen": -54.20761489868164, "logps/ref_rejected": -84.93669128417969, "logps/rejected": -301.07635498046875, "loss": 1.1751, "margin_dpo/margin_mean": 65.2265853881836, "margin_dpo/margin_std": 127.71304321289062, "step": 625 }, { "KL/chosen_KL_mean": -142.59837341308594, "KL/mean": -187.04966735839844, "KL/rejected_KL_mean": -231.5009765625, "KL/std": 111.5535888671875, "epoch": 0.9463340891912321, "fcm_dpo/beta": 0.0052881138399243355, "fcm_dpo/delta": -0.0737067237496376, "fcm_dpo/margin": 88.90258026123047, "fcm_dpo/q_t": 0.3975789546966553, "grad_norm": 13.397624015808105, "learning_rate": 4.517825684323323e-09, "logits/chosen": 0.8315858840942383, "logits/rejected": 0.6983362436294556, "logps/chosen": -187.660400390625, "logps/ref_chosen": -45.06201934814453, "logps/ref_rejected": -89.66368103027344, "logps/rejected": -321.1646728515625, "loss": 1.0832, "margin_dpo/margin_mean": 88.902587890625, "margin_dpo/margin_std": 135.9188232421875, "step": 626 }, { "KL/chosen_KL_mean": -164.44793701171875, "KL/mean": -211.04330444335938, "KL/rejected_KL_mean": -257.638671875, "KL/std": 116.6589584350586, "epoch": 0.9478458049886621, "fcm_dpo/beta": 0.005192744545638561, "fcm_dpo/delta": -0.08852745592594147, "fcm_dpo/margin": 93.19075012207031, "fcm_dpo/q_t": 0.38831257820129395, "grad_norm": 14.439326286315918, "learning_rate": 4.271028567242818e-09, "logits/chosen": 0.6702276468276978, "logits/rejected": 0.5530549883842468, "logps/chosen": -223.23898315429688, "logps/ref_chosen": -58.791053771972656, "logps/ref_rejected": -94.90802001953125, "logps/rejected": -352.54669189453125, "loss": 1.0465, "margin_dpo/margin_mean": 93.19074249267578, "margin_dpo/margin_std": 117.09492492675781, "step": 627 }, { "KL/chosen_KL_mean": -149.05197143554688, "KL/mean": -193.82723999023438, "KL/rejected_KL_mean": -238.60252380371094, "KL/std": 108.63807678222656, "epoch": 0.9493575207860923, "fcm_dpo/beta": 0.005168614909052849, "fcm_dpo/delta": -0.06649599969387054, "fcm_dpo/margin": 89.55052947998047, "fcm_dpo/q_t": 0.3932734727859497, "grad_norm": 14.440736770629883, "learning_rate": 4.0311050177251895e-09, "logits/chosen": 0.7511168718338013, "logits/rejected": 0.7099975347518921, "logps/chosen": -201.85556030273438, "logps/ref_chosen": -52.80357360839844, "logps/ref_rejected": -76.49468994140625, "logps/rejected": -315.09722900390625, "loss": 1.0868, "margin_dpo/margin_mean": 89.55052185058594, "margin_dpo/margin_std": 124.22722625732422, "step": 628 }, { "KL/chosen_KL_mean": -161.83309936523438, "KL/mean": -194.438720703125, "KL/rejected_KL_mean": -227.04434204101562, "KL/std": 101.754150390625, "epoch": 0.9508692365835223, "fcm_dpo/beta": 0.005171940661966801, "fcm_dpo/delta": 0.06475324183702469, "fcm_dpo/margin": 65.21124267578125, "fcm_dpo/q_t": 0.4241940975189209, "grad_norm": 13.463421821594238, "learning_rate": 3.798061746947995e-09, "logits/chosen": 0.7015402317047119, "logits/rejected": 0.7001262307167053, "logps/chosen": -232.55059814453125, "logps/ref_chosen": -70.71749877929688, "logps/ref_rejected": -78.96273803710938, "logps/rejected": -306.007080078125, "loss": 1.1524, "margin_dpo/margin_mean": 65.21124267578125, "margin_dpo/margin_std": 109.4168930053711, "step": 629 }, { "KL/chosen_KL_mean": -149.88983154296875, "KL/mean": -194.91024780273438, "KL/rejected_KL_mean": -239.9306640625, "KL/std": 109.47689819335938, "epoch": 0.9523809523809523, "fcm_dpo/beta": 0.00514186080545187, "fcm_dpo/delta": -0.06599334627389908, "fcm_dpo/margin": 90.04084014892578, "fcm_dpo/q_t": 0.39410853385925293, "grad_norm": 11.221573829650879, "learning_rate": 3.5719052736323806e-09, "logits/chosen": 0.6838537454605103, "logits/rejected": 0.6431096792221069, "logps/chosen": -206.09124755859375, "logps/ref_chosen": -56.201412200927734, "logps/ref_rejected": -74.69807434082031, "logps/rejected": -314.62872314453125, "loss": 1.061, "margin_dpo/margin_mean": 90.04084777832031, "margin_dpo/margin_std": 117.41181945800781, "step": 630 }, { "KL/chosen_KL_mean": -143.79959106445312, "KL/mean": -194.85324096679688, "KL/rejected_KL_mean": -245.9068603515625, "KL/std": 116.61567687988281, "epoch": 0.9538926681783825, "fcm_dpo/beta": 0.004982981365174055, "fcm_dpo/delta": -0.11698402464389801, "fcm_dpo/margin": 102.10726928710938, "fcm_dpo/q_t": 0.38580942153930664, "grad_norm": 13.442806243896484, "learning_rate": 3.352641923861144e-09, "logits/chosen": 0.798941969871521, "logits/rejected": 0.6885172128677368, "logps/chosen": -202.62020874023438, "logps/ref_chosen": -58.82059860229492, "logps/ref_rejected": -96.51437377929688, "logps/rejected": -342.4212646484375, "loss": 1.0399, "margin_dpo/margin_mean": 102.10726928710938, "margin_dpo/margin_std": 128.56503295898438, "step": 631 }, { "KL/chosen_KL_mean": -146.59799194335938, "KL/mean": -193.16958618164062, "KL/rejected_KL_mean": -239.74114990234375, "KL/std": 102.21504211425781, "epoch": 0.9554043839758125, "fcm_dpo/beta": 0.004959441255778074, "fcm_dpo/delta": -0.06491100043058395, "fcm_dpo/margin": 93.14317321777344, "fcm_dpo/q_t": 0.39288681745529175, "grad_norm": 12.798542022705078, "learning_rate": 3.140277830901428e-09, "logits/chosen": 0.771345853805542, "logits/rejected": 0.7523634433746338, "logps/chosen": -205.38404846191406, "logps/ref_chosen": -58.786048889160156, "logps/ref_rejected": -67.21923828125, "logps/rejected": -306.96038818359375, "loss": 1.0488, "margin_dpo/margin_mean": 93.14315795898438, "margin_dpo/margin_std": 111.02008056640625, "step": 632 }, { "KL/chosen_KL_mean": -141.06906127929688, "KL/mean": -177.5758514404297, "KL/rejected_KL_mean": -214.08261108398438, "KL/std": 98.3267822265625, "epoch": 0.9569160997732427, "fcm_dpo/beta": 0.004975237883627415, "fcm_dpo/delta": 0.03772956132888794, "fcm_dpo/margin": 73.01356506347656, "fcm_dpo/q_t": 0.4177446663379669, "grad_norm": 13.155255317687988, "learning_rate": 2.9348189350335007e-09, "logits/chosen": 0.6996445655822754, "logits/rejected": 0.6450544595718384, "logps/chosen": -193.1992645263672, "logps/ref_chosen": -52.13019561767578, "logps/ref_rejected": -67.23016357421875, "logps/rejected": -281.3127746582031, "loss": 1.1303, "margin_dpo/margin_mean": 73.01355743408203, "margin_dpo/margin_std": 112.07791900634766, "step": 633 }, { "KL/chosen_KL_mean": -169.75709533691406, "KL/mean": -184.82101440429688, "KL/rejected_KL_mean": -199.8849334716797, "KL/std": 102.64701843261719, "epoch": 0.9584278155706727, "fcm_dpo/beta": 0.005162340588867664, "fcm_dpo/delta": 0.24793994426727295, "fcm_dpo/margin": 30.12784194946289, "fcm_dpo/q_t": 0.46613961458206177, "grad_norm": 16.462249755859375, "learning_rate": 2.736270983384276e-09, "logits/chosen": 0.7885172963142395, "logits/rejected": 0.8044285774230957, "logps/chosen": -230.7368927001953, "logps/ref_chosen": -60.97979736328125, "logps/ref_rejected": -58.50825119018555, "logps/rejected": -258.3931884765625, "loss": 1.3256, "margin_dpo/margin_mean": 30.12784194946289, "margin_dpo/margin_std": 118.00814819335938, "step": 634 }, { "KL/chosen_KL_mean": -186.54104614257812, "KL/mean": -216.65505981445312, "KL/rejected_KL_mean": -246.76907348632812, "KL/std": 102.26522827148438, "epoch": 0.9599395313681028, "fcm_dpo/beta": 0.005301056429743767, "fcm_dpo/delta": 0.08293192088603973, "fcm_dpo/margin": 60.2280387878418, "fcm_dpo/q_t": 0.4280802309513092, "grad_norm": 13.260807991027832, "learning_rate": 2.5446395297668287e-09, "logits/chosen": 0.627880871295929, "logits/rejected": 0.5735411047935486, "logps/chosen": -252.51405334472656, "logps/ref_chosen": -65.9730224609375, "logps/ref_rejected": -85.61317443847656, "logps/rejected": -332.3822326660156, "loss": 1.1937, "margin_dpo/margin_mean": 60.2280387878418, "margin_dpo/margin_std": 122.29312896728516, "step": 635 }, { "KL/chosen_KL_mean": -144.86769104003906, "KL/mean": -186.0552978515625, "KL/rejected_KL_mean": -227.24293518066406, "KL/std": 103.58889770507812, "epoch": 0.9614512471655329, "fcm_dpo/beta": 0.005266260355710983, "fcm_dpo/delta": -0.03554299846291542, "fcm_dpo/margin": 82.37523651123047, "fcm_dpo/q_t": 0.39980411529541016, "grad_norm": 11.369194984436035, "learning_rate": 2.359929934524829e-09, "logits/chosen": 0.7130604386329651, "logits/rejected": 0.6156246662139893, "logps/chosen": -194.0078582763672, "logps/ref_chosen": -49.140167236328125, "logps/ref_rejected": -81.26971435546875, "logps/rejected": -308.51263427734375, "loss": 1.0788, "margin_dpo/margin_mean": 82.37523651123047, "margin_dpo/margin_std": 110.31642150878906, "step": 636 }, { "KL/chosen_KL_mean": -194.6241455078125, "KL/mean": -224.41693115234375, "KL/rejected_KL_mean": -254.209716796875, "KL/std": 104.2559814453125, "epoch": 0.9629629629629629, "fcm_dpo/beta": 0.005353373475372791, "fcm_dpo/delta": 0.08315958827733994, "fcm_dpo/margin": 59.5855712890625, "fcm_dpo/q_t": 0.4289915859699249, "grad_norm": 14.297188758850098, "learning_rate": 2.1821473643827137e-09, "logits/chosen": 0.6850694417953491, "logits/rejected": 0.6259503364562988, "logps/chosen": -268.32073974609375, "logps/ref_chosen": -73.69658660888672, "logps/ref_rejected": -83.01487731933594, "logps/rejected": -337.224609375, "loss": 1.2052, "margin_dpo/margin_mean": 59.5855712890625, "margin_dpo/margin_std": 130.34597778320312, "step": 637 }, { "KL/chosen_KL_mean": -168.71023559570312, "KL/mean": -205.56460571289062, "KL/rejected_KL_mean": -242.4189910888672, "KL/std": 103.81689453125, "epoch": 0.9644746787603931, "fcm_dpo/beta": 0.005360821262001991, "fcm_dpo/delta": 0.005060809198766947, "fcm_dpo/margin": 73.70875549316406, "fcm_dpo/q_t": 0.4109768569469452, "grad_norm": 13.191207885742188, "learning_rate": 2.0112967923011646e-09, "logits/chosen": 0.7055521011352539, "logits/rejected": 0.6617870330810547, "logps/chosen": -231.49180603027344, "logps/ref_chosen": -62.78158187866211, "logps/ref_rejected": -85.40478515625, "logps/rejected": -327.82379150390625, "loss": 1.1151, "margin_dpo/margin_mean": 73.70875549316406, "margin_dpo/margin_std": 113.07898712158203, "step": 638 }, { "KL/chosen_KL_mean": -150.687744140625, "KL/mean": -193.58897399902344, "KL/rejected_KL_mean": -236.49020385742188, "KL/std": 110.87168884277344, "epoch": 0.9659863945578231, "fcm_dpo/beta": 0.00532121304422617, "fcm_dpo/delta": -0.059241220355033875, "fcm_dpo/margin": 85.80245208740234, "fcm_dpo/q_t": 0.39776647090911865, "grad_norm": 16.12027359008789, "learning_rate": 1.847382997337943e-09, "logits/chosen": 0.7033539414405823, "logits/rejected": 0.614127516746521, "logps/chosen": -204.45433044433594, "logps/ref_chosen": -53.76658630371094, "logps/ref_rejected": -72.30009460449219, "logps/rejected": -308.790283203125, "loss": 1.0754, "margin_dpo/margin_mean": 85.80245971679688, "margin_dpo/margin_std": 120.51637268066406, "step": 639 }, { "KL/chosen_KL_mean": -157.09376525878906, "KL/mean": -196.0372772216797, "KL/rejected_KL_mean": -234.98080444335938, "KL/std": 106.0317611694336, "epoch": 0.9674981103552532, "fcm_dpo/beta": 0.005276700481772423, "fcm_dpo/delta": -0.011631457135081291, "fcm_dpo/margin": 77.88703918457031, "fcm_dpo/q_t": 0.4065815806388855, "grad_norm": 13.552083015441895, "learning_rate": 1.690410564514244e-09, "logits/chosen": 0.7871816158294678, "logits/rejected": 0.724848747253418, "logps/chosen": -208.51153564453125, "logps/ref_chosen": -51.41777801513672, "logps/ref_rejected": -77.27879333496094, "logps/rejected": -312.25958251953125, "loss": 1.1062, "margin_dpo/margin_mean": 77.88703918457031, "margin_dpo/margin_std": 116.83767700195312, "step": 640 }, { "KL/chosen_KL_mean": -164.87841796875, "KL/mean": -199.95181274414062, "KL/rejected_KL_mean": -235.02520751953125, "KL/std": 97.74984741210938, "epoch": 0.9690098261526833, "fcm_dpo/beta": 0.0052994233556091785, "fcm_dpo/delta": 0.02932356297969818, "fcm_dpo/margin": 70.14678192138672, "fcm_dpo/q_t": 0.4133235216140747, "grad_norm": 14.538914680480957, "learning_rate": 1.5403838846864692e-09, "logits/chosen": 0.7304531335830688, "logits/rejected": 0.7157402038574219, "logps/chosen": -235.93307495117188, "logps/ref_chosen": -71.0546646118164, "logps/ref_rejected": -82.2440185546875, "logps/rejected": -317.26922607421875, "loss": 1.1067, "margin_dpo/margin_mean": 70.14678192138672, "margin_dpo/margin_std": 93.35204315185547, "step": 641 }, { "KL/chosen_KL_mean": -172.72177124023438, "KL/mean": -199.08563232421875, "KL/rejected_KL_mean": -225.44949340820312, "KL/std": 95.5352783203125, "epoch": 0.9705215419501134, "fcm_dpo/beta": 0.005387555807828903, "fcm_dpo/delta": 0.11940450966358185, "fcm_dpo/margin": 52.727725982666016, "fcm_dpo/q_t": 0.43623262643814087, "grad_norm": 16.527711868286133, "learning_rate": 1.3973071544233218e-09, "logits/chosen": 0.6687928438186646, "logits/rejected": 0.689499020576477, "logps/chosen": -241.65106201171875, "logps/ref_chosen": -68.92927551269531, "logps/ref_rejected": -70.85682678222656, "logps/rejected": -296.30633544921875, "loss": 1.2159, "margin_dpo/margin_mean": 52.727718353271484, "margin_dpo/margin_std": 114.44686889648438, "step": 642 }, { "KL/chosen_KL_mean": -157.55726623535156, "KL/mean": -198.06973266601562, "KL/rejected_KL_mean": -238.58216857910156, "KL/std": 103.79234313964844, "epoch": 0.9720332577475435, "fcm_dpo/beta": 0.005408759694546461, "fcm_dpo/delta": -0.04006705433130264, "fcm_dpo/margin": 81.02490234375, "fcm_dpo/q_t": 0.4013475775718689, "grad_norm": 18.932897567749023, "learning_rate": 1.261184375888541e-09, "logits/chosen": 0.6582962274551392, "logits/rejected": 0.5775290727615356, "logps/chosen": -222.86630249023438, "logps/ref_chosen": -65.30903625488281, "logps/ref_rejected": -83.61613464355469, "logps/rejected": -322.19830322265625, "loss": 1.0966, "margin_dpo/margin_mean": 81.02490234375, "margin_dpo/margin_std": 121.24118041992188, "step": 643 }, { "KL/chosen_KL_mean": -147.66122436523438, "KL/mean": -175.08193969726562, "KL/rejected_KL_mean": -202.50262451171875, "KL/std": 93.31697082519531, "epoch": 0.9735449735449735, "fcm_dpo/beta": 0.005520361475646496, "fcm_dpo/delta": 0.0987476035952568, "fcm_dpo/margin": 54.841392517089844, "fcm_dpo/q_t": 0.4305710196495056, "grad_norm": 12.613712310791016, "learning_rate": 1.1320193567288527e-09, "logits/chosen": 0.7914489507675171, "logits/rejected": 0.7597800493240356, "logps/chosen": -198.66384887695312, "logps/ref_chosen": -51.002601623535156, "logps/ref_rejected": -64.46372985839844, "logps/rejected": -266.96636962890625, "loss": 1.204, "margin_dpo/margin_mean": 54.841392517089844, "margin_dpo/margin_std": 113.74671936035156, "step": 644 }, { "KL/chosen_KL_mean": -161.51620483398438, "KL/mean": -195.10723876953125, "KL/rejected_KL_mean": -228.69830322265625, "KL/std": 101.00505828857422, "epoch": 0.9750566893424036, "fcm_dpo/beta": 0.0055358922109007835, "fcm_dpo/delta": 0.02916993945837021, "fcm_dpo/margin": 67.1821060180664, "fcm_dpo/q_t": 0.41418036818504333, "grad_norm": 16.324495315551758, "learning_rate": 1.0098157099674987e-09, "logits/chosen": 0.6692662835121155, "logits/rejected": 0.6541492938995361, "logps/chosen": -222.4796142578125, "logps/ref_chosen": -60.963409423828125, "logps/ref_rejected": -69.73353576660156, "logps/rejected": -298.43182373046875, "loss": 1.1259, "margin_dpo/margin_mean": 67.1821060180664, "margin_dpo/margin_std": 103.56056213378906, "step": 645 }, { "KL/chosen_KL_mean": -177.9462890625, "KL/mean": -210.1239013671875, "KL/rejected_KL_mean": -242.301513671875, "KL/std": 117.13391876220703, "epoch": 0.9765684051398337, "fcm_dpo/beta": 0.0055798254907131195, "fcm_dpo/delta": 0.04242541640996933, "fcm_dpo/margin": 64.35521697998047, "fcm_dpo/q_t": 0.41991138458251953, "grad_norm": 12.960869789123535, "learning_rate": 8.945768539031783e-10, "logits/chosen": 0.7192285060882568, "logits/rejected": 0.6622798442840576, "logps/chosen": -240.23635864257812, "logps/ref_chosen": -62.290069580078125, "logps/ref_rejected": -85.54812622070312, "logps/rejected": -327.8496398925781, "loss": 1.1636, "margin_dpo/margin_mean": 64.35521697998047, "margin_dpo/margin_std": 120.24072265625, "step": 646 }, { "KL/chosen_KL_mean": -175.16905212402344, "KL/mean": -225.5684051513672, "KL/rejected_KL_mean": -275.9677734375, "KL/std": 116.21209716796875, "epoch": 0.9780801209372638, "fcm_dpo/beta": 0.005468328017741442, "fcm_dpo/delta": -0.1598251312971115, "fcm_dpo/margin": 100.79869842529297, "fcm_dpo/q_t": 0.3744158446788788, "grad_norm": 15.467833518981934, "learning_rate": 7.863060120144316e-10, "logits/chosen": 0.7321820259094238, "logits/rejected": 0.6378551721572876, "logps/chosen": -242.6849365234375, "logps/ref_chosen": -67.515869140625, "logps/ref_rejected": -101.50871276855469, "logps/rejected": -377.4764404296875, "loss": 0.9981, "margin_dpo/margin_mean": 100.79869079589844, "margin_dpo/margin_std": 113.80744934082031, "step": 647 }, { "KL/chosen_KL_mean": -175.15017700195312, "KL/mean": -209.93740844726562, "KL/rejected_KL_mean": -244.72459411621094, "KL/std": 101.48037719726562, "epoch": 0.9795918367346939, "fcm_dpo/beta": 0.005409829318523407, "fcm_dpo/delta": 0.0242181196808815, "fcm_dpo/margin": 69.57443237304688, "fcm_dpo/q_t": 0.41655397415161133, "grad_norm": 14.243342399597168, "learning_rate": 6.850062128694045e-10, "logits/chosen": 0.6601248383522034, "logits/rejected": 0.6005022525787354, "logps/chosen": -239.74610900878906, "logps/ref_chosen": -64.59593963623047, "logps/ref_rejected": -83.384033203125, "logps/rejected": -328.108642578125, "loss": 1.1685, "margin_dpo/margin_mean": 69.57443237304688, "margin_dpo/margin_std": 131.31674194335938, "step": 648 }, { "KL/chosen_KL_mean": -156.38809204101562, "KL/mean": -193.30819702148438, "KL/rejected_KL_mean": -230.2283172607422, "KL/std": 100.22056579589844, "epoch": 0.981103552532124, "fcm_dpo/beta": 0.0054200924932956696, "fcm_dpo/delta": -0.0005342736840248108, "fcm_dpo/margin": 73.8402328491211, "fcm_dpo/q_t": 0.4088748097419739, "grad_norm": 17.24158477783203, "learning_rate": 5.906802900412788e-10, "logits/chosen": 0.7577875256538391, "logits/rejected": 0.696466326713562, "logps/chosen": -205.69773864746094, "logps/ref_chosen": -49.30964660644531, "logps/ref_rejected": -73.73710632324219, "logps/rejected": -303.9654235839844, "loss": 1.1238, "margin_dpo/margin_mean": 73.84024047851562, "margin_dpo/margin_std": 118.47221374511719, "step": 649 }, { "KL/chosen_KL_mean": -154.02386474609375, "KL/mean": -191.7683868408203, "KL/rejected_KL_mean": -229.51287841796875, "KL/std": 102.37055969238281, "epoch": 0.982615268329554, "fcm_dpo/beta": 0.0054198214784264565, "fcm_dpo/delta": -0.009670298546552658, "fcm_dpo/margin": 75.4890365600586, "fcm_dpo/q_t": 0.40784576535224915, "grad_norm": 13.829277038574219, "learning_rate": 5.033308820289184e-10, "logits/chosen": 0.7949092388153076, "logits/rejected": 0.7321877479553223, "logps/chosen": -209.08712768554688, "logps/ref_chosen": -55.06325912475586, "logps/ref_rejected": -77.39610290527344, "logps/rejected": -306.90899658203125, "loss": 1.1243, "margin_dpo/margin_mean": 75.48902893066406, "margin_dpo/margin_std": 123.73271179199219, "step": 650 }, { "KL/chosen_KL_mean": -151.89053344726562, "KL/mean": -184.58779907226562, "KL/rejected_KL_mean": -217.28509521484375, "KL/std": 101.58538818359375, "epoch": 0.9841269841269841, "fcm_dpo/beta": 0.005477217026054859, "fcm_dpo/delta": 0.043183956295251846, "fcm_dpo/margin": 65.39456939697266, "fcm_dpo/q_t": 0.42023220658302307, "grad_norm": 13.305492401123047, "learning_rate": 4.2296043218295606e-10, "logits/chosen": 0.775327205657959, "logits/rejected": 0.703120768070221, "logps/chosen": -205.9556884765625, "logps/ref_chosen": -54.065162658691406, "logps/ref_rejected": -77.79080200195312, "logps/rejected": -295.0758972167969, "loss": 1.1409, "margin_dpo/margin_mean": 65.39457702636719, "margin_dpo/margin_std": 107.34722900390625, "step": 651 }, { "KL/chosen_KL_mean": -180.54974365234375, "KL/mean": -211.1258544921875, "KL/rejected_KL_mean": -241.70196533203125, "KL/std": 111.74209594726562, "epoch": 0.9856386999244142, "fcm_dpo/beta": 0.0055029550567269325, "fcm_dpo/delta": 0.06549069285392761, "fcm_dpo/margin": 61.152244567871094, "fcm_dpo/q_t": 0.4275883436203003, "grad_norm": 15.900915145874023, "learning_rate": 3.4957118863768176e-10, "logits/chosen": 0.7466914653778076, "logits/rejected": 0.6997571587562561, "logps/chosen": -244.19003295898438, "logps/ref_chosen": -63.64030456542969, "logps/ref_rejected": -78.86882019042969, "logps/rejected": -320.57080078125, "loss": 1.2034, "margin_dpo/margin_mean": 61.15224075317383, "margin_dpo/margin_std": 134.0066680908203, "step": 652 }, { "KL/chosen_KL_mean": -158.57315063476562, "KL/mean": -195.68978881835938, "KL/rejected_KL_mean": -232.806396484375, "KL/std": 99.21732330322266, "epoch": 0.9871504157218443, "fcm_dpo/beta": 0.005554167553782463, "fcm_dpo/delta": -0.012951435521245003, "fcm_dpo/margin": 74.23324584960938, "fcm_dpo/q_t": 0.4068150818347931, "grad_norm": 15.474823951721191, "learning_rate": 2.831652042480093e-10, "logits/chosen": 0.677382230758667, "logits/rejected": 0.6351486444473267, "logps/chosen": -220.2415313720703, "logps/ref_chosen": -61.668373107910156, "logps/ref_rejected": -73.83012390136719, "logps/rejected": -306.63653564453125, "loss": 1.1146, "margin_dpo/margin_mean": 74.23324584960938, "margin_dpo/margin_std": 116.3241958618164, "step": 653 }, { "KL/chosen_KL_mean": -166.5314178466797, "KL/mean": -201.0532989501953, "KL/rejected_KL_mean": -235.5751953125, "KL/std": 117.71331787109375, "epoch": 0.9886621315192744, "fcm_dpo/beta": 0.0054998332634568214, "fcm_dpo/delta": 0.019461914896965027, "fcm_dpo/margin": 69.04376220703125, "fcm_dpo/q_t": 0.41467058658599854, "grad_norm": 14.01585865020752, "learning_rate": 2.2374433653205016e-10, "logits/chosen": 0.6830437183380127, "logits/rejected": 0.5889010429382324, "logps/chosen": -224.0996856689453, "logps/ref_chosen": -57.568267822265625, "logps/ref_rejected": -87.74789428710938, "logps/rejected": -323.3230895996094, "loss": 1.1355, "margin_dpo/margin_mean": 69.04376220703125, "margin_dpo/margin_std": 108.8658447265625, "step": 654 }, { "KL/chosen_KL_mean": -131.15521240234375, "KL/mean": -182.16342163085938, "KL/rejected_KL_mean": -233.171630859375, "KL/std": 102.94525146484375, "epoch": 0.9901738473167044, "fcm_dpo/beta": 0.005375571083277464, "fcm_dpo/delta": -0.16030044853687286, "fcm_dpo/margin": 102.01641082763672, "fcm_dpo/q_t": 0.37344735860824585, "grad_norm": 12.33569049835205, "learning_rate": 1.7131024761923852e-10, "logits/chosen": 0.7126660346984863, "logits/rejected": 0.627194881439209, "logps/chosen": -183.3023681640625, "logps/ref_chosen": -52.14714813232422, "logps/ref_rejected": -80.85014343261719, "logps/rejected": -314.0217590332031, "loss": 0.9847, "margin_dpo/margin_mean": 102.01641082763672, "margin_dpo/margin_std": 100.66226196289062, "step": 655 }, { "KL/chosen_KL_mean": -172.86846923828125, "KL/mean": -209.77020263671875, "KL/rejected_KL_mean": -246.67193603515625, "KL/std": 104.92657470703125, "epoch": 0.9916855631141346, "fcm_dpo/beta": 0.0053861262276768684, "fcm_dpo/delta": 0.0025822424795478582, "fcm_dpo/margin": 73.80348205566406, "fcm_dpo/q_t": 0.40978139638900757, "grad_norm": 11.625679969787598, "learning_rate": 1.2586440420372934e-10, "logits/chosen": 0.6522502899169922, "logits/rejected": 0.6087788343429565, "logps/chosen": -246.12518310546875, "logps/ref_chosen": -73.25672912597656, "logps/ref_rejected": -85.35127258300781, "logps/rejected": -332.023193359375, "loss": 1.1199, "margin_dpo/margin_mean": 73.80347442626953, "margin_dpo/margin_std": 116.68196105957031, "step": 656 }, { "KL/chosen_KL_mean": -153.27273559570312, "KL/mean": -199.0772247314453, "KL/rejected_KL_mean": -244.88169860839844, "KL/std": 116.36036682128906, "epoch": 0.9931972789115646, "fcm_dpo/beta": 0.005317493341863155, "fcm_dpo/delta": -0.09151517599821091, "fcm_dpo/margin": 91.60895538330078, "fcm_dpo/q_t": 0.3905050456523895, "grad_norm": 11.552481651306152, "learning_rate": 8.740807750345913e-11, "logits/chosen": 0.7955504059791565, "logits/rejected": 0.7159940004348755, "logps/chosen": -202.99612426757812, "logps/ref_chosen": -49.72339630126953, "logps/ref_rejected": -75.1568603515625, "logps/rejected": -320.03857421875, "loss": 1.0617, "margin_dpo/margin_mean": 91.60896301269531, "margin_dpo/margin_std": 126.87873840332031, "step": 657 }, { "KL/chosen_KL_mean": -163.15029907226562, "KL/mean": -200.5810089111328, "KL/rejected_KL_mean": -238.01170349121094, "KL/std": 119.92159271240234, "epoch": 0.9947089947089947, "fcm_dpo/beta": 0.005321793258190155, "fcm_dpo/delta": 0.0008049830794334412, "fcm_dpo/margin": 74.86140441894531, "fcm_dpo/q_t": 0.41124677658081055, "grad_norm": 12.123971939086914, "learning_rate": 5.594234322453539e-11, "logits/chosen": 0.7718138098716736, "logits/rejected": 0.7259960770606995, "logps/chosen": -226.1966552734375, "logps/ref_chosen": -63.04634094238281, "logps/ref_rejected": -83.44963073730469, "logps/rejected": -321.4613342285156, "loss": 1.1559, "margin_dpo/margin_mean": 74.86140441894531, "margin_dpo/margin_std": 139.01248168945312, "step": 658 }, { "KL/chosen_KL_mean": -165.0323028564453, "KL/mean": -194.35629272460938, "KL/rejected_KL_mean": -223.68023681640625, "KL/std": 104.22109985351562, "epoch": 0.9962207105064248, "fcm_dpo/beta": 0.005332798697054386, "fcm_dpo/delta": 0.09000739455223083, "fcm_dpo/margin": 58.64795684814453, "fcm_dpo/q_t": 0.42909497022628784, "grad_norm": 17.333219528198242, "learning_rate": 3.146808153123293e-11, "logits/chosen": 0.7888455390930176, "logits/rejected": 0.7255823612213135, "logps/chosen": -220.1125030517578, "logps/ref_chosen": -55.0802001953125, "logps/ref_rejected": -71.91049194335938, "logps/rejected": -295.59075927734375, "loss": 1.195, "margin_dpo/margin_mean": 58.6479606628418, "margin_dpo/margin_std": 118.41319274902344, "step": 659 }, { "KL/chosen_KL_mean": -156.80108642578125, "KL/mean": -203.4031982421875, "KL/rejected_KL_mean": -250.00535583496094, "KL/std": 112.88031768798828, "epoch": 0.9977324263038548, "fcm_dpo/beta": 0.005306331440806389, "fcm_dpo/delta": -0.09946230053901672, "fcm_dpo/margin": 93.20425415039062, "fcm_dpo/q_t": 0.38910984992980957, "grad_norm": 13.725273132324219, "learning_rate": 1.3985977021235829e-11, "logits/chosen": 0.8498274087905884, "logits/rejected": 0.7831590175628662, "logps/chosen": -211.32699584960938, "logps/ref_chosen": -54.525917053222656, "logps/ref_rejected": -81.23604583740234, "logps/rejected": -331.24139404296875, "loss": 1.04, "margin_dpo/margin_mean": 93.20425415039062, "margin_dpo/margin_std": 120.00505828857422, "step": 660 }, { "KL/chosen_KL_mean": -181.01803588867188, "KL/mean": -210.42062377929688, "KL/rejected_KL_mean": -239.8232421875, "KL/std": 113.02836608886719, "epoch": 0.999244142101285, "fcm_dpo/beta": 0.005359668284654617, "fcm_dpo/delta": 0.08725185692310333, "fcm_dpo/margin": 58.80520248413086, "fcm_dpo/q_t": 0.4290045499801636, "grad_norm": 12.07127857208252, "learning_rate": 3.4965187065971735e-12, "logits/chosen": 0.7182981967926025, "logits/rejected": 0.6424450874328613, "logps/chosen": -241.3906707763672, "logps/ref_chosen": -60.37263870239258, "logps/ref_rejected": -77.42874145507812, "logps/rejected": -317.25201416015625, "loss": 1.2084, "margin_dpo/margin_mean": 58.80520248413086, "margin_dpo/margin_std": 129.78515625, "step": 661 }, { "epoch": 0.999244142101285, "step": 661, "total_flos": 0.0, "train_loss": 1.1488945873710643, "train_runtime": 1653.0477, "train_samples_per_second": 25.611, "train_steps_per_second": 0.4 } ], "logging_steps": 1, "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 }