{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.999244142101285, "eval_steps": 200, "global_step": 661, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0015117157974300832, "fcm_dpo/beta": 0.11051072180271149, "fcm_dpo/delta": 0.49971169233322144, "fcm_dpo/margin": -0.0013532638549804688, "fcm_dpo/q_t": 0.5000448226928711, "grad_norm": 31.18895149230957, "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 }, { "epoch": 0.0030234315948601664, "fcm_dpo/beta": 0.11625976860523224, "fcm_dpo/delta": 0.49490365386009216, "fcm_dpo/margin": 0.037450045347213745, "fcm_dpo/q_t": 0.4989655911922455, "grad_norm": 32.422725677490234, "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": 1.3819, "margin_dpo/margin_mean": 0.03744968771934509, "margin_dpo/margin_std": 0.27811938524246216, "step": 2 }, { "epoch": 0.0045351473922902496, "fcm_dpo/beta": 0.12837310135364532, "fcm_dpo/delta": 0.4961715638637543, "fcm_dpo/margin": 0.002266407012939453, "fcm_dpo/q_t": 0.4999309778213501, "grad_norm": 40.0953483581543, "learning_rate": 1.4925373134328357e-08, "logits/chosen": 0.0993770956993103, "logits/rejected": 0.06136491894721985, "logps/chosen": -65.4115219116211, "logps/ref_chosen": -65.45726013183594, "logps/ref_rejected": -90.82853698730469, "logps/rejected": -90.78506469726562, "loss": 1.3859, "margin_dpo/margin_mean": 0.0022667646408081055, "margin_dpo/margin_std": 0.26775944232940674, "step": 3 }, { "epoch": 0.006046863189720333, "fcm_dpo/beta": 0.14178214967250824, "fcm_dpo/delta": 0.4972817301750183, "fcm_dpo/margin": -0.031194627285003662, "fcm_dpo/q_t": 0.5010493993759155, "grad_norm": 48.728790283203125, "learning_rate": 2.2388059701492534e-08, "logits/chosen": 0.10073457658290863, "logits/rejected": 0.08476720750331879, "logps/chosen": -76.87289428710938, "logps/ref_chosen": -76.86018371582031, "logps/ref_rejected": -79.91523742675781, "logps/rejected": -79.8967514038086, "loss": 1.3908, "margin_dpo/margin_mean": -0.031194984912872314, "margin_dpo/margin_std": 0.3357463479042053, "step": 4 }, { "epoch": 0.007558578987150416, "fcm_dpo/beta": 0.1644459217786789, "fcm_dpo/delta": 0.49898889660835266, "fcm_dpo/margin": -0.022104412317276, "fcm_dpo/q_t": 0.5009359121322632, "grad_norm": 48.499725341796875, "learning_rate": 2.9850746268656714e-08, "logits/chosen": 0.08101461827754974, "logits/rejected": 0.04222995042800903, "logps/chosen": -62.999996185302734, "logps/ref_chosen": -62.97134017944336, "logps/ref_rejected": -79.9192123413086, "logps/rejected": -79.92576599121094, "loss": 1.3905, "margin_dpo/margin_mean": -0.02210336923599243, "margin_dpo/margin_std": 0.283627986907959, "step": 5 }, { "epoch": 0.009070294784580499, "fcm_dpo/beta": 0.1644459217786789, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.07127216458320618, "fcm_dpo/q_t": 0.5029286742210388, "grad_norm": 48.75336456298828, "learning_rate": 3.731343283582089e-08, "logits/chosen": 0.13917648792266846, "logits/rejected": 0.09997415542602539, "logps/chosen": -51.33795166015625, "logps/ref_chosen": -51.30736541748047, "logps/ref_rejected": -82.77239227294922, "logps/rejected": -82.73170471191406, "loss": 1.3986, "margin_dpo/margin_mean": -0.07127270102500916, "margin_dpo/margin_std": 0.29276320338249207, "step": 6 }, { "epoch": 0.010582010582010581, "fcm_dpo/beta": 0.1816796362400055, "fcm_dpo/delta": 0.4983155131340027, "fcm_dpo/margin": -0.0058057308197021484, "fcm_dpo/q_t": 0.500307023525238, "grad_norm": 49.4698600769043, "learning_rate": 4.477611940298507e-08, "logits/chosen": 0.017587212845683098, "logits/rejected": -0.02612943761050701, "logps/chosen": -51.4460334777832, "logps/ref_chosen": -51.45941162109375, "logps/ref_rejected": -66.3828125, "logps/rejected": -66.3636245727539, "loss": 1.388, "margin_dpo/margin_mean": -0.005805850028991699, "margin_dpo/margin_std": 0.2854662537574768, "step": 7 }, { "epoch": 0.012093726379440665, "fcm_dpo/beta": 0.1816796362400055, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.029055893421173096, "fcm_dpo/q_t": 0.5013211965560913, "grad_norm": 52.10679244995117, "learning_rate": 5.223880597014925e-08, "logits/chosen": 0.09742860496044159, "logits/rejected": 0.07472395896911621, "logps/chosen": -62.214691162109375, "logps/ref_chosen": -62.197547912597656, "logps/ref_rejected": -74.66180419921875, "logps/rejected": -74.64989471435547, "loss": 1.3925, "margin_dpo/margin_mean": -0.02905610203742981, "margin_dpo/margin_std": 0.3362266421318054, "step": 8 }, { "epoch": 0.013605442176870748, "fcm_dpo/beta": 0.21118226647377014, "fcm_dpo/delta": 0.9957462549209595, "fcm_dpo/margin": 0.02298596501350403, "fcm_dpo/q_t": 0.498937726020813, "grad_norm": 66.31742858886719, "learning_rate": 5.970149253731343e-08, "logits/chosen": 0.17290404438972473, "logits/rejected": 0.1132848858833313, "logps/chosen": -55.65576934814453, "logps/ref_chosen": -55.629722595214844, "logps/ref_rejected": -86.21221923828125, "logps/rejected": -86.2612533569336, "loss": 1.3826, "margin_dpo/margin_mean": 0.02298620343208313, "margin_dpo/margin_std": 0.30049267411231995, "step": 9 }, { "epoch": 0.015117157974300832, "fcm_dpo/beta": 0.24479255080223083, "fcm_dpo/delta": 0.4950849413871765, "fcm_dpo/margin": 0.01253315806388855, "fcm_dpo/q_t": 0.4993648827075958, "grad_norm": 72.88284301757812, "learning_rate": 6.71641791044776e-08, "logits/chosen": 0.11379828304052353, "logits/rejected": 0.08385583758354187, "logps/chosen": -62.68762969970703, "logps/ref_chosen": -62.69060134887695, "logps/ref_rejected": -90.610107421875, "logps/rejected": -90.61967468261719, "loss": 1.3853, "margin_dpo/margin_mean": 0.01253288984298706, "margin_dpo/margin_std": 0.3701857328414917, "step": 10 }, { "epoch": 0.016628873771730914, "fcm_dpo/beta": 0.26968804001808167, "fcm_dpo/delta": 0.484273761510849, "fcm_dpo/margin": 0.04408371448516846, "fcm_dpo/q_t": 0.4974249601364136, "grad_norm": 79.30926513671875, "learning_rate": 7.462686567164178e-08, "logits/chosen": 0.10774752497673035, "logits/rejected": 0.10090956091880798, "logps/chosen": -65.75579071044922, "logps/ref_chosen": -65.76712036132812, "logps/ref_rejected": -72.4764633178711, "logps/rejected": -72.50921630859375, "loss": 1.376, "margin_dpo/margin_mean": 0.04408392310142517, "margin_dpo/margin_std": 0.2787271738052368, "step": 11 }, { "epoch": 0.018140589569160998, "fcm_dpo/beta": 0.26968804001808167, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.013816118240356445, "fcm_dpo/q_t": 0.5009292364120483, "grad_norm": 76.66712951660156, "learning_rate": 8.208955223880596e-08, "logits/chosen": 0.03750212490558624, "logits/rejected": 0.02106173150241375, "logps/chosen": -60.710899353027344, "logps/ref_chosen": -60.704891204833984, "logps/ref_rejected": -69.41564178466797, "logps/rejected": -69.4078369140625, "loss": 1.3914, "margin_dpo/margin_mean": -0.013815999031066895, "margin_dpo/margin_std": 0.2728922367095947, "step": 12 }, { "epoch": 0.019652305366591082, "fcm_dpo/beta": 0.26968804001808167, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.035725027322769165, "fcm_dpo/q_t": 0.5023995637893677, "grad_norm": 78.57530212402344, "learning_rate": 8.955223880597014e-08, "logits/chosen": 0.10675495862960815, "logits/rejected": 0.044550854712724686, "logps/chosen": -49.91514587402344, "logps/ref_chosen": -49.90925598144531, "logps/ref_rejected": -92.37818145751953, "logps/rejected": -92.34834289550781, "loss": 1.3971, "margin_dpo/margin_mean": -0.03572601079940796, "margin_dpo/margin_std": 0.24918314814567566, "step": 13 }, { "epoch": 0.021164021164021163, "fcm_dpo/beta": 0.31274157762527466, "fcm_dpo/delta": 0.9786568880081177, "fcm_dpo/margin": 0.075018972158432, "fcm_dpo/q_t": 0.49468034505844116, "grad_norm": 91.86367797851562, "learning_rate": 9.701492537313432e-08, "logits/chosen": 0.08078277111053467, "logits/rejected": 0.06322959065437317, "logps/chosen": -60.599849700927734, "logps/ref_chosen": -60.61879348754883, "logps/ref_rejected": -71.79306030273438, "logps/rejected": -71.84913635253906, "loss": 1.3649, "margin_dpo/margin_mean": 0.07501909136772156, "margin_dpo/margin_std": 0.2812075614929199, "step": 14 }, { "epoch": 0.022675736961451247, "fcm_dpo/beta": 0.3279946446418762, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02524346113204956, "fcm_dpo/q_t": 0.5020579099655151, "grad_norm": 109.66602325439453, "learning_rate": 1.044776119402985e-07, "logits/chosen": 0.06694771349430084, "logits/rejected": 0.023975659161806107, "logps/chosen": -63.49495315551758, "logps/ref_chosen": -63.46953582763672, "logps/ref_rejected": -88.88951110839844, "logps/rejected": -88.88968658447266, "loss": 1.3976, "margin_dpo/margin_mean": -0.025244086980819702, "margin_dpo/margin_std": 0.3348177969455719, "step": 15 }, { "epoch": 0.02418745275888133, "fcm_dpo/beta": 0.34503084421157837, "fcm_dpo/delta": 0.4941604435443878, "fcm_dpo/margin": 0.014522925019264221, "fcm_dpo/q_t": 0.4988110661506653, "grad_norm": 93.19413757324219, "learning_rate": 1.1194029850746268e-07, "logits/chosen": 0.11576000601053238, "logits/rejected": 0.07864248752593994, "logps/chosen": -46.55487823486328, "logps/ref_chosen": -46.53229904174805, "logps/ref_rejected": -74.27533721923828, "logps/rejected": -74.31243896484375, "loss": 1.3833, "margin_dpo/margin_mean": 0.014522776007652283, "margin_dpo/margin_std": 0.27633634209632874, "step": 16 }, { "epoch": 0.025699168556311415, "fcm_dpo/beta": 0.39912450313568115, "fcm_dpo/delta": 0.487219899892807, "fcm_dpo/margin": -0.00017780065536499023, "fcm_dpo/q_t": 0.5003235936164856, "grad_norm": 134.72512817382812, "learning_rate": 1.1940298507462686e-07, "logits/chosen": 0.04781803488731384, "logits/rejected": 0.02929597906768322, "logps/chosen": -64.07421875, "logps/ref_chosen": -64.07783508300781, "logps/ref_rejected": -86.40876770019531, "logps/rejected": -86.40497589111328, "loss": 1.3906, "margin_dpo/margin_mean": -0.00017789006233215332, "margin_dpo/margin_std": 0.3190717101097107, "step": 17 }, { "epoch": 0.027210884353741496, "fcm_dpo/beta": 0.4197884202003479, "fcm_dpo/delta": 0.4926441013813019, "fcm_dpo/margin": -0.029925107955932617, "fcm_dpo/q_t": 0.5029613971710205, "grad_norm": 120.34202575683594, "learning_rate": 1.2686567164179106e-07, "logits/chosen": 0.10955735296010971, "logits/rejected": 0.06258425116539001, "logps/chosen": -44.86344528198242, "logps/ref_chosen": -44.87433624267578, "logps/ref_rejected": -70.97604370117188, "logps/rejected": -70.93523406982422, "loss": 1.4019, "margin_dpo/margin_mean": -0.029924869537353516, "margin_dpo/margin_std": 0.30913057923316956, "step": 18 }, { "epoch": 0.02872260015117158, "fcm_dpo/beta": 0.5113855004310608, "fcm_dpo/delta": 0.9829479455947876, "fcm_dpo/margin": 0.03535567224025726, "fcm_dpo/q_t": 0.4958151578903198, "grad_norm": 156.30233764648438, "learning_rate": 1.343283582089552e-07, "logits/chosen": 0.07092909514904022, "logits/rejected": 0.05735887587070465, "logps/chosen": -68.16175842285156, "logps/ref_chosen": -68.1598129272461, "logps/ref_rejected": -81.17138671875, "logps/rejected": -81.20869445800781, "loss": 1.3751, "margin_dpo/margin_mean": 0.03535632789134979, "margin_dpo/margin_std": 0.33560460805892944, "step": 19 }, { "epoch": 0.030234315948601664, "fcm_dpo/beta": 0.621636152267456, "fcm_dpo/delta": 0.9787266254425049, "fcm_dpo/margin": 0.03790883719921112, "fcm_dpo/q_t": 0.49468833208084106, "grad_norm": 184.15403747558594, "learning_rate": 1.4179104477611938e-07, "logits/chosen": 0.11737000942230225, "logits/rejected": 0.09504500031471252, "logps/chosen": -53.65413284301758, "logps/ref_chosen": -53.67856216430664, "logps/ref_rejected": -74.16911315917969, "logps/rejected": -74.18260192871094, "loss": 1.3704, "margin_dpo/margin_mean": 0.03790910542011261, "margin_dpo/margin_std": 0.27916550636291504, "step": 20 }, { "epoch": 0.031746031746031744, "fcm_dpo/beta": 0.719708263874054, "fcm_dpo/delta": 0.49354374408721924, "fcm_dpo/margin": 0.006038039922714233, "fcm_dpo/q_t": 0.49931585788726807, "grad_norm": 208.18948364257812, "learning_rate": 1.4925373134328355e-07, "logits/chosen": 0.09322724491357803, "logits/rejected": 0.06828559935092926, "logps/chosen": -64.68141174316406, "logps/ref_chosen": -64.70155334472656, "logps/ref_rejected": -81.02095031738281, "logps/rejected": -81.0068359375, "loss": 1.397, "margin_dpo/margin_mean": 0.0060374438762664795, "margin_dpo/margin_std": 0.34163737297058105, "step": 21 }, { "epoch": 0.03325774754346183, "fcm_dpo/beta": 0.7948847413063049, "fcm_dpo/delta": 0.4967557489871979, "fcm_dpo/margin": -0.008793145418167114, "fcm_dpo/q_t": 0.501756489276886, "grad_norm": 235.72225952148438, "learning_rate": 1.5671641791044775e-07, "logits/chosen": 0.020929213613271713, "logits/rejected": -0.0007745649782009423, "logps/chosen": -58.05523681640625, "logps/ref_chosen": -58.03599166870117, "logps/ref_rejected": -80.72721862792969, "logps/rejected": -80.7376708984375, "loss": 1.4021, "margin_dpo/margin_mean": -0.008793413639068604, "margin_dpo/margin_std": 0.23543663322925568, "step": 22 }, { "epoch": 0.03476946334089191, "fcm_dpo/beta": 0.8356242179870605, "fcm_dpo/delta": 0.4879206120967865, "fcm_dpo/margin": -0.007578670978546143, "fcm_dpo/q_t": 0.5014785528182983, "grad_norm": 285.7133483886719, "learning_rate": 1.6417910447761193e-07, "logits/chosen": 0.1404346227645874, "logits/rejected": 0.11506737768650055, "logps/chosen": -66.38188934326172, "logps/ref_chosen": -66.35608673095703, "logps/ref_rejected": -93.02769470214844, "logps/rejected": -93.04591369628906, "loss": 1.4157, "margin_dpo/margin_mean": -0.007578998804092407, "margin_dpo/margin_std": 0.3775022029876709, "step": 23 }, { "epoch": 0.036281179138321996, "fcm_dpo/beta": 0.8763637542724609, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.023401737213134766, "fcm_dpo/q_t": 0.5050784349441528, "grad_norm": 237.49981689453125, "learning_rate": 1.716417910447761e-07, "logits/chosen": 0.13775691390037537, "logits/rejected": 0.10474318265914917, "logps/chosen": -54.4842529296875, "logps/ref_chosen": -54.461238861083984, "logps/ref_rejected": -68.33817291259766, "logps/rejected": -68.33778381347656, "loss": 1.4189, "margin_dpo/margin_mean": -0.02340218424797058, "margin_dpo/margin_std": 0.2496114820241928, "step": 24 }, { "epoch": 0.03779289493575208, "fcm_dpo/beta": 1.0109740495681763, "fcm_dpo/delta": 0.933335542678833, "fcm_dpo/margin": 0.07055863738059998, "fcm_dpo/q_t": 0.48359498381614685, "grad_norm": 284.9603271484375, "learning_rate": 1.7910447761194027e-07, "logits/chosen": 0.09323587268590927, "logits/rejected": 0.04254044592380524, "logps/chosen": -60.0122184753418, "logps/ref_chosen": -60.00420379638672, "logps/ref_rejected": -90.47376251220703, "logps/rejected": -90.55233764648438, "loss": 1.3345, "margin_dpo/margin_mean": 0.07055890560150146, "margin_dpo/margin_std": 0.27643194794654846, "step": 25 }, { "epoch": 0.039304610733182165, "fcm_dpo/beta": 1.2268295288085938, "fcm_dpo/delta": 0.9872031211853027, "fcm_dpo/margin": 0.011123806238174438, "fcm_dpo/q_t": 0.49779731035232544, "grad_norm": 363.8032531738281, "learning_rate": 1.8656716417910447e-07, "logits/chosen": 0.12300634384155273, "logits/rejected": 0.1041463315486908, "logps/chosen": -56.83332061767578, "logps/ref_chosen": -56.81915283203125, "logps/ref_rejected": -77.84333038330078, "logps/rejected": -77.86862182617188, "loss": 1.418, "margin_dpo/margin_mean": 0.011123299598693848, "margin_dpo/margin_std": 0.3511189818382263, "step": 26 }, { "epoch": 0.04081632653061224, "fcm_dpo/beta": 1.4186409711837769, "fcm_dpo/delta": 0.4878283143043518, "fcm_dpo/margin": -0.018930166959762573, "fcm_dpo/q_t": 0.5060181617736816, "grad_norm": 426.4020690917969, "learning_rate": 1.9402985074626865e-07, "logits/chosen": 0.10950794070959091, "logits/rejected": 0.08432013541460037, "logps/chosen": -62.87256622314453, "logps/ref_chosen": -62.87702560424805, "logps/ref_rejected": -71.34437561035156, "logps/rejected": -71.32098388671875, "loss": 1.465, "margin_dpo/margin_mean": -0.0189303457736969, "margin_dpo/margin_std": 0.3256291151046753, "step": 27 }, { "epoch": 0.042328042328042326, "fcm_dpo/beta": 1.5392485857009888, "fcm_dpo/delta": 0.40797513723373413, "fcm_dpo/margin": 0.019986987113952637, "fcm_dpo/q_t": 0.49472981691360474, "grad_norm": 428.8187561035156, "learning_rate": 2.0149253731343282e-07, "logits/chosen": 0.05481361597776413, "logits/rejected": 0.04620601236820221, "logps/chosen": -59.83418273925781, "logps/ref_chosen": -59.8333740234375, "logps/ref_rejected": -70.39804077148438, "logps/rejected": -70.4188232421875, "loss": 1.4194, "margin_dpo/margin_mean": 0.019986748695373535, "margin_dpo/margin_std": 0.31196290254592896, "step": 28 }, { "epoch": 0.04383975812547241, "fcm_dpo/beta": 1.7656760215759277, "fcm_dpo/delta": 0.9155327081680298, "fcm_dpo/margin": 0.05319638550281525, "fcm_dpo/q_t": 0.47983771562576294, "grad_norm": 544.4055786132812, "learning_rate": 2.08955223880597e-07, "logits/chosen": 0.13988614082336426, "logits/rejected": 0.12211866676807404, "logps/chosen": -74.13853454589844, "logps/ref_chosen": -74.12020111083984, "logps/ref_rejected": -83.33099365234375, "logps/rejected": -83.40251159667969, "loss": 1.3481, "margin_dpo/margin_mean": 0.053196460008621216, "margin_dpo/margin_std": 0.2625643312931061, "step": 29 }, { "epoch": 0.045351473922902494, "fcm_dpo/beta": 2.0389435291290283, "fcm_dpo/delta": 0.4901547431945801, "fcm_dpo/margin": -0.004268288612365723, "fcm_dpo/q_t": 0.5039973258972168, "grad_norm": 641.551025390625, "learning_rate": 2.1641791044776117e-07, "logits/chosen": 0.13575318455696106, "logits/rejected": 0.08044232428073883, "logps/chosen": -50.786643981933594, "logps/ref_chosen": -50.75128936767578, "logps/ref_rejected": -89.29063415527344, "logps/rejected": -89.32171630859375, "loss": 1.4942, "margin_dpo/margin_mean": -0.004268676042556763, "margin_dpo/margin_std": 0.32202666997909546, "step": 30 }, { "epoch": 0.04686318972033258, "fcm_dpo/beta": 2.3525500297546387, "fcm_dpo/delta": 0.93133544921875, "fcm_dpo/margin": 0.0310114324092865, "fcm_dpo/q_t": 0.4873350262641907, "grad_norm": 821.974365234375, "learning_rate": 2.2388059701492537e-07, "logits/chosen": 0.10764593631029129, "logits/rejected": 0.061387479305267334, "logps/chosen": -65.36897277832031, "logps/ref_chosen": -65.33675384521484, "logps/ref_rejected": -100.76666259765625, "logps/rejected": -100.82989501953125, "loss": 1.4747, "margin_dpo/margin_mean": 0.03101155161857605, "margin_dpo/margin_std": 0.35841095447540283, "step": 31 }, { "epoch": 0.04837490551776266, "fcm_dpo/beta": 2.5499143600463867, "fcm_dpo/delta": 0.3668806552886963, "fcm_dpo/margin": 0.045269906520843506, "fcm_dpo/q_t": 0.47517523169517517, "grad_norm": 814.4436645507812, "learning_rate": 2.3134328358208954e-07, "logits/chosen": 0.10604210197925568, "logits/rejected": 0.09800020605325699, "logps/chosen": -67.19169616699219, "logps/ref_chosen": -67.18333435058594, "logps/ref_rejected": -82.80763244628906, "logps/rejected": -82.86127471923828, "loss": 1.4012, "margin_dpo/margin_mean": 0.045270055532455444, "margin_dpo/margin_std": 0.2860063314437866, "step": 32 }, { "epoch": 0.049886621315192746, "fcm_dpo/beta": 2.7437379360198975, "fcm_dpo/delta": 0.3657742738723755, "fcm_dpo/margin": 0.042653635144233704, "fcm_dpo/q_t": 0.4748280644416809, "grad_norm": 934.1159057617188, "learning_rate": 2.388059701492537e-07, "logits/chosen": 0.0492779016494751, "logits/rejected": 0.022379783913493156, "logps/chosen": -64.0654067993164, "logps/ref_chosen": -64.03948211669922, "logps/ref_rejected": -75.68357849121094, "logps/rejected": -75.75216674804688, "loss": 1.4557, "margin_dpo/margin_mean": 0.042654380202293396, "margin_dpo/margin_std": 0.3218376338481903, "step": 33 }, { "epoch": 0.05139833711262283, "fcm_dpo/beta": 3.2462871074676514, "fcm_dpo/delta": 0.868577241897583, "fcm_dpo/margin": 0.04391145706176758, "fcm_dpo/q_t": 0.47198355197906494, "grad_norm": 959.0977172851562, "learning_rate": 2.4626865671641786e-07, "logits/chosen": 0.0971483588218689, "logits/rejected": 0.0673779547214508, "logps/chosen": -53.695762634277344, "logps/ref_chosen": -53.6642951965332, "logps/ref_rejected": -65.77989959716797, "logps/rejected": -65.85527038574219, "loss": 1.4312, "margin_dpo/margin_mean": 0.04391142725944519, "margin_dpo/margin_std": 0.2787018120288849, "step": 34 }, { "epoch": 0.05291005291005291, "fcm_dpo/beta": 3.5613765716552734, "fcm_dpo/delta": 0.49954742193222046, "fcm_dpo/margin": 0.0079115629196167, "fcm_dpo/q_t": 0.4954897165298462, "grad_norm": 1141.895751953125, "learning_rate": 2.537313432835821e-07, "logits/chosen": 0.04493723437190056, "logits/rejected": 0.022909432649612427, "logps/chosen": -61.07041549682617, "logps/ref_chosen": -61.01686096191406, "logps/ref_rejected": -72.78598022460938, "logps/rejected": -72.84745025634766, "loss": 1.6389, "margin_dpo/margin_mean": 0.0079115629196167, "margin_dpo/margin_std": 0.32450151443481445, "step": 35 }, { "epoch": 0.05442176870748299, "fcm_dpo/beta": 3.89394474029541, "fcm_dpo/delta": 0.39777103066444397, "fcm_dpo/margin": 0.013609737157821655, "fcm_dpo/q_t": 0.5010133981704712, "grad_norm": 1230.4146728515625, "learning_rate": 2.611940298507462e-07, "logits/chosen": 0.1099657341837883, "logits/rejected": 0.056441109627485275, "logps/chosen": -50.61913299560547, "logps/ref_chosen": -50.53736114501953, "logps/ref_rejected": -78.11678314208984, "logps/rejected": -78.212158203125, "loss": 1.6768, "margin_dpo/margin_mean": 0.013609647750854492, "margin_dpo/margin_std": 0.32924020290374756, "step": 36 }, { "epoch": 0.055933484504913075, "fcm_dpo/beta": 4.411220550537109, "fcm_dpo/delta": 0.5480049252510071, "fcm_dpo/margin": 0.10784178972244263, "fcm_dpo/q_t": 0.4541955292224884, "grad_norm": 1664.73095703125, "learning_rate": 2.686567164179104e-07, "logits/chosen": 0.08195307105779648, "logits/rejected": 0.004701277241110802, "logps/chosen": -59.608680725097656, "logps/ref_chosen": -59.55394744873047, "logps/ref_rejected": -108.27702331542969, "logps/rejected": -108.43960571289062, "loss": 1.4659, "margin_dpo/margin_mean": 0.10784146189689636, "margin_dpo/margin_std": 0.4072269797325134, "step": 37 }, { "epoch": 0.05744520030234316, "fcm_dpo/beta": 4.725587844848633, "fcm_dpo/delta": 0.22491098940372467, "fcm_dpo/margin": 0.04019525647163391, "fcm_dpo/q_t": 0.4804103374481201, "grad_norm": 1523.71044921875, "learning_rate": 2.761194029850746e-07, "logits/chosen": 0.056998323649168015, "logits/rejected": 0.04311235621571541, "logps/chosen": -65.85867309570312, "logps/ref_chosen": -65.78836059570312, "logps/ref_rejected": -76.1619873046875, "logps/rejected": -76.27249908447266, "loss": 1.7981, "margin_dpo/margin_mean": 0.040194928646087646, "margin_dpo/margin_std": 0.36567050218582153, "step": 38 }, { "epoch": 0.05895691609977324, "fcm_dpo/beta": 5.4121809005737305, "fcm_dpo/delta": 0.8187992572784424, "fcm_dpo/margin": 0.03373938798904419, "fcm_dpo/q_t": 0.4426850378513336, "grad_norm": 1836.44189453125, "learning_rate": 2.8358208955223876e-07, "logits/chosen": 0.14140446484088898, "logits/rejected": 0.11523914337158203, "logps/chosen": -57.26402282714844, "logps/ref_chosen": -57.17681121826172, "logps/ref_rejected": -79.486328125, "logps/rejected": -79.60729217529297, "loss": 1.9268, "margin_dpo/margin_mean": 0.033740073442459106, "margin_dpo/margin_std": 0.3646671772003174, "step": 39 }, { "epoch": 0.06046863189720333, "fcm_dpo/beta": 5.907527923583984, "fcm_dpo/delta": 0.297378271818161, "fcm_dpo/margin": 0.006361484527587891, "fcm_dpo/q_t": 0.4902653098106384, "grad_norm": 2190.686767578125, "learning_rate": 2.9104477611940296e-07, "logits/chosen": 0.10774768888950348, "logits/rejected": 0.058216311037540436, "logps/chosen": -61.42626953125, "logps/ref_chosen": -61.33416748046875, "logps/ref_rejected": -79.10697174072266, "logps/rejected": -79.20543670654297, "loss": 2.1253, "margin_dpo/margin_mean": 0.00636136531829834, "margin_dpo/margin_std": 0.3456147611141205, "step": 40 }, { "epoch": 0.06198034769463341, "fcm_dpo/beta": 6.096949100494385, "fcm_dpo/delta": 0.3107817769050598, "fcm_dpo/margin": 0.03788435459136963, "fcm_dpo/q_t": 0.4918346107006073, "grad_norm": 2313.122314453125, "learning_rate": 2.985074626865671e-07, "logits/chosen": 0.06469070911407471, "logits/rejected": 0.04410509765148163, "logps/chosen": -67.65518188476562, "logps/ref_chosen": -67.5467300415039, "logps/ref_rejected": -83.87788391113281, "logps/rejected": -84.02423095703125, "loss": 2.1107, "margin_dpo/margin_mean": 0.03788486123085022, "margin_dpo/margin_std": 0.3886951506137848, "step": 41 }, { "epoch": 0.06349206349206349, "fcm_dpo/beta": 6.579242706298828, "fcm_dpo/delta": 0.22767893970012665, "fcm_dpo/margin": 0.00848454236984253, "fcm_dpo/q_t": 0.4764997959136963, "grad_norm": 2254.65283203125, "learning_rate": 3.059701492537313e-07, "logits/chosen": 0.04841721057891846, "logits/rejected": 0.026929516345262527, "logps/chosen": -61.369564056396484, "logps/ref_chosen": -61.26485824584961, "logps/ref_rejected": -76.3629150390625, "logps/rejected": -76.47610473632812, "loss": 2.2164, "margin_dpo/margin_mean": 0.008484512567520142, "margin_dpo/margin_std": 0.33282727003097534, "step": 42 }, { "epoch": 0.06500377928949358, "fcm_dpo/beta": 7.226245880126953, "fcm_dpo/delta": 0.5711226463317871, "fcm_dpo/margin": 0.06204667687416077, "fcm_dpo/q_t": 0.44025903940200806, "grad_norm": 2534.08984375, "learning_rate": 3.134328358208955e-07, "logits/chosen": 0.062497012317180634, "logits/rejected": 0.051983729004859924, "logps/chosen": -71.88862609863281, "logps/ref_chosen": -71.80902862548828, "logps/ref_rejected": -81.12464141845703, "logps/rejected": -81.26628875732422, "loss": 1.9347, "margin_dpo/margin_mean": 0.0620463490486145, "margin_dpo/margin_std": 0.32832396030426025, "step": 43 }, { "epoch": 0.06651549508692366, "fcm_dpo/beta": 7.3753557205200195, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.003143906593322754, "fcm_dpo/q_t": 0.5123412609100342, "grad_norm": 3110.311279296875, "learning_rate": 3.2089552238805965e-07, "logits/chosen": 0.058967474848032, "logits/rejected": 0.02790246158838272, "logps/chosen": -66.68571472167969, "logps/ref_chosen": -66.55043029785156, "logps/ref_rejected": -85.06198120117188, "logps/rejected": -85.19412994384766, "loss": 2.7173, "margin_dpo/margin_mean": -0.00314408540725708, "margin_dpo/margin_std": 0.39732399582862854, "step": 44 }, { "epoch": 0.06802721088435375, "fcm_dpo/beta": 7.334336280822754, "fcm_dpo/delta": -0.0020287036895751953, "fcm_dpo/margin": 0.13652004301548004, "fcm_dpo/q_t": 0.35875630378723145, "grad_norm": 1952.19775390625, "learning_rate": 3.2835820895522385e-07, "logits/chosen": 0.10706457495689392, "logits/rejected": 0.05497971177101135, "logps/chosen": -62.34738540649414, "logps/ref_chosen": -62.24385452270508, "logps/ref_rejected": -92.96665954589844, "logps/rejected": -93.20671081542969, "loss": 1.5806, "margin_dpo/margin_mean": 0.1365204155445099, "margin_dpo/margin_std": 0.33906716108322144, "step": 45 }, { "epoch": 0.06953892668178382, "fcm_dpo/beta": 7.5740742683410645, "fcm_dpo/delta": 0.09216267615556717, "fcm_dpo/margin": 0.12037345767021179, "fcm_dpo/q_t": 0.4014553129673004, "grad_norm": 2483.462646484375, "learning_rate": 3.3582089552238805e-07, "logits/chosen": 0.10916443914175034, "logits/rejected": 0.06387359648942947, "logps/chosen": -61.58867263793945, "logps/ref_chosen": -61.498905181884766, "logps/ref_rejected": -78.91172790527344, "logps/rejected": -79.12187194824219, "loss": 1.8488, "margin_dpo/margin_mean": 0.12037333846092224, "margin_dpo/margin_std": 0.37311580777168274, "step": 46 }, { "epoch": 0.0710506424792139, "fcm_dpo/beta": 7.277153968811035, "fcm_dpo/delta": -0.1604071408510208, "fcm_dpo/margin": 0.15663331747055054, "fcm_dpo/q_t": 0.3513562083244324, "grad_norm": 1870.4586181640625, "learning_rate": 3.432835820895522e-07, "logits/chosen": 0.04342661425471306, "logits/rejected": -0.00011028675362467766, "logps/chosen": -51.65697479248047, "logps/ref_chosen": -51.578346252441406, "logps/ref_rejected": -68.2215576171875, "logps/rejected": -68.45681762695312, "loss": 1.3277, "margin_dpo/margin_mean": 0.1566331386566162, "margin_dpo/margin_std": 0.3138054609298706, "step": 47 }, { "epoch": 0.07256235827664399, "fcm_dpo/beta": 8.184296607971191, "fcm_dpo/delta": 0.7369337677955627, "fcm_dpo/margin": 0.03382202982902527, "fcm_dpo/q_t": 0.485114187002182, "grad_norm": 2785.64453125, "learning_rate": 3.507462686567164e-07, "logits/chosen": 0.16853903234004974, "logits/rejected": 0.1380309760570526, "logps/chosen": -51.94841003417969, "logps/ref_chosen": -51.79365158081055, "logps/ref_rejected": -64.22503662109375, "logps/rejected": -64.41361999511719, "loss": 2.617, "margin_dpo/margin_mean": 0.03382223844528198, "margin_dpo/margin_std": 0.38840365409851074, "step": 48 }, { "epoch": 0.07407407407407407, "fcm_dpo/beta": 9.334107398986816, "fcm_dpo/delta": 0.5954843163490295, "fcm_dpo/margin": 0.044813498854637146, "fcm_dpo/q_t": 0.4577757716178894, "grad_norm": 2963.8173828125, "learning_rate": 3.5820895522388055e-07, "logits/chosen": 0.038059771060943604, "logits/rejected": 0.01622236706316471, "logps/chosen": -58.26384353637695, "logps/ref_chosen": -58.13460159301758, "logps/ref_rejected": -64.63206481933594, "logps/rejected": -64.80612182617188, "loss": 2.5832, "margin_dpo/margin_mean": 0.04481416940689087, "margin_dpo/margin_std": 0.3690647482872009, "step": 49 }, { "epoch": 0.07558578987150416, "fcm_dpo/beta": 9.334890365600586, "fcm_dpo/delta": -0.06085062772035599, "fcm_dpo/margin": 0.11278587579727173, "fcm_dpo/q_t": 0.4165031909942627, "grad_norm": 2980.169921875, "learning_rate": 3.6567164179104475e-07, "logits/chosen": 0.11696229875087738, "logits/rejected": 0.08666031062602997, "logps/chosen": -52.99673080444336, "logps/ref_chosen": -52.85643768310547, "logps/ref_rejected": -72.17460632324219, "logps/rejected": -72.42768859863281, "loss": 2.2159, "margin_dpo/margin_mean": 0.11278638243675232, "margin_dpo/margin_std": 0.39860716462135315, "step": 50 }, { "epoch": 0.07709750566893424, "fcm_dpo/beta": 8.363540649414062, "fcm_dpo/delta": -0.5726056098937988, "fcm_dpo/margin": 0.15886437892913818, "fcm_dpo/q_t": 0.4000622630119324, "grad_norm": 2550.51953125, "learning_rate": 3.7313432835820895e-07, "logits/chosen": 0.08093120157718658, "logits/rejected": 0.05302312225103378, "logps/chosen": -63.80950927734375, "logps/ref_chosen": -63.65644073486328, "logps/ref_rejected": -86.13229370117188, "logps/rejected": -86.44422912597656, "loss": 1.8986, "margin_dpo/margin_mean": 0.15886464715003967, "margin_dpo/margin_std": 0.420296311378479, "step": 51 }, { "epoch": 0.07860922146636433, "fcm_dpo/beta": 8.45156478881836, "fcm_dpo/delta": 0.09679965674877167, "fcm_dpo/margin": 0.10791899263858795, "fcm_dpo/q_t": 0.4053017497062683, "grad_norm": 3315.59912109375, "learning_rate": 3.805970149253731e-07, "logits/chosen": 0.07245868444442749, "logits/rejected": 0.023480474948883057, "logps/chosen": -68.0364990234375, "logps/ref_chosen": -67.8402099609375, "logps/ref_rejected": -96.97090911865234, "logps/rejected": -97.27511596679688, "loss": 2.2996, "margin_dpo/margin_mean": 0.10791890323162079, "margin_dpo/margin_std": 0.4656970500946045, "step": 52 }, { "epoch": 0.0801209372637944, "fcm_dpo/beta": 8.50004768371582, "fcm_dpo/delta": 0.018193505704402924, "fcm_dpo/margin": 0.11565050482749939, "fcm_dpo/q_t": 0.371160626411438, "grad_norm": 2222.445556640625, "learning_rate": 3.880597014925373e-07, "logits/chosen": 0.07711566984653473, "logits/rejected": 0.06644274294376373, "logps/chosen": -57.06134796142578, "logps/ref_chosen": -56.87813949584961, "logps/ref_rejected": -60.75569152832031, "logps/rejected": -61.05455017089844, "loss": 1.7873, "margin_dpo/margin_mean": 0.11565083265304565, "margin_dpo/margin_std": 0.32278263568878174, "step": 53 }, { "epoch": 0.08163265306122448, "fcm_dpo/beta": 8.553638458251953, "fcm_dpo/delta": 0.20764021575450897, "fcm_dpo/margin": 0.09295859932899475, "fcm_dpo/q_t": 0.40855568647384644, "grad_norm": 2532.174072265625, "learning_rate": 3.9552238805970144e-07, "logits/chosen": 0.03949524462223053, "logits/rejected": 0.024343054741621017, "logps/chosen": -47.497901916503906, "logps/ref_chosen": -47.26692199707031, "logps/ref_rejected": -62.19426727294922, "logps/rejected": -62.51820373535156, "loss": 2.2098, "margin_dpo/margin_mean": 0.09295853972434998, "margin_dpo/margin_std": 0.3836787939071655, "step": 54 }, { "epoch": 0.08314436885865457, "fcm_dpo/beta": 7.794929027557373, "fcm_dpo/delta": -0.9819808006286621, "fcm_dpo/margin": 0.23201557993888855, "fcm_dpo/q_t": 0.3249555230140686, "grad_norm": 2186.456298828125, "learning_rate": 4.0298507462686564e-07, "logits/chosen": 0.029716331511735916, "logits/rejected": -0.04705657809972763, "logps/chosen": -50.511959075927734, "logps/ref_chosen": -50.32619094848633, "logps/ref_rejected": -92.44389343261719, "logps/rejected": -92.8616714477539, "loss": 1.4207, "margin_dpo/margin_mean": 0.2320151925086975, "margin_dpo/margin_std": 0.42721158266067505, "step": 55 }, { "epoch": 0.08465608465608465, "fcm_dpo/beta": 7.265107154846191, "fcm_dpo/delta": -0.08848509937524796, "fcm_dpo/margin": 0.1485108733177185, "fcm_dpo/q_t": 0.32371559739112854, "grad_norm": 1776.4510498046875, "learning_rate": 4.1044776119402984e-07, "logits/chosen": 0.12203441560268402, "logits/rejected": 0.09960527718067169, "logps/chosen": -56.979713439941406, "logps/ref_chosen": -56.766971588134766, "logps/ref_rejected": -66.30504608154297, "logps/rejected": -66.66629791259766, "loss": 1.5337, "margin_dpo/margin_mean": 0.1485109031200409, "margin_dpo/margin_std": 0.36086180806159973, "step": 56 }, { "epoch": 0.08616780045351474, "fcm_dpo/beta": 7.139953136444092, "fcm_dpo/delta": -0.0027963966131210327, "fcm_dpo/margin": 0.14022627472877502, "fcm_dpo/q_t": 0.3702242970466614, "grad_norm": 2208.6005859375, "learning_rate": 4.17910447761194e-07, "logits/chosen": 0.09269669651985168, "logits/rejected": 0.02815322019159794, "logps/chosen": -58.0312614440918, "logps/ref_chosen": -57.76774597167969, "logps/ref_rejected": -82.75698852539062, "logps/rejected": -83.16073608398438, "loss": 1.9812, "margin_dpo/margin_mean": 0.14022645354270935, "margin_dpo/margin_std": 0.4531812369823456, "step": 57 }, { "epoch": 0.08767951625094482, "fcm_dpo/beta": 6.859474182128906, "fcm_dpo/delta": -0.0021596550941467285, "fcm_dpo/margin": 0.13939353823661804, "fcm_dpo/q_t": 0.4166187345981598, "grad_norm": 2155.36083984375, "learning_rate": 4.253731343283582e-07, "logits/chosen": 0.04018617421388626, "logits/rejected": 0.024990694597363472, "logps/chosen": -72.99790954589844, "logps/ref_chosen": -72.76408386230469, "logps/ref_rejected": -84.49275207519531, "logps/rejected": -84.865966796875, "loss": 2.1325, "margin_dpo/margin_mean": 0.13939306139945984, "margin_dpo/margin_std": 0.4941212832927704, "step": 58 }, { "epoch": 0.08919123204837491, "fcm_dpo/beta": 7.301891326904297, "fcm_dpo/delta": -0.22430884838104248, "fcm_dpo/margin": 0.15986737608909607, "fcm_dpo/q_t": 0.36295658349990845, "grad_norm": 1929.1551513671875, "learning_rate": 4.3283582089552234e-07, "logits/chosen": 0.10927902162075043, "logits/rejected": 0.04400138556957245, "logps/chosen": -50.06452178955078, "logps/ref_chosen": -49.820777893066406, "logps/ref_rejected": -77.14368438720703, "logps/rejected": -77.54730224609375, "loss": 1.5376, "margin_dpo/margin_mean": 0.15986764430999756, "margin_dpo/margin_std": 0.36004385352134705, "step": 59 }, { "epoch": 0.09070294784580499, "fcm_dpo/beta": 7.388426780700684, "fcm_dpo/delta": 0.45878517627716064, "fcm_dpo/margin": 0.07655715942382812, "fcm_dpo/q_t": 0.44889310002326965, "grad_norm": 2771.480712890625, "learning_rate": 4.4029850746268654e-07, "logits/chosen": 0.09602642804384232, "logits/rejected": 0.09458990395069122, "logps/chosen": -63.469207763671875, "logps/ref_chosen": -63.22477340698242, "logps/ref_rejected": -61.360477447509766, "logps/rejected": -61.68146896362305, "loss": 2.1317, "margin_dpo/margin_mean": 0.07655695080757141, "margin_dpo/margin_std": 0.3793250024318695, "step": 60 }, { "epoch": 0.09221466364323508, "fcm_dpo/beta": 7.838181495666504, "fcm_dpo/delta": 0.23324143886566162, "fcm_dpo/margin": 0.09999506175518036, "fcm_dpo/q_t": 0.40534743666648865, "grad_norm": 2478.5703125, "learning_rate": 4.4776119402985074e-07, "logits/chosen": 0.12898309528827667, "logits/rejected": 0.0962858498096466, "logps/chosen": -49.336029052734375, "logps/ref_chosen": -49.01679992675781, "logps/ref_rejected": -74.90817260742188, "logps/rejected": -75.327392578125, "loss": 2.1133, "margin_dpo/margin_mean": 0.09999510645866394, "margin_dpo/margin_std": 0.3832091987133026, "step": 61 }, { "epoch": 0.09372637944066516, "fcm_dpo/beta": 7.966916084289551, "fcm_dpo/delta": -0.13704264163970947, "fcm_dpo/margin": 0.13938570022583008, "fcm_dpo/q_t": 0.36962825059890747, "grad_norm": 2647.16162109375, "learning_rate": 4.552238805970149e-07, "logits/chosen": 0.11322169005870819, "logits/rejected": 0.07340162247419357, "logps/chosen": -63.07339859008789, "logps/ref_chosen": -62.751869201660156, "logps/ref_rejected": -78.93360900878906, "logps/rejected": -79.39452362060547, "loss": 2.0748, "margin_dpo/margin_mean": 0.13938573002815247, "margin_dpo/margin_std": 0.46554332971572876, "step": 62 }, { "epoch": 0.09523809523809523, "fcm_dpo/beta": 7.405724048614502, "fcm_dpo/delta": -0.43855780363082886, "fcm_dpo/margin": 0.1857583224773407, "fcm_dpo/q_t": 0.3208463490009308, "grad_norm": 2495.121826171875, "learning_rate": 4.626865671641791e-07, "logits/chosen": 0.17758890986442566, "logits/rejected": 0.15252208709716797, "logps/chosen": -60.78285598754883, "logps/ref_chosen": -60.51525115966797, "logps/ref_rejected": -85.11021423339844, "logps/rejected": -85.56358337402344, "loss": 1.4936, "margin_dpo/margin_mean": 0.1857585310935974, "margin_dpo/margin_std": 0.3698121905326843, "step": 63 }, { "epoch": 0.09674981103552532, "fcm_dpo/beta": 7.346306324005127, "fcm_dpo/delta": 0.42287638783454895, "fcm_dpo/margin": 0.08198145031929016, "fcm_dpo/q_t": 0.44880834221839905, "grad_norm": 2452.107421875, "learning_rate": 4.701492537313433e-07, "logits/chosen": 0.07220865786075592, "logits/rejected": 0.047520771622657776, "logps/chosen": -51.5291748046875, "logps/ref_chosen": -51.20684814453125, "logps/ref_rejected": -66.93081665039062, "logps/rejected": -67.33512878417969, "loss": 2.1847, "margin_dpo/margin_mean": 0.08198148012161255, "margin_dpo/margin_std": 0.39635199308395386, "step": 64 }, { "epoch": 0.0982615268329554, "fcm_dpo/beta": 6.500675201416016, "fcm_dpo/delta": -1.11879301071167, "fcm_dpo/margin": 0.2926085889339447, "fcm_dpo/q_t": 0.2865545451641083, "grad_norm": 2000.08251953125, "learning_rate": 4.776119402985074e-07, "logits/chosen": 0.1627321094274521, "logits/rejected": 0.13373470306396484, "logps/chosen": -67.60956573486328, "logps/ref_chosen": -67.2886962890625, "logps/ref_rejected": -74.44281005859375, "logps/rejected": -75.05628967285156, "loss": 1.3075, "margin_dpo/margin_mean": 0.29260820150375366, "margin_dpo/margin_std": 0.4580235481262207, "step": 65 }, { "epoch": 0.09977324263038549, "fcm_dpo/beta": 6.545133590698242, "fcm_dpo/delta": 0.5023878216743469, "fcm_dpo/margin": 0.08048596978187561, "fcm_dpo/q_t": 0.435330331325531, "grad_norm": 2383.31005859375, "learning_rate": 4.850746268656717e-07, "logits/chosen": 0.08348944783210754, "logits/rejected": 0.05969405546784401, "logps/chosen": -71.08229064941406, "logps/ref_chosen": -70.743408203125, "logps/ref_rejected": -77.26499938964844, "logps/rejected": -77.68437194824219, "loss": 1.9852, "margin_dpo/margin_mean": 0.08048596978187561, "margin_dpo/margin_std": 0.4329206943511963, "step": 66 }, { "epoch": 0.10128495842781557, "fcm_dpo/beta": 6.768010139465332, "fcm_dpo/delta": 0.0033745458349585533, "fcm_dpo/margin": 0.1473006308078766, "fcm_dpo/q_t": 0.39285334944725037, "grad_norm": 2144.995849609375, "learning_rate": 4.925373134328357e-07, "logits/chosen": 0.06257347017526627, "logits/rejected": 0.007243716157972813, "logps/chosen": -60.883270263671875, "logps/ref_chosen": -60.60260009765625, "logps/ref_rejected": -75.22235870361328, "logps/rejected": -75.65032958984375, "loss": 1.6865, "margin_dpo/margin_mean": 0.1473003625869751, "margin_dpo/margin_std": 0.44022125005722046, "step": 67 }, { "epoch": 0.10279667422524566, "fcm_dpo/beta": 6.251596450805664, "fcm_dpo/delta": -0.4169546067714691, "fcm_dpo/margin": 0.21536040306091309, "fcm_dpo/q_t": 0.33271753787994385, "grad_norm": 1849.2083740234375, "learning_rate": 5e-07, "logits/chosen": 0.04055653512477875, "logits/rejected": 0.01153562217950821, "logps/chosen": -77.89004516601562, "logps/ref_chosen": -77.52836608886719, "logps/ref_rejected": -93.17778015136719, "logps/rejected": -93.75480651855469, "loss": 1.5844, "margin_dpo/margin_mean": 0.21536031365394592, "margin_dpo/margin_std": 0.4642139673233032, "step": 68 }, { "epoch": 0.10430839002267574, "fcm_dpo/beta": 6.393548488616943, "fcm_dpo/delta": -0.062498897314071655, "fcm_dpo/margin": 0.16213825345039368, "fcm_dpo/q_t": 0.3610564172267914, "grad_norm": 1912.0982666015625, "learning_rate": 4.999965034812934e-07, "logits/chosen": 0.06711920350790024, "logits/rejected": 0.02523168735206127, "logps/chosen": -66.31640625, "logps/ref_chosen": -65.94305419921875, "logps/ref_rejected": -89.7735595703125, "logps/rejected": -90.30905151367188, "loss": 1.8845, "margin_dpo/margin_mean": 0.16213801503181458, "margin_dpo/margin_std": 0.45384150743484497, "step": 69 }, { "epoch": 0.10582010582010581, "fcm_dpo/beta": 5.698063850402832, "fcm_dpo/delta": -0.3884986937046051, "fcm_dpo/margin": 0.13598540425300598, "fcm_dpo/q_t": 0.3947943449020386, "grad_norm": 1745.823486328125, "learning_rate": 4.999860140229787e-07, "logits/chosen": 0.11856390535831451, "logits/rejected": 0.09516175091266632, "logps/chosen": -62.308258056640625, "logps/ref_chosen": -61.95791244506836, "logps/ref_rejected": -75.80945587158203, "logps/rejected": -76.29579162597656, "loss": 1.7483, "margin_dpo/margin_mean": 0.1359853446483612, "margin_dpo/margin_std": 0.397053986787796, "step": 70 }, { "epoch": 0.1073318216175359, "fcm_dpo/beta": 6.111842632293701, "fcm_dpo/delta": 0.5105581283569336, "fcm_dpo/margin": 0.08478209376335144, "fcm_dpo/q_t": 0.4329409897327423, "grad_norm": 2060.398193359375, "learning_rate": 4.999685319184688e-07, "logits/chosen": 0.0789206326007843, "logits/rejected": 0.06302116811275482, "logps/chosen": -63.77684020996094, "logps/ref_chosen": -63.34757995605469, "logps/ref_rejected": -67.49658203125, "logps/rejected": -68.0106201171875, "loss": 2.1644, "margin_dpo/margin_mean": 0.08478212356567383, "margin_dpo/margin_std": 0.4536302983760834, "step": 71 }, { "epoch": 0.10884353741496598, "fcm_dpo/beta": 5.628780364990234, "fcm_dpo/delta": -0.7731190919876099, "fcm_dpo/margin": 0.29211413860321045, "fcm_dpo/q_t": 0.3161046504974365, "grad_norm": 1577.560302734375, "learning_rate": 4.999440576567755e-07, "logits/chosen": 0.0920741856098175, "logits/rejected": 0.029652319848537445, "logps/chosen": -56.157230377197266, "logps/ref_chosen": -55.85929870605469, "logps/ref_rejected": -68.45423889160156, "logps/rejected": -69.04428100585938, "loss": 1.1926, "margin_dpo/margin_mean": 0.29211464524269104, "margin_dpo/margin_std": 0.4848480224609375, "step": 72 }, { "epoch": 0.11035525321239607, "fcm_dpo/beta": 5.599390029907227, "fcm_dpo/delta": 0.03130987286567688, "fcm_dpo/margin": 0.17031516134738922, "fcm_dpo/q_t": 0.40101659297943115, "grad_norm": 1934.7177734375, "learning_rate": 4.999125919224965e-07, "logits/chosen": 0.06832102686166763, "logits/rejected": 0.05396275222301483, "logps/chosen": -69.59184265136719, "logps/ref_chosen": -69.13880920410156, "logps/ref_rejected": -79.04586791992188, "logps/rejected": -79.66921997070312, "loss": 1.8098, "margin_dpo/margin_mean": 0.17031550407409668, "margin_dpo/margin_std": 0.48382318019866943, "step": 73 }, { "epoch": 0.11186696900982615, "fcm_dpo/beta": 5.374726295471191, "fcm_dpo/delta": -0.2641603350639343, "fcm_dpo/margin": 0.22797469794750214, "fcm_dpo/q_t": 0.36758503317832947, "grad_norm": 1349.6485595703125, "learning_rate": 4.998741355957963e-07, "logits/chosen": 0.09487976133823395, "logits/rejected": 0.04454671964049339, "logps/chosen": -50.26209259033203, "logps/ref_chosen": -49.923736572265625, "logps/ref_rejected": -81.73213958740234, "logps/rejected": -82.2984619140625, "loss": 1.4011, "margin_dpo/margin_mean": 0.22797417640686035, "margin_dpo/margin_std": 0.46810248494148254, "step": 74 }, { "epoch": 0.11337868480725624, "fcm_dpo/beta": 4.875063896179199, "fcm_dpo/delta": -0.2485802173614502, "fcm_dpo/margin": 0.24743963778018951, "fcm_dpo/q_t": 0.34337544441223145, "grad_norm": 1073.2872314453125, "learning_rate": 4.998286897523808e-07, "logits/chosen": 0.07122410833835602, "logits/rejected": 0.04160505533218384, "logps/chosen": -46.41517639160156, "logps/ref_chosen": -46.06875228881836, "logps/ref_rejected": -66.1181411743164, "logps/rejected": -66.71200561523438, "loss": 1.2434, "margin_dpo/margin_mean": 0.24743930995464325, "margin_dpo/margin_std": 0.4698425531387329, "step": 75 }, { "epoch": 0.11489040060468632, "fcm_dpo/beta": 5.057272911071777, "fcm_dpo/delta": 0.1727055013179779, "fcm_dpo/margin": 0.16627338528633118, "fcm_dpo/q_t": 0.3987014889717102, "grad_norm": 1406.1661376953125, "learning_rate": 4.997762556634679e-07, "logits/chosen": 0.10183432698249817, "logits/rejected": 0.058754947036504745, "logps/chosen": -54.42407989501953, "logps/ref_chosen": -54.06275177001953, "logps/ref_rejected": -74.87464141845703, "logps/rejected": -75.4022445678711, "loss": 1.5553, "margin_dpo/margin_mean": 0.16627269983291626, "margin_dpo/margin_std": 0.4644964933395386, "step": 76 }, { "epoch": 0.1164021164021164, "fcm_dpo/beta": 5.224140167236328, "fcm_dpo/delta": 0.3271293044090271, "fcm_dpo/margin": 0.1326499581336975, "fcm_dpo/q_t": 0.41528427600860596, "grad_norm": 1620.2890625, "learning_rate": 4.99716834795752e-07, "logits/chosen": 0.10804985463619232, "logits/rejected": 0.06888192892074585, "logps/chosen": -53.52935028076172, "logps/ref_chosen": -53.07609176635742, "logps/ref_rejected": -74.45601654052734, "logps/rejected": -75.04192352294922, "loss": 1.6511, "margin_dpo/margin_mean": 0.1326504349708557, "margin_dpo/margin_std": 0.44195854663848877, "step": 77 }, { "epoch": 0.11791383219954649, "fcm_dpo/beta": 5.861191749572754, "fcm_dpo/delta": 0.4837532937526703, "fcm_dpo/margin": 0.09206165373325348, "fcm_dpo/q_t": 0.40167397260665894, "grad_norm": 2161.646240234375, "learning_rate": 4.996504288113623e-07, "logits/chosen": 0.039431821554899216, "logits/rejected": 0.01997038722038269, "logps/chosen": -68.12590026855469, "logps/ref_chosen": -67.72541809082031, "logps/ref_rejected": -79.03926849365234, "logps/rejected": -79.53181457519531, "loss": 2.4878, "margin_dpo/margin_mean": 0.09206125140190125, "margin_dpo/margin_std": 0.583328366279602, "step": 78 }, { "epoch": 0.11942554799697656, "fcm_dpo/beta": 6.234781265258789, "fcm_dpo/delta": 0.15417495369911194, "fcm_dpo/margin": 0.13618767261505127, "fcm_dpo/q_t": 0.389265775680542, "grad_norm": 1901.2857666015625, "learning_rate": 4.995770395678171e-07, "logits/chosen": 0.150520920753479, "logits/rejected": 0.09114135801792145, "logps/chosen": -52.598140716552734, "logps/ref_chosen": -52.16064453125, "logps/ref_rejected": -83.31062316894531, "logps/rejected": -83.8843002319336, "loss": 1.8135, "margin_dpo/margin_mean": 0.13618725538253784, "margin_dpo/margin_std": 0.4358598589897156, "step": 79 }, { "epoch": 0.12093726379440665, "fcm_dpo/beta": 6.20821475982666, "fcm_dpo/delta": -0.13912838697433472, "fcm_dpo/margin": 0.17933352291584015, "fcm_dpo/q_t": 0.37054312229156494, "grad_norm": 1953.0830078125, "learning_rate": 4.994966691179711e-07, "logits/chosen": 0.09082512557506561, "logits/rejected": 0.035024385899305344, "logps/chosen": -61.85608673095703, "logps/ref_chosen": -61.410560607910156, "logps/ref_rejected": -78.66004943847656, "logps/rejected": -79.284912109375, "loss": 1.7668, "margin_dpo/margin_mean": 0.1793329417705536, "margin_dpo/margin_std": 0.4744713306427002, "step": 80 }, { "epoch": 0.12244897959183673, "fcm_dpo/beta": 5.641842842102051, "fcm_dpo/delta": -0.34320950508117676, "fcm_dpo/margin": 0.2292810082435608, "fcm_dpo/q_t": 0.34941136837005615, "grad_norm": 1636.6348876953125, "learning_rate": 4.994093197099587e-07, "logits/chosen": 0.07053244113922119, "logits/rejected": 0.03839043155312538, "logps/chosen": -64.20402526855469, "logps/ref_chosen": -63.80437088012695, "logps/ref_rejected": -79.3484115600586, "logps/rejected": -79.97734069824219, "loss": 1.4575, "margin_dpo/margin_mean": 0.22928106784820557, "margin_dpo/margin_std": 0.4831709563732147, "step": 81 }, { "epoch": 0.12396069538926682, "fcm_dpo/beta": 5.125918388366699, "fcm_dpo/delta": -0.6294834017753601, "fcm_dpo/margin": 0.29933592677116394, "fcm_dpo/q_t": 0.2602464258670807, "grad_norm": 1189.437255859375, "learning_rate": 4.993149937871306e-07, "logits/chosen": 0.07535186409950256, "logits/rejected": 0.013600241392850876, "logps/chosen": -49.17131805419922, "logps/ref_chosen": -48.817893981933594, "logps/ref_rejected": -70.31497955322266, "logps/rejected": -70.96774291992188, "loss": 1.0653, "margin_dpo/margin_mean": 0.2993359863758087, "margin_dpo/margin_std": 0.41738319396972656, "step": 82 }, { "epoch": 0.1254724111866969, "fcm_dpo/beta": 4.8786211013793945, "fcm_dpo/delta": -0.19270329177379608, "fcm_dpo/margin": 0.23940634727478027, "fcm_dpo/q_t": 0.3403007388114929, "grad_norm": 1305.392333984375, "learning_rate": 4.992136939879856e-07, "logits/chosen": 0.1480909138917923, "logits/rejected": 0.09921949356794357, "logps/chosen": -57.55104064941406, "logps/ref_chosen": -57.15077209472656, "logps/ref_rejected": -75.1710205078125, "logps/rejected": -75.81069946289062, "loss": 1.3796, "margin_dpo/margin_mean": 0.23940622806549072, "margin_dpo/margin_std": 0.48909199237823486, "step": 83 }, { "epoch": 0.12698412698412698, "fcm_dpo/beta": 5.00314998626709, "fcm_dpo/delta": 0.3583996295928955, "fcm_dpo/margin": 0.13308590650558472, "fcm_dpo/q_t": 0.4098473787307739, "grad_norm": 1782.189208984375, "learning_rate": 4.991054231460969e-07, "logits/chosen": 0.12158288061618805, "logits/rejected": 0.08126094937324524, "logps/chosen": -65.28231048583984, "logps/ref_chosen": -64.77729797363281, "logps/ref_rejected": -84.71949768066406, "logps/rejected": -85.35758972167969, "loss": 2.005, "margin_dpo/margin_mean": 0.1330859661102295, "margin_dpo/margin_std": 0.5464334487915039, "step": 84 }, { "epoch": 0.12849584278155707, "fcm_dpo/beta": 4.760544300079346, "fcm_dpo/delta": -0.5031009912490845, "fcm_dpo/margin": 0.3008922040462494, "fcm_dpo/q_t": 0.3266737163066864, "grad_norm": 1381.7164306640625, "learning_rate": 4.989901842900325e-07, "logits/chosen": 0.10575494170188904, "logits/rejected": 0.06379462033510208, "logps/chosen": -50.625328063964844, "logps/ref_chosen": -50.25169372558594, "logps/ref_rejected": -66.55439758300781, "logps/rejected": -67.22891235351562, "loss": 1.3231, "margin_dpo/margin_mean": 0.30089178681373596, "margin_dpo/margin_std": 0.5666717290878296, "step": 85 }, { "epoch": 0.13000755857898716, "fcm_dpo/beta": 4.623910427093506, "fcm_dpo/delta": -0.011702943593263626, "fcm_dpo/margin": 0.21848827600479126, "fcm_dpo/q_t": 0.3801780939102173, "grad_norm": 1446.2276611328125, "learning_rate": 4.988679806432711e-07, "logits/chosen": 0.11094961315393448, "logits/rejected": 0.09372542053461075, "logps/chosen": -61.207275390625, "logps/ref_chosen": -60.72917938232422, "logps/ref_rejected": -72.30961608886719, "logps/rejected": -73.0062026977539, "loss": 1.5695, "margin_dpo/margin_mean": 0.2184884250164032, "margin_dpo/margin_std": 0.5658543109893799, "step": 86 }, { "epoch": 0.13151927437641722, "fcm_dpo/beta": 4.357776641845703, "fcm_dpo/delta": -0.31460562348365784, "fcm_dpo/margin": 0.2916297912597656, "fcm_dpo/q_t": 0.3396541476249695, "grad_norm": 1350.0787353515625, "learning_rate": 4.987388156241114e-07, "logits/chosen": 0.08588653057813644, "logits/rejected": 0.03219534084200859, "logps/chosen": -66.21820831298828, "logps/ref_chosen": -65.75796508789062, "logps/ref_rejected": -84.81159973144531, "logps/rejected": -85.56346130371094, "loss": 1.3305, "margin_dpo/margin_mean": 0.29163050651550293, "margin_dpo/margin_std": 0.5484292507171631, "step": 87 }, { "epoch": 0.1330309901738473, "fcm_dpo/beta": 4.051568984985352, "fcm_dpo/delta": -0.3438121974468231, "fcm_dpo/margin": 0.3187229037284851, "fcm_dpo/q_t": 0.3379303812980652, "grad_norm": 1034.83447265625, "learning_rate": 4.986026928455767e-07, "logits/chosen": 0.1511228084564209, "logits/rejected": 0.12557803094387054, "logps/chosen": -63.23377227783203, "logps/ref_chosen": -62.82402801513672, "logps/ref_rejected": -74.9607162475586, "logps/rejected": -75.68917846679688, "loss": 1.2325, "margin_dpo/margin_mean": 0.31872305274009705, "margin_dpo/margin_std": 0.5725570917129517, "step": 88 }, { "epoch": 0.1345427059712774, "fcm_dpo/beta": 4.111684799194336, "fcm_dpo/delta": -0.005461782217025757, "fcm_dpo/margin": 0.2426835298538208, "fcm_dpo/q_t": 0.36708956956863403, "grad_norm": 1158.6134033203125, "learning_rate": 4.984596161153135e-07, "logits/chosen": 0.18977168202400208, "logits/rejected": 0.11028344929218292, "logps/chosen": -41.57792663574219, "logps/ref_chosen": -41.191436767578125, "logps/ref_rejected": -85.44769287109375, "logps/rejected": -86.07687377929688, "loss": 1.5568, "margin_dpo/margin_mean": 0.24268493056297302, "margin_dpo/margin_std": 0.5871646404266357, "step": 89 }, { "epoch": 0.1360544217687075, "fcm_dpo/beta": 3.9873642921447754, "fcm_dpo/delta": -0.07796984910964966, "fcm_dpo/margin": 0.26826444268226624, "fcm_dpo/q_t": 0.3403598964214325, "grad_norm": 1097.566650390625, "learning_rate": 4.983095894354857e-07, "logits/chosen": 0.08516630530357361, "logits/rejected": 0.03322757035493851, "logps/chosen": -56.98159408569336, "logps/ref_chosen": -56.58390808105469, "logps/ref_rejected": -86.86978149414062, "logps/rejected": -87.53573608398438, "loss": 1.3846, "margin_dpo/margin_mean": 0.26826387643814087, "margin_dpo/margin_std": 0.6303993463516235, "step": 90 }, { "epoch": 0.13756613756613756, "fcm_dpo/beta": 3.869006633758545, "fcm_dpo/delta": -0.14137369394302368, "fcm_dpo/margin": 0.2906471788883209, "fcm_dpo/q_t": 0.34270262718200684, "grad_norm": 1045.7557373046875, "learning_rate": 4.98152617002662e-07, "logits/chosen": 0.10475227236747742, "logits/rejected": 0.06226480007171631, "logps/chosen": -52.82066345214844, "logps/ref_chosen": -52.38234329223633, "logps/ref_rejected": -72.17642211914062, "logps/rejected": -72.90538787841797, "loss": 1.3782, "margin_dpo/margin_mean": 0.29064705967903137, "margin_dpo/margin_std": 0.6195999383926392, "step": 91 }, { "epoch": 0.13907785336356765, "fcm_dpo/beta": 4.068203926086426, "fcm_dpo/delta": 0.24122354388237, "fcm_dpo/margin": 0.1886722892522812, "fcm_dpo/q_t": 0.40757930278778076, "grad_norm": 1103.524658203125, "learning_rate": 4.979887032076988e-07, "logits/chosen": 0.13855835795402527, "logits/rejected": 0.10122767090797424, "logps/chosen": -53.529884338378906, "logps/ref_chosen": -53.00870132446289, "logps/ref_rejected": -79.77812957763672, "logps/rejected": -80.48798370361328, "loss": 1.5458, "margin_dpo/margin_mean": 0.18867191672325134, "margin_dpo/margin_std": 0.5168828368186951, "step": 92 }, { "epoch": 0.14058956916099774, "fcm_dpo/beta": 4.217402458190918, "fcm_dpo/delta": 0.23529532551765442, "fcm_dpo/margin": 0.1851963996887207, "fcm_dpo/q_t": 0.3896501064300537, "grad_norm": 1094.4332275390625, "learning_rate": 4.978178526356172e-07, "logits/chosen": 0.10360229760408401, "logits/rejected": 0.07716604322195053, "logps/chosen": -45.443016052246094, "logps/ref_chosen": -44.90705108642578, "logps/ref_rejected": -58.7879524230957, "logps/rejected": -59.50910949707031, "loss": 1.5698, "margin_dpo/margin_mean": 0.185196191072464, "margin_dpo/margin_std": 0.5235726237297058, "step": 93 }, { "epoch": 0.1421012849584278, "fcm_dpo/beta": 3.851269006729126, "fcm_dpo/delta": -0.4641948938369751, "fcm_dpo/margin": 0.3565562069416046, "fcm_dpo/q_t": 0.2976396679878235, "grad_norm": 884.6483764648438, "learning_rate": 4.976400700654751e-07, "logits/chosen": 0.16318684816360474, "logits/rejected": 0.12688644230365753, "logps/chosen": -60.265010833740234, "logps/ref_chosen": -59.93777084350586, "logps/ref_rejected": -79.3138427734375, "logps/rejected": -79.99763488769531, "loss": 1.2145, "margin_dpo/margin_mean": 0.3565560281276703, "margin_dpo/margin_std": 0.6176527142524719, "step": 94 }, { "epoch": 0.1436130007558579, "fcm_dpo/beta": 3.51151704788208, "fcm_dpo/delta": -0.556129515171051, "fcm_dpo/margin": 0.41667065024375916, "fcm_dpo/q_t": 0.27452635765075684, "grad_norm": 725.6941528320312, "learning_rate": 4.974553604702332e-07, "logits/chosen": 0.04962404444813728, "logits/rejected": -0.009035153314471245, "logps/chosen": -60.609527587890625, "logps/ref_chosen": -60.168487548828125, "logps/ref_rejected": -90.73665618896484, "logps/rejected": -91.59436798095703, "loss": 0.8553, "margin_dpo/margin_mean": 0.4166697561740875, "margin_dpo/margin_std": 0.5119737386703491, "step": 95 }, { "epoch": 0.14512471655328799, "fcm_dpo/beta": 3.3476004600524902, "fcm_dpo/delta": -0.3292371928691864, "fcm_dpo/margin": 0.38447052240371704, "fcm_dpo/q_t": 0.3057492971420288, "grad_norm": 765.777587890625, "learning_rate": 4.972637290166157e-07, "logits/chosen": 0.09734475612640381, "logits/rejected": 0.05707583576440811, "logps/chosen": -61.152687072753906, "logps/ref_chosen": -60.66877746582031, "logps/ref_rejected": -88.30673217773438, "logps/rejected": -89.17510986328125, "loss": 1.0612, "margin_dpo/margin_mean": 0.3844701647758484, "margin_dpo/margin_std": 0.554520845413208, "step": 96 }, { "epoch": 0.14663643235071808, "fcm_dpo/beta": 3.2319469451904297, "fcm_dpo/delta": 0.1403380036354065, "fcm_dpo/margin": 0.2673853933811188, "fcm_dpo/q_t": 0.3827268183231354, "grad_norm": 1068.170654296875, "learning_rate": 4.970651810649666e-07, "logits/chosen": 0.04420812800526619, "logits/rejected": 0.002325967885553837, "logps/chosen": -65.61512756347656, "logps/ref_chosen": -65.04412078857422, "logps/ref_rejected": -78.42092895507812, "logps/rejected": -79.25931549072266, "loss": 1.3514, "margin_dpo/margin_mean": 0.2673855423927307, "margin_dpo/margin_std": 0.6004109382629395, "step": 97 }, { "epoch": 0.14814814814814814, "fcm_dpo/beta": 3.42587947845459, "fcm_dpo/delta": 0.1983877718448639, "fcm_dpo/margin": 0.23879370093345642, "fcm_dpo/q_t": 0.3823162317276001, "grad_norm": 922.5922241210938, "learning_rate": 4.968597221690985e-07, "logits/chosen": 0.13439376652240753, "logits/rejected": 0.10842472314834595, "logps/chosen": -55.90753936767578, "logps/ref_chosen": -55.503231048583984, "logps/ref_rejected": -72.81553649902344, "logps/rejected": -73.45864868164062, "loss": 1.2262, "margin_dpo/margin_mean": 0.23879370093345642, "margin_dpo/margin_std": 0.4960702657699585, "step": 98 }, { "epoch": 0.14965986394557823, "fcm_dpo/beta": 3.3395771980285645, "fcm_dpo/delta": -0.2737107276916504, "fcm_dpo/margin": 0.3713855743408203, "fcm_dpo/q_t": 0.31995880603790283, "grad_norm": 838.7493286132812, "learning_rate": 4.966473580761389e-07, "logits/chosen": 0.14741893112659454, "logits/rejected": 0.11150997132062912, "logps/chosen": -58.98797607421875, "logps/ref_chosen": -58.57563781738281, "logps/ref_rejected": -78.693603515625, "logps/rejected": -79.47733306884766, "loss": 1.0331, "margin_dpo/margin_mean": 0.37138599157333374, "margin_dpo/margin_std": 0.5738701820373535, "step": 99 }, { "epoch": 0.15117157974300832, "fcm_dpo/beta": 3.413600206375122, "fcm_dpo/delta": 0.0910910964012146, "fcm_dpo/margin": 0.2643635869026184, "fcm_dpo/q_t": 0.3801841139793396, "grad_norm": 1048.221923828125, "learning_rate": 4.964280947263676e-07, "logits/chosen": 0.12782281637191772, "logits/rejected": 0.12069296091794968, "logps/chosen": -80.02874755859375, "logps/ref_chosen": -79.58343505859375, "logps/ref_rejected": -92.152587890625, "logps/rejected": -92.86225891113281, "loss": 1.5245, "margin_dpo/margin_mean": 0.264363557100296, "margin_dpo/margin_std": 0.6884479522705078, "step": 100 }, { "epoch": 0.15268329554043839, "fcm_dpo/beta": 3.2258787155151367, "fcm_dpo/delta": -0.34300971031188965, "fcm_dpo/margin": 0.40265652537345886, "fcm_dpo/q_t": 0.29698413610458374, "grad_norm": 702.3856811523438, "learning_rate": 4.96201938253052e-07, "logits/chosen": 0.1229773610830307, "logits/rejected": 0.08786555379629135, "logps/chosen": -52.74365234375, "logps/ref_chosen": -52.332786560058594, "logps/ref_rejected": -69.55589294433594, "logps/rejected": -70.36941528320312, "loss": 0.9957, "margin_dpo/margin_mean": 0.40265610814094543, "margin_dpo/margin_std": 0.5584487318992615, "step": 101 }, { "epoch": 0.15419501133786848, "fcm_dpo/beta": 3.221522808074951, "fcm_dpo/delta": 0.1677694320678711, "fcm_dpo/margin": 0.26210707426071167, "fcm_dpo/q_t": 0.3745608925819397, "grad_norm": 933.2718505859375, "learning_rate": 4.959688949822748e-07, "logits/chosen": 0.05017256736755371, "logits/rejected": 0.012172428891062737, "logps/chosen": -65.21527099609375, "logps/ref_chosen": -64.74348449707031, "logps/ref_rejected": -69.06132507324219, "logps/rejected": -69.79522705078125, "loss": 1.3826, "margin_dpo/margin_mean": 0.26210689544677734, "margin_dpo/margin_std": 0.6122475862503052, "step": 102 }, { "epoch": 0.15570672713529857, "fcm_dpo/beta": 3.343921661376953, "fcm_dpo/delta": 0.14063423871994019, "fcm_dpo/margin": 0.25882646441459656, "fcm_dpo/q_t": 0.3664790987968445, "grad_norm": 876.5048828125, "learning_rate": 4.957289714327572e-07, "logits/chosen": 0.14401525259017944, "logits/rejected": 0.11314252763986588, "logps/chosen": -64.34033203125, "logps/ref_chosen": -63.83664321899414, "logps/ref_rejected": -79.32362365722656, "logps/rejected": -80.08615112304688, "loss": 1.2541, "margin_dpo/margin_mean": 0.2588259279727936, "margin_dpo/margin_std": 0.537617564201355, "step": 103 }, { "epoch": 0.15721844293272866, "fcm_dpo/beta": 3.3820950984954834, "fcm_dpo/delta": 0.11869892477989197, "fcm_dpo/margin": 0.2636609375476837, "fcm_dpo/q_t": 0.37176138162612915, "grad_norm": 1155.879150390625, "learning_rate": 4.954821743156767e-07, "logits/chosen": 0.10751787573099136, "logits/rejected": 0.03126327693462372, "logps/chosen": -61.48141098022461, "logps/ref_chosen": -60.99920654296875, "logps/ref_rejected": -98.84645080566406, "logps/rejected": -99.59231567382812, "loss": 1.4641, "margin_dpo/margin_mean": 0.26366138458251953, "margin_dpo/margin_std": 0.6468815803527832, "step": 104 }, { "epoch": 0.15873015873015872, "fcm_dpo/beta": 3.3653788566589355, "fcm_dpo/delta": -0.06612719595432281, "fcm_dpo/margin": 0.31470757722854614, "fcm_dpo/q_t": 0.3323326110839844, "grad_norm": 1119.451171875, "learning_rate": 4.952285105344791e-07, "logits/chosen": 0.09568033367395401, "logits/rejected": 0.044659968465566635, "logps/chosen": -71.36830139160156, "logps/ref_chosen": -70.95027160644531, "logps/ref_rejected": -87.88340759277344, "logps/rejected": -88.61614227294922, "loss": 1.3043, "margin_dpo/margin_mean": 0.3147069811820984, "margin_dpo/margin_std": 0.6669665575027466, "step": 105 }, { "epoch": 0.1602418745275888, "fcm_dpo/beta": 3.4193921089172363, "fcm_dpo/delta": 0.15833157300949097, "fcm_dpo/margin": 0.25018543004989624, "fcm_dpo/q_t": 0.3685312867164612, "grad_norm": 1039.7274169921875, "learning_rate": 4.949679871846857e-07, "logits/chosen": 0.1333768665790558, "logits/rejected": 0.1204344779253006, "logps/chosen": -62.8930549621582, "logps/ref_chosen": -62.45933151245117, "logps/ref_rejected": -67.00595092773438, "logps/rejected": -67.68985748291016, "loss": 1.3112, "margin_dpo/margin_mean": 0.2501852214336395, "margin_dpo/margin_std": 0.5423716306686401, "step": 106 }, { "epoch": 0.1617535903250189, "fcm_dpo/beta": 3.652026653289795, "fcm_dpo/delta": 0.4439771771430969, "fcm_dpo/margin": 0.15993273258209229, "fcm_dpo/q_t": 0.43477770686149597, "grad_norm": 1451.044677734375, "learning_rate": 4.947006115536947e-07, "logits/chosen": 0.06545466929674149, "logits/rejected": 0.046112120151519775, "logps/chosen": -76.338623046875, "logps/ref_chosen": -75.83796691894531, "logps/ref_rejected": -87.74038696289062, "logps/rejected": -88.40097045898438, "loss": 1.8206, "margin_dpo/margin_mean": 0.15993207693099976, "margin_dpo/margin_std": 0.673768162727356, "step": 107 }, { "epoch": 0.16326530612244897, "fcm_dpo/beta": 3.626925468444824, "fcm_dpo/delta": -0.3162718117237091, "fcm_dpo/margin": 0.3520002067089081, "fcm_dpo/q_t": 0.33733585476875305, "grad_norm": 953.0491943359375, "learning_rate": 4.944263911205772e-07, "logits/chosen": 0.07331952452659607, "logits/rejected": 0.047261983156204224, "logps/chosen": -68.7996826171875, "logps/ref_chosen": -68.39323425292969, "logps/ref_rejected": -83.24267578125, "logps/rejected": -84.00111389160156, "loss": 1.2468, "margin_dpo/margin_mean": 0.35200008749961853, "margin_dpo/margin_std": 0.6720170974731445, "step": 108 }, { "epoch": 0.16477702191987906, "fcm_dpo/beta": 3.5566816329956055, "fcm_dpo/delta": 0.07203048467636108, "fcm_dpo/margin": 0.2625024616718292, "fcm_dpo/q_t": 0.38299477100372314, "grad_norm": 1039.4810791015625, "learning_rate": 4.941453335558681e-07, "logits/chosen": 0.06151915714144707, "logits/rejected": 0.016085410490632057, "logps/chosen": -55.96140670776367, "logps/ref_chosen": -55.52748107910156, "logps/ref_rejected": -83.55218505859375, "logps/rejected": -84.24861907958984, "loss": 1.3522, "margin_dpo/margin_mean": 0.26250216364860535, "margin_dpo/margin_std": 0.6019136309623718, "step": 109 }, { "epoch": 0.16628873771730915, "fcm_dpo/beta": 3.8665976524353027, "fcm_dpo/delta": 0.39112499356269836, "fcm_dpo/margin": 0.16184496879577637, "fcm_dpo/q_t": 0.41690564155578613, "grad_norm": 1377.6337890625, "learning_rate": 4.938574467213517e-07, "logits/chosen": 0.033917784690856934, "logits/rejected": 0.04065680876374245, "logps/chosen": -81.66863250732422, "logps/ref_chosen": -81.15874481201172, "logps/ref_rejected": -72.56021118164062, "logps/rejected": -73.23194885253906, "loss": 1.6791, "margin_dpo/margin_mean": 0.16184476017951965, "margin_dpo/margin_std": 0.5766524076461792, "step": 110 }, { "epoch": 0.16780045351473924, "fcm_dpo/beta": 3.9139037132263184, "fcm_dpo/delta": -0.052766673266887665, "fcm_dpo/margin": 0.2670096457004547, "fcm_dpo/q_t": 0.36203914880752563, "grad_norm": 1263.3140869140625, "learning_rate": 4.935627386698418e-07, "logits/chosen": 0.13472726941108704, "logits/rejected": 0.10240040719509125, "logps/chosen": -52.87812042236328, "logps/ref_chosen": -52.358985900878906, "logps/ref_rejected": -77.06150817871094, "logps/rejected": -77.84764862060547, "loss": 1.6275, "margin_dpo/margin_mean": 0.26701000332832336, "margin_dpo/margin_std": 0.6607059836387634, "step": 111 }, { "epoch": 0.1693121693121693, "fcm_dpo/beta": 3.622741937637329, "fcm_dpo/delta": -0.296929270029068, "fcm_dpo/margin": 0.34460121393203735, "fcm_dpo/q_t": 0.33765172958374023, "grad_norm": 1077.2196044921875, "learning_rate": 4.932612176449559e-07, "logits/chosen": 0.05540486425161362, "logits/rejected": 0.0011176479747518897, "logps/chosen": -63.441078186035156, "logps/ref_chosen": -63.02006530761719, "logps/ref_rejected": -111.36941528320312, "logps/rejected": -112.13502502441406, "loss": 1.3207, "margin_dpo/margin_mean": 0.34459996223449707, "margin_dpo/margin_std": 0.6546406745910645, "step": 112 }, { "epoch": 0.1708238851095994, "fcm_dpo/beta": 3.7972545623779297, "fcm_dpo/delta": 0.12773901224136353, "fcm_dpo/margin": 0.22868876159191132, "fcm_dpo/q_t": 0.37673860788345337, "grad_norm": 1496.3253173828125, "learning_rate": 4.929528920808854e-07, "logits/chosen": 0.09782901406288147, "logits/rejected": 0.06331203132867813, "logps/chosen": -56.33560562133789, "logps/ref_chosen": -55.80766296386719, "logps/ref_rejected": -69.84014129638672, "logps/rejected": -70.59677124023438, "loss": 1.5844, "margin_dpo/margin_mean": 0.22868850827217102, "margin_dpo/margin_std": 0.5964616537094116, "step": 113 }, { "epoch": 0.17233560090702948, "fcm_dpo/beta": 3.3383381366729736, "fcm_dpo/delta": -0.5678998231887817, "fcm_dpo/margin": 0.4367901682853699, "fcm_dpo/q_t": 0.29958969354629517, "grad_norm": 712.6404418945312, "learning_rate": 4.92637770602159e-07, "logits/chosen": 0.13647404313087463, "logits/rejected": 0.08034436404705048, "logps/chosen": -66.70956420898438, "logps/ref_chosen": -66.33277130126953, "logps/ref_rejected": -71.61489868164062, "logps/rejected": -72.42848205566406, "loss": 1.0132, "margin_dpo/margin_mean": 0.43679025769233704, "margin_dpo/margin_std": 0.6181018352508545, "step": 114 }, { "epoch": 0.17384731670445955, "fcm_dpo/beta": 3.3931922912597656, "fcm_dpo/delta": 0.006447508931159973, "fcm_dpo/margin": 0.291039377450943, "fcm_dpo/q_t": 0.3765663206577301, "grad_norm": 1062.108154296875, "learning_rate": 4.923158620234019e-07, "logits/chosen": 0.10150371491909027, "logits/rejected": 0.047993022948503494, "logps/chosen": -56.26490783691406, "logps/ref_chosen": -55.74903869628906, "logps/ref_rejected": -79.59849548339844, "logps/rejected": -80.40541076660156, "loss": 1.3013, "margin_dpo/margin_mean": 0.29103943705558777, "margin_dpo/margin_std": 0.6191039085388184, "step": 115 }, { "epoch": 0.17535903250188964, "fcm_dpo/beta": 3.4025328159332275, "fcm_dpo/delta": -0.009448423981666565, "fcm_dpo/margin": 0.29369187355041504, "fcm_dpo/q_t": 0.3570387065410614, "grad_norm": 851.5135498046875, "learning_rate": 4.91987175349089e-07, "logits/chosen": 0.1060943752527237, "logits/rejected": 0.04715292900800705, "logps/chosen": -49.830238342285156, "logps/ref_chosen": -49.36516571044922, "logps/ref_rejected": -72.84671020507812, "logps/rejected": -73.60546875, "loss": 1.224, "margin_dpo/margin_mean": 0.29369184374809265, "margin_dpo/margin_std": 0.5422225594520569, "step": 116 }, { "epoch": 0.17687074829931973, "fcm_dpo/beta": 3.344947338104248, "fcm_dpo/delta": 0.09896770864725113, "fcm_dpo/margin": 0.27185767889022827, "fcm_dpo/q_t": 0.3576691150665283, "grad_norm": 870.2111206054688, "learning_rate": 4.916517197732933e-07, "logits/chosen": 0.129032701253891, "logits/rejected": 0.09620364010334015, "logps/chosen": -58.112640380859375, "logps/ref_chosen": -57.710899353027344, "logps/ref_rejected": -69.77253723144531, "logps/rejected": -70.4461441040039, "loss": 1.3806, "margin_dpo/margin_mean": 0.27185723185539246, "margin_dpo/margin_std": 0.6084860563278198, "step": 117 }, { "epoch": 0.17838246409674982, "fcm_dpo/beta": 3.271368980407715, "fcm_dpo/delta": -0.09690429270267487, "fcm_dpo/margin": 0.32874226570129395, "fcm_dpo/q_t": 0.34997400641441345, "grad_norm": 930.0485229492188, "learning_rate": 4.913095046794281e-07, "logits/chosen": 0.13153867423534393, "logits/rejected": 0.09707070142030716, "logps/chosen": -52.88352966308594, "logps/ref_chosen": -52.479896545410156, "logps/ref_rejected": -81.359130859375, "logps/rejected": -82.09149169921875, "loss": 1.2167, "margin_dpo/margin_mean": 0.32874205708503723, "margin_dpo/margin_std": 0.582424521446228, "step": 118 }, { "epoch": 0.17989417989417988, "fcm_dpo/beta": 3.331033229827881, "fcm_dpo/delta": 0.008004724979400635, "fcm_dpo/margin": 0.2980421185493469, "fcm_dpo/q_t": 0.35352998971939087, "grad_norm": 886.5343017578125, "learning_rate": 4.909605396399855e-07, "logits/chosen": 0.08484401553869247, "logits/rejected": 0.05185426026582718, "logps/chosen": -61.905765533447266, "logps/ref_chosen": -61.35767364501953, "logps/ref_rejected": -75.71510314941406, "logps/rejected": -76.56123352050781, "loss": 1.2981, "margin_dpo/margin_mean": 0.298042356967926, "margin_dpo/margin_std": 0.5981870293617249, "step": 119 }, { "epoch": 0.18140589569160998, "fcm_dpo/beta": 3.182232618331909, "fcm_dpo/delta": -0.24365851283073425, "fcm_dpo/margin": 0.3802499771118164, "fcm_dpo/q_t": 0.3237895369529724, "grad_norm": 770.2077026367188, "learning_rate": 4.906048344162676e-07, "logits/chosen": 0.07990750670433044, "logits/rejected": 0.02834871970117092, "logps/chosen": -60.297908782958984, "logps/ref_chosen": -59.907569885253906, "logps/ref_rejected": -79.6910629272461, "logps/rejected": -80.46165466308594, "loss": 1.0967, "margin_dpo/margin_mean": 0.3802502751350403, "margin_dpo/margin_std": 0.6159936189651489, "step": 120 }, { "epoch": 0.18291761148904007, "fcm_dpo/beta": 3.1936514377593994, "fcm_dpo/delta": 0.05449778214097023, "fcm_dpo/margin": 0.29768818616867065, "fcm_dpo/q_t": 0.3674450218677521, "grad_norm": 816.0641479492188, "learning_rate": 4.902423989581143e-07, "logits/chosen": 0.17014677822589874, "logits/rejected": 0.09852974861860275, "logps/chosen": -56.18149948120117, "logps/ref_chosen": -55.66604232788086, "logps/ref_rejected": -101.56233978271484, "logps/rejected": -102.37548828125, "loss": 1.2968, "margin_dpo/margin_mean": 0.2976876497268677, "margin_dpo/margin_std": 0.6306780576705933, "step": 121 }, { "epoch": 0.18442932728647016, "fcm_dpo/beta": 3.019984722137451, "fcm_dpo/delta": -0.44598841667175293, "fcm_dpo/margin": 0.4587884843349457, "fcm_dpo/q_t": 0.29104509949684143, "grad_norm": 804.1510009765625, "learning_rate": 4.898732434036243e-07, "logits/chosen": 0.09989838302135468, "logits/rejected": 0.06974966824054718, "logps/chosen": -63.821380615234375, "logps/ref_chosen": -63.334373474121094, "logps/ref_rejected": -73.67523193359375, "logps/rejected": -74.62103271484375, "loss": 0.969, "margin_dpo/margin_mean": 0.45878836512565613, "margin_dpo/margin_std": 0.6608290672302246, "step": 122 }, { "epoch": 0.18594104308390022, "fcm_dpo/beta": 2.9597277641296387, "fcm_dpo/delta": -0.11771807074546814, "fcm_dpo/margin": 0.37006598711013794, "fcm_dpo/q_t": 0.31938499212265015, "grad_norm": 801.2340698242188, "learning_rate": 4.894973780788722e-07, "logits/chosen": 0.13346600532531738, "logits/rejected": 0.09592346101999283, "logps/chosen": -57.292945861816406, "logps/ref_chosen": -56.89874267578125, "logps/ref_rejected": -78.97028350830078, "logps/rejected": -79.73455810546875, "loss": 1.2025, "margin_dpo/margin_mean": 0.37006592750549316, "margin_dpo/margin_std": 0.640432596206665, "step": 123 }, { "epoch": 0.1874527588813303, "fcm_dpo/beta": 2.767918109893799, "fcm_dpo/delta": -0.2092204988002777, "fcm_dpo/margin": 0.4275299310684204, "fcm_dpo/q_t": 0.31604132056236267, "grad_norm": 655.9889526367188, "learning_rate": 4.89114813497619e-07, "logits/chosen": 0.1320020854473114, "logits/rejected": 0.08239568769931793, "logps/chosen": -57.52501678466797, "logps/ref_chosen": -57.116085052490234, "logps/ref_rejected": -87.93074035644531, "logps/rejected": -88.76720428466797, "loss": 0.9876, "margin_dpo/margin_mean": 0.4275299310684204, "margin_dpo/margin_std": 0.6369043588638306, "step": 124 }, { "epoch": 0.1889644746787604, "fcm_dpo/beta": 2.691709518432617, "fcm_dpo/delta": -0.04524332284927368, "fcm_dpo/margin": 0.3849112093448639, "fcm_dpo/q_t": 0.32585692405700684, "grad_norm": 650.68212890625, "learning_rate": 4.887255603610184e-07, "logits/chosen": 0.18828628957271576, "logits/rejected": 0.1363983303308487, "logps/chosen": -66.21900939941406, "logps/ref_chosen": -65.7061767578125, "logps/ref_rejected": -91.72711944580078, "logps/rejected": -92.62486267089844, "loss": 1.0744, "margin_dpo/margin_mean": 0.3849112391471863, "margin_dpo/margin_std": 0.5876812934875488, "step": 125 }, { "epoch": 0.19047619047619047, "fcm_dpo/beta": 2.652081251144409, "fcm_dpo/delta": -0.10823916643857956, "fcm_dpo/margin": 0.41218724846839905, "fcm_dpo/q_t": 0.3507199287414551, "grad_norm": 565.577880859375, "learning_rate": 4.883296295573176e-07, "logits/chosen": -0.01404772698879242, "logits/rejected": -0.02040482684969902, "logps/chosen": -68.63446807861328, "logps/ref_chosen": -68.17608642578125, "logps/ref_rejected": -65.1175537109375, "logps/rejected": -65.98812103271484, "loss": 1.143, "margin_dpo/margin_mean": 0.41218748688697815, "margin_dpo/margin_std": 0.8156576156616211, "step": 126 }, { "epoch": 0.19198790627362056, "fcm_dpo/beta": 2.6257121562957764, "fcm_dpo/delta": -0.039457425475120544, "fcm_dpo/margin": 0.39357566833496094, "fcm_dpo/q_t": 0.32776835560798645, "grad_norm": 675.0282592773438, "learning_rate": 4.87927032161552e-07, "logits/chosen": 0.06537148356437683, "logits/rejected": 0.037958111613988876, "logps/chosen": -62.41735076904297, "logps/ref_chosen": -61.88023376464844, "logps/ref_rejected": -68.46012878417969, "logps/rejected": -69.39082336425781, "loss": 1.1334, "margin_dpo/margin_mean": 0.3935753107070923, "margin_dpo/margin_std": 0.6591010093688965, "step": 127 }, { "epoch": 0.19349962207105065, "fcm_dpo/beta": 2.6458208560943604, "fcm_dpo/delta": 0.08855466544628143, "fcm_dpo/margin": 0.3467669188976288, "fcm_dpo/q_t": 0.36068370938301086, "grad_norm": 708.3317260742188, "learning_rate": 4.875177794352363e-07, "logits/chosen": 0.09111534804105759, "logits/rejected": 0.045390479266643524, "logps/chosen": -67.32196044921875, "logps/ref_chosen": -66.708984375, "logps/ref_rejected": -94.97969055175781, "logps/rejected": -95.9394302368164, "loss": 1.2229, "margin_dpo/margin_mean": 0.34676679968833923, "margin_dpo/margin_std": 0.687119722366333, "step": 128 }, { "epoch": 0.19501133786848074, "fcm_dpo/beta": 2.789608955383301, "fcm_dpo/delta": 0.23963144421577454, "fcm_dpo/margin": 0.2792533040046692, "fcm_dpo/q_t": 0.38132244348526, "grad_norm": 765.4723510742188, "learning_rate": 4.871018828260491e-07, "logits/chosen": 0.1117800921201706, "logits/rejected": 0.10394299030303955, "logps/chosen": -65.94437408447266, "logps/ref_chosen": -65.33882904052734, "logps/ref_rejected": -68.06109619140625, "logps/rejected": -68.94589233398438, "loss": 1.2034, "margin_dpo/margin_mean": 0.279253751039505, "margin_dpo/margin_std": 0.5657248497009277, "step": 129 }, { "epoch": 0.1965230536659108, "fcm_dpo/beta": 2.7729220390319824, "fcm_dpo/delta": 0.050821587443351746, "fcm_dpo/margin": 0.33975258469581604, "fcm_dpo/q_t": 0.35864949226379395, "grad_norm": 768.4454956054688, "learning_rate": 4.866793539675126e-07, "logits/chosen": 0.09036006778478622, "logits/rejected": 0.04790624603629112, "logps/chosen": -59.219017028808594, "logps/ref_chosen": -58.660743713378906, "logps/ref_rejected": -79.24510192871094, "logps/rejected": -80.14312744140625, "loss": 1.1136, "margin_dpo/margin_mean": 0.3397524058818817, "margin_dpo/margin_std": 0.5575762987136841, "step": 130 }, { "epoch": 0.1980347694633409, "fcm_dpo/beta": 2.782139301300049, "fcm_dpo/delta": -0.2223011553287506, "fcm_dpo/margin": 0.4299342930316925, "fcm_dpo/q_t": 0.3284626305103302, "grad_norm": 713.0172119140625, "learning_rate": 4.86250204678667e-07, "logits/chosen": 0.10117419809103012, "logits/rejected": 0.048455823212862015, "logps/chosen": -52.9766845703125, "logps/ref_chosen": -52.51453399658203, "logps/ref_rejected": -85.18299865722656, "logps/rejected": -86.07508087158203, "loss": 1.152, "margin_dpo/margin_mean": 0.42993444204330444, "margin_dpo/margin_std": 0.7103478908538818, "step": 131 }, { "epoch": 0.19954648526077098, "fcm_dpo/beta": 2.728858470916748, "fcm_dpo/delta": 0.014488308690488338, "fcm_dpo/margin": 0.3616468608379364, "fcm_dpo/q_t": 0.3465641736984253, "grad_norm": 803.3519897460938, "learning_rate": 4.858144469637408e-07, "logits/chosen": 0.16957558691501617, "logits/rejected": 0.1406971514225006, "logps/chosen": -66.2645263671875, "logps/ref_chosen": -65.68513488769531, "logps/ref_rejected": -69.54120635986328, "logps/rejected": -70.48225402832031, "loss": 1.3716, "margin_dpo/margin_mean": 0.36164700984954834, "margin_dpo/margin_std": 0.8076159358024597, "step": 132 }, { "epoch": 0.20105820105820105, "fcm_dpo/beta": 2.8398377895355225, "fcm_dpo/delta": 0.20342613756656647, "fcm_dpo/margin": 0.2855343222618103, "fcm_dpo/q_t": 0.3706758916378021, "grad_norm": 806.1981811523438, "learning_rate": 4.853720930118138e-07, "logits/chosen": 0.07018555700778961, "logits/rejected": 0.06098049134016037, "logps/chosen": -64.16683959960938, "logps/ref_chosen": -63.598114013671875, "logps/ref_rejected": -73.72798156738281, "logps/rejected": -74.58223724365234, "loss": 1.3296, "margin_dpo/margin_mean": 0.2855341136455536, "margin_dpo/margin_std": 0.6436434984207153, "step": 133 }, { "epoch": 0.20256991685563114, "fcm_dpo/beta": 2.7438831329345703, "fcm_dpo/delta": -0.14177896082401276, "fcm_dpo/margin": 0.40749433636665344, "fcm_dpo/q_t": 0.32042205333709717, "grad_norm": 637.2561645507812, "learning_rate": 4.849231551964771e-07, "logits/chosen": 0.17743632197380066, "logits/rejected": 0.129373237490654, "logps/chosen": -54.34275817871094, "logps/ref_chosen": -53.79457092285156, "logps/ref_rejected": -74.16741943359375, "logps/rejected": -75.12309265136719, "loss": 1.0797, "margin_dpo/margin_mean": 0.40749499201774597, "margin_dpo/margin_std": 0.6834430694580078, "step": 134 }, { "epoch": 0.20408163265306123, "fcm_dpo/beta": 2.8520781993865967, "fcm_dpo/delta": 0.2403937578201294, "fcm_dpo/margin": 0.2731159031391144, "fcm_dpo/q_t": 0.3797072768211365, "grad_norm": 688.9449462890625, "learning_rate": 4.844676460754862e-07, "logits/chosen": 0.09586119651794434, "logits/rejected": 0.06667510420084, "logps/chosen": -49.97583770751953, "logps/ref_chosen": -49.441078186035156, "logps/ref_rejected": -65.96878051757812, "logps/rejected": -66.77665710449219, "loss": 1.3673, "margin_dpo/margin_mean": 0.27311572432518005, "margin_dpo/margin_std": 0.6469433307647705, "step": 135 }, { "epoch": 0.20559334845049132, "fcm_dpo/beta": 2.90311861038208, "fcm_dpo/delta": 0.004882900044322014, "fcm_dpo/margin": 0.3429165482521057, "fcm_dpo/q_t": 0.36461225152015686, "grad_norm": 1105.5750732421875, "learning_rate": 4.840055783904106e-07, "logits/chosen": 0.10260805487632751, "logits/rejected": 0.04400411248207092, "logps/chosen": -67.37596893310547, "logps/ref_chosen": -66.75926208496094, "logps/ref_rejected": -94.61787414550781, "logps/rejected": -95.57749938964844, "loss": 1.5811, "margin_dpo/margin_mean": 0.34291741251945496, "margin_dpo/margin_std": 0.9085370302200317, "step": 136 }, { "epoch": 0.20710506424792138, "fcm_dpo/beta": 2.891000747680664, "fcm_dpo/delta": -0.11807064712047577, "fcm_dpo/margin": 0.38187700510025024, "fcm_dpo/q_t": 0.3429448902606964, "grad_norm": 694.5245971679688, "learning_rate": 4.835369650662767e-07, "logits/chosen": 0.10072774440050125, "logits/rejected": 0.07752367854118347, "logps/chosen": -57.34507751464844, "logps/ref_chosen": -56.78379821777344, "logps/ref_rejected": -69.89952087402344, "logps/rejected": -70.8426742553711, "loss": 1.2, "margin_dpo/margin_mean": 0.3818773031234741, "margin_dpo/margin_std": 0.6857679486274719, "step": 137 }, { "epoch": 0.20861678004535147, "fcm_dpo/beta": 2.905198574066162, "fcm_dpo/delta": 0.12694688141345978, "fcm_dpo/margin": 0.3041801452636719, "fcm_dpo/q_t": 0.3584628403186798, "grad_norm": 783.094482421875, "learning_rate": 4.830618192112065e-07, "logits/chosen": 0.0714266374707222, "logits/rejected": 0.04139017313718796, "logps/chosen": -59.46550750732422, "logps/ref_chosen": -58.766014099121094, "logps/ref_rejected": -68.12371826171875, "logps/rejected": -69.12739562988281, "loss": 1.3424, "margin_dpo/margin_mean": 0.3041801452636719, "margin_dpo/margin_std": 0.6721060276031494, "step": 138 }, { "epoch": 0.21012849584278157, "fcm_dpo/beta": 2.842034101486206, "fcm_dpo/delta": -0.29515254497528076, "fcm_dpo/margin": 0.4422586262226105, "fcm_dpo/q_t": 0.32127201557159424, "grad_norm": 698.6984252929688, "learning_rate": 4.825801541160509e-07, "logits/chosen": 0.058577846735715866, "logits/rejected": 0.034485623240470886, "logps/chosen": -71.8461685180664, "logps/ref_chosen": -71.2255859375, "logps/ref_rejected": -82.1834716796875, "logps/rejected": -83.2463150024414, "loss": 1.0673, "margin_dpo/margin_mean": 0.44225820899009705, "margin_dpo/margin_std": 0.6527875661849976, "step": 139 }, { "epoch": 0.21164021164021163, "fcm_dpo/beta": 2.5584716796875, "fcm_dpo/delta": -0.5142702460289001, "fcm_dpo/margin": 0.5633730888366699, "fcm_dpo/q_t": 0.28819897770881653, "grad_norm": 720.5505981445312, "learning_rate": 4.820919832540181e-07, "logits/chosen": 0.06424537301063538, "logits/rejected": 0.027711138129234314, "logps/chosen": -63.82072830200195, "logps/ref_chosen": -63.27766418457031, "logps/ref_rejected": -83.30647277832031, "logps/rejected": -84.41291809082031, "loss": 1.1217, "margin_dpo/margin_mean": 0.5633726119995117, "margin_dpo/margin_std": 0.8722689151763916, "step": 140 }, { "epoch": 0.21315192743764172, "fcm_dpo/beta": 2.474776029586792, "fcm_dpo/delta": 0.05119156837463379, "fcm_dpo/margin": 0.3848419487476349, "fcm_dpo/q_t": 0.3656679391860962, "grad_norm": 698.813720703125, "learning_rate": 4.815973202802966e-07, "logits/chosen": 0.09540177881717682, "logits/rejected": 0.06088024377822876, "logps/chosen": -62.38557052612305, "logps/ref_chosen": -61.76676940917969, "logps/ref_rejected": -88.60601806640625, "logps/rejected": -89.60966491699219, "loss": 1.2344, "margin_dpo/margin_mean": 0.38484299182891846, "margin_dpo/margin_std": 0.7531988620758057, "step": 141 }, { "epoch": 0.2146636432350718, "fcm_dpo/beta": 2.6312007904052734, "fcm_dpo/delta": 0.25451576709747314, "fcm_dpo/margin": 0.2886815369129181, "fcm_dpo/q_t": 0.39209288358688354, "grad_norm": 677.07763671875, "learning_rate": 4.810961790316729e-07, "logits/chosen": 0.08383051306009293, "logits/rejected": 0.06255074590444565, "logps/chosen": -65.8593521118164, "logps/ref_chosen": -65.2747802734375, "logps/ref_rejected": -81.1378173828125, "logps/rejected": -82.01107788085938, "loss": 1.3684, "margin_dpo/margin_mean": 0.2886812686920166, "margin_dpo/margin_std": 0.7156628370285034, "step": 142 }, { "epoch": 0.2161753590325019, "fcm_dpo/beta": 2.587679386138916, "fcm_dpo/delta": -0.12541311979293823, "fcm_dpo/margin": 0.42952513694763184, "fcm_dpo/q_t": 0.3180665075778961, "grad_norm": 654.9125366210938, "learning_rate": 4.805885735261454e-07, "logits/chosen": 0.12669947743415833, "logits/rejected": 0.11181487888097763, "logps/chosen": -63.13290023803711, "logps/ref_chosen": -62.617828369140625, "logps/ref_rejected": -70.39239501953125, "logps/rejected": -71.33699035644531, "loss": 1.0608, "margin_dpo/margin_mean": 0.4295256435871124, "margin_dpo/margin_std": 0.6682602167129517, "step": 143 }, { "epoch": 0.21768707482993196, "fcm_dpo/beta": 2.556525707244873, "fcm_dpo/delta": -0.07277373969554901, "fcm_dpo/margin": 0.4165218472480774, "fcm_dpo/q_t": 0.34666940569877625, "grad_norm": 743.940185546875, "learning_rate": 4.800745179625307e-07, "logits/chosen": 0.11285848915576935, "logits/rejected": 0.0883728414773941, "logps/chosen": -61.44839096069336, "logps/ref_chosen": -60.80268859863281, "logps/ref_rejected": -79.07284545898438, "logps/rejected": -80.13507080078125, "loss": 1.1906, "margin_dpo/margin_mean": 0.4165222942829132, "margin_dpo/margin_std": 0.7579972743988037, "step": 144 }, { "epoch": 0.21919879062736206, "fcm_dpo/beta": 2.5957999229431152, "fcm_dpo/delta": 0.1555352658033371, "fcm_dpo/margin": 0.33037135004997253, "fcm_dpo/q_t": 0.3795892000198364, "grad_norm": 965.2445678710938, "learning_rate": 4.795540267200686e-07, "logits/chosen": 0.07255662977695465, "logits/rejected": 0.08949023485183716, "logps/chosen": -75.23291015625, "logps/ref_chosen": -74.61146545410156, "logps/ref_rejected": -83.24461364746094, "logps/rejected": -84.19642639160156, "loss": 1.456, "margin_dpo/margin_mean": 0.3303707540035248, "margin_dpo/margin_std": 0.8115462064743042, "step": 145 }, { "epoch": 0.22071050642479215, "fcm_dpo/beta": 2.5885000228881836, "fcm_dpo/delta": -0.06581529229879379, "fcm_dpo/margin": 0.40906020998954773, "fcm_dpo/q_t": 0.33817818760871887, "grad_norm": 619.466064453125, "learning_rate": 4.790271143580173e-07, "logits/chosen": 0.05093229562044144, "logits/rejected": 0.03536106273531914, "logps/chosen": -58.38066101074219, "logps/ref_chosen": -57.84098434448242, "logps/ref_rejected": -67.47422790527344, "logps/rejected": -68.42295837402344, "loss": 1.1586, "margin_dpo/margin_mean": 0.40906035900115967, "margin_dpo/margin_std": 0.7477720975875854, "step": 146 }, { "epoch": 0.2222222222222222, "fcm_dpo/beta": 2.669675588607788, "fcm_dpo/delta": 0.24072621762752533, "fcm_dpo/margin": 0.29151690006256104, "fcm_dpo/q_t": 0.38209617137908936, "grad_norm": 970.9570922851562, "learning_rate": 4.784937956152489e-07, "logits/chosen": 0.05634861811995506, "logits/rejected": 0.02099587954580784, "logps/chosen": -67.4126968383789, "logps/ref_chosen": -66.81346893310547, "logps/ref_rejected": -81.1796875, "logps/rejected": -82.0704345703125, "loss": 1.4796, "margin_dpo/margin_mean": 0.29151687026023865, "margin_dpo/margin_std": 0.7656582593917847, "step": 147 }, { "epoch": 0.2237339380196523, "fcm_dpo/beta": 2.6029810905456543, "fcm_dpo/delta": -0.31015288829803467, "fcm_dpo/margin": 0.48854613304138184, "fcm_dpo/q_t": 0.3191307783126831, "grad_norm": 515.7787475585938, "learning_rate": 4.779540854098347e-07, "logits/chosen": 0.19165629148483276, "logits/rejected": 0.12959185242652893, "logps/chosen": -49.28028106689453, "logps/ref_chosen": -48.6877555847168, "logps/ref_rejected": -67.50503540039062, "logps/rejected": -68.58610534667969, "loss": 1.0437, "margin_dpo/margin_mean": 0.488546222448349, "margin_dpo/margin_std": 0.7724089622497559, "step": 148 }, { "epoch": 0.2252456538170824, "fcm_dpo/beta": 2.533602476119995, "fcm_dpo/delta": -0.09391121566295624, "fcm_dpo/margin": 0.42736145853996277, "fcm_dpo/q_t": 0.34449946880340576, "grad_norm": 640.553466796875, "learning_rate": 4.774079988386296e-07, "logits/chosen": 0.05926530063152313, "logits/rejected": 0.01661105640232563, "logps/chosen": -55.83122253417969, "logps/ref_chosen": -55.143775939941406, "logps/ref_rejected": -64.79888916015625, "logps/rejected": -65.9136962890625, "loss": 1.1159, "margin_dpo/margin_mean": 0.4273618459701538, "margin_dpo/margin_std": 0.7076586484909058, "step": 149 }, { "epoch": 0.22675736961451248, "fcm_dpo/beta": 2.3987488746643066, "fcm_dpo/delta": -0.2589126229286194, "fcm_dpo/margin": 0.5113659501075745, "fcm_dpo/q_t": 0.2972312569618225, "grad_norm": 609.1364135742188, "learning_rate": 4.768555511768486e-07, "logits/chosen": 0.10247902572154999, "logits/rejected": 0.06602032482624054, "logps/chosen": -67.88574981689453, "logps/ref_chosen": -67.47074890136719, "logps/ref_rejected": -89.21170806884766, "logps/rejected": -90.13807678222656, "loss": 0.9804, "margin_dpo/margin_mean": 0.5113657712936401, "margin_dpo/margin_std": 0.6983498334884644, "step": 150 }, { "epoch": 0.22826908541194255, "fcm_dpo/beta": 2.3255553245544434, "fcm_dpo/delta": -0.19754400849342346, "fcm_dpo/margin": 0.5049396753311157, "fcm_dpo/q_t": 0.3148772716522217, "grad_norm": 509.62213134765625, "learning_rate": 4.762967578776406e-07, "logits/chosen": 0.06744161248207092, "logits/rejected": 0.025958776473999023, "logps/chosen": -52.93949890136719, "logps/ref_chosen": -52.45954132080078, "logps/ref_rejected": -79.0630111694336, "logps/rejected": -80.04790496826172, "loss": 0.9958, "margin_dpo/margin_mean": 0.5049391388893127, "margin_dpo/margin_std": 0.7245649099349976, "step": 151 }, { "epoch": 0.22978080120937264, "fcm_dpo/beta": 2.2168989181518555, "fcm_dpo/delta": -0.2083718478679657, "fcm_dpo/margin": 0.5341185927391052, "fcm_dpo/q_t": 0.30489107966423035, "grad_norm": 501.7793884277344, "learning_rate": 4.757316345716553e-07, "logits/chosen": 0.1456744223833084, "logits/rejected": 0.10498102009296417, "logps/chosen": -57.103904724121094, "logps/ref_chosen": -56.5538330078125, "logps/ref_rejected": -76.55074310302734, "logps/rejected": -77.63492584228516, "loss": 0.9032, "margin_dpo/margin_mean": 0.5341184139251709, "margin_dpo/margin_std": 0.6747971773147583, "step": 152 }, { "epoch": 0.23129251700680273, "fcm_dpo/beta": 2.214939832687378, "fcm_dpo/delta": 0.11927812546491623, "fcm_dpo/margin": 0.40247973799705505, "fcm_dpo/q_t": 0.3638674020767212, "grad_norm": 600.3435668945312, "learning_rate": 4.751601970666064e-07, "logits/chosen": 0.05303303897380829, "logits/rejected": 0.01978529989719391, "logps/chosen": -68.58604431152344, "logps/ref_chosen": -68.00689697265625, "logps/ref_rejected": -74.83482360839844, "logps/rejected": -75.81645202636719, "loss": 1.2482, "margin_dpo/margin_mean": 0.4024793207645416, "margin_dpo/margin_std": 0.7963600158691406, "step": 153 }, { "epoch": 0.2328042328042328, "fcm_dpo/beta": 2.284209728240967, "fcm_dpo/delta": 0.13816551864147186, "fcm_dpo/margin": 0.38259416818618774, "fcm_dpo/q_t": 0.3632936179637909, "grad_norm": 581.8963623046875, "learning_rate": 4.745824613468292e-07, "logits/chosen": 0.13186918199062347, "logits/rejected": 0.12810632586479187, "logps/chosen": -59.913002014160156, "logps/ref_chosen": -59.222537994384766, "logps/ref_rejected": -64.19131469726562, "logps/rejected": -65.2643814086914, "loss": 1.2855, "margin_dpo/margin_mean": 0.3825940191745758, "margin_dpo/margin_std": 0.8229261636734009, "step": 154 }, { "epoch": 0.23431594860166288, "fcm_dpo/beta": 2.359236717224121, "fcm_dpo/delta": 0.17262253165245056, "fcm_dpo/margin": 0.3568933308124542, "fcm_dpo/q_t": 0.3739526867866516, "grad_norm": 687.4069213867188, "learning_rate": 4.7399844357283393e-07, "logits/chosen": 0.14357620477676392, "logits/rejected": 0.1261100172996521, "logps/chosen": -69.06387329101562, "logps/ref_chosen": -68.45469665527344, "logps/ref_rejected": -77.91763305664062, "logps/rejected": -78.88371276855469, "loss": 1.4692, "margin_dpo/margin_mean": 0.35689258575439453, "margin_dpo/margin_std": 0.879474401473999, "step": 155 }, { "epoch": 0.23582766439909297, "fcm_dpo/beta": 2.3249125480651855, "fcm_dpo/delta": -0.2846854329109192, "fcm_dpo/margin": 0.5367815494537354, "fcm_dpo/q_t": 0.33714932203292847, "grad_norm": 715.7948608398438, "learning_rate": 4.7340816008085305e-07, "logits/chosen": 0.09127533435821533, "logits/rejected": 0.05366864800453186, "logps/chosen": -67.89020538330078, "logps/ref_chosen": -67.26959991455078, "logps/ref_rejected": -86.95914459228516, "logps/rejected": -88.11653137207031, "loss": 1.1585, "margin_dpo/margin_mean": 0.5367816686630249, "margin_dpo/margin_std": 0.9481757879257202, "step": 156 }, { "epoch": 0.23733938019652306, "fcm_dpo/beta": 2.2261717319488525, "fcm_dpo/delta": -0.006286881864070892, "fcm_dpo/margin": 0.45090270042419434, "fcm_dpo/q_t": 0.3336307406425476, "grad_norm": 524.041259765625, "learning_rate": 4.728116273823847e-07, "logits/chosen": 0.08175022900104523, "logits/rejected": 0.062264494597911835, "logps/chosen": -55.3340950012207, "logps/ref_chosen": -54.77287292480469, "logps/ref_rejected": -63.87866973876953, "logps/rejected": -64.89079284667969, "loss": 1.0499, "margin_dpo/margin_mean": 0.45090246200561523, "margin_dpo/margin_std": 0.7311956286430359, "step": 157 }, { "epoch": 0.23885109599395313, "fcm_dpo/beta": 2.172646999359131, "fcm_dpo/delta": -0.2092001736164093, "fcm_dpo/margin": 0.5448204278945923, "fcm_dpo/q_t": 0.3070847988128662, "grad_norm": 502.65093994140625, "learning_rate": 4.7220886216373085e-07, "logits/chosen": 0.1196097731590271, "logits/rejected": 0.0907188355922699, "logps/chosen": -65.47095489501953, "logps/ref_chosen": -64.92271423339844, "logps/ref_rejected": -82.23789978027344, "logps/rejected": -83.3309555053711, "loss": 0.9403, "margin_dpo/margin_mean": 0.5448204278945923, "margin_dpo/margin_std": 0.7522009015083313, "step": 158 }, { "epoch": 0.24036281179138322, "fcm_dpo/beta": 2.2735085487365723, "fcm_dpo/delta": 0.19475619494915009, "fcm_dpo/margin": 0.3540440499782562, "fcm_dpo/q_t": 0.35417577624320984, "grad_norm": 675.0926513671875, "learning_rate": 4.715998812855304e-07, "logits/chosen": 0.13686862587928772, "logits/rejected": 0.10717260837554932, "logps/chosen": -57.696720123291016, "logps/ref_chosen": -57.046993255615234, "logps/ref_rejected": -73.32441711425781, "logps/rejected": -74.32818603515625, "loss": 1.2425, "margin_dpo/margin_mean": 0.3540443778038025, "margin_dpo/margin_std": 0.6996503472328186, "step": 159 }, { "epoch": 0.2418745275888133, "fcm_dpo/beta": 2.3114399909973145, "fcm_dpo/delta": 0.17240478098392487, "fcm_dpo/margin": 0.3636714816093445, "fcm_dpo/q_t": 0.3722858130931854, "grad_norm": 635.1336669921875, "learning_rate": 4.7098470178228755e-07, "logits/chosen": -0.020404599606990814, "logits/rejected": -0.04949381574988365, "logps/chosen": -50.52197265625, "logps/ref_chosen": -49.806915283203125, "logps/ref_rejected": -68.3370132446289, "logps/rejected": -69.41574096679688, "loss": 1.2751, "margin_dpo/margin_mean": 0.3636714518070221, "margin_dpo/margin_std": 0.8036404848098755, "step": 160 }, { "epoch": 0.24338624338624337, "fcm_dpo/beta": 2.285512924194336, "fcm_dpo/delta": -0.16357703506946564, "fcm_dpo/margin": 0.5008035898208618, "fcm_dpo/q_t": 0.35015690326690674, "grad_norm": 537.9451293945312, "learning_rate": 4.703633408618955e-07, "logits/chosen": 0.10273732244968414, "logits/rejected": 0.07215458899736404, "logps/chosen": -53.11811065673828, "logps/ref_chosen": -52.50048828125, "logps/ref_rejected": -66.04540252685547, "logps/rejected": -67.16382598876953, "loss": 1.1267, "margin_dpo/margin_mean": 0.5008042454719543, "margin_dpo/margin_std": 0.8614367246627808, "step": 161 }, { "epoch": 0.24489795918367346, "fcm_dpo/beta": 2.08780574798584, "fcm_dpo/delta": -0.3560563623905182, "fcm_dpo/margin": 0.6207355260848999, "fcm_dpo/q_t": 0.2931872010231018, "grad_norm": 511.5837097167969, "learning_rate": 4.697358159051549e-07, "logits/chosen": 0.16846126317977905, "logits/rejected": 0.12896160781383514, "logps/chosen": -70.21339416503906, "logps/ref_chosen": -69.46919250488281, "logps/ref_rejected": -92.00952911376953, "logps/rejected": -93.37446594238281, "loss": 0.9736, "margin_dpo/margin_mean": 0.6207360029220581, "margin_dpo/margin_std": 0.8480439186096191, "step": 162 }, { "epoch": 0.24640967498110355, "fcm_dpo/beta": 2.0367884635925293, "fcm_dpo/delta": -0.2981437146663666, "fcm_dpo/margin": 0.618397057056427, "fcm_dpo/q_t": 0.29892927408218384, "grad_norm": 515.961181640625, "learning_rate": 4.691021444652876e-07, "logits/chosen": 0.10065104067325592, "logits/rejected": 0.06360255181789398, "logps/chosen": -51.2899055480957, "logps/ref_chosen": -50.613834381103516, "logps/ref_rejected": -74.62033081054688, "logps/rejected": -75.914794921875, "loss": 0.9615, "margin_dpo/margin_mean": 0.6183971762657166, "margin_dpo/margin_std": 0.8240780830383301, "step": 163 }, { "epoch": 0.24792139077853365, "fcm_dpo/beta": 1.9806370735168457, "fcm_dpo/delta": -0.047169312834739685, "fcm_dpo/margin": 0.5249905586242676, "fcm_dpo/q_t": 0.32894620299339294, "grad_norm": 464.32489013671875, "learning_rate": 4.6846234426744624e-07, "logits/chosen": 0.08540582656860352, "logits/rejected": 0.03592706099152565, "logps/chosen": -55.629249572753906, "logps/ref_chosen": -54.848114013671875, "logps/ref_rejected": -79.0630111694336, "logps/rejected": -80.369140625, "loss": 1.0554, "margin_dpo/margin_mean": 0.524990439414978, "margin_dpo/margin_std": 0.7865326404571533, "step": 164 }, { "epoch": 0.2494331065759637, "fcm_dpo/beta": 1.966191291809082, "fcm_dpo/delta": -0.0075155869126319885, "fcm_dpo/margin": 0.5113043785095215, "fcm_dpo/q_t": 0.3119150400161743, "grad_norm": 391.06072998046875, "learning_rate": 4.678164332082175e-07, "logits/chosen": 0.1477801650762558, "logits/rejected": 0.10125482082366943, "logps/chosen": -51.86711883544922, "logps/ref_chosen": -51.089210510253906, "logps/ref_rejected": -71.23370361328125, "logps/rejected": -72.52291870117188, "loss": 0.9392, "margin_dpo/margin_mean": 0.5113040804862976, "margin_dpo/margin_std": 0.6569217443466187, "step": 165 }, { "epoch": 0.2509448223733938, "fcm_dpo/beta": 2.0378761291503906, "fcm_dpo/delta": 0.31097179651260376, "fcm_dpo/margin": 0.34910979866981506, "fcm_dpo/q_t": 0.38307300209999084, "grad_norm": 556.363037109375, "learning_rate": 4.6716442935512214e-07, "logits/chosen": 0.10038108378648758, "logits/rejected": 0.036018554121255875, "logps/chosen": -63.89585876464844, "logps/ref_chosen": -63.19081115722656, "logps/ref_rejected": -93.8402099609375, "logps/rejected": -94.89436340332031, "loss": 1.2454, "margin_dpo/margin_mean": 0.3491097092628479, "margin_dpo/margin_std": 0.7657175064086914, "step": 166 }, { "epoch": 0.25245653817082386, "fcm_dpo/beta": 2.0208253860473633, "fcm_dpo/delta": -0.17621225118637085, "fcm_dpo/margin": 0.5719941854476929, "fcm_dpo/q_t": 0.2979215979576111, "grad_norm": 404.3883056640625, "learning_rate": 4.6650635094610966e-07, "logits/chosen": 0.06507319211959839, "logits/rejected": 0.03547991067171097, "logps/chosen": -59.553977966308594, "logps/ref_chosen": -58.92427062988281, "logps/ref_rejected": -72.97377014160156, "logps/rejected": -74.17547607421875, "loss": 0.8796, "margin_dpo/margin_mean": 0.5719939470291138, "margin_dpo/margin_std": 0.6960855722427368, "step": 167 }, { "epoch": 0.25396825396825395, "fcm_dpo/beta": 2.060161828994751, "fcm_dpo/delta": 0.1698514223098755, "fcm_dpo/margin": 0.4094662666320801, "fcm_dpo/q_t": 0.3545387387275696, "grad_norm": 583.7806396484375, "learning_rate": 4.6584221638904767e-07, "logits/chosen": 0.07120160013437271, "logits/rejected": 0.04992123693227768, "logps/chosen": -66.4698486328125, "logps/ref_chosen": -65.65138244628906, "logps/ref_rejected": -79.71418762207031, "logps/rejected": -80.94212341308594, "loss": 1.1174, "margin_dpo/margin_mean": 0.40946611762046814, "margin_dpo/margin_std": 0.7269895076751709, "step": 168 }, { "epoch": 0.25547996976568405, "fcm_dpo/beta": 2.0494308471679688, "fcm_dpo/delta": -0.08093604445457458, "fcm_dpo/margin": 0.5232309699058533, "fcm_dpo/q_t": 0.34376293420791626, "grad_norm": 552.8328857421875, "learning_rate": 4.651720442612075e-07, "logits/chosen": 0.15064923465251923, "logits/rejected": 0.12327395379543304, "logps/chosen": -62.13090515136719, "logps/ref_chosen": -61.425865173339844, "logps/ref_rejected": -76.09590148925781, "logps/rejected": -77.32416534423828, "loss": 1.0984, "margin_dpo/margin_mean": 0.523231029510498, "margin_dpo/margin_std": 0.913813591003418, "step": 169 }, { "epoch": 0.25699168556311414, "fcm_dpo/beta": 2.1044416427612305, "fcm_dpo/delta": 0.22101661562919617, "fcm_dpo/margin": 0.3785492181777954, "fcm_dpo/q_t": 0.3640963137149811, "grad_norm": 517.9861450195312, "learning_rate": 4.6449585330874425e-07, "logits/chosen": 0.05246744677424431, "logits/rejected": 0.051534149795770645, "logps/chosen": -57.42051696777344, "logps/ref_chosen": -56.65319061279297, "logps/ref_rejected": -63.45965576171875, "logps/rejected": -64.60552215576172, "loss": 1.2398, "margin_dpo/margin_mean": 0.3785494565963745, "margin_dpo/margin_std": 0.7789652943611145, "step": 170 }, { "epoch": 0.2585034013605442, "fcm_dpo/beta": 2.0837273597717285, "fcm_dpo/delta": -0.041708558797836304, "fcm_dpo/margin": 0.49539345502853394, "fcm_dpo/q_t": 0.31717920303344727, "grad_norm": 528.3797607421875, "learning_rate": 4.6381366244617224e-07, "logits/chosen": 0.12783187627792358, "logits/rejected": 0.08980339765548706, "logps/chosen": -64.4964599609375, "logps/ref_chosen": -63.73476028442383, "logps/ref_rejected": -78.50328063964844, "logps/rejected": -79.76036834716797, "loss": 1.0854, "margin_dpo/margin_mean": 0.49539363384246826, "margin_dpo/margin_std": 0.778314471244812, "step": 171 }, { "epoch": 0.2600151171579743, "fcm_dpo/beta": 2.0559816360473633, "fcm_dpo/delta": -0.2410603016614914, "fcm_dpo/margin": 0.589635968208313, "fcm_dpo/q_t": 0.2946144938468933, "grad_norm": 428.92987060546875, "learning_rate": 4.631254907558365e-07, "logits/chosen": 0.17183159291744232, "logits/rejected": 0.12959660589694977, "logps/chosen": -52.974788665771484, "logps/ref_chosen": -52.201759338378906, "logps/ref_rejected": -82.85285949707031, "logps/rejected": -84.21553039550781, "loss": 0.949, "margin_dpo/margin_mean": 0.5896360874176025, "margin_dpo/margin_std": 0.7388289570808411, "step": 172 }, { "epoch": 0.2615268329554044, "fcm_dpo/beta": 1.8781511783599854, "fcm_dpo/delta": -0.25002074241638184, "fcm_dpo/margin": 0.6354281902313232, "fcm_dpo/q_t": 0.32253411412239075, "grad_norm": 381.1091613769531, "learning_rate": 4.624313574873786e-07, "logits/chosen": 0.14371845126152039, "logits/rejected": 0.07756569981575012, "logps/chosen": -56.25102233886719, "logps/ref_chosen": -55.434722900390625, "logps/ref_rejected": -77.81967163085938, "logps/rejected": -79.27140045166016, "loss": 0.9918, "margin_dpo/margin_mean": 0.6354283094406128, "margin_dpo/margin_std": 0.9134526252746582, "step": 173 }, { "epoch": 0.26303854875283444, "fcm_dpo/beta": 1.8751147985458374, "fcm_dpo/delta": -0.21836894750595093, "fcm_dpo/margin": 0.6357536315917969, "fcm_dpo/q_t": 0.3092048168182373, "grad_norm": 498.1719665527344, "learning_rate": 4.61731282057198e-07, "logits/chosen": 0.12083408981561661, "logits/rejected": 0.07103556394577026, "logps/chosen": -57.99762725830078, "logps/ref_chosen": -57.17195129394531, "logps/ref_rejected": -85.47578430175781, "logps/rejected": -86.93720245361328, "loss": 1.0066, "margin_dpo/margin_mean": 0.6357530355453491, "margin_dpo/margin_std": 0.8986474275588989, "step": 174 }, { "epoch": 0.26455026455026454, "fcm_dpo/beta": 1.770094394683838, "fcm_dpo/delta": -0.23501265048980713, "fcm_dpo/margin": 0.6816864013671875, "fcm_dpo/q_t": 0.30375754833221436, "grad_norm": 463.9217529296875, "learning_rate": 4.6102528404790965e-07, "logits/chosen": 0.18367326259613037, "logits/rejected": 0.15908128023147583, "logps/chosen": -68.52366638183594, "logps/ref_chosen": -67.6656265258789, "logps/ref_rejected": -84.36766815185547, "logps/rejected": -85.90739440917969, "loss": 0.9656, "margin_dpo/margin_mean": 0.6816866397857666, "margin_dpo/margin_std": 0.915657639503479, "step": 175 }, { "epoch": 0.2660619803476946, "fcm_dpo/beta": 1.7365822792053223, "fcm_dpo/delta": 0.01569700986146927, "fcm_dpo/margin": 0.5668948888778687, "fcm_dpo/q_t": 0.37061506509780884, "grad_norm": 517.9691162109375, "learning_rate": 4.603133832077953e-07, "logits/chosen": 0.08695434033870697, "logits/rejected": 0.06711474061012268, "logps/chosen": -78.7764892578125, "logps/ref_chosen": -77.8587646484375, "logps/ref_rejected": -81.08732604980469, "logps/rejected": -82.57195281982422, "loss": 1.2356, "margin_dpo/margin_mean": 0.5668948292732239, "margin_dpo/margin_std": 1.1584200859069824, "step": 176 }, { "epoch": 0.2675736961451247, "fcm_dpo/beta": 1.6431140899658203, "fcm_dpo/delta": -0.37270694971084595, "fcm_dpo/margin": 0.8022236227989197, "fcm_dpo/q_t": 0.2722761631011963, "grad_norm": 448.8039245605469, "learning_rate": 4.5959559945025183e-07, "logits/chosen": 0.24163630604743958, "logits/rejected": 0.1636749804019928, "logps/chosen": -56.08782958984375, "logps/ref_chosen": -55.22039794921875, "logps/ref_rejected": -92.54973602294922, "logps/rejected": -94.21939086914062, "loss": 0.8902, "margin_dpo/margin_mean": 0.8022229671478271, "margin_dpo/margin_std": 1.0025627613067627, "step": 177 }, { "epoch": 0.2690854119425548, "fcm_dpo/beta": 1.6737594604492188, "fcm_dpo/delta": 0.21938863396644592, "fcm_dpo/margin": 0.47716161608695984, "fcm_dpo/q_t": 0.3556697368621826, "grad_norm": 458.8934020996094, "learning_rate": 4.588719528532341e-07, "logits/chosen": 0.08861234784126282, "logits/rejected": 0.05083230137825012, "logps/chosen": -61.76763916015625, "logps/ref_chosen": -60.81049346923828, "logps/ref_rejected": -81.12973022460938, "logps/rejected": -82.56403350830078, "loss": 1.0621, "margin_dpo/margin_mean": 0.47716209292411804, "margin_dpo/margin_std": 0.7696354985237122, "step": 178 }, { "epoch": 0.2705971277399849, "fcm_dpo/beta": 1.7257657051086426, "fcm_dpo/delta": 0.10213658213615417, "fcm_dpo/margin": 0.5256574153900146, "fcm_dpo/q_t": 0.35149049758911133, "grad_norm": 419.9453125, "learning_rate": 4.581424636586928e-07, "logits/chosen": 0.16226297616958618, "logits/rejected": 0.1488857865333557, "logps/chosen": -66.7244644165039, "logps/ref_chosen": -65.67171478271484, "logps/ref_rejected": -75.32586669921875, "logps/rejected": -76.9042739868164, "loss": 1.0632, "margin_dpo/margin_mean": 0.5256578922271729, "margin_dpo/margin_std": 0.8762655258178711, "step": 179 }, { "epoch": 0.272108843537415, "fcm_dpo/beta": 1.8206181526184082, "fcm_dpo/delta": 0.2291109263896942, "fcm_dpo/margin": 0.42961639165878296, "fcm_dpo/q_t": 0.36887508630752563, "grad_norm": 520.1433715820312, "learning_rate": 4.5740715227200897e-07, "logits/chosen": -0.0028491299599409103, "logits/rejected": -0.017982792109251022, "logps/chosen": -57.53323745727539, "logps/ref_chosen": -56.68280792236328, "logps/ref_rejected": -64.94414520263672, "logps/rejected": -66.22418975830078, "loss": 1.2665, "margin_dpo/margin_mean": 0.4296168386936188, "margin_dpo/margin_std": 0.9101868867874146, "step": 180 }, { "epoch": 0.273620559334845, "fcm_dpo/beta": 1.7677171230316162, "fcm_dpo/delta": -0.24141666293144226, "fcm_dpo/margin": 0.6860055923461914, "fcm_dpo/q_t": 0.2977555990219116, "grad_norm": 448.171630859375, "learning_rate": 4.566660392614228e-07, "logits/chosen": 0.13813161849975586, "logits/rejected": 0.10955438017845154, "logps/chosen": -61.58479309082031, "logps/ref_chosen": -60.77604675292969, "logps/ref_rejected": -83.98361206054688, "logps/rejected": -85.47836303710938, "loss": 0.9069, "margin_dpo/margin_mean": 0.686005711555481, "margin_dpo/margin_std": 0.8478412628173828, "step": 181 }, { "epoch": 0.2751322751322751, "fcm_dpo/beta": 1.698185920715332, "fcm_dpo/delta": -0.05800933390855789, "fcm_dpo/margin": 0.6177934408187866, "fcm_dpo/q_t": 0.3166268467903137, "grad_norm": 430.59722900390625, "learning_rate": 4.5591914535745817e-07, "logits/chosen": 0.1499433070421219, "logits/rejected": 0.09003090858459473, "logps/chosen": -61.18524932861328, "logps/ref_chosen": -60.2537841796875, "logps/ref_rejected": -89.7706298828125, "logps/rejected": -91.31988525390625, "loss": 1.0536, "margin_dpo/margin_mean": 0.6177935004234314, "margin_dpo/margin_std": 0.9228367209434509, "step": 182 }, { "epoch": 0.2766439909297052, "fcm_dpo/beta": 1.8208626508712769, "fcm_dpo/delta": 0.4041329324245453, "fcm_dpo/margin": 0.34058958292007446, "fcm_dpo/q_t": 0.3961232900619507, "grad_norm": 510.5823059082031, "learning_rate": 4.551664914523433e-07, "logits/chosen": 0.1353759616613388, "logits/rejected": 0.12078934907913208, "logps/chosen": -62.945030212402344, "logps/ref_chosen": -61.76142120361328, "logps/ref_rejected": -72.54627990722656, "logps/rejected": -74.07048034667969, "loss": 1.3364, "margin_dpo/margin_mean": 0.3405901789665222, "margin_dpo/margin_std": 0.8557813763618469, "step": 183 }, { "epoch": 0.2781557067271353, "fcm_dpo/beta": 1.8339712619781494, "fcm_dpo/delta": 0.05328105390071869, "fcm_dpo/margin": 0.5154128074645996, "fcm_dpo/q_t": 0.34025606513023376, "grad_norm": 346.7197265625, "learning_rate": 4.544080985994258e-07, "logits/chosen": 0.21997570991516113, "logits/rejected": 0.1718028038740158, "logps/chosen": -47.82099151611328, "logps/ref_chosen": -46.840721130371094, "logps/ref_rejected": -69.3609390258789, "logps/rejected": -70.85662841796875, "loss": 0.989, "margin_dpo/margin_mean": 0.5154126882553101, "margin_dpo/margin_std": 0.7232804298400879, "step": 184 }, { "epoch": 0.2796674225245654, "fcm_dpo/beta": 1.8281052112579346, "fcm_dpo/delta": -0.15045057237148285, "fcm_dpo/margin": 0.6196208000183105, "fcm_dpo/q_t": 0.321952223777771, "grad_norm": 413.64068603515625, "learning_rate": 4.5364398801258394e-07, "logits/chosen": 0.14673639833927155, "logits/rejected": 0.10907743126153946, "logps/chosen": -53.348968505859375, "logps/ref_chosen": -52.32114028930664, "logps/ref_rejected": -68.3885726928711, "logps/rejected": -70.03601837158203, "loss": 1.0869, "margin_dpo/margin_mean": 0.6196208000183105, "margin_dpo/margin_std": 1.0106725692749023, "step": 185 }, { "epoch": 0.2811791383219955, "fcm_dpo/beta": 1.8034803867340088, "fcm_dpo/delta": -0.005721554160118103, "fcm_dpo/margin": 0.5566083788871765, "fcm_dpo/q_t": 0.3407011032104492, "grad_norm": 465.7325744628906, "learning_rate": 4.5287418106563354e-07, "logits/chosen": 0.07695234566926956, "logits/rejected": 0.04599303752183914, "logps/chosen": -68.40414428710938, "logps/ref_chosen": -67.42012786865234, "logps/ref_rejected": -82.50968933105469, "logps/rejected": -84.0503158569336, "loss": 1.0608, "margin_dpo/margin_mean": 0.5566080808639526, "margin_dpo/margin_std": 0.8762015104293823, "step": 186 }, { "epoch": 0.28269085411942557, "fcm_dpo/beta": 1.7329106330871582, "fcm_dpo/delta": -0.18373380601406097, "fcm_dpo/margin": 0.6642186641693115, "fcm_dpo/q_t": 0.32172733545303345, "grad_norm": 530.1708374023438, "learning_rate": 4.520986992917297e-07, "logits/chosen": 0.14111362397670746, "logits/rejected": 0.09435050934553146, "logps/chosen": -76.567138671875, "logps/ref_chosen": -75.52549743652344, "logps/ref_rejected": -94.76289367675781, "logps/rejected": -96.46875, "loss": 1.1758, "margin_dpo/margin_mean": 0.6642183661460876, "margin_dpo/margin_std": 1.1894935369491577, "step": 187 }, { "epoch": 0.2842025699168556, "fcm_dpo/beta": 1.744195818901062, "fcm_dpo/delta": 0.005925014615058899, "fcm_dpo/margin": 0.5699671506881714, "fcm_dpo/q_t": 0.3193795382976532, "grad_norm": 516.5187377929688, "learning_rate": 4.5131756438276466e-07, "logits/chosen": 0.151597797870636, "logits/rejected": 0.11677326261997223, "logps/chosen": -72.44251251220703, "logps/ref_chosen": -71.52333068847656, "logps/ref_rejected": -78.29949951171875, "logps/rejected": -79.78865814208984, "loss": 1.1632, "margin_dpo/margin_mean": 0.56996750831604, "margin_dpo/margin_std": 1.0281386375427246, "step": 188 }, { "epoch": 0.2857142857142857, "fcm_dpo/beta": 1.6932382583618164, "fcm_dpo/delta": -0.15139150619506836, "fcm_dpo/margin": 0.6669385433197021, "fcm_dpo/q_t": 0.30612969398498535, "grad_norm": 391.1927795410156, "learning_rate": 4.5053079818876096e-07, "logits/chosen": 0.12143361568450928, "logits/rejected": 0.12991394102573395, "logps/chosen": -73.05665588378906, "logps/ref_chosen": -72.17626953125, "logps/ref_rejected": -75.26313781738281, "logps/rejected": -76.81045532226562, "loss": 0.8928, "margin_dpo/margin_mean": 0.6669397950172424, "margin_dpo/margin_std": 0.832381010055542, "step": 189 }, { "epoch": 0.2872260015117158, "fcm_dpo/beta": 1.6875749826431274, "fcm_dpo/delta": -0.09505629539489746, "fcm_dpo/margin": 0.642721951007843, "fcm_dpo/q_t": 0.32302048802375793, "grad_norm": 424.7405090332031, "learning_rate": 4.4973842271726024e-07, "logits/chosen": 0.19074919819831848, "logits/rejected": 0.081771120429039, "logps/chosen": -55.586639404296875, "logps/ref_chosen": -54.624271392822266, "logps/ref_rejected": -101.47068786621094, "logps/rejected": -103.0757827758789, "loss": 1.0135, "margin_dpo/margin_mean": 0.6427220106124878, "margin_dpo/margin_std": 0.9737996459007263, "step": 190 }, { "epoch": 0.2887377173091459, "fcm_dpo/beta": 1.6515135765075684, "fcm_dpo/delta": -0.04676612466573715, "fcm_dpo/margin": 0.6305712461471558, "fcm_dpo/q_t": 0.31939807534217834, "grad_norm": 493.140625, "learning_rate": 4.48940460132708e-07, "logits/chosen": 0.1993117779493332, "logits/rejected": 0.1801232248544693, "logps/chosen": -74.03681945800781, "logps/ref_chosen": -72.93251037597656, "logps/ref_rejected": -89.95103454589844, "logps/rejected": -91.68590545654297, "loss": 1.047, "margin_dpo/margin_mean": 0.6305709481239319, "margin_dpo/margin_std": 0.9657796621322632, "step": 191 }, { "epoch": 0.29024943310657597, "fcm_dpo/beta": 1.7161282300949097, "fcm_dpo/delta": 0.19982855021953583, "fcm_dpo/margin": 0.47440922260284424, "fcm_dpo/q_t": 0.3627406060695648, "grad_norm": 372.83734130859375, "learning_rate": 4.481369327558329e-07, "logits/chosen": 0.16536489129066467, "logits/rejected": 0.1462571918964386, "logps/chosen": -55.00927734375, "logps/ref_chosen": -54.001121520996094, "logps/ref_rejected": -63.531551361083984, "logps/rejected": -65.01411437988281, "loss": 1.1335, "margin_dpo/margin_mean": 0.4744090139865875, "margin_dpo/margin_std": 0.8528145551681519, "step": 192 }, { "epoch": 0.29176114890400606, "fcm_dpo/beta": 1.6729130744934082, "fcm_dpo/delta": -0.11725394427776337, "fcm_dpo/margin": 0.6579139828681946, "fcm_dpo/q_t": 0.3185346722602844, "grad_norm": 344.4507751464844, "learning_rate": 4.47327863063023e-07, "logits/chosen": 0.09357620775699615, "logits/rejected": 0.07541916519403458, "logps/chosen": -57.7445182800293, "logps/ref_chosen": -56.74927520751953, "logps/ref_rejected": -58.80629348754883, "logps/rejected": -60.459449768066406, "loss": 0.9244, "margin_dpo/margin_mean": 0.6579139232635498, "margin_dpo/margin_std": 0.9124239087104797, "step": 193 }, { "epoch": 0.29327286470143615, "fcm_dpo/beta": 1.6954293251037598, "fcm_dpo/delta": 0.1348000019788742, "fcm_dpo/margin": 0.5165129899978638, "fcm_dpo/q_t": 0.36234208941459656, "grad_norm": 409.556396484375, "learning_rate": 4.4651327368569684e-07, "logits/chosen": 0.1781534105539322, "logits/rejected": 0.15505832433700562, "logps/chosen": -57.62710189819336, "logps/ref_chosen": -56.64944076538086, "logps/ref_rejected": -69.98954772949219, "logps/rejected": -71.48371887207031, "loss": 1.1709, "margin_dpo/margin_mean": 0.5165130496025085, "margin_dpo/margin_std": 0.9614365100860596, "step": 194 }, { "epoch": 0.2947845804988662, "fcm_dpo/beta": 1.7598028182983398, "fcm_dpo/delta": 0.22711437940597534, "fcm_dpo/margin": 0.4485671818256378, "fcm_dpo/q_t": 0.3598003685474396, "grad_norm": 496.6513977050781, "learning_rate": 4.4569318740967043e-07, "logits/chosen": 0.09457789361476898, "logits/rejected": 0.09252005815505981, "logps/chosen": -71.65326690673828, "logps/ref_chosen": -70.40977478027344, "logps/ref_rejected": -74.39448547363281, "logps/rejected": -76.08654022216797, "loss": 1.2491, "margin_dpo/margin_mean": 0.44856685400009155, "margin_dpo/margin_std": 0.9060893058776855, "step": 195 }, { "epoch": 0.2962962962962963, "fcm_dpo/beta": 1.8433669805526733, "fcm_dpo/delta": 0.033393874764442444, "fcm_dpo/margin": 0.5237653255462646, "fcm_dpo/q_t": 0.34850770235061646, "grad_norm": 410.0362854003906, "learning_rate": 4.448676271745197e-07, "logits/chosen": 0.1736685037612915, "logits/rejected": 0.14138346910476685, "logps/chosen": -60.27099609375, "logps/ref_chosen": -59.227577209472656, "logps/ref_rejected": -83.54757690429688, "logps/rejected": -85.11476135253906, "loss": 1.193, "margin_dpo/margin_mean": 0.5237653255462646, "margin_dpo/margin_std": 0.9456428289413452, "step": 196 }, { "epoch": 0.29780801209372637, "fcm_dpo/beta": 1.8353400230407715, "fcm_dpo/delta": -0.15863925218582153, "fcm_dpo/margin": 0.6141480803489685, "fcm_dpo/q_t": 0.33761459589004517, "grad_norm": 520.2391967773438, "learning_rate": 4.440366160729392e-07, "logits/chosen": 0.23479902744293213, "logits/rejected": 0.1985134482383728, "logps/chosen": -52.59620666503906, "logps/ref_chosen": -51.52912902832031, "logps/ref_rejected": -73.70631408691406, "logps/rejected": -75.38753509521484, "loss": 1.2601, "margin_dpo/margin_mean": 0.6141484975814819, "margin_dpo/margin_std": 1.1263779401779175, "step": 197 }, { "epoch": 0.29931972789115646, "fcm_dpo/beta": 1.6845048666000366, "fcm_dpo/delta": -0.392536461353302, "fcm_dpo/margin": 0.796295702457428, "fcm_dpo/q_t": 0.2878139615058899, "grad_norm": 392.1756591796875, "learning_rate": 4.432001773500957e-07, "logits/chosen": 0.19247442483901978, "logits/rejected": 0.16175703704357147, "logps/chosen": -60.77484893798828, "logps/ref_chosen": -59.78268051147461, "logps/ref_rejected": -72.24533081054688, "logps/rejected": -74.03379821777344, "loss": 0.8946, "margin_dpo/margin_mean": 0.796296238899231, "margin_dpo/margin_std": 1.0002985000610352, "step": 198 }, { "epoch": 0.30083144368858655, "fcm_dpo/beta": 1.6466844081878662, "fcm_dpo/delta": 0.005560420453548431, "fcm_dpo/margin": 0.6032355427742004, "fcm_dpo/q_t": 0.3430374562740326, "grad_norm": 385.1544494628906, "learning_rate": 4.4235833440297856e-07, "logits/chosen": 0.13208839297294617, "logits/rejected": 0.0626702532172203, "logps/chosen": -57.460811614990234, "logps/ref_chosen": -56.38677215576172, "logps/ref_rejected": -74.56779479980469, "logps/rejected": -76.24507141113281, "loss": 1.0833, "margin_dpo/margin_mean": 0.6032348275184631, "margin_dpo/margin_std": 0.970983624458313, "step": 199 }, { "epoch": 0.30234315948601664, "fcm_dpo/beta": 1.5712354183197021, "fcm_dpo/delta": -0.1618885099887848, "fcm_dpo/margin": 0.7240477800369263, "fcm_dpo/q_t": 0.327957421541214, "grad_norm": 412.8072204589844, "learning_rate": 4.415111107797445e-07, "logits/chosen": 0.1844407021999359, "logits/rejected": 0.12967121601104736, "logps/chosen": -58.72947692871094, "logps/ref_chosen": -57.82432556152344, "logps/ref_rejected": -89.28246307373047, "logps/rejected": -90.91166687011719, "loss": 1.0309, "margin_dpo/margin_mean": 0.7240477800369263, "margin_dpo/margin_std": 1.1023731231689453, "step": 200 }, { "epoch": 0.30234315948601664, "eval_fcm_dpo/beta": 1.5800285339355469, "eval_logits/chosen": 0.16845357418060303, "eval_logits/rejected": 0.13373498618602753, "eval_logps/chosen": -75.8633041381836, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -81.05730438232422, "eval_loss": 0.6114334464073181, "eval_margin_dpo/margin_mean": 0.5044752359390259, "eval_margin_dpo/margin_std": 1.028841495513916, "eval_runtime": 38.0264, "eval_samples_per_second": 60.563, "eval_steps_per_second": 1.893, "step": 200 }, { "epoch": 0.30385487528344673, "fcm_dpo/beta": 1.5763221979141235, "fcm_dpo/delta": 0.00995655357837677, "fcm_dpo/margin": 0.6285428404808044, "fcm_dpo/q_t": 0.33469393849372864, "grad_norm": 368.6453552246094, "learning_rate": 4.4065853017905953e-07, "logits/chosen": 0.20987267792224884, "logits/rejected": 0.17573854327201843, "logps/chosen": -60.036251068115234, "logps/ref_chosen": -58.999759674072266, "logps/ref_rejected": -84.67575073242188, "logps/rejected": -86.34077453613281, "loss": 0.9778, "margin_dpo/margin_mean": 0.6285424828529358, "margin_dpo/margin_std": 0.8790519833564758, "step": 201 }, { "epoch": 0.30536659108087677, "fcm_dpo/beta": 1.5742418766021729, "fcm_dpo/delta": -0.20220552384853363, "fcm_dpo/margin": 0.7447031736373901, "fcm_dpo/q_t": 0.306610643863678, "grad_norm": 324.744384765625, "learning_rate": 4.3980061644943575e-07, "logits/chosen": 0.08034056425094604, "logits/rejected": 0.031506434082984924, "logps/chosen": -48.6182861328125, "logps/ref_chosen": -47.660648345947266, "logps/ref_rejected": -73.63249969482422, "logps/rejected": -75.3348388671875, "loss": 0.918, "margin_dpo/margin_mean": 0.7447031736373901, "margin_dpo/margin_std": 0.9296808242797852, "step": 202 }, { "epoch": 0.30687830687830686, "fcm_dpo/beta": 1.5995709896087646, "fcm_dpo/delta": 0.29017671942710876, "fcm_dpo/margin": 0.45493584871292114, "fcm_dpo/q_t": 0.3810551166534424, "grad_norm": 469.91229248046875, "learning_rate": 4.3893739358856455e-07, "logits/chosen": 0.19242677092552185, "logits/rejected": 0.14212460815906525, "logps/chosen": -63.31700134277344, "logps/ref_chosen": -62.32553482055664, "logps/ref_rejected": -99.37226104736328, "logps/rejected": -100.81866455078125, "loss": 1.2402, "margin_dpo/margin_mean": 0.454935759305954, "margin_dpo/margin_std": 0.9751724004745483, "step": 203 }, { "epoch": 0.30839002267573695, "fcm_dpo/beta": 1.5674870014190674, "fcm_dpo/delta": -0.029434487223625183, "fcm_dpo/margin": 0.6477770209312439, "fcm_dpo/q_t": 0.3433707654476166, "grad_norm": 368.2581787109375, "learning_rate": 4.380688857426449e-07, "logits/chosen": 0.0801263153553009, "logits/rejected": 0.030903467908501625, "logps/chosen": -51.68293762207031, "logps/ref_chosen": -50.62931823730469, "logps/ref_rejected": -66.60475158691406, "logps/rejected": -68.30615234375, "loss": 1.1195, "margin_dpo/margin_mean": 0.6477770805358887, "margin_dpo/margin_std": 1.0862679481506348, "step": 204 }, { "epoch": 0.30990173847316704, "fcm_dpo/beta": 1.653544306755066, "fcm_dpo/delta": 0.27683955430984497, "fcm_dpo/margin": 0.450222909450531, "fcm_dpo/q_t": 0.3771224319934845, "grad_norm": 536.0993041992188, "learning_rate": 4.3719511720570814e-07, "logits/chosen": 0.16879329085350037, "logits/rejected": 0.12696264684200287, "logps/chosen": -71.54264831542969, "logps/ref_chosen": -70.3561782836914, "logps/ref_rejected": -93.39848327636719, "logps/rejected": -95.03517150878906, "loss": 1.3927, "margin_dpo/margin_mean": 0.4502222239971161, "margin_dpo/margin_std": 1.1024572849273682, "step": 205 }, { "epoch": 0.31141345427059713, "fcm_dpo/beta": 1.7623982429504395, "fcm_dpo/delta": 0.12155643105506897, "fcm_dpo/margin": 0.49780306220054626, "fcm_dpo/q_t": 0.3561771512031555, "grad_norm": 486.98162841796875, "learning_rate": 4.363161124189387e-07, "logits/chosen": 0.18706491589546204, "logits/rejected": 0.17341138422489166, "logps/chosen": -68.82878112792969, "logps/ref_chosen": -67.64547729492188, "logps/ref_rejected": -79.89584350585938, "logps/rejected": -81.57694244384766, "loss": 1.2778, "margin_dpo/margin_mean": 0.497803270816803, "margin_dpo/margin_std": 0.9891307353973389, "step": 206 }, { "epoch": 0.3129251700680272, "fcm_dpo/beta": 1.7180607318878174, "fcm_dpo/delta": 0.0706307590007782, "fcm_dpo/margin": 0.5409280061721802, "fcm_dpo/q_t": 0.35449251532554626, "grad_norm": 428.377685546875, "learning_rate": 4.3543189596998986e-07, "logits/chosen": 0.10714876651763916, "logits/rejected": 0.05734197795391083, "logps/chosen": -68.83715057373047, "logps/ref_chosen": -67.66419219970703, "logps/ref_rejected": -85.10249328613281, "logps/rejected": -86.81637573242188, "loss": 1.1493, "margin_dpo/margin_mean": 0.5409282445907593, "margin_dpo/margin_std": 1.005875825881958, "step": 207 }, { "epoch": 0.3144368858654573, "fcm_dpo/beta": 1.7887952327728271, "fcm_dpo/delta": 0.09666138887405396, "fcm_dpo/margin": 0.509753942489624, "fcm_dpo/q_t": 0.3555990755558014, "grad_norm": 445.3408508300781, "learning_rate": 4.3454249259229664e-07, "logits/chosen": 0.13969993591308594, "logits/rejected": 0.11949601769447327, "logps/chosen": -58.69970703125, "logps/ref_chosen": -57.731712341308594, "logps/ref_rejected": -74.19276428222656, "logps/rejected": -75.6705093383789, "loss": 1.2387, "margin_dpo/margin_mean": 0.509753406047821, "margin_dpo/margin_std": 0.9776418209075928, "step": 208 }, { "epoch": 0.31594860166288735, "fcm_dpo/beta": 1.7703063488006592, "fcm_dpo/delta": -0.3099019527435303, "fcm_dpo/margin": 0.7123583555221558, "fcm_dpo/q_t": 0.3118298351764679, "grad_norm": 475.9559020996094, "learning_rate": 4.336479271643833e-07, "logits/chosen": 0.08468753099441528, "logits/rejected": 0.04272126033902168, "logps/chosen": -69.55844116210938, "logps/ref_chosen": -68.55007934570312, "logps/ref_rejected": -87.90541076660156, "logps/rejected": -89.62612915039062, "loss": 1.059, "margin_dpo/margin_mean": 0.7123589515686035, "margin_dpo/margin_std": 1.0521396398544312, "step": 209 }, { "epoch": 0.31746031746031744, "fcm_dpo/beta": 1.604590892791748, "fcm_dpo/delta": -0.3447558283805847, "fcm_dpo/margin": 0.8102937340736389, "fcm_dpo/q_t": 0.31004101037979126, "grad_norm": 367.3415832519531, "learning_rate": 4.327482247091679e-07, "logits/chosen": 0.17499999701976776, "logits/rejected": 0.10814614593982697, "logps/chosen": -58.340545654296875, "logps/ref_chosen": -57.268272399902344, "logps/ref_rejected": -85.72807312011719, "logps/rejected": -87.61064147949219, "loss": 0.9243, "margin_dpo/margin_mean": 0.8102930784225464, "margin_dpo/margin_std": 1.078457236289978, "step": 210 }, { "epoch": 0.31897203325774753, "fcm_dpo/beta": 1.5751144886016846, "fcm_dpo/delta": -0.008975658565759659, "fcm_dpo/margin": 0.6399465799331665, "fcm_dpo/q_t": 0.33030977845191956, "grad_norm": 443.0260009765625, "learning_rate": 4.3184341039326217e-07, "logits/chosen": 0.1592363566160202, "logits/rejected": 0.09597043693065643, "logps/chosen": -54.56357192993164, "logps/ref_chosen": -53.640708923339844, "logps/ref_rejected": -93.0387954711914, "logps/rejected": -94.60160827636719, "loss": 0.9671, "margin_dpo/margin_mean": 0.6399465799331665, "margin_dpo/margin_std": 0.9168812036514282, "step": 211 }, { "epoch": 0.3204837490551776, "fcm_dpo/beta": 1.564136028289795, "fcm_dpo/delta": -0.06847534328699112, "fcm_dpo/margin": 0.678282618522644, "fcm_dpo/q_t": 0.3206981420516968, "grad_norm": 370.3670349121094, "learning_rate": 4.309335095262675e-07, "logits/chosen": 0.17273937165737152, "logits/rejected": 0.11869757622480392, "logps/chosen": -58.41630554199219, "logps/ref_chosen": -57.36674499511719, "logps/ref_rejected": -79.89643096923828, "logps/rejected": -81.62427520751953, "loss": 1.027, "margin_dpo/margin_mean": 0.678282618522644, "margin_dpo/margin_std": 1.0015318393707275, "step": 212 }, { "epoch": 0.3219954648526077, "fcm_dpo/beta": 1.527937889099121, "fcm_dpo/delta": -0.042714398354291916, "fcm_dpo/margin": 0.6786133050918579, "fcm_dpo/q_t": 0.3305957615375519, "grad_norm": 391.22064208984375, "learning_rate": 4.3001854756006724e-07, "logits/chosen": 0.15932638943195343, "logits/rejected": 0.138889878988266, "logps/chosen": -66.10162353515625, "logps/ref_chosen": -65.22111511230469, "logps/ref_rejected": -80.1810302734375, "logps/rejected": -81.74015045166016, "loss": 1.0605, "margin_dpo/margin_mean": 0.6786131858825684, "margin_dpo/margin_std": 1.066502332687378, "step": 213 }, { "epoch": 0.3235071806500378, "fcm_dpo/beta": 1.542710781097412, "fcm_dpo/delta": -0.057289645075798035, "fcm_dpo/margin": 0.6799356937408447, "fcm_dpo/q_t": 0.33246564865112305, "grad_norm": 433.1566467285156, "learning_rate": 4.290985500881143e-07, "logits/chosen": 0.06008949503302574, "logits/rejected": 0.0402056947350502, "logps/chosen": -62.26348114013672, "logps/ref_chosen": -61.292327880859375, "logps/ref_rejected": -67.69841003417969, "logps/rejected": -69.3494873046875, "loss": 1.0116, "margin_dpo/margin_mean": 0.679936408996582, "margin_dpo/margin_std": 1.0450165271759033, "step": 214 }, { "epoch": 0.3250188964474679, "fcm_dpo/beta": 1.5210988521575928, "fcm_dpo/delta": -0.0631207600235939, "fcm_dpo/margin": 0.6934947371482849, "fcm_dpo/q_t": 0.33989161252975464, "grad_norm": 409.9744567871094, "learning_rate": 4.281735428447157e-07, "logits/chosen": 0.08377814292907715, "logits/rejected": 0.010123915039002895, "logps/chosen": -64.82623291015625, "logps/ref_chosen": -63.869136810302734, "logps/ref_rejected": -98.7657241821289, "logps/rejected": -100.41632080078125, "loss": 1.0766, "margin_dpo/margin_mean": 0.6934951543807983, "margin_dpo/margin_std": 1.1064403057098389, "step": 215 }, { "epoch": 0.32653061224489793, "fcm_dpo/beta": 1.4754486083984375, "fcm_dpo/delta": -0.2396281659603119, "fcm_dpo/margin": 0.8188655972480774, "fcm_dpo/q_t": 0.34051093459129333, "grad_norm": 398.4416198730469, "learning_rate": 4.2724355170431247e-07, "logits/chosen": 0.16970112919807434, "logits/rejected": 0.10684916377067566, "logps/chosen": -68.89656829833984, "logps/ref_chosen": -67.824951171875, "logps/ref_rejected": -96.40231323242188, "logps/rejected": -98.29280090332031, "loss": 1.1031, "margin_dpo/margin_mean": 0.8188657760620117, "margin_dpo/margin_std": 1.4882557392120361, "step": 216 }, { "epoch": 0.328042328042328, "fcm_dpo/beta": 1.344929575920105, "fcm_dpo/delta": -0.35066401958465576, "fcm_dpo/margin": 0.9668929576873779, "fcm_dpo/q_t": 0.28799083828926086, "grad_norm": 300.48565673828125, "learning_rate": 4.26308602680756e-07, "logits/chosen": 0.11938208341598511, "logits/rejected": 0.03593681752681732, "logps/chosen": -61.565162658691406, "logps/ref_chosen": -60.5049934387207, "logps/ref_rejected": -84.26618194580078, "logps/rejected": -86.29324340820312, "loss": 0.8155, "margin_dpo/margin_mean": 0.966893196105957, "margin_dpo/margin_std": 1.1881346702575684, "step": 217 }, { "epoch": 0.3295540438397581, "fcm_dpo/beta": 1.4142457246780396, "fcm_dpo/delta": 0.2700718343257904, "fcm_dpo/margin": 0.5216431617736816, "fcm_dpo/q_t": 0.3781411647796631, "grad_norm": 394.3163757324219, "learning_rate": 4.253687219265803e-07, "logits/chosen": 0.024805322289466858, "logits/rejected": 0.01834661327302456, "logps/chosen": -71.73900604248047, "logps/ref_chosen": -70.59431457519531, "logps/ref_rejected": -73.89038848876953, "logps/rejected": -75.55671691894531, "loss": 1.2642, "margin_dpo/margin_mean": 0.5216437578201294, "margin_dpo/margin_std": 1.1325141191482544, "step": 218 }, { "epoch": 0.3310657596371882, "fcm_dpo/beta": 1.4622104167938232, "fcm_dpo/delta": 0.27110588550567627, "fcm_dpo/margin": 0.5126116275787354, "fcm_dpo/q_t": 0.36937472224235535, "grad_norm": 361.056884765625, "learning_rate": 4.2442393573227043e-07, "logits/chosen": 0.08768868446350098, "logits/rejected": 0.05643084645271301, "logps/chosen": -61.44779586791992, "logps/ref_chosen": -60.490943908691406, "logps/ref_rejected": -75.85001373291016, "logps/rejected": -77.31947326660156, "loss": 1.0874, "margin_dpo/margin_mean": 0.512610912322998, "margin_dpo/margin_std": 0.8701074123382568, "step": 219 }, { "epoch": 0.3325774754346183, "fcm_dpo/beta": 1.4802911281585693, "fcm_dpo/delta": 0.026991277933120728, "fcm_dpo/margin": 0.6585407257080078, "fcm_dpo/q_t": 0.33476772904396057, "grad_norm": 283.6431579589844, "learning_rate": 4.234742705255272e-07, "logits/chosen": 0.17591653764247894, "logits/rejected": 0.13129651546478271, "logps/chosen": -46.09846496582031, "logps/ref_chosen": -45.013397216796875, "logps/ref_rejected": -70.49369812011719, "logps/rejected": -72.23731231689453, "loss": 0.9934, "margin_dpo/margin_mean": 0.6585406064987183, "margin_dpo/margin_std": 0.9634271264076233, "step": 220 }, { "epoch": 0.3340891912320484, "fcm_dpo/beta": 1.486309289932251, "fcm_dpo/delta": -0.12030621618032455, "fcm_dpo/margin": 0.7435950636863708, "fcm_dpo/q_t": 0.32344067096710205, "grad_norm": 338.2693176269531, "learning_rate": 4.22519752870528e-07, "logits/chosen": 0.16085302829742432, "logits/rejected": 0.11432051658630371, "logps/chosen": -60.045074462890625, "logps/ref_chosen": -59.09584045410156, "logps/ref_rejected": -88.64388275146484, "logps/rejected": -90.33671569824219, "loss": 0.9562, "margin_dpo/margin_mean": 0.7435950636863708, "margin_dpo/margin_std": 1.0672008991241455, "step": 221 }, { "epoch": 0.3356009070294785, "fcm_dpo/beta": 1.4272222518920898, "fcm_dpo/delta": -0.1927495002746582, "fcm_dpo/margin": 0.8200865983963013, "fcm_dpo/q_t": 0.2971838712692261, "grad_norm": 337.138671875, "learning_rate": 4.2156040946718343e-07, "logits/chosen": 0.16604246199131012, "logits/rejected": 0.10780028998851776, "logps/chosen": -57.0385627746582, "logps/ref_chosen": -55.9976921081543, "logps/ref_rejected": -111.94727325439453, "logps/rejected": -113.8082275390625, "loss": 0.856, "margin_dpo/margin_mean": 0.8200874328613281, "margin_dpo/margin_std": 0.9770439863204956, "step": 222 }, { "epoch": 0.3371126228269085, "fcm_dpo/beta": 1.3298912048339844, "fcm_dpo/delta": -0.2783888578414917, "fcm_dpo/margin": 0.9309906363487244, "fcm_dpo/q_t": 0.2817830443382263, "grad_norm": 263.7952880859375, "learning_rate": 4.2059626715039065e-07, "logits/chosen": 0.18012914061546326, "logits/rejected": 0.13743728399276733, "logps/chosen": -60.831424713134766, "logps/ref_chosen": -59.891422271728516, "logps/ref_rejected": -86.28954315185547, "logps/rejected": -88.16053771972656, "loss": 0.7994, "margin_dpo/margin_mean": 0.9309903383255005, "margin_dpo/margin_std": 0.9911828637123108, "step": 223 }, { "epoch": 0.3386243386243386, "fcm_dpo/beta": 1.3959991931915283, "fcm_dpo/delta": 0.3523348867893219, "fcm_dpo/margin": 0.48118603229522705, "fcm_dpo/q_t": 0.3791411519050598, "grad_norm": 364.8034362792969, "learning_rate": 4.1962735288928304e-07, "logits/chosen": 0.19306407868862152, "logits/rejected": 0.17875471711158752, "logps/chosen": -65.12818908691406, "logps/ref_chosen": -64.04463195800781, "logps/ref_rejected": -75.05450439453125, "logps/rejected": -76.61923217773438, "loss": 1.0923, "margin_dpo/margin_mean": 0.4811859726905823, "margin_dpo/margin_std": 0.8308462500572205, "step": 224 }, { "epoch": 0.3401360544217687, "fcm_dpo/beta": 1.395388126373291, "fcm_dpo/delta": -0.05797035992145538, "fcm_dpo/margin": 0.7516753673553467, "fcm_dpo/q_t": 0.3332204222679138, "grad_norm": 468.0265197753906, "learning_rate": 4.186536937864752e-07, "logits/chosen": 0.1659896969795227, "logits/rejected": 0.07978636771440506, "logps/chosen": -67.14776611328125, "logps/ref_chosen": -66.0958251953125, "logps/ref_rejected": -97.68675231933594, "logps/rejected": -99.49036407470703, "loss": 1.0067, "margin_dpo/margin_mean": 0.7516759634017944, "margin_dpo/margin_std": 1.1150200366973877, "step": 225 }, { "epoch": 0.3416477702191988, "fcm_dpo/beta": 1.4121769666671753, "fcm_dpo/delta": 0.040757764130830765, "fcm_dpo/margin": 0.6820341348648071, "fcm_dpo/q_t": 0.3408370018005371, "grad_norm": 292.9568176269531, "learning_rate": 4.176753170773052e-07, "logits/chosen": 0.18655280768871307, "logits/rejected": 0.15081270039081573, "logps/chosen": -52.478660583496094, "logps/ref_chosen": -51.4168701171875, "logps/ref_rejected": -66.30068969726562, "logps/rejected": -68.04450988769531, "loss": 1.0394, "margin_dpo/margin_mean": 0.6820334792137146, "margin_dpo/margin_std": 1.0649542808532715, "step": 226 }, { "epoch": 0.3431594860166289, "fcm_dpo/beta": 1.4467318058013916, "fcm_dpo/delta": 0.14849498867988586, "fcm_dpo/margin": 0.5975882411003113, "fcm_dpo/q_t": 0.3471581041812897, "grad_norm": 412.5190734863281, "learning_rate": 4.166922501290729e-07, "logits/chosen": 0.20183053612709045, "logits/rejected": 0.17074134945869446, "logps/chosen": -59.01850891113281, "logps/ref_chosen": -57.989776611328125, "logps/ref_rejected": -75.05464172363281, "logps/rejected": -76.68096160888672, "loss": 1.1633, "margin_dpo/margin_mean": 0.5975878238677979, "margin_dpo/margin_std": 1.1138098239898682, "step": 227 }, { "epoch": 0.34467120181405897, "fcm_dpo/beta": 1.4542537927627563, "fcm_dpo/delta": -0.03622462600469589, "fcm_dpo/margin": 0.7099840641021729, "fcm_dpo/q_t": 0.3378611207008362, "grad_norm": 351.0310974121094, "learning_rate": 4.1570452044027405e-07, "logits/chosen": 0.17106792330741882, "logits/rejected": 0.11330675333738327, "logps/chosen": -56.669921875, "logps/ref_chosen": -55.55936813354492, "logps/ref_rejected": -77.02364349365234, "logps/rejected": -78.84417724609375, "loss": 1.1439, "margin_dpo/margin_mean": 0.709984540939331, "margin_dpo/margin_std": 1.241539716720581, "step": 228 }, { "epoch": 0.34618291761148906, "fcm_dpo/beta": 1.4833769798278809, "fcm_dpo/delta": 0.18256068229675293, "fcm_dpo/margin": 0.5615620613098145, "fcm_dpo/q_t": 0.3450517952442169, "grad_norm": 629.395263671875, "learning_rate": 4.147121556398312e-07, "logits/chosen": 0.23741164803504944, "logits/rejected": 0.19696751236915588, "logps/chosen": -51.810546875, "logps/ref_chosen": -50.79466247558594, "logps/ref_rejected": -78.4474105834961, "logps/rejected": -80.02485656738281, "loss": 1.0995, "margin_dpo/margin_mean": 0.5615620613098145, "margin_dpo/margin_std": 0.9383633732795715, "step": 229 }, { "epoch": 0.3476946334089191, "fcm_dpo/beta": 1.5409061908721924, "fcm_dpo/delta": 0.06196488440036774, "fcm_dpo/margin": 0.6093405485153198, "fcm_dpo/q_t": 0.35204917192459106, "grad_norm": 399.1847839355469, "learning_rate": 4.137151834863213e-07, "logits/chosen": 0.1482779085636139, "logits/rejected": 0.14588508009910583, "logps/chosen": -57.8225212097168, "logps/ref_chosen": -56.729225158691406, "logps/ref_rejected": -62.99180603027344, "logps/rejected": -64.69444274902344, "loss": 1.1512, "margin_dpo/margin_mean": 0.6093416213989258, "margin_dpo/margin_std": 1.131590485572815, "step": 230 }, { "epoch": 0.3492063492063492, "fcm_dpo/beta": 1.4233132600784302, "fcm_dpo/delta": -0.39863085746765137, "fcm_dpo/margin": 0.9415854215621948, "fcm_dpo/q_t": 0.27584952116012573, "grad_norm": 315.82147216796875, "learning_rate": 4.1271363186719835e-07, "logits/chosen": 0.11343254894018173, "logits/rejected": 0.09590326249599457, "logps/chosen": -73.67927551269531, "logps/ref_chosen": -72.59709930419922, "logps/ref_rejected": -86.2322998046875, "logps/rejected": -88.25605773925781, "loss": 0.7891, "margin_dpo/margin_mean": 0.9415853023529053, "margin_dpo/margin_std": 1.024916172027588, "step": 231 }, { "epoch": 0.3507180650037793, "fcm_dpo/beta": 1.3663169145584106, "fcm_dpo/delta": -0.1273672878742218, "fcm_dpo/margin": 0.8124350309371948, "fcm_dpo/q_t": 0.33132317662239075, "grad_norm": 345.0347595214844, "learning_rate": 4.1170752879801436e-07, "logits/chosen": 0.14546144008636475, "logits/rejected": 0.12237675487995148, "logps/chosen": -69.10570526123047, "logps/ref_chosen": -68.1185302734375, "logps/ref_rejected": -83.79415893554688, "logps/rejected": -85.5937728881836, "loss": 1.0339, "margin_dpo/margin_mean": 0.812435507774353, "margin_dpo/margin_std": 1.2407793998718262, "step": 232 }, { "epoch": 0.35222978080120937, "fcm_dpo/beta": 1.4025087356567383, "fcm_dpo/delta": 0.2927402853965759, "fcm_dpo/margin": 0.5172353386878967, "fcm_dpo/q_t": 0.3894173204898834, "grad_norm": 380.82867431640625, "learning_rate": 4.106969024216348e-07, "logits/chosen": 0.10903730243444443, "logits/rejected": 0.06897353380918503, "logps/chosen": -56.4796142578125, "logps/ref_chosen": -55.070152282714844, "logps/ref_rejected": -66.61845397949219, "logps/rejected": -68.5451431274414, "loss": 1.212, "margin_dpo/margin_mean": 0.5172350406646729, "margin_dpo/margin_std": 1.0877723693847656, "step": 233 }, { "epoch": 0.35374149659863946, "fcm_dpo/beta": 1.5282173156738281, "fcm_dpo/delta": 0.27392610907554626, "fcm_dpo/margin": 0.48579320311546326, "fcm_dpo/q_t": 0.3870254158973694, "grad_norm": 388.5447082519531, "learning_rate": 4.09681781007452e-07, "logits/chosen": 0.09753985702991486, "logits/rejected": 0.08481541275978088, "logps/chosen": -57.07015609741211, "logps/ref_chosen": -55.92589569091797, "logps/ref_rejected": -51.11608123779297, "logps/rejected": -52.746131896972656, "loss": 1.2257, "margin_dpo/margin_mean": 0.48579323291778564, "margin_dpo/margin_std": 1.005543828010559, "step": 234 }, { "epoch": 0.35525321239606955, "fcm_dpo/beta": 1.4440486431121826, "fcm_dpo/delta": -0.3470792770385742, "fcm_dpo/margin": 0.8979411721229553, "fcm_dpo/q_t": 0.2706488370895386, "grad_norm": 331.3047180175781, "learning_rate": 4.08662192950594e-07, "logits/chosen": 0.18871155381202698, "logits/rejected": 0.17248067259788513, "logps/chosen": -65.49934387207031, "logps/ref_chosen": -64.53972625732422, "logps/ref_rejected": -77.69151306152344, "logps/rejected": -79.549072265625, "loss": 0.7827, "margin_dpo/margin_mean": 0.897940993309021, "margin_dpo/margin_std": 0.9550020694732666, "step": 235 }, { "epoch": 0.35676492819349964, "fcm_dpo/beta": 1.4077339172363281, "fcm_dpo/delta": -0.08833800256252289, "fcm_dpo/margin": 0.7657995223999023, "fcm_dpo/q_t": 0.34863966703414917, "grad_norm": 387.498291015625, "learning_rate": 4.076381667711306e-07, "logits/chosen": 0.10846032202243805, "logits/rejected": 0.09852010011672974, "logps/chosen": -72.46954345703125, "logps/ref_chosen": -71.15473937988281, "logps/ref_rejected": -84.88541412353516, "logps/rejected": -86.96601867675781, "loss": 1.1004, "margin_dpo/margin_mean": 0.7657992839813232, "margin_dpo/margin_std": 1.3017526865005493, "step": 236 }, { "epoch": 0.35827664399092973, "fcm_dpo/beta": 1.3855292797088623, "fcm_dpo/delta": -0.0012986212968826294, "fcm_dpo/margin": 0.7191513776779175, "fcm_dpo/q_t": 0.3367508053779602, "grad_norm": 383.1600036621094, "learning_rate": 4.066097311132753e-07, "logits/chosen": 0.20629596710205078, "logits/rejected": 0.19376662373542786, "logps/chosen": -77.21894836425781, "logps/ref_chosen": -76.14201354980469, "logps/ref_rejected": -80.88479614257812, "logps/rejected": -82.6808853149414, "loss": 1.1096, "margin_dpo/margin_mean": 0.7191513180732727, "margin_dpo/margin_std": 1.1922106742858887, "step": 237 }, { "epoch": 0.35978835978835977, "fcm_dpo/beta": 1.488319993019104, "fcm_dpo/delta": 0.24613891541957855, "fcm_dpo/margin": 0.5112044811248779, "fcm_dpo/q_t": 0.3709501624107361, "grad_norm": 2477.388916015625, "learning_rate": 4.0557691474458414e-07, "logits/chosen": 0.13462495803833008, "logits/rejected": 0.11732495576143265, "logps/chosen": -70.04851531982422, "logps/ref_chosen": -68.88484954833984, "logps/ref_rejected": -75.8946304321289, "logps/rejected": -77.56949615478516, "loss": 1.4557, "margin_dpo/margin_mean": 0.5112046003341675, "margin_dpo/margin_std": 1.32561194896698, "step": 238 }, { "epoch": 0.36130007558578986, "fcm_dpo/beta": 1.4520866870880127, "fcm_dpo/delta": -0.12432458996772766, "fcm_dpo/margin": 0.7647981643676758, "fcm_dpo/q_t": 0.31985604763031006, "grad_norm": 413.1396789550781, "learning_rate": 4.045397465551513e-07, "logits/chosen": 0.19768695533275604, "logits/rejected": 0.11818195134401321, "logps/chosen": -58.036109924316406, "logps/ref_chosen": -56.771827697753906, "logps/ref_rejected": -116.23050689697266, "logps/rejected": -118.25959014892578, "loss": 1.0398, "margin_dpo/margin_mean": 0.764798641204834, "margin_dpo/margin_std": 1.1497983932495117, "step": 239 }, { "epoch": 0.36281179138321995, "fcm_dpo/beta": 1.4255015850067139, "fcm_dpo/delta": -0.19121024012565613, "fcm_dpo/margin": 0.8184474110603333, "fcm_dpo/q_t": 0.30751824378967285, "grad_norm": 311.9652404785156, "learning_rate": 4.0349825555680045e-07, "logits/chosen": 0.14041496813297272, "logits/rejected": 0.07365534454584122, "logps/chosen": -54.64623260498047, "logps/ref_chosen": -53.35411071777344, "logps/ref_rejected": -80.12019348144531, "logps/rejected": -82.23076629638672, "loss": 0.9905, "margin_dpo/margin_mean": 0.8184475898742676, "margin_dpo/margin_std": 1.1477313041687012, "step": 240 }, { "epoch": 0.36432350718065004, "fcm_dpo/beta": 1.5252119302749634, "fcm_dpo/delta": 0.4934987425804138, "fcm_dpo/margin": 0.336672842502594, "fcm_dpo/q_t": 0.4018644094467163, "grad_norm": 527.2469482421875, "learning_rate": 4.0245247088227377e-07, "logits/chosen": 0.12555167078971863, "logits/rejected": 0.09838816523551941, "logps/chosen": -73.14974975585938, "logps/ref_chosen": -71.89541625976562, "logps/ref_rejected": -83.03492736816406, "logps/rejected": -84.62593078613281, "loss": 1.4036, "margin_dpo/margin_mean": 0.33667343854904175, "margin_dpo/margin_std": 0.9868639707565308, "step": 241 }, { "epoch": 0.36583522297808013, "fcm_dpo/beta": 1.443713903427124, "fcm_dpo/delta": -0.2312009036540985, "fcm_dpo/margin": 0.8218022584915161, "fcm_dpo/q_t": 0.31463247537612915, "grad_norm": 314.4766845703125, "learning_rate": 4.0140242178441665e-07, "logits/chosen": 0.1157296746969223, "logits/rejected": 0.09765278548002243, "logps/chosen": -58.899356842041016, "logps/ref_chosen": -57.927433013916016, "logps/ref_rejected": -67.838623046875, "logps/rejected": -69.6323471069336, "loss": 0.9264, "margin_dpo/margin_mean": 0.8218023180961609, "margin_dpo/margin_std": 1.0703468322753906, "step": 242 }, { "epoch": 0.3673469387755102, "fcm_dpo/beta": 1.4689741134643555, "fcm_dpo/delta": 0.049783095717430115, "fcm_dpo/margin": 0.6500656604766846, "fcm_dpo/q_t": 0.3428027033805847, "grad_norm": 388.419921875, "learning_rate": 4.003481376353596e-07, "logits/chosen": 0.10413776338100433, "logits/rejected": 0.09902875125408173, "logps/chosen": -75.4229965209961, "logps/ref_chosen": -74.27667236328125, "logps/ref_rejected": -73.24340057373047, "logps/rejected": -75.03977966308594, "loss": 1.1186, "margin_dpo/margin_mean": 0.650065541267395, "margin_dpo/margin_std": 1.1026396751403809, "step": 243 }, { "epoch": 0.3688586545729403, "fcm_dpo/beta": 1.4141755104064941, "fcm_dpo/delta": -0.2628709375858307, "fcm_dpo/margin": 0.8689752817153931, "fcm_dpo/q_t": 0.28472983837127686, "grad_norm": 275.4247131347656, "learning_rate": 3.9928964792569654e-07, "logits/chosen": 0.15633343160152435, "logits/rejected": 0.09156134724617004, "logps/chosen": -54.57572555541992, "logps/ref_chosen": -53.36390686035156, "logps/ref_rejected": -71.10276794433594, "logps/rejected": -73.18356323242188, "loss": 0.7643, "margin_dpo/margin_mean": 0.8689748048782349, "margin_dpo/margin_std": 0.9518204927444458, "step": 244 }, { "epoch": 0.37037037037037035, "fcm_dpo/beta": 1.3580291271209717, "fcm_dpo/delta": -0.23086267709732056, "fcm_dpo/margin": 0.8861613273620605, "fcm_dpo/q_t": 0.3000994920730591, "grad_norm": 464.6539611816406, "learning_rate": 3.982269822636601e-07, "logits/chosen": 0.16047267615795135, "logits/rejected": 0.14088091254234314, "logps/chosen": -72.37004089355469, "logps/ref_chosen": -71.19510650634766, "logps/ref_rejected": -80.76235961914062, "logps/rejected": -82.82345581054688, "loss": 0.9896, "margin_dpo/margin_mean": 0.8861616849899292, "margin_dpo/margin_std": 1.2600171566009521, "step": 245 }, { "epoch": 0.37188208616780044, "fcm_dpo/beta": 1.2877942323684692, "fcm_dpo/delta": -0.19463737308979034, "fcm_dpo/margin": 0.9072257280349731, "fcm_dpo/q_t": 0.3076491057872772, "grad_norm": 311.3534240722656, "learning_rate": 3.971601703742932e-07, "logits/chosen": 0.15642526745796204, "logits/rejected": 0.11311867088079453, "logps/chosen": -72.98123168945312, "logps/ref_chosen": -71.62104797363281, "logps/ref_rejected": -94.03392028808594, "logps/rejected": -96.30133056640625, "loss": 0.8849, "margin_dpo/margin_mean": 0.9072257876396179, "margin_dpo/margin_std": 1.1439390182495117, "step": 246 }, { "epoch": 0.37339380196523053, "fcm_dpo/beta": 1.385887622833252, "fcm_dpo/delta": 0.3913596272468567, "fcm_dpo/margin": 0.44914084672927856, "fcm_dpo/q_t": 0.3937835395336151, "grad_norm": 441.7315673828125, "learning_rate": 3.960892420986177e-07, "logits/chosen": 0.14804767072200775, "logits/rejected": 0.13864608108997345, "logps/chosen": -81.47723388671875, "logps/ref_chosen": -80.02254486083984, "logps/ref_rejected": -89.22705841064453, "logps/rejected": -91.13088989257812, "loss": 1.3732, "margin_dpo/margin_mean": 0.44914010167121887, "margin_dpo/margin_std": 1.1773467063903809, "step": 247 }, { "epoch": 0.3749055177626606, "fcm_dpo/beta": 1.4095741510391235, "fcm_dpo/delta": 0.022829867899417877, "fcm_dpo/margin": 0.6935802698135376, "fcm_dpo/q_t": 0.3401643633842468, "grad_norm": 382.5201110839844, "learning_rate": 3.9501422739279953e-07, "logits/chosen": 0.1225530132651329, "logits/rejected": 0.14110150933265686, "logps/chosen": -66.71062469482422, "logps/ref_chosen": -65.37796020507812, "logps/ref_rejected": -61.365787506103516, "logps/rejected": -63.392032623291016, "loss": 1.0538, "margin_dpo/margin_mean": 0.6935799717903137, "margin_dpo/margin_std": 1.1022088527679443, "step": 248 }, { "epoch": 0.3764172335600907, "fcm_dpo/beta": 1.5373433828353882, "fcm_dpo/delta": 0.5078557133674622, "fcm_dpo/margin": 0.3260525166988373, "fcm_dpo/q_t": 0.40423983335494995, "grad_norm": 464.44989013671875, "learning_rate": 3.9393515632731094e-07, "logits/chosen": 0.10547161102294922, "logits/rejected": 0.12796463072299957, "logps/chosen": -76.10067749023438, "logps/ref_chosen": -74.60145568847656, "logps/ref_rejected": -63.79338455200195, "logps/rejected": -65.61865234375, "loss": 1.4834, "margin_dpo/margin_mean": 0.3260522484779358, "margin_dpo/margin_std": 1.0597259998321533, "step": 249 }, { "epoch": 0.3779289493575208, "fcm_dpo/beta": 1.4959361553192139, "fcm_dpo/delta": -0.35001400113105774, "fcm_dpo/margin": 0.871913731098175, "fcm_dpo/q_t": 0.305349200963974, "grad_norm": 331.92315673828125, "learning_rate": 3.9285205908608934e-07, "logits/chosen": 0.21487398445606232, "logits/rejected": 0.18293528258800507, "logps/chosen": -63.18694305419922, "logps/ref_chosen": -61.938209533691406, "logps/ref_rejected": -72.21602630615234, "logps/rejected": -74.336669921875, "loss": 0.9596, "margin_dpo/margin_mean": 0.8719134330749512, "margin_dpo/margin_std": 1.247259259223938, "step": 250 }, { "epoch": 0.3794406651549509, "fcm_dpo/beta": 1.4492324590682983, "fcm_dpo/delta": 0.03792502358555794, "fcm_dpo/margin": 0.6663841009140015, "fcm_dpo/q_t": 0.3597896099090576, "grad_norm": 376.5051574707031, "learning_rate": 3.9176496596569265e-07, "logits/chosen": 0.2115197777748108, "logits/rejected": 0.1815691590309143, "logps/chosen": -68.17942810058594, "logps/ref_chosen": -66.85694885253906, "logps/ref_rejected": -84.83396911621094, "logps/rejected": -86.82284545898438, "loss": 1.1816, "margin_dpo/margin_mean": 0.6663837432861328, "margin_dpo/margin_std": 1.2780930995941162, "step": 251 }, { "epoch": 0.38095238095238093, "fcm_dpo/beta": 1.496967077255249, "fcm_dpo/delta": -0.03972265124320984, "fcm_dpo/margin": 0.6779400110244751, "fcm_dpo/q_t": 0.32841235399246216, "grad_norm": 365.8186340332031, "learning_rate": 3.9067390737445254e-07, "logits/chosen": 0.12412711977958679, "logits/rejected": 0.0818972960114479, "logps/chosen": -57.51960372924805, "logps/ref_chosen": -56.22393035888672, "logps/ref_rejected": -77.1136245727539, "logps/rejected": -79.08723449707031, "loss": 1.2777, "margin_dpo/margin_mean": 0.6779407262802124, "margin_dpo/margin_std": 1.3109935522079468, "step": 252 }, { "epoch": 0.382464096749811, "fcm_dpo/beta": 1.3989202976226807, "fcm_dpo/delta": -0.0833391547203064, "fcm_dpo/margin": 0.7623114585876465, "fcm_dpo/q_t": 0.3369859755039215, "grad_norm": 290.3008117675781, "learning_rate": 3.8957891383162304e-07, "logits/chosen": 0.17171865701675415, "logits/rejected": 0.14320127665996552, "logps/chosen": -53.48502731323242, "logps/ref_chosen": -52.21001434326172, "logps/ref_rejected": -58.75764846801758, "logps/rejected": -60.79496765136719, "loss": 1.0203, "margin_dpo/margin_mean": 0.762311577796936, "margin_dpo/margin_std": 1.1594964265823364, "step": 253 }, { "epoch": 0.3839758125472411, "fcm_dpo/beta": 1.3931760787963867, "fcm_dpo/delta": -0.08251707255840302, "fcm_dpo/margin": 0.7698103189468384, "fcm_dpo/q_t": 0.33657705783843994, "grad_norm": 385.88311767578125, "learning_rate": 3.884800159665276e-07, "logits/chosen": 0.13405011594295502, "logits/rejected": 0.09684339165687561, "logps/chosen": -66.99877166748047, "logps/ref_chosen": -65.63632202148438, "logps/ref_rejected": -82.34425354003906, "logps/rejected": -84.47650146484375, "loss": 1.1088, "margin_dpo/margin_mean": 0.7698097229003906, "margin_dpo/margin_std": 1.271782398223877, "step": 254 }, { "epoch": 0.3854875283446712, "fcm_dpo/beta": 1.3778660297393799, "fcm_dpo/delta": -0.019536815583705902, "fcm_dpo/margin": 0.7369774580001831, "fcm_dpo/q_t": 0.33573752641677856, "grad_norm": 347.2344055175781, "learning_rate": 3.873772445177015e-07, "logits/chosen": 0.1548142433166504, "logits/rejected": 0.12779046595096588, "logps/chosen": -69.03644561767578, "logps/ref_chosen": -67.91108703613281, "logps/ref_rejected": -83.89114379882812, "logps/rejected": -85.75347900390625, "loss": 1.068, "margin_dpo/margin_mean": 0.7369774580001831, "margin_dpo/margin_std": 1.1885807514190674, "step": 255 }, { "epoch": 0.3869992441421013, "fcm_dpo/beta": 1.396597146987915, "fcm_dpo/delta": -0.011127792298793793, "fcm_dpo/margin": 0.7227224111557007, "fcm_dpo/q_t": 0.3369947075843811, "grad_norm": 397.8918762207031, "learning_rate": 3.862706303320329e-07, "logits/chosen": 0.1215638667345047, "logits/rejected": 0.08205322921276093, "logps/chosen": -64.94349670410156, "logps/ref_chosen": -63.49998474121094, "logps/ref_rejected": -90.77104187011719, "logps/rejected": -92.93727111816406, "loss": 1.0738, "margin_dpo/margin_mean": 0.7227222323417664, "margin_dpo/margin_std": 1.203334927558899, "step": 256 }, { "epoch": 0.3885109599395314, "fcm_dpo/beta": 1.2768826484680176, "fcm_dpo/delta": -0.461418092250824, "fcm_dpo/margin": 1.086154580116272, "fcm_dpo/q_t": 0.28833481669425964, "grad_norm": 322.4739685058594, "learning_rate": 3.851602043638994e-07, "logits/chosen": 0.1613893210887909, "logits/rejected": 0.11499130725860596, "logps/chosen": -71.91011810302734, "logps/ref_chosen": -70.60064697265625, "logps/ref_rejected": -108.58313751220703, "logps/rejected": -110.978759765625, "loss": 0.9387, "margin_dpo/margin_mean": 1.0861549377441406, "margin_dpo/margin_std": 1.5388684272766113, "step": 257 }, { "epoch": 0.3900226757369615, "fcm_dpo/beta": 1.2841336727142334, "fcm_dpo/delta": 0.00753195583820343, "fcm_dpo/margin": 0.7705293297767639, "fcm_dpo/q_t": 0.32425814867019653, "grad_norm": 300.5933837890625, "learning_rate": 3.840459976743023e-07, "logits/chosen": 0.1647588312625885, "logits/rejected": 0.1312163770198822, "logps/chosen": -60.74152374267578, "logps/ref_chosen": -59.25416564941406, "logps/ref_rejected": -85.58709716796875, "logps/rejected": -87.84498596191406, "loss": 0.933, "margin_dpo/margin_mean": 0.7705295085906982, "margin_dpo/margin_std": 1.0195441246032715, "step": 258 }, { "epoch": 0.3915343915343915, "fcm_dpo/beta": 1.1445305347442627, "fcm_dpo/delta": -0.5759162902832031, "fcm_dpo/margin": 1.2901654243469238, "fcm_dpo/q_t": 0.2562459111213684, "grad_norm": 223.67605590820312, "learning_rate": 3.8292804142999796e-07, "logits/chosen": 0.09316843003034592, "logits/rejected": 0.021822050213813782, "logps/chosen": -66.56085205078125, "logps/ref_chosen": -65.43487548828125, "logps/ref_rejected": -95.41731262207031, "logps/rejected": -97.83345031738281, "loss": 0.7207, "margin_dpo/margin_mean": 1.2901657819747925, "margin_dpo/margin_std": 1.2492828369140625, "step": 259 }, { "epoch": 0.3930461073318216, "fcm_dpo/beta": 1.1257474422454834, "fcm_dpo/delta": 0.017584767192602158, "fcm_dpo/margin": 0.8737805485725403, "fcm_dpo/q_t": 0.33795487880706787, "grad_norm": 259.8592834472656, "learning_rate": 3.818063669026256e-07, "logits/chosen": 0.127943217754364, "logits/rejected": 0.06927517056465149, "logps/chosen": -50.53019714355469, "logps/ref_chosen": -49.08958435058594, "logps/ref_rejected": -79.01708221435547, "logps/rejected": -81.33147430419922, "loss": 0.9979, "margin_dpo/margin_mean": 0.873779296875, "margin_dpo/margin_std": 1.2833609580993652, "step": 260 }, { "epoch": 0.3945578231292517, "fcm_dpo/beta": 1.1455121040344238, "fcm_dpo/delta": 0.08530843257904053, "fcm_dpo/margin": 0.8054367303848267, "fcm_dpo/q_t": 0.34108513593673706, "grad_norm": 306.086181640625, "learning_rate": 3.806810054678331e-07, "logits/chosen": 0.03654761239886284, "logits/rejected": 0.04507092386484146, "logps/chosen": -72.11337280273438, "logps/ref_chosen": -70.87239074707031, "logps/ref_rejected": -65.01522064208984, "logps/rejected": -67.06163787841797, "loss": 0.9947, "margin_dpo/margin_mean": 0.805436909198761, "margin_dpo/margin_std": 1.1771018505096436, "step": 261 }, { "epoch": 0.3960695389266818, "fcm_dpo/beta": 1.1686980724334717, "fcm_dpo/delta": 0.04035666957497597, "fcm_dpo/margin": 0.8231496810913086, "fcm_dpo/q_t": 0.3458126187324524, "grad_norm": 323.2743225097656, "learning_rate": 3.7955198860439887e-07, "logits/chosen": 0.17141227424144745, "logits/rejected": 0.13052189350128174, "logps/chosen": -69.32987976074219, "logps/ref_chosen": -67.8706283569336, "logps/ref_rejected": -88.7205810546875, "logps/rejected": -91.00297546386719, "loss": 0.9832, "margin_dpo/margin_mean": 0.8231501579284668, "margin_dpo/margin_std": 1.2316184043884277, "step": 262 }, { "epoch": 0.3975812547241119, "fcm_dpo/beta": 1.23178231716156, "fcm_dpo/delta": 0.2991076111793518, "fcm_dpo/margin": 0.5800197124481201, "fcm_dpo/q_t": 0.38507431745529175, "grad_norm": 314.1044616699219, "learning_rate": 3.784193478933516e-07, "logits/chosen": 0.13119199872016907, "logits/rejected": 0.05917968600988388, "logps/chosen": -56.671165466308594, "logps/ref_chosen": -55.194583892822266, "logps/ref_rejected": -80.54048156738281, "logps/rejected": -82.59709167480469, "loss": 1.231, "margin_dpo/margin_mean": 0.5800192356109619, "margin_dpo/margin_std": 1.2319090366363525, "step": 263 }, { "epoch": 0.39909297052154197, "fcm_dpo/beta": 1.2216218709945679, "fcm_dpo/delta": -0.13029904663562775, "fcm_dpo/margin": 0.9128764867782593, "fcm_dpo/q_t": 0.34319591522216797, "grad_norm": 366.9609069824219, "learning_rate": 3.7728311501708674e-07, "logits/chosen": 0.08489110320806503, "logits/rejected": 0.04814103990793228, "logps/chosen": -84.4659652709961, "logps/ref_chosen": -83.17068481445312, "logps/ref_rejected": -88.33625793457031, "logps/rejected": -90.54441833496094, "loss": 1.0999, "margin_dpo/margin_mean": 0.9128766655921936, "margin_dpo/margin_std": 1.5033926963806152, "step": 264 }, { "epoch": 0.40060468631897206, "fcm_dpo/beta": 1.1946429014205933, "fcm_dpo/delta": -0.042388565838336945, "fcm_dpo/margin": 0.8688783049583435, "fcm_dpo/q_t": 0.31362420320510864, "grad_norm": 336.9339904785156, "learning_rate": 3.7614332175848027e-07, "logits/chosen": 0.1749192476272583, "logits/rejected": 0.12444747984409332, "logps/chosen": -53.23479080200195, "logps/ref_chosen": -51.66284942626953, "logps/ref_rejected": -67.1720962524414, "logps/rejected": -69.6129150390625, "loss": 0.9924, "margin_dpo/margin_mean": 0.8688779473304749, "margin_dpo/margin_std": 1.2264655828475952, "step": 265 }, { "epoch": 0.4021164021164021, "fcm_dpo/beta": 1.200698971748352, "fcm_dpo/delta": 0.018015079200267792, "fcm_dpo/margin": 0.8188613653182983, "fcm_dpo/q_t": 0.342385858297348, "grad_norm": 329.1114807128906, "learning_rate": 3.75e-07, "logits/chosen": 0.1544215977191925, "logits/rejected": 0.10259807109832764, "logps/chosen": -58.89800262451172, "logps/ref_chosen": -57.45049285888672, "logps/ref_rejected": -77.60826110839844, "logps/rejected": -79.87462615966797, "loss": 1.0917, "margin_dpo/margin_mean": 0.8188612461090088, "margin_dpo/margin_std": 1.41795015335083, "step": 266 }, { "epoch": 0.4036281179138322, "fcm_dpo/beta": 1.2707109451293945, "fcm_dpo/delta": 0.28088656067848206, "fcm_dpo/margin": 0.5725541710853577, "fcm_dpo/q_t": 0.3735983073711395, "grad_norm": 291.73394775390625, "learning_rate": 3.738531817228131e-07, "logits/chosen": 0.15856996178627014, "logits/rejected": 0.1470004916191101, "logps/chosen": -56.46894836425781, "logps/ref_chosen": -55.03535079956055, "logps/ref_rejected": -66.0953369140625, "logps/rejected": -68.10148620605469, "loss": 1.2705, "margin_dpo/margin_mean": 0.5725547075271606, "margin_dpo/margin_std": 1.2551192045211792, "step": 267 }, { "epoch": 0.4051398337112623, "fcm_dpo/beta": 1.2686963081359863, "fcm_dpo/delta": -0.053165629506111145, "fcm_dpo/margin": 0.8244317770004272, "fcm_dpo/q_t": 0.34110718965530396, "grad_norm": 308.4896240234375, "learning_rate": 3.7270289900589204e-07, "logits/chosen": 0.07009106129407883, "logits/rejected": 0.05312522128224373, "logps/chosen": -66.37588500976562, "logps/ref_chosen": -65.07174682617188, "logps/ref_rejected": -71.42485809326172, "logps/rejected": -73.5534439086914, "loss": 0.9779, "margin_dpo/margin_mean": 0.8244317770004272, "margin_dpo/margin_std": 1.2355961799621582, "step": 268 }, { "epoch": 0.40665154950869237, "fcm_dpo/beta": 1.2246897220611572, "fcm_dpo/delta": -0.29109132289886475, "fcm_dpo/margin": 1.0223791599273682, "fcm_dpo/q_t": 0.3000904321670532, "grad_norm": 285.43951416015625, "learning_rate": 3.7154918402511714e-07, "logits/chosen": 0.22535249590873718, "logits/rejected": 0.1888647824525833, "logps/chosen": -68.56861114501953, "logps/ref_chosen": -67.1362075805664, "logps/ref_rejected": -82.55778503417969, "logps/rejected": -85.01256561279297, "loss": 0.9234, "margin_dpo/margin_mean": 1.0223793983459473, "margin_dpo/margin_std": 1.2574834823608398, "step": 269 }, { "epoch": 0.40816326530612246, "fcm_dpo/beta": 1.2294461727142334, "fcm_dpo/delta": 0.26967453956604004, "fcm_dpo/margin": 0.6104166507720947, "fcm_dpo/q_t": 0.3626842498779297, "grad_norm": 348.1169738769531, "learning_rate": 3.7039206905237656e-07, "logits/chosen": 0.18059831857681274, "logits/rejected": 0.13146328926086426, "logps/chosen": -68.18730163574219, "logps/ref_chosen": -66.6886978149414, "logps/ref_rejected": -85.16129302978516, "logps/rejected": -87.27030944824219, "loss": 1.1816, "margin_dpo/margin_mean": 0.6104167103767395, "margin_dpo/margin_std": 1.19452965259552, "step": 270 }, { "epoch": 0.40967498110355255, "fcm_dpo/beta": 1.307576060295105, "fcm_dpo/delta": 0.36521047353744507, "fcm_dpo/margin": 0.5049761533737183, "fcm_dpo/q_t": 0.4147690534591675, "grad_norm": 440.6063537597656, "learning_rate": 3.692315864546635e-07, "logits/chosen": 0.1853274405002594, "logits/rejected": 0.14235132932662964, "logps/chosen": -73.94186401367188, "logps/ref_chosen": -72.40754699707031, "logps/ref_rejected": -92.06311798095703, "logps/rejected": -94.10240173339844, "loss": 1.5095, "margin_dpo/margin_mean": 0.5049762725830078, "margin_dpo/margin_std": 1.5028947591781616, "step": 271 }, { "epoch": 0.41118669690098264, "fcm_dpo/beta": 1.266921043395996, "fcm_dpo/delta": -0.4322693943977356, "fcm_dpo/margin": 1.0848881006240845, "fcm_dpo/q_t": 0.28500691056251526, "grad_norm": 282.21990966796875, "learning_rate": 3.6806776869317067e-07, "logits/chosen": 0.15356749296188354, "logits/rejected": 0.15307673811912537, "logps/chosen": -67.91374206542969, "logps/ref_chosen": -66.60140228271484, "logps/ref_rejected": -67.74340057373047, "logps/rejected": -70.140625, "loss": 0.7882, "margin_dpo/margin_mean": 1.0848881006240845, "margin_dpo/margin_std": 1.259714126586914, "step": 272 }, { "epoch": 0.4126984126984127, "fcm_dpo/beta": 1.2134504318237305, "fcm_dpo/delta": -0.202475443482399, "fcm_dpo/margin": 0.9696700572967529, "fcm_dpo/q_t": 0.3056218922138214, "grad_norm": 257.2859191894531, "learning_rate": 3.669006483223828e-07, "logits/chosen": 0.18354235589504242, "logits/rejected": 0.14025747776031494, "logps/chosen": -58.94585037231445, "logps/ref_chosen": -57.35487747192383, "logps/ref_rejected": -84.17168426513672, "logps/rejected": -86.73233032226562, "loss": 0.885, "margin_dpo/margin_mean": 0.9696696996688843, "margin_dpo/margin_std": 1.2579290866851807, "step": 273 }, { "epoch": 0.41421012849584277, "fcm_dpo/beta": 1.17547607421875, "fcm_dpo/delta": -0.022935807704925537, "fcm_dpo/margin": 0.8681968450546265, "fcm_dpo/q_t": 0.3227364718914032, "grad_norm": 261.911376953125, "learning_rate": 3.657302579891656e-07, "logits/chosen": 0.0659763365983963, "logits/rejected": 0.04620751738548279, "logps/chosen": -61.098419189453125, "logps/ref_chosen": -59.64149475097656, "logps/ref_rejected": -68.29348754882812, "logps/rejected": -70.61860656738281, "loss": 1.004, "margin_dpo/margin_mean": 0.8681962490081787, "margin_dpo/margin_std": 1.2701618671417236, "step": 274 }, { "epoch": 0.41572184429327286, "fcm_dpo/beta": 1.157713532447815, "fcm_dpo/delta": -0.10604125261306763, "fcm_dpo/margin": 0.9453713893890381, "fcm_dpo/q_t": 0.32869789004325867, "grad_norm": 274.95843505859375, "learning_rate": 3.645566304318526e-07, "logits/chosen": 0.1414085328578949, "logits/rejected": 0.07877371460199356, "logps/chosen": -54.83768844604492, "logps/ref_chosen": -53.26664352416992, "logps/ref_rejected": -73.84062194824219, "logps/rejected": -76.3570327758789, "loss": 0.9802, "margin_dpo/margin_mean": 0.9453713893890381, "margin_dpo/margin_std": 1.3858097791671753, "step": 275 }, { "epoch": 0.41723356009070295, "fcm_dpo/beta": 1.1194026470184326, "fcm_dpo/delta": -0.17233465611934662, "fcm_dpo/margin": 1.0291403532028198, "fcm_dpo/q_t": 0.29955726861953735, "grad_norm": 242.73602294921875, "learning_rate": 3.633797984793294e-07, "logits/chosen": 0.11177192628383636, "logits/rejected": 0.08380501717329025, "logps/chosen": -54.313621520996094, "logps/ref_chosen": -53.02079772949219, "logps/ref_rejected": -61.56678771972656, "logps/rejected": -63.888755798339844, "loss": 0.8804, "margin_dpo/margin_mean": 1.0291404724121094, "margin_dpo/margin_std": 1.2605907917022705, "step": 276 }, { "epoch": 0.41874527588813304, "fcm_dpo/beta": 1.16245436668396, "fcm_dpo/delta": 0.17784440517425537, "fcm_dpo/margin": 0.7118735313415527, "fcm_dpo/q_t": 0.3743218183517456, "grad_norm": 298.37750244140625, "learning_rate": 3.6219979505011555e-07, "logits/chosen": 0.2183254361152649, "logits/rejected": 0.23113352060317993, "logps/chosen": -73.12422180175781, "logps/ref_chosen": -71.43299102783203, "logps/ref_rejected": -67.65852355957031, "logps/rejected": -70.06163024902344, "loss": 1.1743, "margin_dpo/margin_mean": 0.7118737697601318, "margin_dpo/margin_std": 1.4327894449234009, "step": 277 }, { "epoch": 0.42025699168556313, "fcm_dpo/beta": 1.172609806060791, "fcm_dpo/delta": -0.03564952313899994, "fcm_dpo/margin": 0.8724105358123779, "fcm_dpo/q_t": 0.32611083984375, "grad_norm": 296.2506408691406, "learning_rate": 3.6101665315144353e-07, "logits/chosen": 0.1023668497800827, "logits/rejected": 0.07007478922605515, "logps/chosen": -68.64065551757812, "logps/ref_chosen": -67.11076354980469, "logps/ref_rejected": -88.74851989746094, "logps/rejected": -91.15081787109375, "loss": 1.0143, "margin_dpo/margin_mean": 0.8724101781845093, "margin_dpo/margin_std": 1.2294014692306519, "step": 278 }, { "epoch": 0.4217687074829932, "fcm_dpo/beta": 1.1110622882843018, "fcm_dpo/delta": -0.26436370611190796, "fcm_dpo/margin": 1.1091269254684448, "fcm_dpo/q_t": 0.276674747467041, "grad_norm": 208.7923583984375, "learning_rate": 3.5983040587833563e-07, "logits/chosen": 0.1121751144528389, "logits/rejected": 0.0790662094950676, "logps/chosen": -55.73821258544922, "logps/ref_chosen": -54.49748611450195, "logps/ref_rejected": -70.42373657226562, "logps/rejected": -72.77357482910156, "loss": 0.778, "margin_dpo/margin_mean": 1.1091272830963135, "margin_dpo/margin_std": 1.1377835273742676, "step": 279 }, { "epoch": 0.42328042328042326, "fcm_dpo/beta": 1.045201063156128, "fcm_dpo/delta": -0.19996249675750732, "fcm_dpo/margin": 1.1231290102005005, "fcm_dpo/q_t": 0.2829288840293884, "grad_norm": 192.96217346191406, "learning_rate": 3.586410864126781e-07, "logits/chosen": 0.15486222505569458, "logits/rejected": 0.123613640666008, "logps/chosen": -61.731178283691406, "logps/ref_chosen": -60.43281173706055, "logps/ref_rejected": -78.39051818847656, "logps/rejected": -80.81201171875, "loss": 0.7481, "margin_dpo/margin_mean": 1.1231298446655273, "margin_dpo/margin_std": 1.1084861755371094, "step": 280 }, { "epoch": 0.42479213907785335, "fcm_dpo/beta": 1.0173039436340332, "fcm_dpo/delta": -0.09447715431451797, "fcm_dpo/margin": 1.0629010200500488, "fcm_dpo/q_t": 0.3139858841896057, "grad_norm": 206.9331817626953, "learning_rate": 3.574487280222929e-07, "logits/chosen": 0.15210747718811035, "logits/rejected": 0.15197323262691498, "logps/chosen": -61.607582092285156, "logps/ref_chosen": -60.2820930480957, "logps/ref_rejected": -62.04009246826172, "logps/rejected": -64.42848205566406, "loss": 0.9316, "margin_dpo/margin_mean": 1.0629009008407593, "margin_dpo/margin_std": 1.3708744049072266, "step": 281 }, { "epoch": 0.42630385487528344, "fcm_dpo/beta": 1.054863452911377, "fcm_dpo/delta": 0.04717801511287689, "fcm_dpo/margin": 0.8964927196502686, "fcm_dpo/q_t": 0.33858174085617065, "grad_norm": 262.0866394042969, "learning_rate": 3.562533640600075e-07, "logits/chosen": 0.1025453507900238, "logits/rejected": 0.06392862647771835, "logps/chosen": -62.19837951660156, "logps/ref_chosen": -60.623924255371094, "logps/ref_rejected": -68.67400360107422, "logps/rejected": -71.14495849609375, "loss": 1.0039, "margin_dpo/margin_mean": 0.8964922428131104, "margin_dpo/margin_std": 1.2645740509033203, "step": 282 }, { "epoch": 0.42781557067271353, "fcm_dpo/beta": 1.0366630554199219, "fcm_dpo/delta": -0.0020070038735866547, "fcm_dpo/margin": 0.9660577774047852, "fcm_dpo/q_t": 0.3374154269695282, "grad_norm": 295.226318359375, "learning_rate": 3.550550279627215e-07, "logits/chosen": 0.11783361434936523, "logits/rejected": 0.054866328835487366, "logps/chosen": -69.20338439941406, "logps/ref_chosen": -67.64775085449219, "logps/ref_rejected": -99.96835327148438, "logps/rejected": -102.49005126953125, "loss": 1.0551, "margin_dpo/margin_mean": 0.9660578370094299, "margin_dpo/margin_std": 1.526496171951294, "step": 283 }, { "epoch": 0.4293272864701436, "fcm_dpo/beta": 1.0322619676589966, "fcm_dpo/delta": 0.00461952667683363, "fcm_dpo/margin": 0.9647125005722046, "fcm_dpo/q_t": 0.33289098739624023, "grad_norm": 231.38433837890625, "learning_rate": 3.5385375325047163e-07, "logits/chosen": 0.17530453205108643, "logits/rejected": 0.12596732378005981, "logps/chosen": -58.40034103393555, "logps/ref_chosen": -56.96742630004883, "logps/ref_rejected": -86.36236572265625, "logps/rejected": -88.75999450683594, "loss": 0.9914, "margin_dpo/margin_mean": 0.9647125005722046, "margin_dpo/margin_std": 1.445831298828125, "step": 284 }, { "epoch": 0.4308390022675737, "fcm_dpo/beta": 1.0661512613296509, "fcm_dpo/delta": 0.16971619427204132, "fcm_dpo/margin": 0.7905272245407104, "fcm_dpo/q_t": 0.34532633423805237, "grad_norm": 249.1006317138672, "learning_rate": 3.5264957352549375e-07, "logits/chosen": 0.18527567386627197, "logits/rejected": 0.1653136909008026, "logps/chosen": -73.34848022460938, "logps/ref_chosen": -71.65611267089844, "logps/ref_rejected": -81.63829803466797, "logps/rejected": -84.12120056152344, "loss": 0.9726, "margin_dpo/margin_mean": 0.7905269861221313, "margin_dpo/margin_std": 1.0835275650024414, "step": 285 }, { "epoch": 0.4323507180650038, "fcm_dpo/beta": 1.005662441253662, "fcm_dpo/delta": -0.34280824661254883, "fcm_dpo/margin": 1.2858762741088867, "fcm_dpo/q_t": 0.27013134956359863, "grad_norm": 211.03579711914062, "learning_rate": 3.514425224712835e-07, "logits/chosen": 0.10800629109144211, "logits/rejected": 0.0394493006169796, "logps/chosen": -62.53429412841797, "logps/ref_chosen": -61.07952117919922, "logps/ref_rejected": -91.28128051757812, "logps/rejected": -94.02192687988281, "loss": 0.8318, "margin_dpo/margin_mean": 1.2858755588531494, "margin_dpo/margin_std": 1.5294857025146484, "step": 286 }, { "epoch": 0.43386243386243384, "fcm_dpo/beta": 0.992131769657135, "fcm_dpo/delta": -0.06625291705131531, "fcm_dpo/margin": 1.0675835609436035, "fcm_dpo/q_t": 0.3116587698459625, "grad_norm": 213.88555908203125, "learning_rate": 3.502326338516534e-07, "logits/chosen": 0.13146105408668518, "logits/rejected": 0.10076682269573212, "logps/chosen": -47.52275848388672, "logps/ref_chosen": -46.035789489746094, "logps/ref_rejected": -59.95293426513672, "logps/rejected": -62.507484436035156, "loss": 0.8751, "margin_dpo/margin_mean": 1.067583441734314, "margin_dpo/margin_std": 1.295511245727539, "step": 287 }, { "epoch": 0.43537414965986393, "fcm_dpo/beta": 0.9990655183792114, "fcm_dpo/delta": 0.13602013885974884, "fcm_dpo/margin": 0.8767856359481812, "fcm_dpo/q_t": 0.35159996151924133, "grad_norm": 293.45867919921875, "learning_rate": 3.490199415097892e-07, "logits/chosen": 0.053915057331323624, "logits/rejected": 0.014950074255466461, "logps/chosen": -67.08319854736328, "logps/ref_chosen": -65.3908462524414, "logps/ref_rejected": -88.53607940673828, "logps/rejected": -91.10520935058594, "loss": 1.0811, "margin_dpo/margin_mean": 0.8767852187156677, "margin_dpo/margin_std": 1.4756031036376953, "step": 288 }, { "epoch": 0.436885865457294, "fcm_dpo/beta": 1.0097713470458984, "fcm_dpo/delta": 0.013888869434595108, "fcm_dpo/margin": 0.9776356220245361, "fcm_dpo/q_t": 0.3457440435886383, "grad_norm": 220.5363006591797, "learning_rate": 3.4780447936730247e-07, "logits/chosen": 0.16835784912109375, "logits/rejected": 0.14368662238121033, "logps/chosen": -56.29551696777344, "logps/ref_chosen": -54.5936279296875, "logps/ref_rejected": -67.20855712890625, "logps/rejected": -69.88806915283203, "loss": 1.0415, "margin_dpo/margin_mean": 0.9776356220245361, "margin_dpo/margin_std": 1.556645393371582, "step": 289 }, { "epoch": 0.4383975812547241, "fcm_dpo/beta": 1.0043764114379883, "fcm_dpo/delta": -0.04867362976074219, "fcm_dpo/margin": 1.0387194156646729, "fcm_dpo/q_t": 0.33352506160736084, "grad_norm": 238.85610961914062, "learning_rate": 3.465862814232821e-07, "logits/chosen": 0.19789519906044006, "logits/rejected": 0.14447157084941864, "logps/chosen": -63.15911865234375, "logps/ref_chosen": -61.38457489013672, "logps/ref_rejected": -91.92778015136719, "logps/rejected": -94.74103546142578, "loss": 0.9701, "margin_dpo/margin_mean": 1.038718819618225, "margin_dpo/margin_std": 1.499373435974121, "step": 290 }, { "epoch": 0.4399092970521542, "fcm_dpo/beta": 1.0028091669082642, "fcm_dpo/delta": -0.10185343772172928, "fcm_dpo/margin": 1.0858376026153564, "fcm_dpo/q_t": 0.3095766305923462, "grad_norm": 204.3426055908203, "learning_rate": 3.4536538175334343e-07, "logits/chosen": 0.23501402139663696, "logits/rejected": 0.1915348768234253, "logps/chosen": -52.46550750732422, "logps/ref_chosen": -50.863037109375, "logps/ref_rejected": -82.20868682861328, "logps/rejected": -84.89698791503906, "loss": 0.8974, "margin_dpo/margin_mean": 1.085837960243225, "margin_dpo/margin_std": 1.3475749492645264, "step": 291 }, { "epoch": 0.4414210128495843, "fcm_dpo/beta": 1.0263406038284302, "fcm_dpo/delta": 0.24256381392478943, "fcm_dpo/margin": 0.7545459270477295, "fcm_dpo/q_t": 0.3599158823490143, "grad_norm": 320.62713623046875, "learning_rate": 3.4414181450867465e-07, "logits/chosen": 0.1599317491054535, "logits/rejected": 0.11992324888706207, "logps/chosen": -65.92310333251953, "logps/ref_chosen": -64.34888458251953, "logps/ref_rejected": -72.86434173583984, "logps/rejected": -75.193115234375, "loss": 1.0762, "margin_dpo/margin_mean": 0.7545456886291504, "margin_dpo/margin_std": 1.2547924518585205, "step": 292 }, { "epoch": 0.4429327286470144, "fcm_dpo/beta": 1.0156301259994507, "fcm_dpo/delta": -0.14779676496982574, "fcm_dpo/margin": 1.1138885021209717, "fcm_dpo/q_t": 0.29969820380210876, "grad_norm": 190.7615966796875, "learning_rate": 3.4291561391508185e-07, "logits/chosen": 0.22061100602149963, "logits/rejected": 0.15896283090114594, "logps/chosen": -56.73548889160156, "logps/ref_chosen": -54.869468688964844, "logps/ref_rejected": -81.858642578125, "logps/rejected": -84.83856201171875, "loss": 0.8801, "margin_dpo/margin_mean": 1.1138887405395508, "margin_dpo/margin_std": 1.374760627746582, "step": 293 }, { "epoch": 0.4444444444444444, "fcm_dpo/beta": 0.9778472185134888, "fcm_dpo/delta": -0.08654538542032242, "fcm_dpo/margin": 1.0972121953964233, "fcm_dpo/q_t": 0.309672474861145, "grad_norm": 168.36590576171875, "learning_rate": 3.4168681427203153e-07, "logits/chosen": 0.17619842290878296, "logits/rejected": 0.143341526389122, "logps/chosen": -58.26924133300781, "logps/ref_chosen": -56.670902252197266, "logps/ref_rejected": -70.32819366455078, "logps/rejected": -73.02374267578125, "loss": 0.8444, "margin_dpo/margin_mean": 1.0972115993499756, "margin_dpo/margin_std": 1.2684025764465332, "step": 294 }, { "epoch": 0.4459561602418745, "fcm_dpo/beta": 1.0054044723510742, "fcm_dpo/delta": 0.08222609758377075, "fcm_dpo/margin": 0.9189479947090149, "fcm_dpo/q_t": 0.35169196128845215, "grad_norm": 228.4792938232422, "learning_rate": 3.4045544995169125e-07, "logits/chosen": 0.1701010763645172, "logits/rejected": 0.10789903253316879, "logps/chosen": -52.09330749511719, "logps/ref_chosen": -50.40088653564453, "logps/ref_rejected": -83.43521881103516, "logps/rejected": -86.04659271240234, "loss": 1.0393, "margin_dpo/margin_mean": 0.918948233127594, "margin_dpo/margin_std": 1.4247148036956787, "step": 295 }, { "epoch": 0.4474678760393046, "fcm_dpo/beta": 0.9557000994682312, "fcm_dpo/delta": -0.28874313831329346, "fcm_dpo/margin": 1.3085522651672363, "fcm_dpo/q_t": 0.28932487964630127, "grad_norm": 233.505126953125, "learning_rate": 3.392215553979679e-07, "logits/chosen": 0.12661093473434448, "logits/rejected": 0.09931506216526031, "logps/chosen": -70.89358520507812, "logps/ref_chosen": -69.15034484863281, "logps/ref_rejected": -89.60166931152344, "logps/rejected": -92.65345764160156, "loss": 0.8386, "margin_dpo/margin_mean": 1.3085522651672363, "margin_dpo/margin_std": 1.4957661628723145, "step": 296 }, { "epoch": 0.4489795918367347, "fcm_dpo/beta": 0.9359762668609619, "fcm_dpo/delta": -0.1143687292933464, "fcm_dpo/margin": 1.1771044731140137, "fcm_dpo/q_t": 0.30046796798706055, "grad_norm": 200.27796936035156, "learning_rate": 3.3798516512554485e-07, "logits/chosen": 0.13300062716007233, "logits/rejected": 0.08352112770080566, "logps/chosen": -59.854644775390625, "logps/ref_chosen": -58.01630401611328, "logps/ref_rejected": -69.95780944824219, "logps/rejected": -72.97325134277344, "loss": 0.8157, "margin_dpo/margin_mean": 1.1771044731140137, "margin_dpo/margin_std": 1.2866284847259521, "step": 297 }, { "epoch": 0.4504913076341648, "fcm_dpo/beta": 0.9372185468673706, "fcm_dpo/delta": 0.04445381462574005, "fcm_dpo/margin": 1.0229597091674805, "fcm_dpo/q_t": 0.3366781175136566, "grad_norm": 205.09780883789062, "learning_rate": 3.367463137189156e-07, "logits/chosen": 0.2418043464422226, "logits/rejected": 0.18982425332069397, "logps/chosen": -58.04621505737305, "logps/ref_chosen": -56.1693115234375, "logps/ref_rejected": -68.55052185058594, "logps/rejected": -71.45037841796875, "loss": 1.0065, "margin_dpo/margin_mean": 1.0229599475860596, "margin_dpo/margin_std": 1.4922395944595337, "step": 298 }, { "epoch": 0.4520030234315949, "fcm_dpo/beta": 0.9768006801605225, "fcm_dpo/delta": 0.26853734254837036, "fcm_dpo/margin": 0.7673162817955017, "fcm_dpo/q_t": 0.3727618455886841, "grad_norm": 255.50270080566406, "learning_rate": 3.355050358314172e-07, "logits/chosen": 0.08510833978652954, "logits/rejected": 0.06492967158555984, "logps/chosen": -64.02816009521484, "logps/ref_chosen": -62.31780242919922, "logps/ref_rejected": -72.60028839111328, "logps/rejected": -75.07796478271484, "loss": 1.1587, "margin_dpo/margin_mean": 0.7673170566558838, "margin_dpo/margin_std": 1.4521667957305908, "step": 299 }, { "epoch": 0.45351473922902497, "fcm_dpo/beta": 0.982731282711029, "fcm_dpo/delta": -0.016341693699359894, "fcm_dpo/margin": 1.0325164794921875, "fcm_dpo/q_t": 0.3216491937637329, "grad_norm": 243.2555389404297, "learning_rate": 3.3426136618426043e-07, "logits/chosen": 0.14876267313957214, "logits/rejected": 0.10573962330818176, "logps/chosen": -62.35066223144531, "logps/ref_chosen": -60.38157653808594, "logps/ref_rejected": -75.45442199707031, "logps/rejected": -78.45602416992188, "loss": 0.9532, "margin_dpo/margin_mean": 1.0325164794921875, "margin_dpo/margin_std": 1.3928803205490112, "step": 300 }, { "epoch": 0.455026455026455, "fcm_dpo/beta": 0.9845176935195923, "fcm_dpo/delta": 0.09773456305265427, "fcm_dpo/margin": 0.9239287376403809, "fcm_dpo/q_t": 0.36213570833206177, "grad_norm": 244.90597534179688, "learning_rate": 3.3301533956555885e-07, "logits/chosen": 0.16752877831459045, "logits/rejected": 0.14301586151123047, "logps/chosen": -54.7681999206543, "logps/ref_chosen": -52.85089111328125, "logps/ref_rejected": -69.97584533691406, "logps/rejected": -72.81707763671875, "loss": 1.1465, "margin_dpo/margin_mean": 0.9239292740821838, "margin_dpo/margin_std": 1.6796950101852417, "step": 301 }, { "epoch": 0.4565381708238851, "fcm_dpo/beta": 1.0574851036071777, "fcm_dpo/delta": 0.3823769986629486, "fcm_dpo/margin": 0.607899010181427, "fcm_dpo/q_t": 0.3950553238391876, "grad_norm": 319.1892395019531, "learning_rate": 3.317669908293554e-07, "logits/chosen": 0.06301631033420563, "logits/rejected": 0.02415418066084385, "logps/chosen": -68.90666198730469, "logps/ref_chosen": -66.96650695800781, "logps/ref_rejected": -88.09510803222656, "logps/rejected": -90.64315795898438, "loss": 1.2959, "margin_dpo/margin_mean": 0.6078989505767822, "margin_dpo/margin_std": 1.4545881748199463, "step": 302 }, { "epoch": 0.4580498866213152, "fcm_dpo/beta": 0.9939075112342834, "fcm_dpo/delta": -0.5528866052627563, "fcm_dpo/margin": 1.478973388671875, "fcm_dpo/q_t": 0.2767646014690399, "grad_norm": 199.89747619628906, "learning_rate": 3.3051635489464793e-07, "logits/chosen": 0.17920634150505066, "logits/rejected": 0.1289132535457611, "logps/chosen": -63.91869354248047, "logps/ref_chosen": -62.12152862548828, "logps/ref_rejected": -90.31204223632812, "logps/rejected": -93.58818817138672, "loss": 0.8206, "margin_dpo/margin_mean": 1.4789727926254272, "margin_dpo/margin_std": 1.7652822732925415, "step": 303 }, { "epoch": 0.4595616024187453, "fcm_dpo/beta": 0.9474884271621704, "fcm_dpo/delta": -0.1770581603050232, "fcm_dpo/margin": 1.2210001945495605, "fcm_dpo/q_t": 0.2853131592273712, "grad_norm": 196.20419311523438, "learning_rate": 3.292634667444117e-07, "logits/chosen": 0.14427047967910767, "logits/rejected": 0.10413600504398346, "logps/chosen": -62.46052551269531, "logps/ref_chosen": -60.695091247558594, "logps/ref_rejected": -78.2525405883789, "logps/rejected": -81.23897552490234, "loss": 0.7895, "margin_dpo/margin_mean": 1.2210009098052979, "margin_dpo/margin_std": 1.2481887340545654, "step": 304 }, { "epoch": 0.46107331821617537, "fcm_dpo/beta": 0.9275550842285156, "fcm_dpo/delta": 0.015897810459136963, "fcm_dpo/margin": 1.0611162185668945, "fcm_dpo/q_t": 0.3354414701461792, "grad_norm": 219.49313354492188, "learning_rate": 3.280083614246217e-07, "logits/chosen": 0.09203135967254639, "logits/rejected": 0.09808552265167236, "logps/chosen": -74.84890747070312, "logps/ref_chosen": -72.69914245605469, "logps/ref_rejected": -65.65670776367188, "logps/rejected": -68.86759185791016, "loss": 1.0316, "margin_dpo/margin_mean": 1.0611159801483154, "margin_dpo/margin_std": 1.6295936107635498, "step": 305 }, { "epoch": 0.46258503401360546, "fcm_dpo/beta": 0.9322211742401123, "fcm_dpo/delta": 0.10965774953365326, "fcm_dpo/margin": 0.958516001701355, "fcm_dpo/q_t": 0.34834927320480347, "grad_norm": 212.98146057128906, "learning_rate": 3.267510740432719e-07, "logits/chosen": 0.15844234824180603, "logits/rejected": 0.07754644751548767, "logps/chosen": -56.162513732910156, "logps/ref_chosen": -53.97052764892578, "logps/ref_rejected": -71.02423095703125, "logps/rejected": -74.17473602294922, "loss": 1.0813, "margin_dpo/margin_mean": 0.9585161209106445, "margin_dpo/margin_std": 1.579054355621338, "step": 306 }, { "epoch": 0.46409674981103555, "fcm_dpo/beta": 0.9765808582305908, "fcm_dpo/delta": 0.09200756251811981, "fcm_dpo/margin": 0.9371168613433838, "fcm_dpo/q_t": 0.35316699743270874, "grad_norm": 227.36492919921875, "learning_rate": 3.2549163976939285e-07, "logits/chosen": 0.19748598337173462, "logits/rejected": 0.16263772547245026, "logps/chosen": -59.23182678222656, "logps/ref_chosen": -57.413108825683594, "logps/ref_rejected": -68.68010711669922, "logps/rejected": -71.43594360351562, "loss": 1.1199, "margin_dpo/margin_mean": 0.9371169805526733, "margin_dpo/margin_std": 1.6100661754608154, "step": 307 }, { "epoch": 0.4656084656084656, "fcm_dpo/beta": 0.9509669542312622, "fcm_dpo/delta": -0.15045209228992462, "fcm_dpo/margin": 1.1912171840667725, "fcm_dpo/q_t": 0.30430689454078674, "grad_norm": 198.90879821777344, "learning_rate": 3.2423009383206874e-07, "logits/chosen": 0.14542096853256226, "logits/rejected": 0.1255512833595276, "logps/chosen": -68.51435852050781, "logps/ref_chosen": -66.59879302978516, "logps/ref_rejected": -74.337158203125, "logps/rejected": -77.44393920898438, "loss": 0.8821, "margin_dpo/margin_mean": 1.1912175416946411, "margin_dpo/margin_std": 1.509572982788086, "step": 308 }, { "epoch": 0.4671201814058957, "fcm_dpo/beta": 0.9307016134262085, "fcm_dpo/delta": -0.036133162677288055, "fcm_dpo/margin": 1.1062381267547607, "fcm_dpo/q_t": 0.32213109731674194, "grad_norm": 279.2270812988281, "learning_rate": 3.229664715194511e-07, "logits/chosen": 0.1940372884273529, "logits/rejected": 0.14977452158927917, "logps/chosen": -67.56904602050781, "logps/ref_chosen": -65.39474487304688, "logps/ref_rejected": -75.70930480957031, "logps/rejected": -78.98983764648438, "loss": 0.926, "margin_dpo/margin_mean": 1.1062389612197876, "margin_dpo/margin_std": 1.470036506652832, "step": 309 }, { "epoch": 0.46863189720332576, "fcm_dpo/beta": 0.9882631897926331, "fcm_dpo/delta": 0.319682240486145, "fcm_dpo/margin": 0.7099736928939819, "fcm_dpo/q_t": 0.38903123140335083, "grad_norm": 270.8043518066406, "learning_rate": 3.2170080817777257e-07, "logits/chosen": 0.19459328055381775, "logits/rejected": 0.18906690180301666, "logps/chosen": -76.8685302734375, "logps/ref_chosen": -74.66827392578125, "logps/ref_rejected": -80.5689697265625, "logps/rejected": -83.47919464111328, "loss": 1.2064, "margin_dpo/margin_mean": 0.709973931312561, "margin_dpo/margin_std": 1.481546401977539, "step": 310 }, { "epoch": 0.47014361300075586, "fcm_dpo/beta": 0.997173547744751, "fcm_dpo/delta": -0.014330286532640457, "fcm_dpo/margin": 1.015620231628418, "fcm_dpo/q_t": 0.33949679136276245, "grad_norm": 241.86203002929688, "learning_rate": 3.204331392103574e-07, "logits/chosen": 0.11166363954544067, "logits/rejected": 0.02732861414551735, "logps/chosen": -61.56627655029297, "logps/ref_chosen": -59.738033294677734, "logps/ref_rejected": -93.60757446289062, "logps/rejected": -96.4514389038086, "loss": 1.1047, "margin_dpo/margin_mean": 1.0156208276748657, "margin_dpo/margin_std": 1.744983434677124, "step": 311 }, { "epoch": 0.47165532879818595, "fcm_dpo/beta": 1.0016117095947266, "fcm_dpo/delta": -0.043600842356681824, "fcm_dpo/margin": 1.0361428260803223, "fcm_dpo/q_t": 0.3195294141769409, "grad_norm": 267.0063781738281, "learning_rate": 3.1916350007663176e-07, "logits/chosen": 0.16160011291503906, "logits/rejected": 0.08802653849124908, "logps/chosen": -55.83079528808594, "logps/ref_chosen": -53.816436767578125, "logps/ref_rejected": -68.6575698852539, "logps/rejected": -71.70807647705078, "loss": 1.0571, "margin_dpo/margin_mean": 1.0361424684524536, "margin_dpo/margin_std": 1.636512041091919, "step": 312 }, { "epoch": 0.47316704459561604, "fcm_dpo/beta": 0.9808007478713989, "fcm_dpo/delta": -0.0073838010430336, "fcm_dpo/margin": 1.0249512195587158, "fcm_dpo/q_t": 0.3501819968223572, "grad_norm": 224.44908142089844, "learning_rate": 3.178919262911314e-07, "logits/chosen": 0.19469937682151794, "logits/rejected": 0.17493098974227905, "logps/chosen": -61.82642364501953, "logps/ref_chosen": -59.957359313964844, "logps/ref_rejected": -69.31729888916016, "logps/rejected": -72.21131896972656, "loss": 1.077, "margin_dpo/margin_mean": 1.0249509811401367, "margin_dpo/margin_std": 1.7490208148956299, "step": 313 }, { "epoch": 0.47467876039304613, "fcm_dpo/beta": 0.9450622200965881, "fcm_dpo/delta": -0.23715783655643463, "fcm_dpo/margin": 1.274531364440918, "fcm_dpo/q_t": 0.2915083169937134, "grad_norm": 200.15219116210938, "learning_rate": 3.166184534225087e-07, "logits/chosen": 0.16370022296905518, "logits/rejected": 0.17338353395462036, "logps/chosen": -72.26028442382812, "logps/ref_chosen": -70.26815795898438, "logps/ref_rejected": -69.23971557617188, "logps/rejected": -72.50637817382812, "loss": 0.8243, "margin_dpo/margin_mean": 1.2745311260223389, "margin_dpo/margin_std": 1.4535917043685913, "step": 314 }, { "epoch": 0.47619047619047616, "fcm_dpo/beta": 0.9470343589782715, "fcm_dpo/delta": -0.0004953928291797638, "fcm_dpo/margin": 1.0559828281402588, "fcm_dpo/q_t": 0.328433096408844, "grad_norm": 213.31190490722656, "learning_rate": 3.1534311709253723e-07, "logits/chosen": 0.09684689342975616, "logits/rejected": 0.060719601809978485, "logps/chosen": -69.97232055664062, "logps/ref_chosen": -67.79469299316406, "logps/ref_rejected": -74.55148315429688, "logps/rejected": -77.78509521484375, "loss": 0.9555, "margin_dpo/margin_mean": 1.0559827089309692, "margin_dpo/margin_std": 1.4584524631500244, "step": 315 }, { "epoch": 0.47770219198790626, "fcm_dpo/beta": 0.9314021468162537, "fcm_dpo/delta": -0.23499611020088196, "fcm_dpo/margin": 1.2908413410186768, "fcm_dpo/q_t": 0.3175312876701355, "grad_norm": 208.80650329589844, "learning_rate": 3.1406595297511564e-07, "logits/chosen": 0.08339610695838928, "logits/rejected": 0.005085200071334839, "logps/chosen": -57.207271575927734, "logps/ref_chosen": -55.288482666015625, "logps/ref_rejected": -96.15723419189453, "logps/rejected": -99.36686706542969, "loss": 0.9423, "margin_dpo/margin_mean": 1.290840983390808, "margin_dpo/margin_std": 1.7332630157470703, "step": 316 }, { "epoch": 0.47921390778533635, "fcm_dpo/beta": 0.8467363119125366, "fcm_dpo/delta": -0.33823323249816895, "fcm_dpo/margin": 1.5226337909698486, "fcm_dpo/q_t": 0.2754213809967041, "grad_norm": 168.14048767089844, "learning_rate": 3.1278699679526975e-07, "logits/chosen": 0.2033698856830597, "logits/rejected": 0.1626172512769699, "logps/chosen": -56.44146728515625, "logps/ref_chosen": -54.58137512207031, "logps/ref_rejected": -72.77232360839844, "logps/rejected": -76.15504455566406, "loss": 0.7402, "margin_dpo/margin_mean": 1.5226335525512695, "margin_dpo/margin_std": 1.537990689277649, "step": 317 }, { "epoch": 0.48072562358276644, "fcm_dpo/beta": 0.880409836769104, "fcm_dpo/delta": 0.30436062812805176, "fcm_dpo/margin": 0.8145400285720825, "fcm_dpo/q_t": 0.3881235420703888, "grad_norm": 241.8609619140625, "learning_rate": 3.1150628432815336e-07, "logits/chosen": 0.21144279837608337, "logits/rejected": 0.1659296602010727, "logps/chosen": -55.05265426635742, "logps/ref_chosen": -52.88822937011719, "logps/ref_rejected": -80.63988494873047, "logps/rejected": -83.61885070800781, "loss": 1.3193, "margin_dpo/margin_mean": 0.8145396709442139, "margin_dpo/margin_std": 1.8883013725280762, "step": 318 }, { "epoch": 0.48223733938019653, "fcm_dpo/beta": 0.8839104771614075, "fcm_dpo/delta": -0.04694606736302376, "fcm_dpo/margin": 1.1784477233886719, "fcm_dpo/q_t": 0.3256542682647705, "grad_norm": 204.67640686035156, "learning_rate": 3.1022385139804707e-07, "logits/chosen": 0.14550672471523285, "logits/rejected": 0.12916123867034912, "logps/chosen": -66.23543548583984, "logps/ref_chosen": -64.36333465576172, "logps/ref_rejected": -79.47296142578125, "logps/rejected": -82.52351379394531, "loss": 1.0132, "margin_dpo/margin_mean": 1.1784476041793823, "margin_dpo/margin_std": 1.7645900249481201, "step": 319 }, { "epoch": 0.4837490551776266, "fcm_dpo/beta": 0.9131995439529419, "fcm_dpo/delta": 0.07204453647136688, "fcm_dpo/margin": 1.013108253479004, "fcm_dpo/q_t": 0.37524113059043884, "grad_norm": 217.01268005371094, "learning_rate": 3.0893973387735683e-07, "logits/chosen": 0.055543892085552216, "logits/rejected": 0.023240717127919197, "logps/chosen": -51.213958740234375, "logps/ref_chosen": -49.558746337890625, "logps/ref_rejected": -71.23444366455078, "logps/rejected": -73.90277099609375, "loss": 1.1825, "margin_dpo/margin_mean": 1.0131080150604248, "margin_dpo/margin_std": 1.9926589727401733, "step": 320 }, { "epoch": 0.4852607709750567, "fcm_dpo/beta": 0.8980337381362915, "fcm_dpo/delta": 0.061922501772642136, "fcm_dpo/margin": 1.0499402284622192, "fcm_dpo/q_t": 0.33765465021133423, "grad_norm": 189.1380615234375, "learning_rate": 3.0765396768561004e-07, "logits/chosen": 0.11102212965488434, "logits/rejected": 0.09998691082000732, "logps/chosen": -54.228790283203125, "logps/ref_chosen": -52.08526611328125, "logps/ref_rejected": -55.58674621582031, "logps/rejected": -58.78020477294922, "loss": 1.0637, "margin_dpo/margin_mean": 1.0499403476715088, "margin_dpo/margin_std": 1.613488793373108, "step": 321 }, { "epoch": 0.48677248677248675, "fcm_dpo/beta": 0.8880925178527832, "fcm_dpo/delta": -0.23487280309200287, "fcm_dpo/margin": 1.3585911989212036, "fcm_dpo/q_t": 0.28043854236602783, "grad_norm": 176.82778930664062, "learning_rate": 3.063665887884511e-07, "logits/chosen": 0.17297013103961945, "logits/rejected": 0.1158091127872467, "logps/chosen": -49.436309814453125, "logps/ref_chosen": -47.404109954833984, "logps/ref_rejected": -73.4260025024414, "logps/rejected": -76.81678771972656, "loss": 0.7812, "margin_dpo/margin_mean": 1.3585913181304932, "margin_dpo/margin_std": 1.383784532546997, "step": 322 }, { "epoch": 0.48828420256991684, "fcm_dpo/beta": 0.8809718489646912, "fcm_dpo/delta": -0.021233975887298584, "fcm_dpo/margin": 1.149742841720581, "fcm_dpo/q_t": 0.359465092420578, "grad_norm": 248.77261352539062, "learning_rate": 3.0507763319663517e-07, "logits/chosen": 0.10760500282049179, "logits/rejected": 0.054044321179389954, "logps/chosen": -72.1104736328125, "logps/ref_chosen": -70.00630187988281, "logps/ref_rejected": -86.96690368652344, "logps/rejected": -90.22081756591797, "loss": 1.1742, "margin_dpo/margin_mean": 1.1497416496276855, "margin_dpo/margin_std": 2.1441869735717773, "step": 323 }, { "epoch": 0.4897959183673469, "fcm_dpo/beta": 0.8546582460403442, "fcm_dpo/delta": -0.05345672369003296, "fcm_dpo/margin": 1.2257628440856934, "fcm_dpo/q_t": 0.3085279166698456, "grad_norm": 156.60922241210938, "learning_rate": 3.0378713696502097e-07, "logits/chosen": 0.18619199097156525, "logits/rejected": 0.1414175033569336, "logps/chosen": -57.747230529785156, "logps/ref_chosen": -55.88882064819336, "logps/ref_rejected": -75.23088073730469, "logps/rejected": -78.31505584716797, "loss": 0.8287, "margin_dpo/margin_mean": 1.2257624864578247, "margin_dpo/margin_std": 1.363228678703308, "step": 324 }, { "epoch": 0.491307634164777, "fcm_dpo/beta": 0.8600409626960754, "fcm_dpo/delta": 0.09678801149129868, "fcm_dpo/margin": 1.0598913431167603, "fcm_dpo/q_t": 0.3425368666648865, "grad_norm": 200.6235809326172, "learning_rate": 3.0249513619156206e-07, "logits/chosen": 0.1525503695011139, "logits/rejected": 0.1075083315372467, "logps/chosen": -66.32670593261719, "logps/ref_chosen": -64.14701843261719, "logps/ref_rejected": -79.91143798828125, "logps/rejected": -83.15100860595703, "loss": 0.9779, "margin_dpo/margin_mean": 1.059891700744629, "margin_dpo/margin_std": 1.5498141050338745, "step": 325 }, { "epoch": 0.4928193499622071, "fcm_dpo/beta": 0.9292858242988586, "fcm_dpo/delta": 0.4350077509880066, "fcm_dpo/margin": 0.6357402801513672, "fcm_dpo/q_t": 0.40372180938720703, "grad_norm": 280.2021179199219, "learning_rate": 3.012016670162977e-07, "logits/chosen": 0.11779944598674774, "logits/rejected": 0.11533387750387192, "logps/chosen": -78.00666046142578, "logps/ref_chosen": -75.53131103515625, "logps/ref_rejected": -76.5898666381836, "logps/rejected": -79.70095825195312, "loss": 1.3715, "margin_dpo/margin_mean": 0.6357403993606567, "margin_dpo/margin_std": 1.7088639736175537, "step": 326 }, { "epoch": 0.4943310657596372, "fcm_dpo/beta": 0.9441518783569336, "fcm_dpo/delta": -0.06476020067930222, "fcm_dpo/margin": 1.120490312576294, "fcm_dpo/q_t": 0.32473599910736084, "grad_norm": 212.27267456054688, "learning_rate": 2.99906765620341e-07, "logits/chosen": 0.09260990470647812, "logits/rejected": 0.054583217948675156, "logps/chosen": -71.31211853027344, "logps/ref_chosen": -69.33717346191406, "logps/ref_rejected": -73.37751770019531, "logps/rejected": -76.47296142578125, "loss": 1.0397, "margin_dpo/margin_mean": 1.1204906702041626, "margin_dpo/margin_std": 1.7054201364517212, "step": 327 }, { "epoch": 0.4958427815570673, "fcm_dpo/beta": 0.938388466835022, "fcm_dpo/delta": -0.004852544516324997, "fcm_dpo/margin": 1.0702922344207764, "fcm_dpo/q_t": 0.3356061577796936, "grad_norm": 220.39108276367188, "learning_rate": 2.9861046822486766e-07, "logits/chosen": 0.1284867823123932, "logits/rejected": 0.10963472723960876, "logps/chosen": -63.444252014160156, "logps/ref_chosen": -61.70623016357422, "logps/ref_rejected": -83.73808288574219, "logps/rejected": -86.54638671875, "loss": 0.9952, "margin_dpo/margin_mean": 1.0702919960021973, "margin_dpo/margin_std": 1.561734914779663, "step": 328 }, { "epoch": 0.4973544973544973, "fcm_dpo/beta": 0.926159143447876, "fcm_dpo/delta": 0.030632048845291138, "fcm_dpo/margin": 1.044736385345459, "fcm_dpo/q_t": 0.36076274514198303, "grad_norm": 272.3174743652344, "learning_rate": 2.9731281109010253e-07, "logits/chosen": 0.165533185005188, "logits/rejected": 0.12572559714317322, "logps/chosen": -66.76375579833984, "logps/ref_chosen": -64.4984130859375, "logps/ref_rejected": -83.6591796875, "logps/rejected": -86.96925354003906, "loss": 1.169, "margin_dpo/margin_mean": 1.0447362661361694, "margin_dpo/margin_std": 1.9857655763626099, "step": 329 }, { "epoch": 0.4988662131519274, "fcm_dpo/beta": 0.9391987323760986, "fcm_dpo/delta": -0.06099821627140045, "fcm_dpo/margin": 1.122628927230835, "fcm_dpo/q_t": 0.32373300194740295, "grad_norm": 196.6431427001953, "learning_rate": 2.9601383051430505e-07, "logits/chosen": 0.13443490862846375, "logits/rejected": 0.08146592229604721, "logps/chosen": -56.5888671875, "logps/ref_chosen": -54.80464172363281, "logps/ref_rejected": -75.3194351196289, "logps/rejected": -78.22628784179688, "loss": 1.0489, "margin_dpo/margin_mean": 1.1226279735565186, "margin_dpo/margin_std": 1.7134172916412354, "step": 330 }, { "epoch": 0.5003779289493575, "fcm_dpo/beta": 0.8843910694122314, "fcm_dpo/delta": -0.29690316319465637, "fcm_dpo/margin": 1.4203097820281982, "fcm_dpo/q_t": 0.29872971773147583, "grad_norm": 204.7743377685547, "learning_rate": 2.947135628327544e-07, "logits/chosen": 0.24195344746112823, "logits/rejected": 0.2168186902999878, "logps/chosen": -61.22477340698242, "logps/ref_chosen": -59.242584228515625, "logps/ref_rejected": -69.87483215332031, "logps/rejected": -73.27733612060547, "loss": 0.8837, "margin_dpo/margin_mean": 1.4203091859817505, "margin_dpo/margin_std": 1.860417127609253, "step": 331 }, { "epoch": 0.5018896447467877, "fcm_dpo/beta": 0.8774402141571045, "fcm_dpo/delta": -0.07574308663606644, "fcm_dpo/margin": 1.2151854038238525, "fcm_dpo/q_t": 0.320780485868454, "grad_norm": 200.2390899658203, "learning_rate": 2.934120444167326e-07, "logits/chosen": 0.1037403866648674, "logits/rejected": 0.0657915323972702, "logps/chosen": -69.155517578125, "logps/ref_chosen": -67.10975646972656, "logps/ref_rejected": -77.11839294433594, "logps/rejected": -80.37932586669922, "loss": 0.9145, "margin_dpo/margin_mean": 1.2151854038238525, "margin_dpo/margin_std": 1.6321237087249756, "step": 332 }, { "epoch": 0.5034013605442177, "fcm_dpo/beta": 0.8272979259490967, "fcm_dpo/delta": -0.3091200590133667, "fcm_dpo/margin": 1.535266637802124, "fcm_dpo/q_t": 0.2927742302417755, "grad_norm": 172.9521942138672, "learning_rate": 2.921093116725076e-07, "logits/chosen": 0.1792515516281128, "logits/rejected": 0.12908412516117096, "logps/chosen": -60.283103942871094, "logps/ref_chosen": -58.381134033203125, "logps/ref_rejected": -85.02839660644531, "logps/rejected": -88.46562957763672, "loss": 0.8145, "margin_dpo/margin_mean": 1.5352662801742554, "margin_dpo/margin_std": 1.7731688022613525, "step": 333 }, { "epoch": 0.5049130763416477, "fcm_dpo/beta": 0.8177670240402222, "fcm_dpo/delta": 0.09361538290977478, "fcm_dpo/margin": 1.1182286739349365, "fcm_dpo/q_t": 0.3452100455760956, "grad_norm": 200.41497802734375, "learning_rate": 2.9080540104031484e-07, "logits/chosen": 0.16784140467643738, "logits/rejected": 0.13531029224395752, "logps/chosen": -69.00025939941406, "logps/ref_chosen": -66.89199829101562, "logps/ref_rejected": -91.83695220947266, "logps/rejected": -95.06344604492188, "loss": 1.063, "margin_dpo/margin_mean": 1.118227481842041, "margin_dpo/margin_std": 1.7574753761291504, "step": 334 }, { "epoch": 0.5064247921390779, "fcm_dpo/beta": 0.8364279866218567, "fcm_dpo/delta": -0.04344947636127472, "fcm_dpo/margin": 1.23694908618927, "fcm_dpo/q_t": 0.32282981276512146, "grad_norm": 176.56353759765625, "learning_rate": 2.895003489933375e-07, "logits/chosen": 0.15401214361190796, "logits/rejected": 0.12541456520557404, "logps/chosen": -63.48824691772461, "logps/ref_chosen": -61.51445770263672, "logps/ref_rejected": -75.68916320800781, "logps/rejected": -78.89990234375, "loss": 0.9844, "margin_dpo/margin_mean": 1.2369496822357178, "margin_dpo/margin_std": 1.6888550519943237, "step": 335 }, { "epoch": 0.5079365079365079, "fcm_dpo/beta": 0.8125041127204895, "fcm_dpo/delta": 0.06346192955970764, "fcm_dpo/margin": 1.1519867181777954, "fcm_dpo/q_t": 0.35848677158355713, "grad_norm": 223.40049743652344, "learning_rate": 2.8819419203668675e-07, "logits/chosen": 0.09962709248065948, "logits/rejected": 0.0864500105381012, "logps/chosen": -71.24321746826172, "logps/ref_chosen": -68.85006713867188, "logps/ref_rejected": -92.99603271484375, "logps/rejected": -96.54116821289062, "loss": 1.0724, "margin_dpo/margin_mean": 1.151987075805664, "margin_dpo/margin_std": 1.9282138347625732, "step": 336 }, { "epoch": 0.509448223733938, "fcm_dpo/beta": 0.8386461734771729, "fcm_dpo/delta": 0.09305550158023834, "fcm_dpo/margin": 1.0915064811706543, "fcm_dpo/q_t": 0.3443507254123688, "grad_norm": 215.70370483398438, "learning_rate": 2.8688696670638053e-07, "logits/chosen": 0.08634011447429657, "logits/rejected": 0.05762971565127373, "logps/chosen": -75.31697845458984, "logps/ref_chosen": -73.18783569335938, "logps/ref_rejected": -86.89118957519531, "logps/rejected": -90.11184692382812, "loss": 1.0527, "margin_dpo/margin_mean": 1.0915067195892334, "margin_dpo/margin_std": 1.7208020687103271, "step": 337 }, { "epoch": 0.5109599395313681, "fcm_dpo/beta": 0.8704172372817993, "fcm_dpo/delta": 0.1373847872018814, "fcm_dpo/margin": 1.0027812719345093, "fcm_dpo/q_t": 0.3466625213623047, "grad_norm": 219.7271270751953, "learning_rate": 2.8557870956832133e-07, "logits/chosen": 0.11361702531576157, "logits/rejected": 0.09130830317735672, "logps/chosen": -66.18433380126953, "logps/ref_chosen": -63.939613342285156, "logps/ref_rejected": -75.34243774414062, "logps/rejected": -78.58993530273438, "loss": 1.0112, "margin_dpo/margin_mean": 1.0027809143066406, "margin_dpo/margin_std": 1.4872009754180908, "step": 338 }, { "epoch": 0.5124716553287982, "fcm_dpo/beta": 0.8656154274940491, "fcm_dpo/delta": 0.016111478209495544, "fcm_dpo/margin": 1.137109398841858, "fcm_dpo/q_t": 0.3326491713523865, "grad_norm": 190.49337768554688, "learning_rate": 2.842694572172736e-07, "logits/chosen": 0.19113630056381226, "logits/rejected": 0.12893235683441162, "logps/chosen": -47.86590576171875, "logps/ref_chosen": -45.54913330078125, "logps/ref_rejected": -67.0482177734375, "logps/rejected": -70.50209045410156, "loss": 0.958, "margin_dpo/margin_mean": 1.137109398841858, "margin_dpo/margin_std": 1.5872113704681396, "step": 339 }, { "epoch": 0.5139833711262283, "fcm_dpo/beta": 0.8644633889198303, "fcm_dpo/delta": -0.16399508714675903, "fcm_dpo/margin": 1.3231072425842285, "fcm_dpo/q_t": 0.3184114396572113, "grad_norm": 165.63905334472656, "learning_rate": 2.8295924627584004e-07, "logits/chosen": 0.14443045854568481, "logits/rejected": 0.12298154830932617, "logps/chosen": -56.29311752319336, "logps/ref_chosen": -54.00564956665039, "logps/ref_rejected": -61.314430236816406, "logps/rejected": -64.92501068115234, "loss": 1.0075, "margin_dpo/margin_mean": 1.3231074810028076, "margin_dpo/margin_std": 1.933672547340393, "step": 340 }, { "epoch": 0.5154950869236583, "fcm_dpo/beta": 0.8116443157196045, "fcm_dpo/delta": -0.10012944042682648, "fcm_dpo/margin": 1.3234137296676636, "fcm_dpo/q_t": 0.31874704360961914, "grad_norm": 227.253662109375, "learning_rate": 2.816481133934373e-07, "logits/chosen": 0.1503913700580597, "logits/rejected": 0.11632785201072693, "logps/chosen": -65.45622253417969, "logps/ref_chosen": -63.39509582519531, "logps/ref_rejected": -76.20973205566406, "logps/rejected": -79.59427642822266, "loss": 0.9461, "margin_dpo/margin_mean": 1.3234134912490845, "margin_dpo/margin_std": 1.674351453781128, "step": 341 }, { "epoch": 0.5170068027210885, "fcm_dpo/beta": 0.8011665940284729, "fcm_dpo/delta": -0.21002721786499023, "fcm_dpo/margin": 1.4784516096115112, "fcm_dpo/q_t": 0.30981749296188354, "grad_norm": 152.3086700439453, "learning_rate": 2.8033609524527046e-07, "logits/chosen": 0.17057004570960999, "logits/rejected": 0.13639067113399506, "logps/chosen": -55.32265853881836, "logps/ref_chosen": -53.047813415527344, "logps/ref_rejected": -68.2854232788086, "logps/rejected": -72.03872680664062, "loss": 0.8854, "margin_dpo/margin_mean": 1.4784512519836426, "margin_dpo/margin_std": 1.8578169345855713, "step": 342 }, { "epoch": 0.5185185185185185, "fcm_dpo/beta": 0.8318637609481812, "fcm_dpo/delta": 0.3891153335571289, "fcm_dpo/margin": 0.7660273909568787, "fcm_dpo/q_t": 0.38361668586730957, "grad_norm": 204.89308166503906, "learning_rate": 2.7902322853130753e-07, "logits/chosen": 0.08965672552585602, "logits/rejected": 0.08724290132522583, "logps/chosen": -72.62460327148438, "logps/ref_chosen": -70.57852935791016, "logps/ref_rejected": -84.73873901367188, "logps/rejected": -87.55084228515625, "loss": 1.2263, "margin_dpo/margin_mean": 0.7660267353057861, "margin_dpo/margin_std": 1.6297626495361328, "step": 343 }, { "epoch": 0.5200302343159486, "fcm_dpo/beta": 0.8591570854187012, "fcm_dpo/delta": 0.02412712574005127, "fcm_dpo/margin": 1.1384611129760742, "fcm_dpo/q_t": 0.3305787742137909, "grad_norm": 197.9630584716797, "learning_rate": 2.7770954997525274e-07, "logits/chosen": 0.15758341550827026, "logits/rejected": 0.11625611782073975, "logps/chosen": -58.18259811401367, "logps/ref_chosen": -55.811004638671875, "logps/ref_rejected": -84.77637481689453, "logps/rejected": -88.28643035888672, "loss": 0.982, "margin_dpo/margin_mean": 1.138460636138916, "margin_dpo/margin_std": 1.6236982345581055, "step": 344 }, { "epoch": 0.5215419501133787, "fcm_dpo/beta": 0.8624707460403442, "fcm_dpo/delta": 0.05118731036782265, "fcm_dpo/margin": 1.1047433614730835, "fcm_dpo/q_t": 0.3428131341934204, "grad_norm": 181.71697998046875, "learning_rate": 2.7639509632351927e-07, "logits/chosen": 0.2466953694820404, "logits/rejected": 0.2128785401582718, "logps/chosen": -59.707969665527344, "logps/ref_chosen": -57.78609848022461, "logps/ref_rejected": -78.91847229003906, "logps/rejected": -81.94508361816406, "loss": 1.0228, "margin_dpo/margin_mean": 1.1047430038452148, "margin_dpo/margin_std": 1.6839402914047241, "step": 345 }, { "epoch": 0.5230536659108088, "fcm_dpo/beta": 0.8550155162811279, "fcm_dpo/delta": -0.15863507986068726, "fcm_dpo/margin": 1.334245204925537, "fcm_dpo/q_t": 0.3107675015926361, "grad_norm": 217.05181884765625, "learning_rate": 2.7507990434420123e-07, "logits/chosen": 0.18394121527671814, "logits/rejected": 0.13396984338760376, "logps/chosen": -58.29369354248047, "logps/ref_chosen": -56.285125732421875, "logps/ref_rejected": -91.15303039550781, "logps/rejected": -94.495849609375, "loss": 0.9038, "margin_dpo/margin_mean": 1.334245204925537, "margin_dpo/margin_std": 1.7104822397232056, "step": 346 }, { "epoch": 0.5245653817082389, "fcm_dpo/beta": 0.8462698459625244, "fcm_dpo/delta": 0.07013484090566635, "fcm_dpo/margin": 1.1055729389190674, "fcm_dpo/q_t": 0.34362083673477173, "grad_norm": 187.4880828857422, "learning_rate": 2.737640108260456e-07, "logits/chosen": 0.2319878786802292, "logits/rejected": 0.19426073133945465, "logps/chosen": -55.76850891113281, "logps/ref_chosen": -53.499542236328125, "logps/ref_rejected": -72.52565002441406, "logps/rejected": -75.90019226074219, "loss": 1.0505, "margin_dpo/margin_mean": 1.1055727005004883, "margin_dpo/margin_std": 1.7748844623565674, "step": 347 }, { "epoch": 0.5260770975056689, "fcm_dpo/beta": 0.8080967664718628, "fcm_dpo/delta": -0.3646644949913025, "fcm_dpo/margin": 1.6269646883010864, "fcm_dpo/q_t": 0.3286696970462799, "grad_norm": 163.6527099609375, "learning_rate": 2.724474525774229e-07, "logits/chosen": 0.23792192339897156, "logits/rejected": 0.21459215879440308, "logps/chosen": -52.67319107055664, "logps/ref_chosen": -50.78684997558594, "logps/ref_rejected": -68.63732147216797, "logps/rejected": -72.15061950683594, "loss": 0.9542, "margin_dpo/margin_mean": 1.6269644498825073, "margin_dpo/margin_std": 2.543069362640381, "step": 348 }, { "epoch": 0.527588813303099, "fcm_dpo/beta": 0.779322624206543, "fcm_dpo/delta": -0.11380349844694138, "fcm_dpo/margin": 1.411959171295166, "fcm_dpo/q_t": 0.3116574287414551, "grad_norm": 173.95448303222656, "learning_rate": 2.711302664252973e-07, "logits/chosen": 0.16869422793388367, "logits/rejected": 0.10714869201183319, "logps/chosen": -55.361000061035156, "logps/ref_chosen": -53.325008392333984, "logps/ref_rejected": -83.21236419677734, "logps/rejected": -86.66030883789062, "loss": 0.9061, "margin_dpo/margin_mean": 1.4119596481323242, "margin_dpo/margin_std": 1.8286110162734985, "step": 349 }, { "epoch": 0.5291005291005291, "fcm_dpo/beta": 0.7421770095825195, "fcm_dpo/delta": -0.2771596610546112, "fcm_dpo/margin": 1.6710288524627686, "fcm_dpo/q_t": 0.29638049006462097, "grad_norm": 182.74623107910156, "learning_rate": 2.698124892141971e-07, "logits/chosen": 0.1460535228252411, "logits/rejected": 0.09140360355377197, "logps/chosen": -63.824440002441406, "logps/ref_chosen": -61.625770568847656, "logps/ref_rejected": -87.63627624511719, "logps/rejected": -91.50596618652344, "loss": 0.8304, "margin_dpo/margin_mean": 1.6710278987884521, "margin_dpo/margin_std": 1.9888241291046143, "step": 350 }, { "epoch": 0.5306122448979592, "fcm_dpo/beta": 0.7518589496612549, "fcm_dpo/delta": 0.1678137630224228, "fcm_dpo/margin": 1.1261451244354248, "fcm_dpo/q_t": 0.33188802003860474, "grad_norm": 165.69610595703125, "learning_rate": 2.6849415780518357e-07, "logits/chosen": 0.09675121307373047, "logits/rejected": 0.0414729006588459, "logps/chosen": -58.45286178588867, "logps/ref_chosen": -56.2563362121582, "logps/ref_rejected": -79.11589813232422, "logps/rejected": -82.43856811523438, "loss": 1.0476, "margin_dpo/margin_mean": 1.1261451244354248, "margin_dpo/margin_std": 1.6900222301483154, "step": 351 }, { "epoch": 0.5321239606953893, "fcm_dpo/beta": 0.7760653495788574, "fcm_dpo/delta": 0.16433821618556976, "fcm_dpo/margin": 1.0951128005981445, "fcm_dpo/q_t": 0.35313987731933594, "grad_norm": 193.21861267089844, "learning_rate": 2.6717530907482024e-07, "logits/chosen": 0.16212013363838196, "logits/rejected": 0.12353149801492691, "logps/chosen": -65.12254333496094, "logps/ref_chosen": -63.05195236206055, "logps/ref_rejected": -85.52035522460938, "logps/rejected": -88.68605041503906, "loss": 0.9931, "margin_dpo/margin_mean": 1.0951130390167236, "margin_dpo/margin_std": 1.6728229522705078, "step": 352 }, { "epoch": 0.5336356764928194, "fcm_dpo/beta": 0.7995505332946777, "fcm_dpo/delta": 0.018024399876594543, "fcm_dpo/margin": 1.2259821891784668, "fcm_dpo/q_t": 0.33103376626968384, "grad_norm": 166.7522430419922, "learning_rate": 2.658559799141411e-07, "logits/chosen": 0.1692410409450531, "logits/rejected": 0.16683810949325562, "logps/chosen": -71.08948516845703, "logps/ref_chosen": -69.00918579101562, "logps/ref_rejected": -72.65840148925781, "logps/rejected": -75.96468353271484, "loss": 1.0033, "margin_dpo/margin_mean": 1.2259814739227295, "margin_dpo/margin_std": 1.810058832168579, "step": 353 }, { "epoch": 0.5351473922902494, "fcm_dpo/beta": 0.7726951837539673, "fcm_dpo/delta": -0.05429168790578842, "fcm_dpo/margin": 1.3514502048492432, "fcm_dpo/q_t": 0.319754034280777, "grad_norm": 218.87094116210938, "learning_rate": 2.6453620722761895e-07, "logits/chosen": 0.20377589762210846, "logits/rejected": 0.10091987252235413, "logps/chosen": -41.94265365600586, "logps/ref_chosen": -39.78833770751953, "logps/ref_rejected": -69.56885528564453, "logps/rejected": -73.07461547851562, "loss": 0.93, "margin_dpo/margin_mean": 1.3514502048492432, "margin_dpo/margin_std": 1.7668390274047852, "step": 354 }, { "epoch": 0.5366591080876795, "fcm_dpo/beta": 0.7757099866867065, "fcm_dpo/delta": -0.04570357874035835, "fcm_dpo/margin": 1.3419021368026733, "fcm_dpo/q_t": 0.32356125116348267, "grad_norm": 184.73477172851562, "learning_rate": 2.632160279321328e-07, "logits/chosen": 0.14926283061504364, "logits/rejected": 0.0633564367890358, "logps/chosen": -48.385475158691406, "logps/ref_chosen": -46.25537872314453, "logps/ref_rejected": -78.20236206054688, "logps/rejected": -81.67436218261719, "loss": 0.9503, "margin_dpo/margin_mean": 1.3419021368026733, "margin_dpo/margin_std": 1.8027377128601074, "step": 355 }, { "epoch": 0.5381708238851096, "fcm_dpo/beta": 0.7802422046661377, "fcm_dpo/delta": 0.09977808594703674, "fcm_dpo/margin": 1.165019154548645, "fcm_dpo/q_t": 0.34947988390922546, "grad_norm": 180.77276611328125, "learning_rate": 2.618954789559356e-07, "logits/chosen": 0.15579620003700256, "logits/rejected": 0.1042039543390274, "logps/chosen": -50.003562927246094, "logps/ref_chosen": -47.906158447265625, "logps/ref_rejected": -74.29397583007812, "logps/rejected": -77.556396484375, "loss": 1.1606, "margin_dpo/margin_mean": 1.1650193929672241, "margin_dpo/margin_std": 2.1200852394104004, "step": 356 }, { "epoch": 0.5396825396825397, "fcm_dpo/beta": 0.7977977991104126, "fcm_dpo/delta": 0.14315135776996613, "fcm_dpo/margin": 1.0884279012680054, "fcm_dpo/q_t": 0.3529171347618103, "grad_norm": 215.51937866210938, "learning_rate": 2.6057459723762076e-07, "logits/chosen": 0.16426903009414673, "logits/rejected": 0.13840454816818237, "logps/chosen": -64.85040283203125, "logps/ref_chosen": -62.63500213623047, "logps/ref_rejected": -65.11399841308594, "logps/rejected": -68.41783142089844, "loss": 1.116, "margin_dpo/margin_mean": 1.0884283781051636, "margin_dpo/margin_std": 1.8702547550201416, "step": 357 }, { "epoch": 0.5411942554799698, "fcm_dpo/beta": 0.7857924103736877, "fcm_dpo/delta": -0.2809675335884094, "fcm_dpo/margin": 1.586600661277771, "fcm_dpo/q_t": 0.2850938141345978, "grad_norm": 148.0938720703125, "learning_rate": 2.5925341972508954e-07, "logits/chosen": 0.1491168886423111, "logits/rejected": 0.14593584835529327, "logps/chosen": -69.29679107666016, "logps/ref_chosen": -67.20960998535156, "logps/ref_rejected": -69.34715270996094, "logps/rejected": -73.02093505859375, "loss": 0.798, "margin_dpo/margin_mean": 1.586600661277771, "margin_dpo/margin_std": 1.7209175825119019, "step": 358 }, { "epoch": 0.5427059712773998, "fcm_dpo/beta": 0.7968997955322266, "fcm_dpo/delta": 0.10445237159729004, "fcm_dpo/margin": 1.1224002838134766, "fcm_dpo/q_t": 0.34582918882369995, "grad_norm": 201.40667724609375, "learning_rate": 2.579319833745169e-07, "logits/chosen": 0.14725103974342346, "logits/rejected": 0.12238387763500214, "logps/chosen": -64.69267272949219, "logps/ref_chosen": -62.52578353881836, "logps/ref_rejected": -76.63114929199219, "logps/rejected": -79.9204330444336, "loss": 1.0804, "margin_dpo/margin_mean": 1.1224000453948975, "margin_dpo/margin_std": 1.7543888092041016, "step": 359 }, { "epoch": 0.54421768707483, "fcm_dpo/beta": 0.7883453369140625, "fcm_dpo/delta": 0.023167556151747704, "fcm_dpo/margin": 1.241877794265747, "fcm_dpo/q_t": 0.3393666446208954, "grad_norm": 213.6732635498047, "learning_rate": 2.5661032514931834e-07, "logits/chosen": 0.11307230591773987, "logits/rejected": 0.0418628454208374, "logps/chosen": -65.72219848632812, "logps/ref_chosen": -63.48772048950195, "logps/ref_rejected": -90.6891098022461, "logps/rejected": -94.16546630859375, "loss": 1.0189, "margin_dpo/margin_mean": 1.2418781518936157, "margin_dpo/margin_std": 1.9586660861968994, "step": 360 }, { "epoch": 0.54572940287226, "fcm_dpo/beta": 0.7847793698310852, "fcm_dpo/delta": -0.006262313574552536, "fcm_dpo/margin": 1.2811236381530762, "fcm_dpo/q_t": 0.32984602451324463, "grad_norm": 186.03604125976562, "learning_rate": 2.552884820191154e-07, "logits/chosen": 0.21646341681480408, "logits/rejected": 0.17365378141403198, "logps/chosen": -60.005924224853516, "logps/ref_chosen": -57.917144775390625, "logps/ref_rejected": -72.39089965820312, "logps/rejected": -75.76080322265625, "loss": 0.9667, "margin_dpo/margin_mean": 1.2811236381530762, "margin_dpo/margin_std": 1.8610559701919556, "step": 361 }, { "epoch": 0.54724111866969, "fcm_dpo/beta": 0.7910502552986145, "fcm_dpo/delta": -0.12872016429901123, "fcm_dpo/margin": 1.3991649150848389, "fcm_dpo/q_t": 0.3150370121002197, "grad_norm": 179.85292053222656, "learning_rate": 2.53966490958702e-07, "logits/chosen": 0.20901912450790405, "logits/rejected": 0.14084625244140625, "logps/chosen": -65.4914321899414, "logps/ref_chosen": -63.4434700012207, "logps/ref_rejected": -103.45516967773438, "logps/rejected": -106.90229797363281, "loss": 0.8868, "margin_dpo/margin_mean": 1.3991665840148926, "margin_dpo/margin_std": 1.6471703052520752, "step": 362 }, { "epoch": 0.5487528344671202, "fcm_dpo/beta": 0.7514413595199585, "fcm_dpo/delta": -0.15377500653266907, "fcm_dpo/margin": 1.512410044670105, "fcm_dpo/q_t": 0.3079715967178345, "grad_norm": 172.7432403564453, "learning_rate": 2.526443889470099e-07, "logits/chosen": 0.18472757935523987, "logits/rejected": 0.09980542212724686, "logps/chosen": -51.084228515625, "logps/ref_chosen": -48.65182876586914, "logps/ref_rejected": -88.65904235839844, "logps/rejected": -92.60385131835938, "loss": 0.9557, "margin_dpo/margin_mean": 1.512410044670105, "margin_dpo/margin_std": 2.1208112239837646, "step": 363 }, { "epoch": 0.5502645502645502, "fcm_dpo/beta": 0.7271950840950012, "fcm_dpo/delta": -0.1419481635093689, "fcm_dpo/margin": 1.5479559898376465, "fcm_dpo/q_t": 0.31253015995025635, "grad_norm": 143.4964599609375, "learning_rate": 2.513222129660744e-07, "logits/chosen": 0.09488549828529358, "logits/rejected": 0.020679466426372528, "logps/chosen": -59.90724563598633, "logps/ref_chosen": -57.87107467651367, "logps/ref_rejected": -80.95503234863281, "logps/rejected": -84.53915405273438, "loss": 0.9253, "margin_dpo/margin_mean": 1.5479564666748047, "margin_dpo/margin_std": 2.0348784923553467, "step": 364 }, { "epoch": 0.5517762660619804, "fcm_dpo/beta": 0.7135534882545471, "fcm_dpo/delta": -0.0528385192155838, "fcm_dpo/margin": 1.4665672779083252, "fcm_dpo/q_t": 0.30548185110092163, "grad_norm": 147.20799255371094, "learning_rate": 2.5e-07, "logits/chosen": 0.19886715710163116, "logits/rejected": 0.18875735998153687, "logps/chosen": -66.82514190673828, "logps/ref_chosen": -64.94217681884766, "logps/ref_rejected": -74.8599853515625, "logps/rejected": -78.20951843261719, "loss": 0.868, "margin_dpo/margin_mean": 1.4665677547454834, "margin_dpo/margin_std": 1.6899524927139282, "step": 365 }, { "epoch": 0.5532879818594104, "fcm_dpo/beta": 0.749003529548645, "fcm_dpo/delta": 0.20677639544010162, "fcm_dpo/margin": 1.071431279182434, "fcm_dpo/q_t": 0.36068403720855713, "grad_norm": 177.8300323486328, "learning_rate": 2.486777870339255e-07, "logits/chosen": 0.1121918261051178, "logits/rejected": 0.09326402097940445, "logps/chosen": -56.96438217163086, "logps/ref_chosen": -55.16598129272461, "logps/ref_rejected": -65.26121520996094, "logps/rejected": -68.13104248046875, "loss": 1.1563, "margin_dpo/margin_mean": 1.0714313983917236, "margin_dpo/margin_std": 1.9580434560775757, "step": 366 }, { "epoch": 0.5547996976568406, "fcm_dpo/beta": 0.7712104320526123, "fcm_dpo/delta": 0.2716788947582245, "fcm_dpo/margin": 0.9718486070632935, "fcm_dpo/q_t": 0.3597671687602997, "grad_norm": 178.6884002685547, "learning_rate": 2.4735561105299014e-07, "logits/chosen": 0.1118677407503128, "logits/rejected": 0.03625689074397087, "logps/chosen": -58.22844696044922, "logps/ref_chosen": -56.01046371459961, "logps/ref_rejected": -77.31010437011719, "logps/rejected": -80.49993896484375, "loss": 1.0665, "margin_dpo/margin_mean": 0.971848726272583, "margin_dpo/margin_std": 1.5560146570205688, "step": 367 }, { "epoch": 0.5563114134542706, "fcm_dpo/beta": 0.8040578365325928, "fcm_dpo/delta": 0.1679600328207016, "fcm_dpo/margin": 1.0528676509857178, "fcm_dpo/q_t": 0.3611965775489807, "grad_norm": 214.42361450195312, "learning_rate": 2.46033509041298e-07, "logits/chosen": 0.06508797407150269, "logits/rejected": 0.056503720581531525, "logps/chosen": -76.96674346923828, "logps/ref_chosen": -74.82927703857422, "logps/ref_rejected": -76.11680603027344, "logps/rejected": -79.30712890625, "loss": 1.1432, "margin_dpo/margin_mean": 1.052868127822876, "margin_dpo/margin_std": 1.877270221710205, "step": 368 }, { "epoch": 0.5578231292517006, "fcm_dpo/beta": 0.83278489112854, "fcm_dpo/delta": 0.09441757202148438, "fcm_dpo/margin": 1.095184564590454, "fcm_dpo/q_t": 0.3382675051689148, "grad_norm": 172.21173095703125, "learning_rate": 2.447115179808846e-07, "logits/chosen": 0.122508205473423, "logits/rejected": 0.09032686054706573, "logps/chosen": -60.496971130371094, "logps/ref_chosen": -58.32621765136719, "logps/ref_rejected": -80.92183685302734, "logps/rejected": -84.18777465820312, "loss": 1.0485, "margin_dpo/margin_mean": 1.0951846837997437, "margin_dpo/margin_std": 1.6767044067382812, "step": 369 }, { "epoch": 0.5593348450491308, "fcm_dpo/beta": 0.8102331757545471, "fcm_dpo/delta": -0.10076209902763367, "fcm_dpo/margin": 1.342164397239685, "fcm_dpo/q_t": 0.3211871385574341, "grad_norm": 181.2273406982422, "learning_rate": 2.4338967485068164e-07, "logits/chosen": 0.24069687724113464, "logits/rejected": 0.18847504258155823, "logps/chosen": -55.11189270019531, "logps/ref_chosen": -52.88372039794922, "logps/ref_rejected": -79.43692016601562, "logps/rejected": -83.00725555419922, "loss": 1.0596, "margin_dpo/margin_mean": 1.3421647548675537, "margin_dpo/margin_std": 2.065488576889038, "step": 370 }, { "epoch": 0.5608465608465608, "fcm_dpo/beta": 0.8171911239624023, "fcm_dpo/delta": -0.08116129040718079, "fcm_dpo/margin": 1.3074921369552612, "fcm_dpo/q_t": 0.3368060290813446, "grad_norm": 183.6264190673828, "learning_rate": 2.420680166254831e-07, "logits/chosen": 0.2111239731311798, "logits/rejected": 0.18440525233745575, "logps/chosen": -51.46292495727539, "logps/ref_chosen": -49.224212646484375, "logps/ref_rejected": -63.348472595214844, "logps/rejected": -66.8946762084961, "loss": 1.0896, "margin_dpo/margin_mean": 1.3074921369552612, "margin_dpo/margin_std": 2.0974409580230713, "step": 371 }, { "epoch": 0.562358276643991, "fcm_dpo/beta": 0.8467217683792114, "fcm_dpo/delta": 0.2932736575603485, "fcm_dpo/margin": 0.851443350315094, "fcm_dpo/q_t": 0.3849431276321411, "grad_norm": 214.92108154296875, "learning_rate": 2.4074658027491044e-07, "logits/chosen": 0.18400293588638306, "logits/rejected": 0.12166933715343475, "logps/chosen": -54.62636184692383, "logps/ref_chosen": -52.269554138183594, "logps/ref_rejected": -72.99522399902344, "logps/rejected": -76.20347595214844, "loss": 1.3565, "margin_dpo/margin_mean": 0.8514436483383179, "margin_dpo/margin_std": 2.075626850128174, "step": 372 }, { "epoch": 0.563869992441421, "fcm_dpo/beta": 0.8382232189178467, "fcm_dpo/delta": -0.05690415948629379, "fcm_dpo/margin": 1.2532103061676025, "fcm_dpo/q_t": 0.3378145694732666, "grad_norm": 255.2552947998047, "learning_rate": 2.394254027623792e-07, "logits/chosen": 0.19523033499717712, "logits/rejected": 0.14481961727142334, "logps/chosen": -63.56736755371094, "logps/ref_chosen": -61.112998962402344, "logps/ref_rejected": -76.24851989746094, "logps/rejected": -79.95610046386719, "loss": 1.0989, "margin_dpo/margin_mean": 1.2532098293304443, "margin_dpo/margin_std": 2.067584991455078, "step": 373 }, { "epoch": 0.5653817082388511, "fcm_dpo/beta": 0.788150429725647, "fcm_dpo/delta": -0.36466753482818604, "fcm_dpo/margin": 1.6652493476867676, "fcm_dpo/q_t": 0.2760714292526245, "grad_norm": 186.27012634277344, "learning_rate": 2.381045210440644e-07, "logits/chosen": 0.12782002985477448, "logits/rejected": 0.11331016570329666, "logps/chosen": -74.75846099853516, "logps/ref_chosen": -72.66920471191406, "logps/ref_rejected": -76.83158874511719, "logps/rejected": -80.58609771728516, "loss": 0.7647, "margin_dpo/margin_mean": 1.6652493476867676, "margin_dpo/margin_std": 1.7359645366668701, "step": 374 }, { "epoch": 0.5668934240362812, "fcm_dpo/beta": 0.8175476789474487, "fcm_dpo/delta": 0.3215762972831726, "fcm_dpo/margin": 0.8576102256774902, "fcm_dpo/q_t": 0.37826499342918396, "grad_norm": 234.281494140625, "learning_rate": 2.3678397206786715e-07, "logits/chosen": 0.1810404360294342, "logits/rejected": 0.1419374942779541, "logps/chosen": -59.804378509521484, "logps/ref_chosen": -57.68330383300781, "logps/ref_rejected": -79.34097290039062, "logps/rejected": -82.31965637207031, "loss": 1.1949, "margin_dpo/margin_mean": 0.8576098680496216, "margin_dpo/margin_std": 1.7258224487304688, "step": 375 }, { "epoch": 0.5684051398337112, "fcm_dpo/beta": 0.7721197605133057, "fcm_dpo/delta": -0.40702202916145325, "fcm_dpo/margin": 1.7395694255828857, "fcm_dpo/q_t": 0.28671878576278687, "grad_norm": 153.21202087402344, "learning_rate": 2.3546379277238103e-07, "logits/chosen": 0.19977155327796936, "logits/rejected": 0.15468192100524902, "logps/chosen": -53.76110076904297, "logps/ref_chosen": -51.674072265625, "logps/ref_rejected": -75.69713592529297, "logps/rejected": -79.52372741699219, "loss": 0.8446, "margin_dpo/margin_mean": 1.739569067955017, "margin_dpo/margin_std": 2.094245433807373, "step": 376 }, { "epoch": 0.5699168556311414, "fcm_dpo/beta": 0.7831248641014099, "fcm_dpo/delta": 0.08697693049907684, "fcm_dpo/margin": 1.1734894514083862, "fcm_dpo/q_t": 0.3411799371242523, "grad_norm": 162.50051879882812, "learning_rate": 2.3414402008585886e-07, "logits/chosen": 0.11727416515350342, "logits/rejected": 0.09625902771949768, "logps/chosen": -48.64626693725586, "logps/ref_chosen": -46.17853546142578, "logps/ref_rejected": -57.756500244140625, "logps/rejected": -61.39772415161133, "loss": 1.0109, "margin_dpo/margin_mean": 1.1734893321990967, "margin_dpo/margin_std": 1.7293052673339844, "step": 377 }, { "epoch": 0.5714285714285714, "fcm_dpo/beta": 0.8013092875480652, "fcm_dpo/delta": 0.08626553416252136, "fcm_dpo/margin": 1.1440807580947876, "fcm_dpo/q_t": 0.35015422105789185, "grad_norm": 186.11314392089844, "learning_rate": 2.3282469092517977e-07, "logits/chosen": 0.19343531131744385, "logits/rejected": 0.15595267713069916, "logps/chosen": -61.52544021606445, "logps/ref_chosen": -59.21887969970703, "logps/ref_rejected": -71.24818420410156, "logps/rejected": -74.69883728027344, "loss": 1.045, "margin_dpo/margin_mean": 1.1440809965133667, "margin_dpo/margin_std": 1.8243110179901123, "step": 378 }, { "epoch": 0.5729402872260015, "fcm_dpo/beta": 0.7728543281555176, "fcm_dpo/delta": -0.18132196366786957, "fcm_dpo/margin": 1.5013034343719482, "fcm_dpo/q_t": 0.3048959970474243, "grad_norm": 197.8959197998047, "learning_rate": 2.3150584219481643e-07, "logits/chosen": 0.21861502528190613, "logits/rejected": 0.1745017170906067, "logps/chosen": -78.35917663574219, "logps/ref_chosen": -76.31658935546875, "logps/ref_rejected": -104.26200103759766, "logps/rejected": -107.80589294433594, "loss": 0.865, "margin_dpo/margin_mean": 1.501303791999817, "margin_dpo/margin_std": 1.88877534866333, "step": 379 }, { "epoch": 0.5744520030234316, "fcm_dpo/beta": 0.732434093952179, "fcm_dpo/delta": -0.3159186840057373, "fcm_dpo/margin": 1.7422823905944824, "fcm_dpo/q_t": 0.2831631302833557, "grad_norm": 152.97654724121094, "learning_rate": 2.3018751078580283e-07, "logits/chosen": 0.18231691420078278, "logits/rejected": 0.15374861657619476, "logps/chosen": -63.16365051269531, "logps/ref_chosen": -61.283164978027344, "logps/ref_rejected": -72.38892364501953, "logps/rejected": -76.01168823242188, "loss": 0.8666, "margin_dpo/margin_mean": 1.7422822713851929, "margin_dpo/margin_std": 2.142005443572998, "step": 380 }, { "epoch": 0.5759637188208617, "fcm_dpo/beta": 0.7804316282272339, "fcm_dpo/delta": 0.42041903734207153, "fcm_dpo/margin": 0.7539185285568237, "fcm_dpo/q_t": 0.3955872654914856, "grad_norm": 201.4615936279297, "learning_rate": 2.288697335747027e-07, "logits/chosen": 0.12479900568723679, "logits/rejected": 0.10375261306762695, "logps/chosen": -60.67311096191406, "logps/ref_chosen": -58.2139892578125, "logps/ref_rejected": -60.78669357299805, "logps/rejected": -63.999732971191406, "loss": 1.2184, "margin_dpo/margin_mean": 0.7539188861846924, "margin_dpo/margin_std": 1.6283390522003174, "step": 381 }, { "epoch": 0.5774754346182918, "fcm_dpo/beta": 0.8213146924972534, "fcm_dpo/delta": 0.16648587584495544, "fcm_dpo/margin": 1.0126454830169678, "fcm_dpo/q_t": 0.3555706739425659, "grad_norm": 187.99722290039062, "learning_rate": 2.2755254742257706e-07, "logits/chosen": 0.1935308575630188, "logits/rejected": 0.1620044708251953, "logps/chosen": -64.31198120117188, "logps/ref_chosen": -61.82532501220703, "logps/ref_rejected": -83.0452880859375, "logps/rejected": -86.54458618164062, "loss": 1.0464, "margin_dpo/margin_mean": 1.0126454830169678, "margin_dpo/margin_std": 1.5206871032714844, "step": 382 }, { "epoch": 0.5789871504157218, "fcm_dpo/beta": 0.7823382616043091, "fcm_dpo/delta": -0.15291021764278412, "fcm_dpo/margin": 1.4475116729736328, "fcm_dpo/q_t": 0.3087931275367737, "grad_norm": 219.67236328125, "learning_rate": 2.2623598917395436e-07, "logits/chosen": 0.10844056308269501, "logits/rejected": 0.11368384212255478, "logps/chosen": -82.67160034179688, "logps/ref_chosen": -80.56326293945312, "logps/ref_rejected": -74.62922668457031, "logps/rejected": -78.18507385253906, "loss": 0.9182, "margin_dpo/margin_mean": 1.4475116729736328, "margin_dpo/margin_std": 1.882810354232788, "step": 383 }, { "epoch": 0.5804988662131519, "fcm_dpo/beta": 0.8081178665161133, "fcm_dpo/delta": 0.1466008424758911, "fcm_dpo/margin": 1.0682569742202759, "fcm_dpo/q_t": 0.3386213779449463, "grad_norm": 210.8586883544922, "learning_rate": 2.2492009565579875e-07, "logits/chosen": 0.15908417105674744, "logits/rejected": 0.1250711977481842, "logps/chosen": -68.10568237304688, "logps/ref_chosen": -65.47514343261719, "logps/ref_rejected": -79.67378234863281, "logps/rejected": -83.37257385253906, "loss": 1.014, "margin_dpo/margin_mean": 1.0682566165924072, "margin_dpo/margin_std": 1.5980737209320068, "step": 384 }, { "epoch": 0.582010582010582, "fcm_dpo/beta": 0.8039337396621704, "fcm_dpo/delta": -0.06407226622104645, "fcm_dpo/margin": 1.3148212432861328, "fcm_dpo/q_t": 0.3122956156730652, "grad_norm": 201.8307647705078, "learning_rate": 2.2360490367648084e-07, "logits/chosen": 0.14007516205310822, "logits/rejected": 0.11297205090522766, "logps/chosen": -68.39266967773438, "logps/ref_chosen": -66.0565185546875, "logps/ref_rejected": -86.68023681640625, "logps/rejected": -90.33121490478516, "loss": 0.9181, "margin_dpo/margin_mean": 1.3148208856582642, "margin_dpo/margin_std": 1.7250840663909912, "step": 385 }, { "epoch": 0.5835222978080121, "fcm_dpo/beta": 0.7948161363601685, "fcm_dpo/delta": 0.02499794214963913, "fcm_dpo/margin": 1.229053020477295, "fcm_dpo/q_t": 0.3376579284667969, "grad_norm": 190.71925354003906, "learning_rate": 2.2229045002474724e-07, "logits/chosen": 0.1209828183054924, "logits/rejected": 0.0802159532904625, "logps/chosen": -78.08750915527344, "logps/ref_chosen": -75.6236572265625, "logps/ref_rejected": -92.62330627441406, "logps/rejected": -96.31620788574219, "loss": 1.0068, "margin_dpo/margin_mean": 1.229053258895874, "margin_dpo/margin_std": 1.8680897951126099, "step": 386 }, { "epoch": 0.5850340136054422, "fcm_dpo/beta": 0.7716137170791626, "fcm_dpo/delta": -0.2335832417011261, "fcm_dpo/margin": 1.5613083839416504, "fcm_dpo/q_t": 0.27384334802627563, "grad_norm": 133.3782501220703, "learning_rate": 2.209767714686924e-07, "logits/chosen": 0.15220069885253906, "logits/rejected": 0.08792141824960709, "logps/chosen": -49.354488372802734, "logps/ref_chosen": -47.22170639038086, "logps/ref_rejected": -87.338134765625, "logps/rejected": -91.03223419189453, "loss": 0.7497, "margin_dpo/margin_mean": 1.5613081455230713, "margin_dpo/margin_std": 1.4777767658233643, "step": 387 }, { "epoch": 0.5865457294028723, "fcm_dpo/beta": 0.763427734375, "fcm_dpo/delta": 0.034832365810871124, "fcm_dpo/margin": 1.267984390258789, "fcm_dpo/q_t": 0.3503814935684204, "grad_norm": 178.84954833984375, "learning_rate": 2.1966390475472954e-07, "logits/chosen": 0.16702687740325928, "logits/rejected": 0.1540418267250061, "logps/chosen": -76.91740417480469, "logps/ref_chosen": -74.5794677734375, "logps/ref_rejected": -79.92558288574219, "logps/rejected": -83.531494140625, "loss": 1.0483, "margin_dpo/margin_mean": 1.2679840326309204, "margin_dpo/margin_std": 2.043673515319824, "step": 388 }, { "epoch": 0.5880574452003023, "fcm_dpo/beta": 0.7542685270309448, "fcm_dpo/delta": -0.14975543320178986, "fcm_dpo/margin": 1.5020861625671387, "fcm_dpo/q_t": 0.3079478144645691, "grad_norm": 161.67514038085938, "learning_rate": 2.1835188660656265e-07, "logits/chosen": 0.17143261432647705, "logits/rejected": 0.1412460058927536, "logps/chosen": -63.959442138671875, "logps/ref_chosen": -61.624366760253906, "logps/ref_rejected": -76.50978088378906, "logps/rejected": -80.3469467163086, "loss": 0.9259, "margin_dpo/margin_mean": 1.5020864009857178, "margin_dpo/margin_std": 1.9342763423919678, "step": 389 }, { "epoch": 0.5895691609977324, "fcm_dpo/beta": 0.7468278408050537, "fcm_dpo/delta": 0.005739331711083651, "fcm_dpo/margin": 1.3320647478103638, "fcm_dpo/q_t": 0.3248102068901062, "grad_norm": 146.38316345214844, "learning_rate": 2.170407537241599e-07, "logits/chosen": 0.2075425386428833, "logits/rejected": 0.16074812412261963, "logps/chosen": -48.055908203125, "logps/ref_chosen": -45.871864318847656, "logps/ref_rejected": -61.305999755859375, "logps/rejected": -64.82211303710938, "loss": 0.9054, "margin_dpo/margin_mean": 1.3320646286010742, "margin_dpo/margin_std": 1.6783784627914429, "step": 390 }, { "epoch": 0.5910808767951625, "fcm_dpo/beta": 0.7326708436012268, "fcm_dpo/delta": -0.18802016973495483, "fcm_dpo/margin": 1.591862440109253, "fcm_dpo/q_t": 0.31688395142555237, "grad_norm": 162.29026794433594, "learning_rate": 2.1573054278272636e-07, "logits/chosen": 0.17586693167686462, "logits/rejected": 0.12610454857349396, "logps/chosen": -60.547569274902344, "logps/ref_chosen": -58.18701171875, "logps/ref_rejected": -83.63442993164062, "logps/rejected": -87.58686065673828, "loss": 1.01, "margin_dpo/margin_mean": 1.5918623208999634, "margin_dpo/margin_std": 2.3288512229919434, "step": 391 }, { "epoch": 0.5925925925925926, "fcm_dpo/beta": 0.6947846412658691, "fcm_dpo/delta": -0.25457581877708435, "fcm_dpo/margin": 1.7612890005111694, "fcm_dpo/q_t": 0.3048982322216034, "grad_norm": 162.5065155029297, "learning_rate": 2.1442129043167873e-07, "logits/chosen": 0.24103191494941711, "logits/rejected": 0.19386835396289825, "logps/chosen": -71.8067626953125, "logps/ref_chosen": -69.7445297241211, "logps/ref_rejected": -94.05877685546875, "logps/rejected": -97.8823013305664, "loss": 0.9129, "margin_dpo/margin_mean": 1.761289358139038, "margin_dpo/margin_std": 2.361508846282959, "step": 392 }, { "epoch": 0.5941043083900227, "fcm_dpo/beta": 0.6695666313171387, "fcm_dpo/delta": -0.061286166310310364, "fcm_dpo/margin": 1.5697075128555298, "fcm_dpo/q_t": 0.30320626497268677, "grad_norm": 139.36740112304688, "learning_rate": 2.131130332936195e-07, "logits/chosen": 0.20045891404151917, "logits/rejected": 0.17006459832191467, "logps/chosen": -54.7346076965332, "logps/ref_chosen": -52.33489990234375, "logps/ref_rejected": -74.33809661865234, "logps/rejected": -78.30751037597656, "loss": 0.8145, "margin_dpo/margin_mean": 1.5697076320648193, "margin_dpo/margin_std": 1.6137369871139526, "step": 393 }, { "epoch": 0.5956160241874527, "fcm_dpo/beta": 0.683825671672821, "fcm_dpo/delta": 0.04361763596534729, "fcm_dpo/margin": 1.4036848545074463, "fcm_dpo/q_t": 0.3363683223724365, "grad_norm": 153.11077880859375, "learning_rate": 2.1180580796331323e-07, "logits/chosen": 0.22308960556983948, "logits/rejected": 0.1974237710237503, "logps/chosen": -63.075340270996094, "logps/ref_chosen": -60.6761360168457, "logps/ref_rejected": -71.36074829101562, "logps/rejected": -75.16364288330078, "loss": 0.9732, "margin_dpo/margin_mean": 1.4036844968795776, "margin_dpo/margin_std": 2.0274405479431152, "step": 394 }, { "epoch": 0.5971277399848829, "fcm_dpo/beta": 0.6959511041641235, "fcm_dpo/delta": 0.1088649109005928, "fcm_dpo/margin": 1.2934165000915527, "fcm_dpo/q_t": 0.3352372646331787, "grad_norm": 158.16360473632812, "learning_rate": 2.104996510066625e-07, "logits/chosen": 0.1776614785194397, "logits/rejected": 0.117831751704216, "logps/chosen": -52.74278259277344, "logps/ref_chosen": -50.60432434082031, "logps/ref_rejected": -77.08731079101562, "logps/rejected": -80.51919555664062, "loss": 0.9631, "margin_dpo/margin_mean": 1.2934160232543945, "margin_dpo/margin_std": 1.7340922355651855, "step": 395 }, { "epoch": 0.5986394557823129, "fcm_dpo/beta": 0.6787519454956055, "fcm_dpo/delta": -0.02714592218399048, "fcm_dpo/margin": 1.4940762519836426, "fcm_dpo/q_t": 0.31732630729675293, "grad_norm": 145.00909423828125, "learning_rate": 2.0919459895968517e-07, "logits/chosen": 0.17002803087234497, "logits/rejected": 0.09879619628190994, "logps/chosen": -53.40643310546875, "logps/ref_chosen": -51.35961151123047, "logps/ref_rejected": -79.89360046386719, "logps/rejected": -83.43449401855469, "loss": 0.8661, "margin_dpo/margin_mean": 1.4940763711929321, "margin_dpo/margin_std": 1.6158559322357178, "step": 396 }, { "epoch": 0.600151171579743, "fcm_dpo/beta": 0.7337859869003296, "fcm_dpo/delta": 0.3556361794471741, "fcm_dpo/margin": 0.9081060886383057, "fcm_dpo/q_t": 0.37586018443107605, "grad_norm": 224.95326232910156, "learning_rate": 2.078906883274924e-07, "logits/chosen": 0.12839001417160034, "logits/rejected": 0.09453357756137848, "logps/chosen": -68.88770294189453, "logps/ref_chosen": -66.45622253417969, "logps/ref_rejected": -85.74736785888672, "logps/rejected": -89.08694458007812, "loss": 1.3125, "margin_dpo/margin_mean": 0.9081062078475952, "margin_dpo/margin_std": 2.2067739963531494, "step": 397 }, { "epoch": 0.6016628873771731, "fcm_dpo/beta": 0.7362475395202637, "fcm_dpo/delta": -0.06381751596927643, "fcm_dpo/margin": 1.4355955123901367, "fcm_dpo/q_t": 0.3111931383609772, "grad_norm": 139.69631958007812, "learning_rate": 2.065879555832674e-07, "logits/chosen": 0.16362245380878448, "logits/rejected": 0.11482103168964386, "logps/chosen": -51.42768859863281, "logps/ref_chosen": -49.244239807128906, "logps/ref_rejected": -75.18949127197266, "logps/rejected": -78.80854034423828, "loss": 0.8542, "margin_dpo/margin_mean": 1.4355957508087158, "margin_dpo/margin_std": 1.6850342750549316, "step": 398 }, { "epoch": 0.6031746031746031, "fcm_dpo/beta": 0.7134385108947754, "fcm_dpo/delta": -0.1938168704509735, "fcm_dpo/margin": 1.6412277221679688, "fcm_dpo/q_t": 0.32044440507888794, "grad_norm": 166.16122436523438, "learning_rate": 2.052864371672457e-07, "logits/chosen": 0.12605230510234833, "logits/rejected": 0.031615402549505234, "logps/chosen": -70.56350708007812, "logps/ref_chosen": -68.30679321289062, "logps/ref_rejected": -113.2708511352539, "logps/rejected": -117.16879272460938, "loss": 0.9153, "margin_dpo/margin_mean": 1.641228437423706, "margin_dpo/margin_std": 2.2904515266418457, "step": 399 }, { "epoch": 0.6046863189720333, "fcm_dpo/beta": 0.7214968204498291, "fcm_dpo/delta": 0.2427944540977478, "fcm_dpo/margin": 1.0737788677215576, "fcm_dpo/q_t": 0.37354040145874023, "grad_norm": 195.83456420898438, "learning_rate": 2.0398616948569493e-07, "logits/chosen": 0.18185412883758545, "logits/rejected": 0.14102932810783386, "logps/chosen": -74.3447036743164, "logps/ref_chosen": -71.62649536132812, "logps/ref_rejected": -90.98765563964844, "logps/rejected": -94.77964782714844, "loss": 1.0799, "margin_dpo/margin_mean": 1.0737799406051636, "margin_dpo/margin_std": 1.769069790840149, "step": 400 }, { "epoch": 0.6046863189720333, "eval_fcm_dpo/beta": 0.7420421838760376, "eval_logits/chosen": 0.20529566705226898, "eval_logits/rejected": 0.1674942821264267, "eval_logps/chosen": -77.24707794189453, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -83.04328918457031, "eval_loss": 0.5736358761787415, "eval_margin_dpo/margin_mean": 1.1066911220550537, "eval_margin_dpo/margin_std": 2.0390639305114746, "eval_runtime": 37.9981, "eval_samples_per_second": 60.608, "eval_steps_per_second": 1.895, "step": 400 }, { "epoch": 0.6061980347694633, "fcm_dpo/beta": 0.7018467783927917, "fcm_dpo/delta": -0.38142985105514526, "fcm_dpo/margin": 1.8959496021270752, "fcm_dpo/q_t": 0.28283387422561646, "grad_norm": 134.66848754882812, "learning_rate": 2.0268718890989752e-07, "logits/chosen": 0.19736307859420776, "logits/rejected": 0.12394518405199051, "logps/chosen": -55.94346618652344, "logps/ref_chosen": -53.72495651245117, "logps/ref_rejected": -75.06304931640625, "logps/rejected": -79.17750549316406, "loss": 0.7958, "margin_dpo/margin_mean": 1.8959496021270752, "margin_dpo/margin_std": 2.1448566913604736, "step": 401 }, { "epoch": 0.6077097505668935, "fcm_dpo/beta": 0.6789320707321167, "fcm_dpo/delta": -0.07589547336101532, "fcm_dpo/margin": 1.5725529193878174, "fcm_dpo/q_t": 0.31703245639801025, "grad_norm": 150.985595703125, "learning_rate": 2.013895317751323e-07, "logits/chosen": 0.18823865056037903, "logits/rejected": 0.155757337808609, "logps/chosen": -64.17735290527344, "logps/ref_chosen": -61.873931884765625, "logps/ref_rejected": -66.15198516845703, "logps/rejected": -70.0279541015625, "loss": 0.8937, "margin_dpo/margin_mean": 1.57255220413208, "margin_dpo/margin_std": 1.990880012512207, "step": 402 }, { "epoch": 0.6092214663643235, "fcm_dpo/beta": 0.6459471583366394, "fcm_dpo/delta": -0.33136266469955444, "fcm_dpo/margin": 1.995465636253357, "fcm_dpo/q_t": 0.2850838303565979, "grad_norm": 122.05958557128906, "learning_rate": 2.0009323437965898e-07, "logits/chosen": 0.254935085773468, "logits/rejected": 0.19566068053245544, "logps/chosen": -53.623878479003906, "logps/ref_chosen": -51.321502685546875, "logps/ref_rejected": -86.54010772705078, "logps/rejected": -90.83795166015625, "loss": 0.822, "margin_dpo/margin_mean": 1.9954662322998047, "margin_dpo/margin_std": 2.2620480060577393, "step": 403 }, { "epoch": 0.6107331821617535, "fcm_dpo/beta": 0.6391547918319702, "fcm_dpo/delta": 0.12487616389989853, "fcm_dpo/margin": 1.3852322101593018, "fcm_dpo/q_t": 0.33947524428367615, "grad_norm": 162.3168487548828, "learning_rate": 1.9879833298370237e-07, "logits/chosen": 0.15734031796455383, "logits/rejected": 0.0940411314368248, "logps/chosen": -64.49229431152344, "logps/ref_chosen": -62.26288604736328, "logps/ref_rejected": -95.19029998779297, "logps/rejected": -98.80493927001953, "loss": 0.9692, "margin_dpo/margin_mean": 1.3852319717407227, "margin_dpo/margin_std": 1.8755829334259033, "step": 404 }, { "epoch": 0.6122448979591837, "fcm_dpo/beta": 0.634131669998169, "fcm_dpo/delta": 0.03272247314453125, "fcm_dpo/margin": 1.5051655769348145, "fcm_dpo/q_t": 0.3402136266231537, "grad_norm": 127.78704071044922, "learning_rate": 1.975048638084379e-07, "logits/chosen": 0.19416359066963196, "logits/rejected": 0.16025002300739288, "logps/chosen": -52.953269958496094, "logps/ref_chosen": -50.5843391418457, "logps/ref_rejected": -65.43156433105469, "logps/rejected": -69.3056640625, "loss": 0.968, "margin_dpo/margin_mean": 1.5051651000976562, "margin_dpo/margin_std": 2.0198493003845215, "step": 405 }, { "epoch": 0.6137566137566137, "fcm_dpo/beta": 0.632315993309021, "fcm_dpo/delta": -0.13053575158119202, "fcm_dpo/margin": 1.755932331085205, "fcm_dpo/q_t": 0.299167662858963, "grad_norm": 111.40565490722656, "learning_rate": 1.9621286303497914e-07, "logits/chosen": 0.19205182790756226, "logits/rejected": 0.0876828134059906, "logps/chosen": -51.30841827392578, "logps/ref_chosen": -48.99560546875, "logps/ref_rejected": -92.47774505615234, "logps/rejected": -96.54649353027344, "loss": 0.8306, "margin_dpo/margin_mean": 1.7559325695037842, "margin_dpo/margin_std": 1.8786392211914062, "step": 406 }, { "epoch": 0.6152683295540439, "fcm_dpo/beta": 0.657385528087616, "fcm_dpo/delta": 0.21200606226921082, "fcm_dpo/margin": 1.224708914756775, "fcm_dpo/q_t": 0.3558533787727356, "grad_norm": 209.7600555419922, "learning_rate": 1.9492236680336483e-07, "logits/chosen": 0.13539977371692657, "logits/rejected": 0.08569268882274628, "logps/chosen": -92.08710479736328, "logps/ref_chosen": -89.40056610107422, "logps/ref_rejected": -99.28775024414062, "logps/rejected": -103.19900512695312, "loss": 1.0517, "margin_dpo/margin_mean": 1.224708914756775, "margin_dpo/margin_std": 2.031216621398926, "step": 407 }, { "epoch": 0.6167800453514739, "fcm_dpo/beta": 0.6533396244049072, "fcm_dpo/delta": -0.2026488482952118, "fcm_dpo/margin": 1.8033205270767212, "fcm_dpo/q_t": 0.2949105501174927, "grad_norm": 119.43750762939453, "learning_rate": 1.9363341121154895e-07, "logits/chosen": 0.1879599392414093, "logits/rejected": 0.1291370391845703, "logps/chosen": -56.83644485473633, "logps/ref_chosen": -54.70391845703125, "logps/ref_rejected": -73.98648834228516, "logps/rejected": -77.92233276367188, "loss": 0.8189, "margin_dpo/margin_mean": 1.8033205270767212, "margin_dpo/margin_std": 1.967519998550415, "step": 408 }, { "epoch": 0.618291761148904, "fcm_dpo/beta": 0.675295352935791, "fcm_dpo/delta": 0.34728795289993286, "fcm_dpo/margin": 0.9984000325202942, "fcm_dpo/q_t": 0.3867419362068176, "grad_norm": 168.26300048828125, "learning_rate": 1.9234603231438994e-07, "logits/chosen": 0.1908985674381256, "logits/rejected": 0.19027680158615112, "logps/chosen": -64.74275970458984, "logps/ref_chosen": -62.11822509765625, "logps/ref_rejected": -61.933509826660156, "logps/rejected": -65.55644226074219, "loss": 1.1776, "margin_dpo/margin_mean": 0.9983994960784912, "margin_dpo/margin_std": 2.0483100414276123, "step": 409 }, { "epoch": 0.6198034769463341, "fcm_dpo/beta": 0.6897294521331787, "fcm_dpo/delta": 0.15048189461231232, "fcm_dpo/margin": 1.2474337816238403, "fcm_dpo/q_t": 0.3310784697532654, "grad_norm": 171.6697998046875, "learning_rate": 1.9106026612264315e-07, "logits/chosen": 0.19930198788642883, "logits/rejected": 0.1775505691766739, "logps/chosen": -64.07925415039062, "logps/ref_chosen": -61.80266189575195, "logps/ref_rejected": -76.60002136230469, "logps/rejected": -80.12403869628906, "loss": 0.9295, "margin_dpo/margin_mean": 1.2474339008331299, "margin_dpo/margin_std": 1.4843469858169556, "step": 410 }, { "epoch": 0.6213151927437641, "fcm_dpo/beta": 0.6963478326797485, "fcm_dpo/delta": -0.008774511516094208, "fcm_dpo/margin": 1.4437086582183838, "fcm_dpo/q_t": 0.3453066349029541, "grad_norm": 181.04974365234375, "learning_rate": 1.8977614860195296e-07, "logits/chosen": 0.15856947004795074, "logits/rejected": 0.1135367900133133, "logps/chosen": -57.06800842285156, "logps/ref_chosen": -54.44539260864258, "logps/ref_rejected": -74.5650863647461, "logps/rejected": -78.63140869140625, "loss": 1.0746, "margin_dpo/margin_mean": 1.4437092542648315, "margin_dpo/margin_std": 2.3472909927368164, "step": 411 }, { "epoch": 0.6228269085411943, "fcm_dpo/beta": 0.6941448450088501, "fcm_dpo/delta": -0.037317849695682526, "fcm_dpo/margin": 1.4864020347595215, "fcm_dpo/q_t": 0.3218909502029419, "grad_norm": 150.66571044921875, "learning_rate": 1.8849371567184662e-07, "logits/chosen": 0.15843887627124786, "logits/rejected": 0.1075374186038971, "logps/chosen": -58.101863861083984, "logps/ref_chosen": -55.248085021972656, "logps/ref_rejected": -68.96623229980469, "logps/rejected": -73.30641174316406, "loss": 0.9281, "margin_dpo/margin_mean": 1.4864020347595215, "margin_dpo/margin_std": 1.928009033203125, "step": 412 }, { "epoch": 0.6243386243386243, "fcm_dpo/beta": 0.7043063640594482, "fcm_dpo/delta": 0.07277508080005646, "fcm_dpo/margin": 1.3262073993682861, "fcm_dpo/q_t": 0.3529996871948242, "grad_norm": 191.52565002441406, "learning_rate": 1.872130032047302e-07, "logits/chosen": 0.07102949917316437, "logits/rejected": 0.04926881939172745, "logps/chosen": -71.4292984008789, "logps/ref_chosen": -68.72074890136719, "logps/ref_rejected": -78.76539611816406, "logps/rejected": -82.80016326904297, "loss": 1.102, "margin_dpo/margin_mean": 1.3262066841125488, "margin_dpo/margin_std": 2.299673557281494, "step": 413 }, { "epoch": 0.6258503401360545, "fcm_dpo/beta": 0.6957840919494629, "fcm_dpo/delta": -0.08349283784627914, "fcm_dpo/margin": 1.542493462562561, "fcm_dpo/q_t": 0.3177918791770935, "grad_norm": 144.82449340820312, "learning_rate": 1.8593404702488436e-07, "logits/chosen": 0.17691153287887573, "logits/rejected": 0.12794461846351624, "logps/chosen": -56.64301300048828, "logps/ref_chosen": -54.138214111328125, "logps/ref_rejected": -74.65741729736328, "logps/rejected": -78.7047119140625, "loss": 0.9185, "margin_dpo/margin_mean": 1.542493462562561, "margin_dpo/margin_std": 2.0337252616882324, "step": 414 }, { "epoch": 0.6273620559334845, "fcm_dpo/beta": 0.7059125900268555, "fcm_dpo/delta": 0.12531909346580505, "fcm_dpo/margin": 1.2547566890716553, "fcm_dpo/q_t": 0.3529791235923767, "grad_norm": 175.94845581054688, "learning_rate": 1.846568829074628e-07, "logits/chosen": 0.17382574081420898, "logits/rejected": 0.15772220492362976, "logps/chosen": -58.64318084716797, "logps/ref_chosen": -55.91856002807617, "logps/ref_rejected": -61.747703552246094, "logps/rejected": -65.72708129882812, "loss": 1.1261, "margin_dpo/margin_mean": 1.2547566890716553, "margin_dpo/margin_std": 2.235924243927002, "step": 415 }, { "epoch": 0.6288737717309146, "fcm_dpo/beta": 0.7500655651092529, "fcm_dpo/delta": 0.13376402854919434, "fcm_dpo/margin": 1.1456831693649292, "fcm_dpo/q_t": 0.356852650642395, "grad_norm": 205.03404235839844, "learning_rate": 1.8338154657749128e-07, "logits/chosen": 0.18582022190093994, "logits/rejected": 0.1482175588607788, "logps/chosen": -57.20629119873047, "logps/ref_chosen": -54.72308349609375, "logps/ref_rejected": -69.17388916015625, "logps/rejected": -72.80278015136719, "loss": 1.154, "margin_dpo/margin_mean": 1.1456834077835083, "margin_dpo/margin_std": 2.031163454055786, "step": 416 }, { "epoch": 0.6303854875283447, "fcm_dpo/beta": 0.7249786853790283, "fcm_dpo/delta": -0.24252735078334808, "fcm_dpo/margin": 1.6672532558441162, "fcm_dpo/q_t": 0.29793938994407654, "grad_norm": 194.7884521484375, "learning_rate": 1.8210807370886849e-07, "logits/chosen": 0.24135205149650574, "logits/rejected": 0.19347181916236877, "logps/chosen": -59.584938049316406, "logps/ref_chosen": -56.791259765625, "logps/ref_rejected": -68.7791748046875, "logps/rejected": -73.2401123046875, "loss": 0.9246, "margin_dpo/margin_mean": 1.6672537326812744, "margin_dpo/margin_std": 2.092496395111084, "step": 417 }, { "epoch": 0.6318972033257747, "fcm_dpo/beta": 0.7283662557601929, "fcm_dpo/delta": 0.3391045331954956, "fcm_dpo/margin": 0.9409202337265015, "fcm_dpo/q_t": 0.3989126980304718, "grad_norm": 244.12945556640625, "learning_rate": 1.8083649992336825e-07, "logits/chosen": 0.19916735589504242, "logits/rejected": 0.2042698860168457, "logps/chosen": -72.2450942993164, "logps/ref_chosen": -69.10798645019531, "logps/ref_rejected": -75.09132385253906, "logps/rejected": -79.16935729980469, "loss": 1.3312, "margin_dpo/margin_mean": 0.9409199357032776, "margin_dpo/margin_std": 2.4089250564575195, "step": 418 }, { "epoch": 0.6334089191232048, "fcm_dpo/beta": 0.7232016324996948, "fcm_dpo/delta": -0.157709002494812, "fcm_dpo/margin": 1.570444941520691, "fcm_dpo/q_t": 0.3125431537628174, "grad_norm": 159.44833374023438, "learning_rate": 1.7956686078964255e-07, "logits/chosen": 0.09053687751293182, "logits/rejected": 0.04949123412370682, "logps/chosen": -60.52400588989258, "logps/ref_chosen": -58.1717643737793, "logps/ref_rejected": -71.67066955566406, "logps/rejected": -75.59335327148438, "loss": 0.9202, "margin_dpo/margin_mean": 1.5704445838928223, "margin_dpo/margin_std": 2.1407108306884766, "step": 419 }, { "epoch": 0.6349206349206349, "fcm_dpo/beta": 0.7380191683769226, "fcm_dpo/delta": 0.1838487833738327, "fcm_dpo/margin": 1.1262156963348389, "fcm_dpo/q_t": 0.37532341480255127, "grad_norm": 189.95645141601562, "learning_rate": 1.782991918222275e-07, "logits/chosen": 0.15650036931037903, "logits/rejected": 0.11561809480190277, "logps/chosen": -60.085960388183594, "logps/ref_chosen": -57.05351257324219, "logps/ref_rejected": -62.670982360839844, "logps/rejected": -66.82964324951172, "loss": 1.2387, "margin_dpo/margin_mean": 1.126215934753418, "margin_dpo/margin_std": 2.343519687652588, "step": 420 }, { "epoch": 0.636432350718065, "fcm_dpo/beta": 0.7731253504753113, "fcm_dpo/delta": 0.1831362247467041, "fcm_dpo/margin": 1.076323390007019, "fcm_dpo/q_t": 0.3773455023765564, "grad_norm": 191.1004180908203, "learning_rate": 1.7703352848054887e-07, "logits/chosen": 0.13414627313613892, "logits/rejected": 0.09018285572528839, "logps/chosen": -60.364906311035156, "logps/ref_chosen": -57.32324981689453, "logps/ref_rejected": -75.33782958984375, "logps/rejected": -79.455810546875, "loss": 1.3191, "margin_dpo/margin_mean": 1.0763235092163086, "margin_dpo/margin_std": 2.4303441047668457, "step": 421 }, { "epoch": 0.6379440665154951, "fcm_dpo/beta": 0.7678795456886292, "fcm_dpo/delta": -0.15709903836250305, "fcm_dpo/margin": 1.4837114810943604, "fcm_dpo/q_t": 0.3305337727069855, "grad_norm": 208.15638732910156, "learning_rate": 1.7576990616793137e-07, "logits/chosen": 0.18905231356620789, "logits/rejected": 0.15734535455703735, "logps/chosen": -69.5533447265625, "logps/ref_chosen": -67.05757141113281, "logps/ref_rejected": -72.12803649902344, "logps/rejected": -76.10751342773438, "loss": 1.0168, "margin_dpo/margin_mean": 1.4837113618850708, "margin_dpo/margin_std": 2.3469998836517334, "step": 422 }, { "epoch": 0.6394557823129252, "fcm_dpo/beta": 0.7414518594741821, "fcm_dpo/delta": -0.12381698191165924, "fcm_dpo/margin": 1.4967904090881348, "fcm_dpo/q_t": 0.3256801962852478, "grad_norm": 163.99639892578125, "learning_rate": 1.745083602306071e-07, "logits/chosen": 0.17420369386672974, "logits/rejected": 0.12361004948616028, "logps/chosen": -56.651309967041016, "logps/ref_chosen": -54.06167221069336, "logps/ref_rejected": -76.64092254638672, "logps/rejected": -80.72735595703125, "loss": 1.0145, "margin_dpo/margin_mean": 1.4967900514602661, "margin_dpo/margin_std": 2.2581710815429688, "step": 423 }, { "epoch": 0.6409674981103552, "fcm_dpo/beta": 0.7214820384979248, "fcm_dpo/delta": -0.13758614659309387, "fcm_dpo/margin": 1.5547800064086914, "fcm_dpo/q_t": 0.3215063512325287, "grad_norm": 178.93624877929688, "learning_rate": 1.7324892595672804e-07, "logits/chosen": 0.12239620089530945, "logits/rejected": 0.09341743588447571, "logps/chosen": -56.10038757324219, "logps/ref_chosen": -53.60887145996094, "logps/ref_rejected": -79.2139892578125, "logps/rejected": -83.26029205322266, "loss": 0.9368, "margin_dpo/margin_mean": 1.554780125617981, "margin_dpo/margin_std": 2.143885612487793, "step": 424 }, { "epoch": 0.6424792139077853, "fcm_dpo/beta": 0.7137551307678223, "fcm_dpo/delta": -0.05110887810587883, "fcm_dpo/margin": 1.4651919603347778, "fcm_dpo/q_t": 0.32126736640930176, "grad_norm": 145.7388153076172, "learning_rate": 1.7199163857537824e-07, "logits/chosen": 0.18603307008743286, "logits/rejected": 0.16199590265750885, "logps/chosen": -60.9288330078125, "logps/ref_chosen": -58.41468048095703, "logps/ref_rejected": -66.59054565429688, "logps/rejected": -70.56989288330078, "loss": 0.9247, "margin_dpo/margin_mean": 1.4651916027069092, "margin_dpo/margin_std": 1.9558470249176025, "step": 425 }, { "epoch": 0.6439909297052154, "fcm_dpo/beta": 0.7687985301017761, "fcm_dpo/delta": 0.5285735130310059, "fcm_dpo/margin": 0.6476625204086304, "fcm_dpo/q_t": 0.4075689911842346, "grad_norm": 223.34030151367188, "learning_rate": 1.7073653325558828e-07, "logits/chosen": 0.14943718910217285, "logits/rejected": 0.15013810992240906, "logps/chosen": -74.68634033203125, "logps/ref_chosen": -71.70822143554688, "logps/ref_rejected": -73.57725524902344, "logps/rejected": -77.20303344726562, "loss": 1.3908, "margin_dpo/margin_mean": 0.6476625800132751, "margin_dpo/margin_std": 1.9961354732513428, "step": 426 }, { "epoch": 0.6455026455026455, "fcm_dpo/beta": 0.7852897644042969, "fcm_dpo/delta": -0.0677163228392601, "fcm_dpo/margin": 1.3504165410995483, "fcm_dpo/q_t": 0.33633100986480713, "grad_norm": 163.32342529296875, "learning_rate": 1.6948364510535218e-07, "logits/chosen": 0.20361992716789246, "logits/rejected": 0.16401880979537964, "logps/chosen": -61.42845916748047, "logps/ref_chosen": -58.64276885986328, "logps/ref_rejected": -86.25437927246094, "logps/rejected": -90.39048767089844, "loss": 1.0482, "margin_dpo/margin_mean": 1.350417137145996, "margin_dpo/margin_std": 2.189502716064453, "step": 427 }, { "epoch": 0.6470143613000756, "fcm_dpo/beta": 0.7506411671638489, "fcm_dpo/delta": -0.2936919033527374, "fcm_dpo/margin": 1.6753690242767334, "fcm_dpo/q_t": 0.29713624715805054, "grad_norm": 168.94140625, "learning_rate": 1.6823300917064458e-07, "logits/chosen": 0.1335376501083374, "logits/rejected": 0.09132213890552521, "logps/chosen": -69.21764373779297, "logps/ref_chosen": -66.5960464477539, "logps/ref_rejected": -82.3941650390625, "logps/rejected": -86.69113159179688, "loss": 0.8538, "margin_dpo/margin_mean": 1.6753690242767334, "margin_dpo/margin_std": 2.044569969177246, "step": 428 }, { "epoch": 0.6485260770975056, "fcm_dpo/beta": 0.730757474899292, "fcm_dpo/delta": -0.06324490904808044, "fcm_dpo/margin": 1.4457752704620361, "fcm_dpo/q_t": 0.3198990225791931, "grad_norm": 177.0709991455078, "learning_rate": 1.669846604344412e-07, "logits/chosen": 0.13211305439472198, "logits/rejected": 0.13196702301502228, "logps/chosen": -59.602317810058594, "logps/ref_chosen": -57.00970458984375, "logps/ref_rejected": -59.86549377441406, "logps/rejected": -63.90388488769531, "loss": 0.969, "margin_dpo/margin_mean": 1.4457753896713257, "margin_dpo/margin_std": 2.0829176902770996, "step": 429 }, { "epoch": 0.6500377928949358, "fcm_dpo/beta": 0.733278751373291, "fcm_dpo/delta": 0.05039960518479347, "fcm_dpo/margin": 1.3011215925216675, "fcm_dpo/q_t": 0.3425254225730896, "grad_norm": 174.14768981933594, "learning_rate": 1.6573863381573954e-07, "logits/chosen": 0.08477595448493958, "logits/rejected": 0.0728166401386261, "logps/chosen": -61.990821838378906, "logps/ref_chosen": -59.563194274902344, "logps/ref_rejected": -70.52289581298828, "logps/rejected": -74.25164794921875, "loss": 1.0004, "margin_dpo/margin_mean": 1.3011209964752197, "margin_dpo/margin_std": 1.9769561290740967, "step": 430 }, { "epoch": 0.6515495086923658, "fcm_dpo/beta": 0.7297165393829346, "fcm_dpo/delta": 0.036733031272888184, "fcm_dpo/margin": 1.3228065967559814, "fcm_dpo/q_t": 0.3405199646949768, "grad_norm": 171.15907287597656, "learning_rate": 1.6449496416858282e-07, "logits/chosen": 0.17870327830314636, "logits/rejected": 0.14549797773361206, "logps/chosen": -52.56354904174805, "logps/ref_chosen": -50.20032501220703, "logps/ref_rejected": -77.81680297851562, "logps/rejected": -81.50283813476562, "loss": 1.0051, "margin_dpo/margin_mean": 1.3228061199188232, "margin_dpo/margin_std": 2.025378704071045, "step": 431 }, { "epoch": 0.6530612244897959, "fcm_dpo/beta": 0.7157025337219238, "fcm_dpo/delta": -0.10452289134263992, "fcm_dpo/margin": 1.518718957901001, "fcm_dpo/q_t": 0.32164376974105835, "grad_norm": 169.30746459960938, "learning_rate": 1.632536862810844e-07, "logits/chosen": 0.17181310057640076, "logits/rejected": 0.1360006034374237, "logps/chosen": -64.20341491699219, "logps/ref_chosen": -61.662757873535156, "logps/ref_rejected": -83.94496154785156, "logps/rejected": -88.00434112548828, "loss": 0.9613, "margin_dpo/margin_mean": 1.5187186002731323, "margin_dpo/margin_std": 2.0636777877807617, "step": 432 }, { "epoch": 0.654572940287226, "fcm_dpo/beta": 0.6925790309906006, "fcm_dpo/delta": -0.28451234102249146, "fcm_dpo/margin": 1.802809476852417, "fcm_dpo/q_t": 0.3154519498348236, "grad_norm": 145.2095947265625, "learning_rate": 1.6201483487445515e-07, "logits/chosen": 0.24505794048309326, "logits/rejected": 0.23090487718582153, "logps/chosen": -66.46299743652344, "logps/ref_chosen": -63.72917938232422, "logps/ref_rejected": -65.8391342163086, "logps/rejected": -70.3757553100586, "loss": 0.9081, "margin_dpo/margin_mean": 1.802809476852417, "margin_dpo/margin_std": 2.475156784057617, "step": 433 }, { "epoch": 0.656084656084656, "fcm_dpo/beta": 0.659028947353363, "fcm_dpo/delta": -0.13485188782215118, "fcm_dpo/margin": 1.6889315843582153, "fcm_dpo/q_t": 0.31248465180397034, "grad_norm": 130.46426391601562, "learning_rate": 1.6077844460203204e-07, "logits/chosen": 0.21205420792102814, "logits/rejected": 0.16366201639175415, "logps/chosen": -50.280784606933594, "logps/ref_chosen": -47.97331619262695, "logps/ref_rejected": -72.51132202148438, "logps/rejected": -76.50772857666016, "loss": 0.9792, "margin_dpo/margin_mean": 1.6889313459396362, "margin_dpo/margin_std": 2.39209246635437, "step": 434 }, { "epoch": 0.6575963718820862, "fcm_dpo/beta": 0.6766383647918701, "fcm_dpo/delta": 0.04958092421293259, "fcm_dpo/margin": 1.4080250263214111, "fcm_dpo/q_t": 0.3341384530067444, "grad_norm": 151.441162109375, "learning_rate": 1.5954455004830878e-07, "logits/chosen": 0.22931721806526184, "logits/rejected": 0.1980956494808197, "logps/chosen": -59.903053283691406, "logps/ref_chosen": -57.06024932861328, "logps/ref_rejected": -71.69146728515625, "logps/rejected": -75.94229888916016, "loss": 0.9779, "margin_dpo/margin_mean": 1.4080252647399902, "margin_dpo/margin_std": 2.0040295124053955, "step": 435 }, { "epoch": 0.6591080876795162, "fcm_dpo/beta": 0.6710116863250732, "fcm_dpo/delta": -0.004951075650751591, "fcm_dpo/margin": 1.4969241619110107, "fcm_dpo/q_t": 0.3257167935371399, "grad_norm": 149.94261169433594, "learning_rate": 1.5831318572796847e-07, "logits/chosen": 0.16502049565315247, "logits/rejected": 0.1188136488199234, "logps/chosen": -58.86479949951172, "logps/ref_chosen": -56.158050537109375, "logps/ref_rejected": -67.63787841796875, "logps/rejected": -71.841552734375, "loss": 0.9499, "margin_dpo/margin_mean": 1.4969241619110107, "margin_dpo/margin_std": 2.013727903366089, "step": 436 }, { "epoch": 0.6606198034769464, "fcm_dpo/beta": 0.6555310487747192, "fcm_dpo/delta": 0.013455048203468323, "fcm_dpo/margin": 1.48863685131073, "fcm_dpo/q_t": 0.3591005504131317, "grad_norm": 169.48338317871094, "learning_rate": 1.5708438608491815e-07, "logits/chosen": 0.15093836188316345, "logits/rejected": 0.05515362694859505, "logps/chosen": -59.8901252746582, "logps/ref_chosen": -56.98578643798828, "logps/ref_rejected": -85.61524963378906, "logps/rejected": -90.00823974609375, "loss": 1.1728, "margin_dpo/margin_mean": 1.4886366128921509, "margin_dpo/margin_std": 2.654061794281006, "step": 437 }, { "epoch": 0.6621315192743764, "fcm_dpo/beta": 0.6835014820098877, "fcm_dpo/delta": 0.04397985339164734, "fcm_dpo/margin": 1.4018324613571167, "fcm_dpo/q_t": 0.3333927392959595, "grad_norm": 121.23139190673828, "learning_rate": 1.558581854913253e-07, "logits/chosen": 0.19322752952575684, "logits/rejected": 0.14423127472400665, "logps/chosen": -43.83991241455078, "logps/ref_chosen": -41.27777862548828, "logps/ref_rejected": -65.33840942382812, "logps/rejected": -69.3023681640625, "loss": 0.9826, "margin_dpo/margin_mean": 1.401832938194275, "margin_dpo/margin_std": 1.9659230709075928, "step": 438 }, { "epoch": 0.6636432350718064, "fcm_dpo/beta": 0.6828280091285706, "fcm_dpo/delta": -0.04300057888031006, "fcm_dpo/margin": 1.5173701047897339, "fcm_dpo/q_t": 0.31167930364608765, "grad_norm": 186.42234802246094, "learning_rate": 1.5463461824665658e-07, "logits/chosen": 0.11366529762744904, "logits/rejected": 0.09049699455499649, "logps/chosen": -83.722900390625, "logps/ref_chosen": -81.41764831542969, "logps/ref_rejected": -94.72309875488281, "logps/rejected": -98.54571533203125, "loss": 0.9309, "margin_dpo/margin_mean": 1.5173696279525757, "margin_dpo/margin_std": 1.9623498916625977, "step": 439 }, { "epoch": 0.6651549508692366, "fcm_dpo/beta": 0.65775465965271, "fcm_dpo/delta": -0.16628237068653107, "fcm_dpo/margin": 1.7444255352020264, "fcm_dpo/q_t": 0.30622392892837524, "grad_norm": 129.40713500976562, "learning_rate": 1.534137185767178e-07, "logits/chosen": 0.1228909119963646, "logits/rejected": 0.04697669297456741, "logps/chosen": -44.876808166503906, "logps/ref_chosen": -42.538185119628906, "logps/ref_rejected": -69.78813934326172, "logps/rejected": -73.87118530273438, "loss": 0.8618, "margin_dpo/margin_mean": 1.744425654411316, "margin_dpo/margin_std": 2.120790481567383, "step": 440 }, { "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.6342558860778809, "fcm_dpo/delta": -0.15066742897033691, "fcm_dpo/margin": 1.7861425876617432, "fcm_dpo/q_t": 0.292005717754364, "grad_norm": 122.53746032714844, "learning_rate": 1.521955206326976e-07, "logits/chosen": 0.14506568014621735, "logits/rejected": 0.07649335265159607, "logps/chosen": -59.81481170654297, "logps/ref_chosen": -57.593223571777344, "logps/ref_rejected": -84.82878875732422, "logps/rejected": -88.83651733398438, "loss": 0.7795, "margin_dpo/margin_mean": 1.7861430644989014, "margin_dpo/margin_std": 1.8194975852966309, "step": 441 }, { "epoch": 0.6681783824640968, "fcm_dpo/beta": 0.6411465406417847, "fcm_dpo/delta": 0.09939359128475189, "fcm_dpo/margin": 1.418702483177185, "fcm_dpo/q_t": 0.34212759137153625, "grad_norm": 168.3711395263672, "learning_rate": 1.5098005849021078e-07, "logits/chosen": 0.19873833656311035, "logits/rejected": 0.16300469636917114, "logps/chosen": -70.19367980957031, "logps/ref_chosen": -67.46121978759766, "logps/ref_rejected": -89.0693588256836, "logps/rejected": -93.22052001953125, "loss": 0.9856, "margin_dpo/margin_mean": 1.418702483177185, "margin_dpo/margin_std": 2.081653594970703, "step": 442 }, { "epoch": 0.6696900982615268, "fcm_dpo/beta": 0.6237589120864868, "fcm_dpo/delta": -0.24319452047348022, "fcm_dpo/margin": 1.9465469121932983, "fcm_dpo/q_t": 0.29789623618125916, "grad_norm": 127.22098541259766, "learning_rate": 1.4976736614834662e-07, "logits/chosen": 0.18308596312999725, "logits/rejected": 0.12943433225154877, "logps/chosen": -57.22307586669922, "logps/ref_chosen": -54.79610061645508, "logps/ref_rejected": -77.80781555175781, "logps/rejected": -82.18133544921875, "loss": 0.9041, "margin_dpo/margin_mean": 1.946547269821167, "margin_dpo/margin_std": 2.4835057258605957, "step": 443 }, { "epoch": 0.671201814058957, "fcm_dpo/beta": 0.661973237991333, "fcm_dpo/delta": 0.5227335691452026, "fcm_dpo/margin": 0.7623451948165894, "fcm_dpo/q_t": 0.4185434579849243, "grad_norm": 196.5355224609375, "learning_rate": 1.4855747752871654e-07, "logits/chosen": 0.18811815977096558, "logits/rejected": 0.1319284737110138, "logps/chosen": -61.6188850402832, "logps/ref_chosen": -58.749061584472656, "logps/ref_rejected": -86.87396240234375, "logps/rejected": -90.50614166259766, "loss": 1.3947, "margin_dpo/margin_mean": 0.7623450756072998, "margin_dpo/margin_std": 2.2845544815063477, "step": 444 }, { "epoch": 0.672713529856387, "fcm_dpo/beta": 0.6808522343635559, "fcm_dpo/delta": -0.0005050599575042725, "fcm_dpo/margin": 1.4694095849990845, "fcm_dpo/q_t": 0.3312973380088806, "grad_norm": 179.10009765625, "learning_rate": 1.473504264745062e-07, "logits/chosen": 0.1777781844139099, "logits/rejected": 0.16238978505134583, "logps/chosen": -64.01307678222656, "logps/ref_chosen": -60.91743850708008, "logps/ref_rejected": -71.5637435913086, "logps/rejected": -76.1287841796875, "loss": 1.0013, "margin_dpo/margin_mean": 1.4694093465805054, "margin_dpo/margin_std": 2.1590332984924316, "step": 445 }, { "epoch": 0.674225245653817, "fcm_dpo/beta": 0.6526922583580017, "fcm_dpo/delta": -0.4451631009578705, "fcm_dpo/margin": 2.119361162185669, "fcm_dpo/q_t": 0.25671201944351196, "grad_norm": 107.62913513183594, "learning_rate": 1.461462467495284e-07, "logits/chosen": 0.19539491832256317, "logits/rejected": 0.13630658388137817, "logps/chosen": -51.429866790771484, "logps/ref_chosen": -48.79924774169922, "logps/ref_rejected": -71.8719482421875, "logps/rejected": -76.62193298339844, "loss": 0.6997, "margin_dpo/margin_mean": 2.1193606853485107, "margin_dpo/margin_std": 1.913917064666748, "step": 446 }, { "epoch": 0.6757369614512472, "fcm_dpo/beta": 0.6015419363975525, "fcm_dpo/delta": -0.2893209457397461, "fcm_dpo/margin": 2.0844779014587402, "fcm_dpo/q_t": 0.27232983708381653, "grad_norm": 112.18714141845703, "learning_rate": 1.4494497203727843e-07, "logits/chosen": 0.1319853812456131, "logits/rejected": 0.05933520570397377, "logps/chosen": -55.8741569519043, "logps/ref_chosen": -53.682716369628906, "logps/ref_rejected": -88.17315673828125, "logps/rejected": -92.44908142089844, "loss": 0.8532, "margin_dpo/margin_mean": 2.084477424621582, "margin_dpo/margin_std": 2.4492878913879395, "step": 447 }, { "epoch": 0.6772486772486772, "fcm_dpo/beta": 0.5884913206100464, "fcm_dpo/delta": -8.67573544383049e-06, "fcm_dpo/margin": 1.6992573738098145, "fcm_dpo/q_t": 0.3187348246574402, "grad_norm": 115.34869384765625, "learning_rate": 1.4374663593999256e-07, "logits/chosen": 0.17938536405563354, "logits/rejected": 0.14003482460975647, "logps/chosen": -56.258331298828125, "logps/ref_chosen": -53.75125503540039, "logps/ref_rejected": -77.17623901367188, "logps/rejected": -81.382568359375, "loss": 0.8966, "margin_dpo/margin_mean": 1.6992576122283936, "margin_dpo/margin_std": 2.116860866546631, "step": 448 }, { "epoch": 0.6787603930461074, "fcm_dpo/beta": 0.6253612041473389, "fcm_dpo/delta": 0.423923134803772, "fcm_dpo/margin": 0.9630190134048462, "fcm_dpo/q_t": 0.38473382592201233, "grad_norm": 156.52899169921875, "learning_rate": 1.4255127197770707e-07, "logits/chosen": 0.07655443251132965, "logits/rejected": 0.061458148062229156, "logps/chosen": -78.82083129882812, "logps/ref_chosen": -75.82737731933594, "logps/ref_rejected": -82.20687866210938, "logps/rejected": -86.16334533691406, "loss": 1.1046, "margin_dpo/margin_mean": 0.9630191326141357, "margin_dpo/margin_std": 1.6913801431655884, "step": 449 }, { "epoch": 0.6802721088435374, "fcm_dpo/beta": 0.6548283100128174, "fcm_dpo/delta": 0.12034881114959717, "fcm_dpo/margin": 1.3581256866455078, "fcm_dpo/q_t": 0.35161659121513367, "grad_norm": 150.02163696289062, "learning_rate": 1.4135891358732205e-07, "logits/chosen": 0.21886947751045227, "logits/rejected": 0.1426752656698227, "logps/chosen": -49.74790954589844, "logps/ref_chosen": -47.11572265625, "logps/ref_rejected": -78.7546615600586, "logps/rejected": -82.7449722290039, "loss": 1.0203, "margin_dpo/margin_mean": 1.3581254482269287, "margin_dpo/margin_std": 2.10986590385437, "step": 450 }, { "epoch": 0.6817838246409675, "fcm_dpo/beta": 0.6705623865127563, "fcm_dpo/delta": 0.17599515616893768, "fcm_dpo/margin": 1.2513468265533447, "fcm_dpo/q_t": 0.3569420576095581, "grad_norm": 179.85128784179688, "learning_rate": 1.4016959412166437e-07, "logits/chosen": 0.18319055438041687, "logits/rejected": 0.14441323280334473, "logps/chosen": -66.11190795898438, "logps/ref_chosen": -63.350440979003906, "logps/ref_rejected": -76.28530883789062, "logps/rejected": -80.29811096191406, "loss": 1.0998, "margin_dpo/margin_mean": 1.251347541809082, "margin_dpo/margin_std": 2.137026786804199, "step": 451 }, { "epoch": 0.6832955404383976, "fcm_dpo/beta": 0.6785616874694824, "fcm_dpo/delta": -0.009446687065064907, "fcm_dpo/margin": 1.486222743988037, "fcm_dpo/q_t": 0.32352200150489807, "grad_norm": 165.91653442382812, "learning_rate": 1.3898334684855645e-07, "logits/chosen": 0.13353146612644196, "logits/rejected": 0.08063468337059021, "logps/chosen": -58.241943359375, "logps/ref_chosen": -55.58583450317383, "logps/ref_rejected": -77.68738555908203, "logps/rejected": -81.8297119140625, "loss": 0.9588, "margin_dpo/margin_mean": 1.4862233400344849, "margin_dpo/margin_std": 2.007995128631592, "step": 452 }, { "epoch": 0.6848072562358276, "fcm_dpo/beta": 0.6866965293884277, "fcm_dpo/delta": 0.07683775573968887, "fcm_dpo/margin": 1.354539394378662, "fcm_dpo/q_t": 0.36106228828430176, "grad_norm": 152.8515167236328, "learning_rate": 1.3780020494988445e-07, "logits/chosen": 0.12297318130731583, "logits/rejected": 0.09657086431980133, "logps/chosen": -64.26089477539062, "logps/ref_chosen": -61.778202056884766, "logps/ref_rejected": -71.51403045654297, "logps/rejected": -75.35125732421875, "loss": 1.083, "margin_dpo/margin_mean": 1.354539394378662, "margin_dpo/margin_std": 2.2785511016845703, "step": 453 }, { "epoch": 0.6863189720332578, "fcm_dpo/beta": 0.6685348749160767, "fcm_dpo/delta": -0.19035013020038605, "fcm_dpo/margin": 1.7468868494033813, "fcm_dpo/q_t": 0.3080099821090698, "grad_norm": 127.66053771972656, "learning_rate": 1.366202015206706e-07, "logits/chosen": 0.16820810735225677, "logits/rejected": 0.13530485332012177, "logps/chosen": -53.909812927246094, "logps/ref_chosen": -51.59515380859375, "logps/ref_rejected": -63.96732711791992, "logps/rejected": -68.02886962890625, "loss": 0.9374, "margin_dpo/margin_mean": 1.7468867301940918, "margin_dpo/margin_std": 2.297898292541504, "step": 454 }, { "epoch": 0.6878306878306878, "fcm_dpo/beta": 0.6465753316879272, "fcm_dpo/delta": -0.1078774556517601, "fcm_dpo/margin": 1.6916618347167969, "fcm_dpo/q_t": 0.31326356530189514, "grad_norm": 150.88607788085938, "learning_rate": 1.354433695681474e-07, "logits/chosen": 0.040207911282777786, "logits/rejected": 0.007178250700235367, "logps/chosen": -73.28399658203125, "logps/ref_chosen": -70.65170288085938, "logps/ref_rejected": -77.44276428222656, "logps/rejected": -81.7667236328125, "loss": 0.8921, "margin_dpo/margin_mean": 1.6916615962982178, "margin_dpo/margin_std": 2.1551051139831543, "step": 455 }, { "epoch": 0.6893424036281179, "fcm_dpo/beta": 0.6534501910209656, "fcm_dpo/delta": 0.01624855026602745, "fcm_dpo/margin": 1.5070427656173706, "fcm_dpo/q_t": 0.32177361845970154, "grad_norm": 149.37515258789062, "learning_rate": 1.3426974201083439e-07, "logits/chosen": 0.11166486144065857, "logits/rejected": 0.07302643358707428, "logps/chosen": -59.19749450683594, "logps/ref_chosen": -56.398284912109375, "logps/ref_rejected": -82.61642456054688, "logps/rejected": -86.92267608642578, "loss": 0.921, "margin_dpo/margin_mean": 1.507042646408081, "margin_dpo/margin_std": 1.9450860023498535, "step": 456 }, { "epoch": 0.690854119425548, "fcm_dpo/beta": 0.6657828092575073, "fcm_dpo/delta": 0.21099001169204712, "fcm_dpo/margin": 1.2110447883605957, "fcm_dpo/q_t": 0.3547195792198181, "grad_norm": 146.60675048828125, "learning_rate": 1.3309935167761717e-07, "logits/chosen": 0.20737716555595398, "logits/rejected": 0.1518602967262268, "logps/chosen": -47.50776290893555, "logps/ref_chosen": -44.72057342529297, "logps/ref_rejected": -68.1158676147461, "logps/rejected": -72.11409759521484, "loss": 1.0168, "margin_dpo/margin_mean": 1.2110450267791748, "margin_dpo/margin_std": 1.8031151294708252, "step": 457 }, { "epoch": 0.6923658352229781, "fcm_dpo/beta": 0.6568002104759216, "fcm_dpo/delta": -0.16649408638477325, "fcm_dpo/margin": 1.7431389093399048, "fcm_dpo/q_t": 0.2945740222930908, "grad_norm": 143.60699462890625, "learning_rate": 1.3193223130682936e-07, "logits/chosen": 0.15916739404201508, "logits/rejected": 0.07576747238636017, "logps/chosen": -52.41571807861328, "logps/ref_chosen": -50.00569152832031, "logps/ref_rejected": -87.50015258789062, "logps/rejected": -91.6533203125, "loss": 0.8903, "margin_dpo/margin_mean": 1.7431399822235107, "margin_dpo/margin_std": 2.1261298656463623, "step": 458 }, { "epoch": 0.6938775510204082, "fcm_dpo/beta": 0.6480120420455933, "fcm_dpo/delta": -0.23831713199615479, "fcm_dpo/margin": 1.8602559566497803, "fcm_dpo/q_t": 0.2993336021900177, "grad_norm": 136.50265502929688, "learning_rate": 1.3076841354533658e-07, "logits/chosen": 0.18282179534435272, "logits/rejected": 0.15275558829307556, "logps/chosen": -67.98956298828125, "logps/ref_chosen": -65.37794494628906, "logps/ref_rejected": -88.19244384765625, "logps/rejected": -92.66431427001953, "loss": 0.8494, "margin_dpo/margin_mean": 1.8602561950683594, "margin_dpo/margin_std": 2.121655225753784, "step": 459 }, { "epoch": 0.6953892668178382, "fcm_dpo/beta": 0.6043037176132202, "fcm_dpo/delta": -0.21189001202583313, "fcm_dpo/margin": 1.9620928764343262, "fcm_dpo/q_t": 0.3046306073665619, "grad_norm": 148.3003692626953, "learning_rate": 1.2960793094762345e-07, "logits/chosen": 0.19387787580490112, "logits/rejected": 0.09904222190380096, "logps/chosen": -67.26277160644531, "logps/ref_chosen": -64.5616683959961, "logps/ref_rejected": -88.67890167236328, "logps/rejected": -93.34209442138672, "loss": 0.8275, "margin_dpo/margin_mean": 1.962093472480774, "margin_dpo/margin_std": 2.332686185836792, "step": 460 }, { "epoch": 0.6969009826152683, "fcm_dpo/beta": 0.5892372131347656, "fcm_dpo/delta": -0.03291664272546768, "fcm_dpo/margin": 1.7409393787384033, "fcm_dpo/q_t": 0.31874844431877136, "grad_norm": 127.28734588623047, "learning_rate": 1.2845081597488286e-07, "logits/chosen": 0.23900935053825378, "logits/rejected": 0.17159438133239746, "logps/chosen": -52.00554275512695, "logps/ref_chosen": -49.4779167175293, "logps/ref_rejected": -72.65262603759766, "logps/rejected": -76.92119598388672, "loss": 0.9043, "margin_dpo/margin_mean": 1.7409393787384033, "margin_dpo/margin_std": 2.132420063018799, "step": 461 }, { "epoch": 0.6984126984126984, "fcm_dpo/beta": 0.577314019203186, "fcm_dpo/delta": -0.1604897826910019, "fcm_dpo/margin": 1.9741871356964111, "fcm_dpo/q_t": 0.2894290089607239, "grad_norm": 113.71949005126953, "learning_rate": 1.27297100994108e-07, "logits/chosen": 0.13943126797676086, "logits/rejected": 0.09149923920631409, "logps/chosen": -63.15895080566406, "logps/ref_chosen": -60.4951171875, "logps/ref_rejected": -74.82136535644531, "logps/rejected": -79.45939636230469, "loss": 0.7873, "margin_dpo/margin_mean": 1.9741871356964111, "margin_dpo/margin_std": 2.1054553985595703, "step": 462 }, { "epoch": 0.6999244142101285, "fcm_dpo/beta": 0.5867961049079895, "fcm_dpo/delta": 0.11635659635066986, "fcm_dpo/margin": 1.523716688156128, "fcm_dpo/q_t": 0.3367578387260437, "grad_norm": 127.23412322998047, "learning_rate": 1.2614681827718695e-07, "logits/chosen": 0.15458309650421143, "logits/rejected": 0.14070303738117218, "logps/chosen": -70.25871276855469, "logps/ref_chosen": -67.68511962890625, "logps/ref_rejected": -71.32196044921875, "logps/rejected": -75.41926574707031, "loss": 0.931, "margin_dpo/margin_mean": 1.523715853691101, "margin_dpo/margin_std": 1.9335708618164062, "step": 463 }, { "epoch": 0.7014361300075586, "fcm_dpo/beta": 0.6116993427276611, "fcm_dpo/delta": 0.15539291501045227, "fcm_dpo/margin": 1.3953006267547607, "fcm_dpo/q_t": 0.3521912395954132, "grad_norm": 161.49534606933594, "learning_rate": 1.2500000000000005e-07, "logits/chosen": 0.12334546446800232, "logits/rejected": 0.10288789868354797, "logps/chosen": -61.994964599609375, "logps/ref_chosen": -59.16564178466797, "logps/ref_rejected": -69.56146240234375, "logps/rejected": -73.78608703613281, "loss": 1.0905, "margin_dpo/margin_mean": 1.3953003883361816, "margin_dpo/margin_std": 2.2576606273651123, "step": 464 }, { "epoch": 0.7029478458049887, "fcm_dpo/beta": 0.6138174533843994, "fcm_dpo/delta": 0.10352025926113129, "fcm_dpo/margin": 1.4749349355697632, "fcm_dpo/q_t": 0.3457157015800476, "grad_norm": 138.18177795410156, "learning_rate": 1.238566782415197e-07, "logits/chosen": 0.2334330677986145, "logits/rejected": 0.18456201255321503, "logps/chosen": -61.38888168334961, "logps/ref_chosen": -58.513671875, "logps/ref_rejected": -84.31745910644531, "logps/rejected": -88.6676025390625, "loss": 1.0361, "margin_dpo/margin_mean": 1.4749336242675781, "margin_dpo/margin_std": 2.255413293838501, "step": 465 }, { "epoch": 0.7044595616024187, "fcm_dpo/beta": 0.6493447422981262, "fcm_dpo/delta": 0.3105998635292053, "fcm_dpo/margin": 1.0969356298446655, "fcm_dpo/q_t": 0.37098121643066406, "grad_norm": 185.26171875, "learning_rate": 1.2271688498291334e-07, "logits/chosen": 0.18046687543392181, "logits/rejected": 0.1757928878068924, "logps/chosen": -76.58619689941406, "logps/ref_chosen": -73.26580810546875, "logps/ref_rejected": -74.83621215820312, "logps/rejected": -79.25353240966797, "loss": 1.0578, "margin_dpo/margin_mean": 1.096935749053955, "margin_dpo/margin_std": 1.7959051132202148, "step": 466 }, { "epoch": 0.7059712773998488, "fcm_dpo/beta": 0.6433865427970886, "fcm_dpo/delta": -0.17267094552516937, "fcm_dpo/margin": 1.7906594276428223, "fcm_dpo/q_t": 0.3067885637283325, "grad_norm": 113.32935333251953, "learning_rate": 1.2158065210664848e-07, "logits/chosen": 0.16693203151226044, "logits/rejected": 0.060477063059806824, "logps/chosen": -50.42617416381836, "logps/ref_chosen": -47.57947540283203, "logps/ref_rejected": -78.68522644042969, "logps/rejected": -83.32258605957031, "loss": 0.8707, "margin_dpo/margin_mean": 1.7906594276428223, "margin_dpo/margin_std": 2.2650554180145264, "step": 467 }, { "epoch": 0.7074829931972789, "fcm_dpo/beta": 0.6142877340316772, "fcm_dpo/delta": -0.2895115911960602, "fcm_dpo/margin": 2.040811538696289, "fcm_dpo/q_t": 0.28548452258110046, "grad_norm": 135.33714294433594, "learning_rate": 1.204480113956011e-07, "logits/chosen": 0.17146506905555725, "logits/rejected": 0.1613762527704239, "logps/chosen": -66.34950256347656, "logps/ref_chosen": -63.92778778076172, "logps/ref_rejected": -76.51626586914062, "logps/rejected": -80.97879028320312, "loss": 0.796, "margin_dpo/margin_mean": 2.040811061859131, "margin_dpo/margin_std": 2.2666611671447754, "step": 468 }, { "epoch": 0.708994708994709, "fcm_dpo/beta": 0.5907766222953796, "fcm_dpo/delta": -0.01965993642807007, "fcm_dpo/margin": 1.710775375366211, "fcm_dpo/q_t": 0.3215448260307312, "grad_norm": 115.97638702392578, "learning_rate": 1.1931899453216697e-07, "logits/chosen": 0.21937254071235657, "logits/rejected": 0.2032082974910736, "logps/chosen": -61.57379150390625, "logps/ref_chosen": -59.05818176269531, "logps/ref_rejected": -75.67672729492188, "logps/rejected": -79.90310668945312, "loss": 0.8873, "margin_dpo/margin_mean": 1.710775375366211, "margin_dpo/margin_std": 2.0356462001800537, "step": 469 }, { "epoch": 0.7105064247921391, "fcm_dpo/beta": 0.6098539233207703, "fcm_dpo/delta": 0.04422697797417641, "fcm_dpo/margin": 1.5718050003051758, "fcm_dpo/q_t": 0.3227683901786804, "grad_norm": 119.2257308959961, "learning_rate": 1.1819363309737438e-07, "logits/chosen": 0.13354477286338806, "logits/rejected": 0.08637814223766327, "logps/chosen": -50.68056106567383, "logps/ref_chosen": -47.86743927001953, "logps/ref_rejected": -65.96859741210938, "logps/rejected": -70.353515625, "loss": 0.9471, "margin_dpo/margin_mean": 1.5718050003051758, "margin_dpo/margin_std": 2.07285475730896, "step": 470 }, { "epoch": 0.7120181405895691, "fcm_dpo/beta": 0.6002909541130066, "fcm_dpo/delta": -0.08815348893404007, "fcm_dpo/margin": 1.7969985008239746, "fcm_dpo/q_t": 0.3020731508731842, "grad_norm": 131.24606323242188, "learning_rate": 1.1707195857000215e-07, "logits/chosen": 0.1816762238740921, "logits/rejected": 0.12994712591171265, "logps/chosen": -60.345733642578125, "logps/ref_chosen": -57.777854919433594, "logps/ref_rejected": -73.81172180175781, "logps/rejected": -78.17660522460938, "loss": 0.9, "margin_dpo/margin_mean": 1.796998381614685, "margin_dpo/margin_std": 2.2218680381774902, "step": 471 }, { "epoch": 0.7135298563869993, "fcm_dpo/beta": 0.6000721454620361, "fcm_dpo/delta": 0.006265308707952499, "fcm_dpo/margin": 1.6558948755264282, "fcm_dpo/q_t": 0.3246491551399231, "grad_norm": 141.67776489257812, "learning_rate": 1.1595400232569768e-07, "logits/chosen": 0.20169669389724731, "logits/rejected": 0.15981845557689667, "logps/chosen": -58.36172866821289, "logps/ref_chosen": -55.908668518066406, "logps/ref_rejected": -74.70294189453125, "logps/rejected": -78.81190490722656, "loss": 0.9713, "margin_dpo/margin_mean": 1.6558947563171387, "margin_dpo/margin_std": 2.2935879230499268, "step": 472 }, { "epoch": 0.7150415721844293, "fcm_dpo/beta": 0.5857222080230713, "fcm_dpo/delta": -0.07999872416257858, "fcm_dpo/margin": 1.8269248008728027, "fcm_dpo/q_t": 0.32797205448150635, "grad_norm": 127.49575805664062, "learning_rate": 1.1483979563610069e-07, "logits/chosen": 0.23042967915534973, "logits/rejected": 0.15509197115898132, "logps/chosen": -56.60422134399414, "logps/ref_chosen": -54.16088104248047, "logps/ref_rejected": -92.76789855957031, "logps/rejected": -97.03816223144531, "loss": 0.9793, "margin_dpo/margin_mean": 1.8269245624542236, "margin_dpo/margin_std": 2.6278867721557617, "step": 473 }, { "epoch": 0.7165532879818595, "fcm_dpo/beta": 0.6013132333755493, "fcm_dpo/delta": 0.11396686732769012, "fcm_dpo/margin": 1.4877792596817017, "fcm_dpo/q_t": 0.34768766164779663, "grad_norm": 151.1398162841797, "learning_rate": 1.1372936966796709e-07, "logits/chosen": 0.210426926612854, "logits/rejected": 0.1586945503950119, "logps/chosen": -49.76041793823242, "logps/ref_chosen": -46.685707092285156, "logps/ref_rejected": -71.44731903076172, "logps/rejected": -76.00980377197266, "loss": 1.0262, "margin_dpo/margin_mean": 1.487779140472412, "margin_dpo/margin_std": 2.2466931343078613, "step": 474 }, { "epoch": 0.7180650037792895, "fcm_dpo/beta": 0.56545090675354, "fcm_dpo/delta": -0.3536713421344757, "fcm_dpo/margin": 2.305065631866455, "fcm_dpo/q_t": 0.26997214555740356, "grad_norm": 108.57855224609375, "learning_rate": 1.126227554822985e-07, "logits/chosen": 0.15623445808887482, "logits/rejected": 0.11391064524650574, "logps/chosen": -61.26993179321289, "logps/ref_chosen": -58.4873046875, "logps/ref_rejected": -87.00187683105469, "logps/rejected": -92.08956909179688, "loss": 0.7329, "margin_dpo/margin_mean": 2.3050661087036133, "margin_dpo/margin_std": 2.278512477874756, "step": 475 }, { "epoch": 0.7195767195767195, "fcm_dpo/beta": 0.5778172016143799, "fcm_dpo/delta": 0.17008031904697418, "fcm_dpo/margin": 1.4578584432601929, "fcm_dpo/q_t": 0.3523421287536621, "grad_norm": 163.50177001953125, "learning_rate": 1.1151998403347243e-07, "logits/chosen": 0.11702927947044373, "logits/rejected": 0.10260109603404999, "logps/chosen": -78.54095458984375, "logps/ref_chosen": -75.38162231445312, "logps/ref_rejected": -76.99822235107422, "logps/rejected": -81.61541748046875, "loss": 1.0833, "margin_dpo/margin_mean": 1.4578593969345093, "margin_dpo/margin_std": 2.4156503677368164, "step": 476 }, { "epoch": 0.7210884353741497, "fcm_dpo/beta": 0.591883659362793, "fcm_dpo/delta": 0.1255907416343689, "fcm_dpo/margin": 1.4951434135437012, "fcm_dpo/q_t": 0.3490224778652191, "grad_norm": 168.64578247070312, "learning_rate": 1.1042108616837692e-07, "logits/chosen": 0.19208115339279175, "logits/rejected": 0.15968218445777893, "logps/chosen": -64.10305786132812, "logps/ref_chosen": -61.073387145996094, "logps/ref_rejected": -81.34375, "logps/rejected": -85.86856079101562, "loss": 1.0765, "margin_dpo/margin_mean": 1.495143175125122, "margin_dpo/margin_std": 2.407855987548828, "step": 477 }, { "epoch": 0.7226001511715797, "fcm_dpo/beta": 0.6088930368423462, "fcm_dpo/delta": 0.1552659273147583, "fcm_dpo/margin": 1.4085874557495117, "fcm_dpo/q_t": 0.35270342230796814, "grad_norm": 148.8921356201172, "learning_rate": 1.0932609262554746e-07, "logits/chosen": 0.12717093527317047, "logits/rejected": 0.1243201345205307, "logps/chosen": -59.892295837402344, "logps/ref_chosen": -57.16731643676758, "logps/ref_rejected": -53.30917739868164, "logps/rejected": -57.44274139404297, "loss": 1.0176, "margin_dpo/margin_mean": 1.408586859703064, "margin_dpo/margin_std": 2.1223185062408447, "step": 478 }, { "epoch": 0.7241118669690099, "fcm_dpo/beta": 0.607843816280365, "fcm_dpo/delta": 0.027871206402778625, "fcm_dpo/margin": 1.5997779369354248, "fcm_dpo/q_t": 0.3466408848762512, "grad_norm": 140.39195251464844, "learning_rate": 1.0823503403430734e-07, "logits/chosen": 0.08278117328882217, "logits/rejected": 0.03745885565876961, "logps/chosen": -62.16324996948242, "logps/ref_chosen": -58.91331481933594, "logps/ref_rejected": -63.7403450012207, "logps/rejected": -68.59005737304688, "loss": 1.0645, "margin_dpo/margin_mean": 1.5997782945632935, "margin_dpo/margin_std": 2.665220260620117, "step": 479 }, { "epoch": 0.7256235827664399, "fcm_dpo/beta": 0.6309263110160828, "fcm_dpo/delta": 0.006531953811645508, "fcm_dpo/margin": 1.5598734617233276, "fcm_dpo/q_t": 0.31962987780570984, "grad_norm": 165.5882568359375, "learning_rate": 1.0714794091391072e-07, "logits/chosen": 0.13291680812835693, "logits/rejected": 0.1215020939707756, "logps/chosen": -65.6745376586914, "logps/ref_chosen": -62.80061340332031, "logps/ref_rejected": -67.58859252929688, "logps/rejected": -72.02239227294922, "loss": 1.0675, "margin_dpo/margin_mean": 1.5598732233047485, "margin_dpo/margin_std": 2.3550682067871094, "step": 480 }, { "epoch": 0.72713529856387, "fcm_dpo/beta": 0.6073616743087769, "fcm_dpo/delta": -0.08397047966718674, "fcm_dpo/margin": 1.7691869735717773, "fcm_dpo/q_t": 0.3220970034599304, "grad_norm": 130.34205627441406, "learning_rate": 1.0606484367268906e-07, "logits/chosen": 0.11158512532711029, "logits/rejected": 0.10091142356395721, "logps/chosen": -67.81123352050781, "logps/ref_chosen": -65.28649139404297, "logps/ref_rejected": -70.78668212890625, "logps/rejected": -75.08061218261719, "loss": 0.9249, "margin_dpo/margin_mean": 1.7691867351531982, "margin_dpo/margin_std": 2.550887107849121, "step": 481 }, { "epoch": 0.7286470143613001, "fcm_dpo/beta": 0.620997428894043, "fcm_dpo/delta": 0.16822174191474915, "fcm_dpo/margin": 1.3626244068145752, "fcm_dpo/q_t": 0.35040992498397827, "grad_norm": 171.70579528808594, "learning_rate": 1.0498577260720048e-07, "logits/chosen": 0.12327564507722855, "logits/rejected": 0.022273845970630646, "logps/chosen": -63.85777282714844, "logps/ref_chosen": -60.906185150146484, "logps/ref_rejected": -103.44656372070312, "logps/rejected": -107.76078033447266, "loss": 1.0802, "margin_dpo/margin_mean": 1.3626246452331543, "margin_dpo/margin_std": 2.2782201766967773, "step": 482 }, { "epoch": 0.7301587301587301, "fcm_dpo/beta": 0.6081717014312744, "fcm_dpo/delta": -0.14480583369731903, "fcm_dpo/margin": 1.8506314754486084, "fcm_dpo/q_t": 0.31974995136260986, "grad_norm": 137.3916473388672, "learning_rate": 1.0391075790138232e-07, "logits/chosen": 0.20215514302253723, "logits/rejected": 0.13051192462444305, "logps/chosen": -56.16196823120117, "logps/ref_chosen": -53.192012786865234, "logps/ref_rejected": -81.83927154541016, "logps/rejected": -86.65986633300781, "loss": 0.9449, "margin_dpo/margin_mean": 1.850631594657898, "margin_dpo/margin_std": 2.5201127529144287, "step": 483 }, { "epoch": 0.7316704459561603, "fcm_dpo/beta": 0.6308771967887878, "fcm_dpo/delta": 0.20309945940971375, "fcm_dpo/margin": 1.2873001098632812, "fcm_dpo/q_t": 0.3507644832134247, "grad_norm": 157.5721435546875, "learning_rate": 1.0283982962570681e-07, "logits/chosen": 0.18846221268177032, "logits/rejected": 0.15292689204216003, "logps/chosen": -60.82984924316406, "logps/ref_chosen": -57.76945877075195, "logps/ref_rejected": -71.6829833984375, "logps/rejected": -76.03067016601562, "loss": 1.0035, "margin_dpo/margin_mean": 1.2873002290725708, "margin_dpo/margin_std": 1.9288554191589355, "step": 484 }, { "epoch": 0.7331821617535903, "fcm_dpo/beta": 0.5989984273910522, "fcm_dpo/delta": -0.20387829840183258, "fcm_dpo/margin": 1.9372670650482178, "fcm_dpo/q_t": 0.30228108167648315, "grad_norm": 129.9238739013672, "learning_rate": 1.0177301773633992e-07, "logits/chosen": 0.1555819809436798, "logits/rejected": 0.13119317591190338, "logps/chosen": -59.23163604736328, "logps/ref_chosen": -56.63584899902344, "logps/ref_rejected": -70.85614013671875, "logps/rejected": -75.38919067382812, "loss": 0.8593, "margin_dpo/margin_mean": 1.9372668266296387, "margin_dpo/margin_std": 2.2261605262756348, "step": 485 }, { "epoch": 0.7346938775510204, "fcm_dpo/beta": 0.6060769557952881, "fcm_dpo/delta": -0.02147604152560234, "fcm_dpo/margin": 1.6815690994262695, "fcm_dpo/q_t": 0.34005385637283325, "grad_norm": 140.710693359375, "learning_rate": 1.007103520743035e-07, "logits/chosen": 0.1885671615600586, "logits/rejected": 0.10879142582416534, "logps/chosen": -59.6024169921875, "logps/ref_chosen": -56.347023010253906, "logps/ref_rejected": -85.97221374511719, "logps/rejected": -90.9091796875, "loss": 1.0424, "margin_dpo/margin_mean": 1.6815693378448486, "margin_dpo/margin_std": 2.760887861251831, "step": 486 }, { "epoch": 0.7362055933484505, "fcm_dpo/beta": 0.6121037006378174, "fcm_dpo/delta": 0.00196036696434021, "fcm_dpo/margin": 1.6285200119018555, "fcm_dpo/q_t": 0.3259222209453583, "grad_norm": 140.3463897705078, "learning_rate": 9.965186236464046e-08, "logits/chosen": 0.1968272179365158, "logits/rejected": 0.15555810928344727, "logps/chosen": -63.40202331542969, "logps/ref_chosen": -60.617218017578125, "logps/ref_rejected": -82.50975036621094, "logps/rejected": -86.9230728149414, "loss": 0.9054, "margin_dpo/margin_mean": 1.6285200119018555, "margin_dpo/margin_std": 2.1146717071533203, "step": 487 }, { "epoch": 0.7377173091458806, "fcm_dpo/beta": 0.6010755300521851, "fcm_dpo/delta": -0.20396147668361664, "fcm_dpo/margin": 1.956758975982666, "fcm_dpo/q_t": 0.29864153265953064, "grad_norm": 125.88450622558594, "learning_rate": 9.859757821558337e-08, "logits/chosen": 0.1706887185573578, "logits/rejected": 0.11105503141880035, "logps/chosen": -65.75413513183594, "logps/ref_chosen": -63.10905075073242, "logps/ref_rejected": -82.49348449707031, "logps/rejected": -87.09532165527344, "loss": 0.829, "margin_dpo/margin_mean": 1.956758737564087, "margin_dpo/margin_std": 2.2036216259002686, "step": 488 }, { "epoch": 0.7392290249433107, "fcm_dpo/beta": 0.6157445311546326, "fcm_dpo/delta": 0.3702337145805359, "fcm_dpo/margin": 1.0620077848434448, "fcm_dpo/q_t": 0.3890204429626465, "grad_norm": 165.01705932617188, "learning_rate": 9.754752911772615e-08, "logits/chosen": 0.19015483558177948, "logits/rejected": 0.15638966858386993, "logps/chosen": -67.7677993774414, "logps/ref_chosen": -64.98896026611328, "logps/ref_rejected": -84.39607238769531, "logps/rejected": -88.23690795898438, "loss": 1.2058, "margin_dpo/margin_mean": 1.0620079040527344, "margin_dpo/margin_std": 2.2157487869262695, "step": 489 }, { "epoch": 0.7407407407407407, "fcm_dpo/beta": 0.6417911648750305, "fcm_dpo/delta": 0.10067185759544373, "fcm_dpo/margin": 1.4126074314117432, "fcm_dpo/q_t": 0.36741340160369873, "grad_norm": 186.93276977539062, "learning_rate": 9.650174444319956e-08, "logits/chosen": 0.22295230627059937, "logits/rejected": 0.2012663632631302, "logps/chosen": -64.97543334960938, "logps/ref_chosen": -61.90874481201172, "logps/ref_rejected": -70.58566284179688, "logps/rejected": -75.06495666503906, "loss": 1.1948, "margin_dpo/margin_mean": 1.4126070737838745, "margin_dpo/margin_std": 2.7005350589752197, "step": 490 }, { "epoch": 0.7422524565381708, "fcm_dpo/beta": 0.6383862495422363, "fcm_dpo/delta": 0.07852260023355484, "fcm_dpo/margin": 1.450089931488037, "fcm_dpo/q_t": 0.33126381039619446, "grad_norm": 139.84234619140625, "learning_rate": 9.546025344484868e-08, "logits/chosen": 0.11358515918254852, "logits/rejected": 0.0668938159942627, "logps/chosen": -58.38109588623047, "logps/ref_chosen": -55.47570037841797, "logps/ref_rejected": -78.70318603515625, "logps/rejected": -83.05867004394531, "loss": 0.9724, "margin_dpo/margin_mean": 1.4500904083251953, "margin_dpo/margin_std": 1.9347925186157227, "step": 491 }, { "epoch": 0.7437641723356009, "fcm_dpo/beta": 0.6746935844421387, "fcm_dpo/delta": 0.11538802087306976, "fcm_dpo/margin": 1.3122856616973877, "fcm_dpo/q_t": 0.3527218699455261, "grad_norm": 183.52841186523438, "learning_rate": 9.442308525541589e-08, "logits/chosen": 0.15461723506450653, "logits/rejected": 0.09914899617433548, "logps/chosen": -70.79060363769531, "logps/ref_chosen": -67.28638458251953, "logps/ref_rejected": -82.78628540039062, "logps/rejected": -87.60279846191406, "loss": 1.1795, "margin_dpo/margin_mean": 1.3122851848602295, "margin_dpo/margin_std": 2.379267692565918, "step": 492 }, { "epoch": 0.745275888133031, "fcm_dpo/beta": 0.6467149257659912, "fcm_dpo/delta": -0.2506517171859741, "fcm_dpo/margin": 1.8874269723892212, "fcm_dpo/q_t": 0.2871362268924713, "grad_norm": 143.90199279785156, "learning_rate": 9.339026888672468e-08, "logits/chosen": 0.13737066090106964, "logits/rejected": 0.0812433660030365, "logps/chosen": -58.82109451293945, "logps/ref_chosen": -55.92750549316406, "logps/ref_rejected": -79.12149810791016, "logps/rejected": -83.90251159667969, "loss": 0.8709, "margin_dpo/margin_mean": 1.8874274492263794, "margin_dpo/margin_std": 2.288954734802246, "step": 493 }, { "epoch": 0.7467876039304611, "fcm_dpo/beta": 0.6492782831192017, "fcm_dpo/delta": 0.15562888979911804, "fcm_dpo/margin": 1.3203234672546387, "fcm_dpo/q_t": 0.3562992513179779, "grad_norm": 196.9374542236328, "learning_rate": 9.236183322886945e-08, "logits/chosen": 0.06576605886220932, "logits/rejected": 0.027181722223758698, "logps/chosen": -70.94686126708984, "logps/ref_chosen": -67.95410919189453, "logps/ref_rejected": -90.50865173339844, "logps/rejected": -94.82173156738281, "loss": 1.1436, "margin_dpo/margin_mean": 1.3203232288360596, "margin_dpo/margin_std": 2.456470251083374, "step": 494 }, { "epoch": 0.7482993197278912, "fcm_dpo/beta": 0.6703627109527588, "fcm_dpo/delta": 0.14631986618041992, "fcm_dpo/margin": 1.2908090353012085, "fcm_dpo/q_t": 0.35358014702796936, "grad_norm": 143.63949584960938, "learning_rate": 9.133780704940594e-08, "logits/chosen": 0.21046996116638184, "logits/rejected": 0.15889891982078552, "logps/chosen": -55.39236068725586, "logps/ref_chosen": -52.62546157836914, "logps/ref_rejected": -72.06781005859375, "logps/rejected": -76.12551879882812, "loss": 1.0334, "margin_dpo/margin_mean": 1.2908086776733398, "margin_dpo/margin_std": 2.0524096488952637, "step": 495 }, { "epoch": 0.7498110355253212, "fcm_dpo/beta": 0.6439297199249268, "fcm_dpo/delta": -0.214058056473732, "fcm_dpo/margin": 1.8385138511657715, "fcm_dpo/q_t": 0.33099353313446045, "grad_norm": 153.26148986816406, "learning_rate": 9.031821899254797e-08, "logits/chosen": 0.17050223052501678, "logits/rejected": 0.08947437256574631, "logps/chosen": -60.537879943847656, "logps/ref_chosen": -57.597320556640625, "logps/ref_rejected": -94.36127471923828, "logps/rejected": -99.14034271240234, "loss": 1.0009, "margin_dpo/margin_mean": 1.838512897491455, "margin_dpo/margin_std": 2.8675289154052734, "step": 496 }, { "epoch": 0.7513227513227513, "fcm_dpo/beta": 0.5975298881530762, "fcm_dpo/delta": -0.43948090076446533, "fcm_dpo/margin": 2.297381639480591, "fcm_dpo/q_t": 0.27814143896102905, "grad_norm": 137.8128204345703, "learning_rate": 8.930309757836516e-08, "logits/chosen": 0.20551130175590515, "logits/rejected": 0.17609372735023499, "logps/chosen": -75.94544982910156, "logps/ref_chosen": -72.78994750976562, "logps/ref_rejected": -89.48483276367188, "logps/rejected": -94.9377212524414, "loss": 0.8075, "margin_dpo/margin_mean": 2.29738187789917, "margin_dpo/margin_std": 2.609776258468628, "step": 497 }, { "epoch": 0.7528344671201814, "fcm_dpo/beta": 0.5923163294792175, "fcm_dpo/delta": -0.017407868057489395, "fcm_dpo/margin": 1.7135266065597534, "fcm_dpo/q_t": 0.3283523619174957, "grad_norm": 158.02391052246094, "learning_rate": 8.829247120198563e-08, "logits/chosen": 0.17303167283535004, "logits/rejected": 0.14456358551979065, "logps/chosen": -71.15170288085938, "logps/ref_chosen": -68.36572265625, "logps/ref_rejected": -71.28846740722656, "logps/rejected": -75.78797912597656, "loss": 0.9063, "margin_dpo/margin_mean": 1.713526964187622, "margin_dpo/margin_std": 2.233605146408081, "step": 498 }, { "epoch": 0.7543461829176115, "fcm_dpo/beta": 0.594528317451477, "fcm_dpo/delta": 0.08794374763965607, "fcm_dpo/margin": 1.5478346347808838, "fcm_dpo/q_t": 0.35329633951187134, "grad_norm": 145.75146484375, "learning_rate": 8.728636813280163e-08, "logits/chosen": 0.16119879484176636, "logits/rejected": 0.10922683030366898, "logps/chosen": -64.74832916259766, "logps/ref_chosen": -61.90882873535156, "logps/ref_rejected": -91.9411392211914, "logps/rejected": -96.32847595214844, "loss": 1.121, "margin_dpo/margin_mean": 1.547835111618042, "margin_dpo/margin_std": 2.651371479034424, "step": 499 }, { "epoch": 0.7558578987150416, "fcm_dpo/beta": 0.6114327311515808, "fcm_dpo/delta": 0.14150665700435638, "fcm_dpo/margin": 1.4240680932998657, "fcm_dpo/q_t": 0.35586458444595337, "grad_norm": 174.01441955566406, "learning_rate": 8.628481651367875e-08, "logits/chosen": 0.12461017072200775, "logits/rejected": 0.11526093631982803, "logps/chosen": -73.16278076171875, "logps/ref_chosen": -70.225830078125, "logps/ref_rejected": -71.72203063964844, "logps/rejected": -76.08305358886719, "loss": 1.1357, "margin_dpo/margin_mean": 1.4240679740905762, "margin_dpo/margin_std": 2.5242857933044434, "step": 500 }, { "epoch": 0.7573696145124716, "fcm_dpo/beta": 0.6245852708816528, "fcm_dpo/delta": 0.013538122177124023, "fcm_dpo/margin": 1.5765228271484375, "fcm_dpo/q_t": 0.3233751058578491, "grad_norm": 121.06085205078125, "learning_rate": 8.528784436016878e-08, "logits/chosen": 0.15389983355998993, "logits/rejected": 0.13935419917106628, "logps/chosen": -67.49052429199219, "logps/ref_chosen": -64.59880828857422, "logps/ref_rejected": -70.59329223632812, "logps/rejected": -75.0615234375, "loss": 0.8765, "margin_dpo/margin_mean": 1.5765225887298584, "margin_dpo/margin_std": 1.870557188987732, "step": 501 }, { "epoch": 0.7588813303099018, "fcm_dpo/beta": 0.6334064602851868, "fcm_dpo/delta": 0.15443843603134155, "fcm_dpo/margin": 1.3549081087112427, "fcm_dpo/q_t": 0.3447534441947937, "grad_norm": 170.02090454101562, "learning_rate": 8.4295479559726e-08, "logits/chosen": 0.17954713106155396, "logits/rejected": 0.14447346329689026, "logps/chosen": -68.44071960449219, "logps/ref_chosen": -65.46662902832031, "logps/ref_rejected": -90.22233581542969, "logps/rejected": -94.55133056640625, "loss": 1.0121, "margin_dpo/margin_mean": 1.3549081087112427, "margin_dpo/margin_std": 2.0818443298339844, "step": 502 }, { "epoch": 0.7603930461073318, "fcm_dpo/beta": 0.6419456005096436, "fcm_dpo/delta": 0.04625112935900688, "fcm_dpo/margin": 1.4924449920654297, "fcm_dpo/q_t": 0.3314594030380249, "grad_norm": 147.28880310058594, "learning_rate": 8.330774987092712e-08, "logits/chosen": 0.16441036760807037, "logits/rejected": 0.14501985907554626, "logps/chosen": -54.53971862792969, "logps/ref_chosen": -51.83476257324219, "logps/ref_rejected": -57.62522506713867, "logps/rejected": -61.82262420654297, "loss": 0.9904, "margin_dpo/margin_mean": 1.4924452304840088, "margin_dpo/margin_std": 2.1151676177978516, "step": 503 }, { "epoch": 0.7619047619047619, "fcm_dpo/beta": 0.6248334646224976, "fcm_dpo/delta": -0.23319105803966522, "fcm_dpo/margin": 1.9295209646224976, "fcm_dpo/q_t": 0.295588880777359, "grad_norm": 145.23736572265625, "learning_rate": 8.232468292269479e-08, "logits/chosen": 0.12036092579364777, "logits/rejected": 0.0984935611486435, "logps/chosen": -71.39968872070312, "logps/ref_chosen": -68.65119934082031, "logps/ref_rejected": -77.91394805908203, "logps/rejected": -82.59195709228516, "loss": 0.7956, "margin_dpo/margin_mean": 1.929521083831787, "margin_dpo/margin_std": 2.150576114654541, "step": 504 }, { "epoch": 0.763416477702192, "fcm_dpo/beta": 0.6214442253112793, "fcm_dpo/delta": 0.14665237069129944, "fcm_dpo/margin": 1.393211007118225, "fcm_dpo/q_t": 0.35855334997177124, "grad_norm": 172.9752960205078, "learning_rate": 8.134630621352483e-08, "logits/chosen": 0.177079439163208, "logits/rejected": 0.14601781964302063, "logps/chosen": -62.98767852783203, "logps/ref_chosen": -59.99884796142578, "logps/ref_rejected": -76.88048553466797, "logps/rejected": -81.26252746582031, "loss": 1.0917, "margin_dpo/margin_mean": 1.3932104110717773, "margin_dpo/margin_std": 2.397246837615967, "step": 505 }, { "epoch": 0.764928193499622, "fcm_dpo/beta": 0.627830982208252, "fcm_dpo/delta": 0.032355912029743195, "fcm_dpo/margin": 1.54330575466156, "fcm_dpo/q_t": 0.3359745740890503, "grad_norm": 169.81752014160156, "learning_rate": 8.037264711071698e-08, "logits/chosen": 0.1882736086845398, "logits/rejected": 0.166158989071846, "logps/chosen": -72.7292251586914, "logps/ref_chosen": -70.07130432128906, "logps/ref_rejected": -82.03775024414062, "logps/rejected": -86.23898315429688, "loss": 1.0458, "margin_dpo/margin_mean": 1.54330575466156, "margin_dpo/margin_std": 2.4680933952331543, "step": 506 }, { "epoch": 0.7664399092970522, "fcm_dpo/beta": 0.6320427656173706, "fcm_dpo/delta": -0.013070136308670044, "fcm_dpo/margin": 1.600289225578308, "fcm_dpo/q_t": 0.3405856490135193, "grad_norm": 162.6297607421875, "learning_rate": 7.940373284960933e-08, "logits/chosen": 0.16370144486427307, "logits/rejected": 0.1263059675693512, "logps/chosen": -74.96257019042969, "logps/ref_chosen": -72.00703430175781, "logps/ref_rejected": -93.94987487792969, "logps/rejected": -98.50569152832031, "loss": 1.032, "margin_dpo/margin_mean": 1.6002895832061768, "margin_dpo/margin_std": 2.5143651962280273, "step": 507 }, { "epoch": 0.7679516250944822, "fcm_dpo/beta": 0.6303126811981201, "fcm_dpo/delta": -0.09110675752162933, "fcm_dpo/margin": 1.7146742343902588, "fcm_dpo/q_t": 0.3268600106239319, "grad_norm": 160.6134033203125, "learning_rate": 7.843959053281663e-08, "logits/chosen": 0.1638413965702057, "logits/rejected": 0.07367105782032013, "logps/chosen": -62.97923278808594, "logps/ref_chosen": -60.21992492675781, "logps/ref_rejected": -95.9200668334961, "logps/rejected": -100.39404296875, "loss": 0.9646, "margin_dpo/margin_mean": 1.7146737575531006, "margin_dpo/margin_std": 2.4505600929260254, "step": 508 }, { "epoch": 0.7694633408919124, "fcm_dpo/beta": 0.6267092227935791, "fcm_dpo/delta": 0.0032510310411453247, "fcm_dpo/margin": 1.5897384881973267, "fcm_dpo/q_t": 0.3286612629890442, "grad_norm": 157.1903839111328, "learning_rate": 7.748024712947204e-08, "logits/chosen": 0.13599814474582672, "logits/rejected": 0.11172134429216385, "logps/chosen": -69.07527160644531, "logps/ref_chosen": -66.27017211914062, "logps/ref_rejected": -71.73065185546875, "logps/rejected": -76.12548828125, "loss": 0.9953, "margin_dpo/margin_mean": 1.589739203453064, "margin_dpo/margin_std": 2.3416152000427246, "step": 509 }, { "epoch": 0.7709750566893424, "fcm_dpo/beta": 0.6256568431854248, "fcm_dpo/delta": 0.07200966775417328, "fcm_dpo/margin": 1.4934487342834473, "fcm_dpo/q_t": 0.3356286585330963, "grad_norm": 166.45079040527344, "learning_rate": 7.652572947447272e-08, "logits/chosen": 0.22737839818000793, "logits/rejected": 0.16066043078899384, "logps/chosen": -56.61336898803711, "logps/ref_chosen": -53.54487609863281, "logps/ref_rejected": -91.36648559570312, "logps/rejected": -95.92843627929688, "loss": 1.005, "margin_dpo/margin_mean": 1.4934483766555786, "margin_dpo/margin_std": 2.2340614795684814, "step": 510 }, { "epoch": 0.7724867724867724, "fcm_dpo/beta": 0.5967855453491211, "fcm_dpo/delta": -0.3847648501396179, "fcm_dpo/margin": 2.233816146850586, "fcm_dpo/q_t": 0.2725888192653656, "grad_norm": 127.65147399902344, "learning_rate": 7.557606426772961e-08, "logits/chosen": 0.19227483868598938, "logits/rejected": 0.1546470820903778, "logps/chosen": -58.66801452636719, "logps/ref_chosen": -55.844383239746094, "logps/ref_rejected": -86.49819946289062, "logps/rejected": -91.55564880371094, "loss": 0.7557, "margin_dpo/margin_mean": 2.233816623687744, "margin_dpo/margin_std": 2.2734181880950928, "step": 511 }, { "epoch": 0.7739984882842026, "fcm_dpo/beta": 0.5729248523712158, "fcm_dpo/delta": -0.2912992537021637, "fcm_dpo/margin": 2.1851539611816406, "fcm_dpo/q_t": 0.3602805733680725, "grad_norm": 148.83749389648438, "learning_rate": 7.463127807341966e-08, "logits/chosen": 0.06669703125953674, "logits/rejected": 0.05016005039215088, "logps/chosen": -64.66706085205078, "logps/ref_chosen": -61.653038024902344, "logps/ref_rejected": -72.83148193359375, "logps/rejected": -78.03065490722656, "loss": 1.1281, "margin_dpo/margin_mean": 2.185153007507324, "margin_dpo/margin_std": 6.0767316818237305, "step": 512 }, { "epoch": 0.7755102040816326, "fcm_dpo/beta": 0.5423198938369751, "fcm_dpo/delta": -0.018883943557739258, "fcm_dpo/margin": 1.8668808937072754, "fcm_dpo/q_t": 0.3136303424835205, "grad_norm": 92.23009490966797, "learning_rate": 7.369139731924401e-08, "logits/chosen": 0.27449309825897217, "logits/rejected": 0.23453694581985474, "logps/chosen": -53.48218536376953, "logps/ref_chosen": -50.85256576538086, "logps/ref_rejected": -69.21754455566406, "logps/rejected": -73.71403503417969, "loss": 0.8408, "margin_dpo/margin_mean": 1.8668807744979858, "margin_dpo/margin_std": 2.0499701499938965, "step": 513 }, { "epoch": 0.7770219198790628, "fcm_dpo/beta": 0.5583192110061646, "fcm_dpo/delta": 0.07093075662851334, "fcm_dpo/margin": 1.6746280193328857, "fcm_dpo/q_t": 0.3376282751560211, "grad_norm": 142.2669677734375, "learning_rate": 7.275644829568747e-08, "logits/chosen": 0.20312434434890747, "logits/rejected": 0.16620582342147827, "logps/chosen": -72.64167785644531, "logps/ref_chosen": -69.38493347167969, "logps/ref_rejected": -83.32447814941406, "logps/rejected": -88.25584411621094, "loss": 1.0344, "margin_dpo/margin_mean": 1.674628496170044, "margin_dpo/margin_std": 2.6116552352905273, "step": 514 }, { "epoch": 0.7785336356764928, "fcm_dpo/beta": 0.5502352118492126, "fcm_dpo/delta": -0.08825686573982239, "fcm_dpo/margin": 1.9603986740112305, "fcm_dpo/q_t": 0.31272023916244507, "grad_norm": 118.43267822265625, "learning_rate": 7.182645715528435e-08, "logits/chosen": 0.18149125576019287, "logits/rejected": 0.11726510524749756, "logps/chosen": -56.91847610473633, "logps/ref_chosen": -53.687034606933594, "logps/ref_rejected": -83.59614562988281, "logps/rejected": -88.78797912597656, "loss": 0.8794, "margin_dpo/margin_mean": 1.9603983163833618, "margin_dpo/margin_std": 2.4689695835113525, "step": 515 }, { "epoch": 0.780045351473923, "fcm_dpo/beta": 0.5598210096359253, "fcm_dpo/delta": 0.140054851770401, "fcm_dpo/margin": 1.5577726364135742, "fcm_dpo/q_t": 0.33335772156715393, "grad_norm": 122.2513427734375, "learning_rate": 7.090144991188568e-08, "logits/chosen": 0.13685812056064606, "logits/rejected": 0.0911111831665039, "logps/chosen": -59.56391143798828, "logps/ref_chosen": -56.9017219543457, "logps/ref_rejected": -67.83477783203125, "logps/rejected": -72.05474090576172, "loss": 0.9382, "margin_dpo/margin_mean": 1.557773470878601, "margin_dpo/margin_std": 2.0453054904937744, "step": 516 }, { "epoch": 0.781557067271353, "fcm_dpo/beta": 0.5620474815368652, "fcm_dpo/delta": 0.013102632015943527, "fcm_dpo/margin": 1.7575607299804688, "fcm_dpo/q_t": 0.3347151577472687, "grad_norm": 115.9723892211914, "learning_rate": 6.998145243993284e-08, "logits/chosen": 0.18582028150558472, "logits/rejected": 0.17517045140266418, "logps/chosen": -64.92308044433594, "logps/ref_chosen": -61.775142669677734, "logps/ref_rejected": -62.88270950317383, "logps/rejected": -67.78820037841797, "loss": 0.9368, "margin_dpo/margin_mean": 1.7575602531433105, "margin_dpo/margin_std": 2.4262712001800537, "step": 517 }, { "epoch": 0.783068783068783, "fcm_dpo/beta": 0.5985446572303772, "fcm_dpo/delta": 0.34289732575416565, "fcm_dpo/margin": 1.131626844406128, "fcm_dpo/q_t": 0.38726097345352173, "grad_norm": 139.290283203125, "learning_rate": 6.906649047373245e-08, "logits/chosen": 0.19278889894485474, "logits/rejected": 0.15261326730251312, "logps/chosen": -64.96379089355469, "logps/ref_chosen": -62.02523422241211, "logps/ref_rejected": -79.06085205078125, "logps/rejected": -83.13102722167969, "loss": 1.1567, "margin_dpo/margin_mean": 1.1316269636154175, "margin_dpo/margin_std": 2.253190517425537, "step": 518 }, { "epoch": 0.7845804988662132, "fcm_dpo/beta": 0.6441134214401245, "fcm_dpo/delta": 0.37903302907943726, "fcm_dpo/margin": 0.9974351525306702, "fcm_dpo/q_t": 0.4084378778934479, "grad_norm": 205.64553833007812, "learning_rate": 6.815658960673781e-08, "logits/chosen": 0.20801779627799988, "logits/rejected": 0.15976354479789734, "logps/chosen": -65.09986877441406, "logps/ref_chosen": -61.60636901855469, "logps/ref_rejected": -74.50727844238281, "logps/rejected": -78.99821472167969, "loss": 1.4602, "margin_dpo/margin_mean": 0.9974346160888672, "margin_dpo/margin_std": 2.888352870941162, "step": 519 }, { "epoch": 0.7860922146636432, "fcm_dpo/beta": 0.6441489458084106, "fcm_dpo/delta": -0.010032668709754944, "fcm_dpo/margin": 1.562050461769104, "fcm_dpo/q_t": 0.3334931433200836, "grad_norm": 156.4764404296875, "learning_rate": 6.725177529083209e-08, "logits/chosen": 0.23202162981033325, "logits/rejected": 0.18750135600566864, "logps/chosen": -65.67581939697266, "logps/ref_chosen": -62.87343215942383, "logps/ref_rejected": -76.505615234375, "logps/rejected": -80.87004852294922, "loss": 0.9578, "margin_dpo/margin_mean": 1.5620505809783936, "margin_dpo/margin_std": 2.2463040351867676, "step": 520 }, { "epoch": 0.7876039304610734, "fcm_dpo/beta": 0.6531983017921448, "fcm_dpo/delta": 0.006205732002854347, "fcm_dpo/margin": 1.5222787857055664, "fcm_dpo/q_t": 0.32361727952957153, "grad_norm": 157.93814086914062, "learning_rate": 6.63520728356167e-08, "logits/chosen": 0.1032562106847763, "logits/rejected": 0.04845578968524933, "logps/chosen": -67.12205505371094, "logps/ref_chosen": -64.20668029785156, "logps/ref_rejected": -92.28083038330078, "logps/rejected": -96.71849060058594, "loss": 0.9371, "margin_dpo/margin_mean": 1.5222779512405396, "margin_dpo/margin_std": 2.023488759994507, "step": 521 }, { "epoch": 0.7891156462585034, "fcm_dpo/beta": 0.6555431485176086, "fcm_dpo/delta": 0.03148447349667549, "fcm_dpo/margin": 1.48207688331604, "fcm_dpo/q_t": 0.32747435569763184, "grad_norm": 147.4585723876953, "learning_rate": 6.545750740770336e-08, "logits/chosen": 0.16423772275447845, "logits/rejected": 0.14849795401096344, "logps/chosen": -61.12641525268555, "logps/ref_chosen": -58.369720458984375, "logps/ref_rejected": -68.79248046875, "logps/rejected": -73.03125, "loss": 1.0698, "margin_dpo/margin_mean": 1.4820764064788818, "margin_dpo/margin_std": 2.3686909675598145, "step": 522 }, { "epoch": 0.7906273620559335, "fcm_dpo/beta": 0.6287499666213989, "fcm_dpo/delta": -0.24582098424434662, "fcm_dpo/margin": 1.9300764799118042, "fcm_dpo/q_t": 0.2935262620449066, "grad_norm": 161.8909454345703, "learning_rate": 6.456810403001012e-08, "logits/chosen": 0.17682617902755737, "logits/rejected": 0.09023305773735046, "logps/chosen": -68.99554443359375, "logps/ref_chosen": -65.71324157714844, "logps/ref_rejected": -91.98896789550781, "logps/rejected": -97.20135498046875, "loss": 0.9505, "margin_dpo/margin_mean": 1.9300763607025146, "margin_dpo/margin_std": 2.5030922889709473, "step": 523 }, { "epoch": 0.7921390778533636, "fcm_dpo/beta": 0.642947256565094, "fcm_dpo/delta": 0.15807999670505524, "fcm_dpo/margin": 1.3285942077636719, "fcm_dpo/q_t": 0.34164804220199585, "grad_norm": 157.54954528808594, "learning_rate": 6.368388758106134e-08, "logits/chosen": 0.14576482772827148, "logits/rejected": 0.12724286317825317, "logps/chosen": -78.79303741455078, "logps/ref_chosen": -76.35124969482422, "logps/ref_rejected": -89.96072387695312, "logps/rejected": -93.73110961914062, "loss": 1.0368, "margin_dpo/margin_mean": 1.328594446182251, "margin_dpo/margin_std": 2.0358855724334717, "step": 524 }, { "epoch": 0.7936507936507936, "fcm_dpo/beta": 0.6652133464813232, "fcm_dpo/delta": 0.20262369513511658, "fcm_dpo/margin": 1.2232202291488647, "fcm_dpo/q_t": 0.3485579490661621, "grad_norm": 169.54148864746094, "learning_rate": 6.280488279429185e-08, "logits/chosen": 0.0526951402425766, "logits/rejected": 0.039441537111997604, "logps/chosen": -78.24242401123047, "logps/ref_chosen": -75.49578857421875, "logps/ref_rejected": -84.04852294921875, "logps/rejected": -88.01837921142578, "loss": 0.9943, "margin_dpo/margin_mean": 1.2232205867767334, "margin_dpo/margin_std": 1.763154149055481, "step": 525 }, { "epoch": 0.7951625094482238, "fcm_dpo/beta": 0.6876204609870911, "fcm_dpo/delta": 0.2604686915874481, "fcm_dpo/margin": 1.1028797626495361, "fcm_dpo/q_t": 0.36391156911849976, "grad_norm": 168.36891174316406, "learning_rate": 6.193111425735515e-08, "logits/chosen": 0.1737247258424759, "logits/rejected": 0.12701740860939026, "logps/chosen": -64.25382995605469, "logps/ref_chosen": -61.29241943359375, "logps/ref_rejected": -82.47763061523438, "logps/rejected": -86.54191589355469, "loss": 1.0833, "margin_dpo/margin_mean": 1.1028800010681152, "margin_dpo/margin_std": 1.7894057035446167, "step": 526 }, { "epoch": 0.7966742252456538, "fcm_dpo/beta": 0.7351027727127075, "fcm_dpo/delta": 0.22121518850326538, "fcm_dpo/margin": 1.081107497215271, "fcm_dpo/q_t": 0.3715837299823761, "grad_norm": 250.8863067626953, "learning_rate": 6.106260641143546e-08, "logits/chosen": 0.2451198697090149, "logits/rejected": 0.1942567229270935, "logps/chosen": -64.66523742675781, "logps/ref_chosen": -61.472625732421875, "logps/ref_rejected": -90.52831268310547, "logps/rejected": -94.8020248413086, "loss": 1.3058, "margin_dpo/margin_mean": 1.0811076164245605, "margin_dpo/margin_std": 2.4224696159362793, "step": 527 }, { "epoch": 0.7981859410430839, "fcm_dpo/beta": 0.7391092777252197, "fcm_dpo/delta": 0.08262480795383453, "fcm_dpo/margin": 1.248426914215088, "fcm_dpo/q_t": 0.33980193734169006, "grad_norm": 163.9738311767578, "learning_rate": 6.019938355056422e-08, "logits/chosen": 0.050329744815826416, "logits/rejected": -0.005134463310241699, "logps/chosen": -61.76133728027344, "logps/ref_chosen": -58.792015075683594, "logps/ref_rejected": -71.82516479492188, "logps/rejected": -76.04290771484375, "loss": 1.1046, "margin_dpo/margin_mean": 1.2484264373779297, "margin_dpo/margin_std": 2.0079092979431152, "step": 528 }, { "epoch": 0.799697656840514, "fcm_dpo/beta": 0.6621348857879639, "fcm_dpo/delta": -0.6627082824707031, "fcm_dpo/margin": 2.306537628173828, "fcm_dpo/q_t": 0.25794824957847595, "grad_norm": 117.09664154052734, "learning_rate": 5.934146982094049e-08, "logits/chosen": 0.10348678380250931, "logits/rejected": 0.05463102087378502, "logps/chosen": -57.58445739746094, "logps/ref_chosen": -55.070960998535156, "logps/ref_rejected": -75.44007873535156, "logps/rejected": -80.26010131835938, "loss": 0.7618, "margin_dpo/margin_mean": 2.306537389755249, "margin_dpo/margin_std": 2.351503372192383, "step": 529 }, { "epoch": 0.8012093726379441, "fcm_dpo/beta": 0.6574649214744568, "fcm_dpo/delta": -0.030184239149093628, "fcm_dpo/margin": 1.5620912313461304, "fcm_dpo/q_t": 0.32785388827323914, "grad_norm": 138.7952880859375, "learning_rate": 5.848888922025552e-08, "logits/chosen": 0.1791817992925644, "logits/rejected": 0.14665716886520386, "logps/chosen": -59.61901092529297, "logps/ref_chosen": -56.743812561035156, "logps/ref_rejected": -76.6692123413086, "logps/rejected": -81.10650634765625, "loss": 0.9455, "margin_dpo/margin_mean": 1.5620914697647095, "margin_dpo/margin_std": 2.117258071899414, "step": 530 }, { "epoch": 0.8027210884353742, "fcm_dpo/beta": 0.6852550506591797, "fcm_dpo/delta": 0.22120189666748047, "fcm_dpo/margin": 1.1534594297409058, "fcm_dpo/q_t": 0.3655146360397339, "grad_norm": 180.98231506347656, "learning_rate": 5.7641665597021435e-08, "logits/chosen": 0.14505237340927124, "logits/rejected": 0.09899041056632996, "logps/chosen": -54.173072814941406, "logps/ref_chosen": -51.116455078125, "logps/ref_rejected": -79.52884674072266, "logps/rejected": -83.73892211914062, "loss": 1.0873, "margin_dpo/margin_mean": 1.1534587144851685, "margin_dpo/margin_std": 1.9886727333068848, "step": 531 }, { "epoch": 0.8042328042328042, "fcm_dpo/beta": 0.6626486778259277, "fcm_dpo/delta": -0.1822899430990219, "fcm_dpo/margin": 1.750182867050171, "fcm_dpo/q_t": 0.31126174330711365, "grad_norm": 145.5829315185547, "learning_rate": 5.679982264990424e-08, "logits/chosen": 0.09937071800231934, "logits/rejected": 0.05832071602344513, "logps/chosen": -61.42143249511719, "logps/ref_chosen": -58.279945373535156, "logps/ref_rejected": -78.05426788330078, "logps/rejected": -82.94593811035156, "loss": 0.9066, "margin_dpo/margin_mean": 1.7501822710037231, "margin_dpo/margin_std": 2.264770030975342, "step": 532 }, { "epoch": 0.8057445200302343, "fcm_dpo/beta": 0.6552125811576843, "fcm_dpo/delta": -0.08179665356874466, "fcm_dpo/margin": 1.6374917030334473, "fcm_dpo/q_t": 0.3215191662311554, "grad_norm": 148.30023193359375, "learning_rate": 5.596338392706076e-08, "logits/chosen": 0.25206273794174194, "logits/rejected": 0.20781204104423523, "logps/chosen": -58.997161865234375, "logps/ref_chosen": -56.41801071166992, "logps/ref_rejected": -73.89324951171875, "logps/rejected": -78.1098861694336, "loss": 0.9886, "margin_dpo/margin_mean": 1.6374918222427368, "margin_dpo/margin_std": 2.354870319366455, "step": 533 }, { "epoch": 0.8072562358276644, "fcm_dpo/beta": 0.649742603302002, "fcm_dpo/delta": 0.05204106122255325, "fcm_dpo/margin": 1.4657820463180542, "fcm_dpo/q_t": 0.32745662331581116, "grad_norm": 144.36602783203125, "learning_rate": 5.513237282548033e-08, "logits/chosen": 0.17291076481342316, "logits/rejected": 0.13234050571918488, "logps/chosen": -63.49784851074219, "logps/ref_chosen": -60.748687744140625, "logps/ref_rejected": -73.8623046875, "logps/rejected": -78.07723999023438, "loss": 0.9544, "margin_dpo/margin_mean": 1.4657821655273438, "margin_dpo/margin_std": 1.9620198011398315, "step": 534 }, { "epoch": 0.8087679516250945, "fcm_dpo/beta": 0.6593036651611328, "fcm_dpo/delta": -0.019535936415195465, "fcm_dpo/margin": 1.5417048931121826, "fcm_dpo/q_t": 0.3396652340888977, "grad_norm": 148.61146545410156, "learning_rate": 5.430681259032957e-08, "logits/chosen": 0.09707458317279816, "logits/rejected": 0.04526631161570549, "logps/chosen": -64.64535522460938, "logps/ref_chosen": -61.637413024902344, "logps/ref_rejected": -80.93138885498047, "logps/rejected": -85.48104095458984, "loss": 1.0638, "margin_dpo/margin_mean": 1.5417053699493408, "margin_dpo/margin_std": 2.5180535316467285, "step": 535 }, { "epoch": 0.8102796674225246, "fcm_dpo/beta": 0.6411547064781189, "fcm_dpo/delta": -0.05495788902044296, "fcm_dpo/margin": 1.6322299242019653, "fcm_dpo/q_t": 0.3149953782558441, "grad_norm": 129.11102294921875, "learning_rate": 5.3486726314303175e-08, "logits/chosen": 0.20541507005691528, "logits/rejected": 0.14478695392608643, "logps/chosen": -54.65433120727539, "logps/ref_chosen": -51.88897705078125, "logps/ref_rejected": -73.34864044189453, "logps/rejected": -77.74623107910156, "loss": 0.8709, "margin_dpo/margin_mean": 1.6322304010391235, "margin_dpo/margin_std": 1.9500904083251953, "step": 536 }, { "epoch": 0.8117913832199547, "fcm_dpo/beta": 0.6352528929710388, "fcm_dpo/delta": -0.06024022772908211, "fcm_dpo/margin": 1.6563501358032227, "fcm_dpo/q_t": 0.32194915413856506, "grad_norm": 129.3059844970703, "learning_rate": 5.267213693697695e-08, "logits/chosen": 0.24240414798259735, "logits/rejected": 0.1752084195613861, "logps/chosen": -57.34803771972656, "logps/ref_chosen": -54.248619079589844, "logps/ref_rejected": -94.94343566894531, "logps/rejected": -99.69920349121094, "loss": 1.0076, "margin_dpo/margin_mean": 1.656351089477539, "margin_dpo/margin_std": 2.3621954917907715, "step": 537 }, { "epoch": 0.8133030990173847, "fcm_dpo/beta": 0.6266754865646362, "fcm_dpo/delta": -0.23747026920318604, "fcm_dpo/margin": 1.9266445636749268, "fcm_dpo/q_t": 0.28824812173843384, "grad_norm": 151.20790100097656, "learning_rate": 5.1863067244167144e-08, "logits/chosen": 0.17140713334083557, "logits/rejected": 0.14127670228481293, "logps/chosen": -73.17599487304688, "logps/ref_chosen": -70.09353637695312, "logps/ref_rejected": -79.49833679199219, "logps/rejected": -84.5074462890625, "loss": 0.8148, "margin_dpo/margin_mean": 1.926644206047058, "margin_dpo/margin_std": 2.1346817016601562, "step": 538 }, { "epoch": 0.8148148148148148, "fcm_dpo/beta": 0.6051099300384521, "fcm_dpo/delta": -0.027577966451644897, "fcm_dpo/margin": 1.6932862997055054, "fcm_dpo/q_t": 0.3206092119216919, "grad_norm": 135.66448974609375, "learning_rate": 5.105953986729195e-08, "logits/chosen": 0.13851284980773926, "logits/rejected": 0.08982232213020325, "logps/chosen": -64.91091918945312, "logps/ref_chosen": -61.93169403076172, "logps/ref_rejected": -84.08946228027344, "logps/rejected": -88.76197052001953, "loss": 0.8703, "margin_dpo/margin_mean": 1.6932868957519531, "margin_dpo/margin_std": 2.0537302494049072, "step": 539 }, { "epoch": 0.8163265306122449, "fcm_dpo/beta": 0.5865040421485901, "fcm_dpo/delta": -0.2764972448348999, "fcm_dpo/margin": 2.119068145751953, "fcm_dpo/q_t": 0.29492413997650146, "grad_norm": 142.99563598632812, "learning_rate": 5.026157728273966e-08, "logits/chosen": 0.1776391565799713, "logits/rejected": 0.11434172093868256, "logps/chosen": -65.55619812011719, "logps/ref_chosen": -62.704254150390625, "logps/ref_rejected": -95.63597106933594, "logps/rejected": -100.60699462890625, "loss": 0.841, "margin_dpo/margin_mean": 2.1190683841705322, "margin_dpo/margin_std": 2.443237781524658, "step": 540 }, { "epoch": 0.817838246409675, "fcm_dpo/beta": 0.5594385266304016, "fcm_dpo/delta": -0.11497347056865692, "fcm_dpo/margin": 1.9675724506378174, "fcm_dpo/q_t": 0.3052162230014801, "grad_norm": 126.23677062988281, "learning_rate": 4.9469201811239035e-08, "logits/chosen": 0.12904441356658936, "logits/rejected": 0.12872368097305298, "logps/chosen": -65.19651794433594, "logps/ref_chosen": -62.48084259033203, "logps/ref_rejected": -57.55541229248047, "logps/rejected": -62.23865509033203, "loss": 0.8973, "margin_dpo/margin_mean": 1.967572569847107, "margin_dpo/margin_std": 2.4288015365600586, "step": 541 }, { "epoch": 0.8193499622071051, "fcm_dpo/beta": 0.5571799278259277, "fcm_dpo/delta": -0.033925510942935944, "fcm_dpo/margin": 1.8493754863739014, "fcm_dpo/q_t": 0.3220483660697937, "grad_norm": 113.93995666503906, "learning_rate": 4.868243561723534e-08, "logits/chosen": 0.2040795087814331, "logits/rejected": 0.1579345464706421, "logps/chosen": -52.065277099609375, "logps/ref_chosen": -49.454891204833984, "logps/ref_rejected": -65.33275604248047, "logps/rejected": -69.79252624511719, "loss": 0.9384, "margin_dpo/margin_mean": 1.8493754863739014, "margin_dpo/margin_std": 2.4480319023132324, "step": 542 }, { "epoch": 0.8208616780045351, "fcm_dpo/beta": 0.5707334280014038, "fcm_dpo/delta": 0.13935577869415283, "fcm_dpo/margin": 1.527040958404541, "fcm_dpo/q_t": 0.33425813913345337, "grad_norm": 109.5389175415039, "learning_rate": 4.790130070827028e-08, "logits/chosen": 0.1735750138759613, "logits/rejected": 0.10716632753610611, "logps/chosen": -53.960105895996094, "logps/ref_chosen": -51.100860595703125, "logps/ref_rejected": -76.06130981445312, "logps/rejected": -80.44760131835938, "loss": 0.9579, "margin_dpo/margin_mean": 1.5270410776138306, "margin_dpo/margin_std": 2.048778533935547, "step": 543 }, { "epoch": 0.8223733938019653, "fcm_dpo/beta": 0.5578969717025757, "fcm_dpo/delta": -0.18755921721458435, "fcm_dpo/margin": 2.0900115966796875, "fcm_dpo/q_t": 0.3165471851825714, "grad_norm": 130.18556213378906, "learning_rate": 4.7125818934366454e-08, "logits/chosen": 0.14418599009513855, "logits/rejected": 0.09046932309865952, "logps/chosen": -63.256656646728516, "logps/ref_chosen": -60.2772331237793, "logps/ref_rejected": -88.40553283691406, "logps/rejected": -93.4749755859375, "loss": 0.9465, "margin_dpo/margin_mean": 2.0900115966796875, "margin_dpo/margin_std": 2.8377792835235596, "step": 544 }, { "epoch": 0.8238851095993953, "fcm_dpo/beta": 0.5586047172546387, "fcm_dpo/delta": 0.11241482198238373, "fcm_dpo/margin": 1.607243299484253, "fcm_dpo/q_t": 0.3465924859046936, "grad_norm": 141.80880737304688, "learning_rate": 4.635601198741607e-08, "logits/chosen": 0.1261473000049591, "logits/rejected": 0.07930372655391693, "logps/chosen": -64.5820541381836, "logps/ref_chosen": -61.61524963378906, "logps/ref_rejected": -78.71266174316406, "logps/rejected": -83.28670501708984, "loss": 1.0031, "margin_dpo/margin_mean": 1.6072428226470947, "margin_dpo/margin_std": 2.3972063064575195, "step": 545 }, { "epoch": 0.8253968253968254, "fcm_dpo/beta": 0.5760623216629028, "fcm_dpo/delta": 0.1294553279876709, "fcm_dpo/margin": 1.5299469232559204, "fcm_dpo/q_t": 0.3359874188899994, "grad_norm": 148.13609313964844, "learning_rate": 4.559190140057428e-08, "logits/chosen": 0.17436569929122925, "logits/rejected": 0.15731996297836304, "logps/chosen": -62.12708282470703, "logps/ref_chosen": -59.313262939453125, "logps/ref_rejected": -64.73631286621094, "logps/rejected": -69.080078125, "loss": 1.0087, "margin_dpo/margin_mean": 1.52994704246521, "margin_dpo/margin_std": 2.212397813796997, "step": 546 }, { "epoch": 0.8269085411942555, "fcm_dpo/beta": 0.5641611814498901, "fcm_dpo/delta": -0.1594116985797882, "fcm_dpo/margin": 2.0226144790649414, "fcm_dpo/q_t": 0.2903903126716614, "grad_norm": 107.82057189941406, "learning_rate": 4.483350854765672e-08, "logits/chosen": 0.1362178921699524, "logits/rejected": 0.08240213245153427, "logps/chosen": -57.526573181152344, "logps/ref_chosen": -54.97674560546875, "logps/ref_rejected": -75.35922241210938, "logps/rejected": -79.9316635131836, "loss": 0.8307, "margin_dpo/margin_mean": 2.0226151943206787, "margin_dpo/margin_std": 2.2596964836120605, "step": 547 }, { "epoch": 0.8284202569916855, "fcm_dpo/beta": 0.5807280540466309, "fcm_dpo/delta": 0.16467252373695374, "fcm_dpo/margin": 1.4542980194091797, "fcm_dpo/q_t": 0.3473031222820282, "grad_norm": 137.8083038330078, "learning_rate": 4.4080854642541826e-08, "logits/chosen": 0.09831206500530243, "logits/rejected": 0.04538644477725029, "logps/chosen": -66.22586822509766, "logps/ref_chosen": -63.21067428588867, "logps/ref_rejected": -81.23347473144531, "logps/rejected": -85.70297241210938, "loss": 0.9561, "margin_dpo/margin_mean": 1.4542980194091797, "margin_dpo/margin_std": 1.9517710208892822, "step": 548 }, { "epoch": 0.8299319727891157, "fcm_dpo/beta": 0.5972309708595276, "fcm_dpo/delta": 0.16186922788619995, "fcm_dpo/margin": 1.4225797653198242, "fcm_dpo/q_t": 0.35338258743286133, "grad_norm": 165.7979736328125, "learning_rate": 4.333396073857723e-08, "logits/chosen": 0.2513388395309448, "logits/rejected": 0.2025221437215805, "logps/chosen": -67.16740417480469, "logps/ref_chosen": -64.27351379394531, "logps/ref_rejected": -92.31663513183594, "logps/rejected": -96.63310241699219, "loss": 1.0582, "margin_dpo/margin_mean": 1.4225800037384033, "margin_dpo/margin_std": 2.26432466506958, "step": 549 }, { "epoch": 0.8314436885865457, "fcm_dpo/beta": 0.6041165590286255, "fcm_dpo/delta": 0.15158365666866302, "fcm_dpo/margin": 1.4242266416549683, "fcm_dpo/q_t": 0.3457239270210266, "grad_norm": 118.26493072509766, "learning_rate": 4.259284772799099e-08, "logits/chosen": 0.17192748188972473, "logits/rejected": 0.14534920454025269, "logps/chosen": -59.13628387451172, "logps/ref_chosen": -56.230438232421875, "logps/ref_rejected": -62.59788513183594, "logps/rejected": -66.92796325683594, "loss": 1.0043, "margin_dpo/margin_mean": 1.4242260456085205, "margin_dpo/margin_std": 2.080674171447754, "step": 550 }, { "epoch": 0.8329554043839759, "fcm_dpo/beta": 0.6244951486587524, "fcm_dpo/delta": 0.035569630563259125, "fcm_dpo/margin": 1.5473127365112305, "fcm_dpo/q_t": 0.33668047189712524, "grad_norm": 132.6877899169922, "learning_rate": 4.1857536341307176e-08, "logits/chosen": 0.17000985145568848, "logits/rejected": 0.14413632452487946, "logps/chosen": -70.88568878173828, "logps/ref_chosen": -67.74720764160156, "logps/ref_rejected": -87.04285430908203, "logps/rejected": -91.72865295410156, "loss": 0.9498, "margin_dpo/margin_mean": 1.5473123788833618, "margin_dpo/margin_std": 2.1508963108062744, "step": 551 }, { "epoch": 0.8344671201814059, "fcm_dpo/beta": 0.6188427209854126, "fcm_dpo/delta": -0.11111941188573837, "fcm_dpo/margin": 1.7723690271377563, "fcm_dpo/q_t": 0.2986357808113098, "grad_norm": 139.71243286132812, "learning_rate": 4.112804714676593e-08, "logits/chosen": 0.14941178262233734, "logits/rejected": 0.10768507421016693, "logps/chosen": -65.74885559082031, "logps/ref_chosen": -62.92625427246094, "logps/ref_rejected": -82.98365783691406, "logps/rejected": -87.57861328125, "loss": 0.887, "margin_dpo/margin_mean": 1.7723690271377563, "margin_dpo/margin_std": 2.1054043769836426, "step": 552 }, { "epoch": 0.8359788359788359, "fcm_dpo/beta": 0.5973865985870361, "fcm_dpo/delta": -0.033724166452884674, "fcm_dpo/margin": 1.7203757762908936, "fcm_dpo/q_t": 0.3467303514480591, "grad_norm": 141.0452880859375, "learning_rate": 4.0404400549748144e-08, "logits/chosen": 0.1345456838607788, "logits/rejected": 0.05725400522351265, "logps/chosen": -59.39005661010742, "logps/ref_chosen": -56.038490295410156, "logps/ref_rejected": -84.48454284667969, "logps/rejected": -89.55648803710938, "loss": 1.1217, "margin_dpo/margin_mean": 1.720375657081604, "margin_dpo/margin_std": 2.994457721710205, "step": 553 }, { "epoch": 0.8374905517762661, "fcm_dpo/beta": 0.61543208360672, "fcm_dpo/delta": 0.10628563910722733, "fcm_dpo/margin": 1.4661433696746826, "fcm_dpo/q_t": 0.3466363847255707, "grad_norm": 153.4082489013672, "learning_rate": 3.968661679220467e-08, "logits/chosen": 0.14405734837055206, "logits/rejected": 0.12058596312999725, "logps/chosen": -67.53119659423828, "logps/ref_chosen": -64.53059387207031, "logps/ref_rejected": -71.2155990600586, "logps/rejected": -75.6823501586914, "loss": 1.032, "margin_dpo/margin_mean": 1.4661433696746826, "margin_dpo/margin_std": 2.351094961166382, "step": 554 }, { "epoch": 0.8390022675736961, "fcm_dpo/beta": 0.6489007472991943, "fcm_dpo/delta": 0.20813655853271484, "fcm_dpo/margin": 1.2255511283874512, "fcm_dpo/q_t": 0.3576173484325409, "grad_norm": 182.52911376953125, "learning_rate": 3.89747159520904e-08, "logits/chosen": 0.15571804344654083, "logits/rejected": 0.12670589983463287, "logps/chosen": -70.10320281982422, "logps/ref_chosen": -66.65191650390625, "logps/ref_rejected": -68.6667251586914, "logps/rejected": -73.34355163574219, "loss": 1.1729, "margin_dpo/margin_mean": 1.225550651550293, "margin_dpo/margin_std": 2.238006591796875, "step": 555 }, { "epoch": 0.8405139833711263, "fcm_dpo/beta": 0.6440955400466919, "fcm_dpo/delta": 0.016805479303002357, "fcm_dpo/margin": 1.5289926528930664, "fcm_dpo/q_t": 0.34293437004089355, "grad_norm": 150.5496063232422, "learning_rate": 3.826871794280192e-08, "logits/chosen": 0.17429228127002716, "logits/rejected": 0.13041889667510986, "logps/chosen": -56.17012405395508, "logps/ref_chosen": -52.832366943359375, "logps/ref_rejected": -64.49044036865234, "logps/rejected": -69.35718536376953, "loss": 1.0381, "margin_dpo/margin_mean": 1.5289928913116455, "margin_dpo/margin_std": 2.3977890014648438, "step": 556 }, { "epoch": 0.8420256991685563, "fcm_dpo/beta": 0.5984865427017212, "fcm_dpo/delta": -0.35813382267951965, "fcm_dpo/margin": 2.1672024726867676, "fcm_dpo/q_t": 0.2816886305809021, "grad_norm": 126.752197265625, "learning_rate": 3.756864251262143e-08, "logits/chosen": 0.21231526136398315, "logits/rejected": 0.1561964750289917, "logps/chosen": -58.17803955078125, "logps/ref_chosen": -55.03598403930664, "logps/ref_rejected": -75.80644989013672, "logps/rejected": -81.11570739746094, "loss": 0.7731, "margin_dpo/margin_mean": 2.1672027111053467, "margin_dpo/margin_std": 2.3840556144714355, "step": 557 }, { "epoch": 0.8435374149659864, "fcm_dpo/beta": 0.5904332399368286, "fcm_dpo/delta": -0.06778506934642792, "fcm_dpo/margin": 1.794862151145935, "fcm_dpo/q_t": 0.32028087973594666, "grad_norm": 146.46681213378906, "learning_rate": 3.687450924416341e-08, "logits/chosen": 0.22109848260879517, "logits/rejected": 0.17496977746486664, "logps/chosen": -66.06083679199219, "logps/ref_chosen": -63.226348876953125, "logps/ref_rejected": -91.46881866455078, "logps/rejected": -96.09817504882812, "loss": 0.9082, "margin_dpo/margin_mean": 1.7948615550994873, "margin_dpo/margin_std": 2.3338146209716797, "step": 558 }, { "epoch": 0.8450491307634165, "fcm_dpo/beta": 0.5790784358978271, "fcm_dpo/delta": -0.06187023967504501, "fcm_dpo/margin": 1.8165602684020996, "fcm_dpo/q_t": 0.33412617444992065, "grad_norm": 138.7181854248047, "learning_rate": 3.6186337553827743e-08, "logits/chosen": 0.11689116060733795, "logits/rejected": 0.06340146064758301, "logps/chosen": -64.47853088378906, "logps/ref_chosen": -61.521644592285156, "logps/ref_rejected": -82.83859252929688, "logps/rejected": -87.61204528808594, "loss": 1.0151, "margin_dpo/margin_mean": 1.8165605068206787, "margin_dpo/margin_std": 2.6852447986602783, "step": 559 }, { "epoch": 0.8465608465608465, "fcm_dpo/beta": 0.6011730432510376, "fcm_dpo/delta": 0.08525849133729935, "fcm_dpo/margin": 1.5243772268295288, "fcm_dpo/q_t": 0.345813125371933, "grad_norm": 146.03248596191406, "learning_rate": 3.550414669125573e-08, "logits/chosen": 0.16141119599342346, "logits/rejected": 0.13059382140636444, "logps/chosen": -63.782081604003906, "logps/ref_chosen": -60.64122009277344, "logps/ref_rejected": -78.75474548339844, "logps/rejected": -83.41998291015625, "loss": 1.0243, "margin_dpo/margin_mean": 1.5243771076202393, "margin_dpo/margin_std": 2.256977081298828, "step": 560 }, { "epoch": 0.8480725623582767, "fcm_dpo/beta": 0.5956799983978271, "fcm_dpo/delta": 0.0049747563898563385, "fcm_dpo/margin": 1.6711280345916748, "fcm_dpo/q_t": 0.3338923752307892, "grad_norm": 122.3641128540039, "learning_rate": 3.482795573879241e-08, "logits/chosen": 0.15920129418373108, "logits/rejected": 0.13256219029426575, "logps/chosen": -65.13980102539062, "logps/ref_chosen": -62.49859619140625, "logps/ref_rejected": -78.72064208984375, "logps/rejected": -83.03296661376953, "loss": 0.954, "margin_dpo/margin_mean": 1.671128273010254, "margin_dpo/margin_std": 2.351602077484131, "step": 561 }, { "epoch": 0.8495842781557067, "fcm_dpo/beta": 0.5753499269485474, "fcm_dpo/delta": -0.1565413773059845, "fcm_dpo/margin": 1.973330020904541, "fcm_dpo/q_t": 0.3080880045890808, "grad_norm": 128.77142333984375, "learning_rate": 3.415778361095226e-08, "logits/chosen": 0.20189374685287476, "logits/rejected": 0.17044463753700256, "logps/chosen": -77.8094482421875, "logps/ref_chosen": -74.78173828125, "logps/ref_rejected": -92.63499450683594, "logps/rejected": -97.63603210449219, "loss": 0.8933, "margin_dpo/margin_mean": 1.9733293056488037, "margin_dpo/margin_std": 2.4358203411102295, "step": 562 }, { "epoch": 0.8510959939531368, "fcm_dpo/beta": 0.5790232419967651, "fcm_dpo/delta": 0.03163836523890495, "fcm_dpo/margin": 1.6777245998382568, "fcm_dpo/q_t": 0.3400580883026123, "grad_norm": 139.59011840820312, "learning_rate": 3.349364905389032e-08, "logits/chosen": 0.2000666856765747, "logits/rejected": 0.15850435197353363, "logps/chosen": -53.24721145629883, "logps/ref_chosen": -50.19850158691406, "logps/ref_rejected": -66.76687622070312, "logps/rejected": -71.4933090209961, "loss": 1.0943, "margin_dpo/margin_mean": 1.677725076675415, "margin_dpo/margin_std": 2.819387912750244, "step": 563 }, { "epoch": 0.8526077097505669, "fcm_dpo/beta": 0.5817031264305115, "fcm_dpo/delta": -0.0036928579211235046, "fcm_dpo/margin": 1.7244809865951538, "fcm_dpo/q_t": 0.3220234513282776, "grad_norm": 143.81246948242188, "learning_rate": 3.283557064487785e-08, "logits/chosen": 0.1359567940235138, "logits/rejected": 0.10976006835699081, "logps/chosen": -58.42416000366211, "logps/ref_chosen": -55.7408447265625, "logps/ref_rejected": -74.82323455810547, "logps/rejected": -79.23103332519531, "loss": 0.9548, "margin_dpo/margin_mean": 1.7244811058044434, "margin_dpo/margin_std": 2.30587100982666, "step": 564 }, { "epoch": 0.854119425547997, "fcm_dpo/beta": 0.5901836156845093, "fcm_dpo/delta": 0.04009624570608139, "fcm_dpo/margin": 1.627647876739502, "fcm_dpo/q_t": 0.32860618829727173, "grad_norm": 144.67605590820312, "learning_rate": 3.218356679178252e-08, "logits/chosen": 0.16707636415958405, "logits/rejected": 0.12742547690868378, "logps/chosen": -61.788658142089844, "logps/ref_chosen": -58.33738327026367, "logps/ref_rejected": -78.31776428222656, "logps/rejected": -83.39668273925781, "loss": 0.9857, "margin_dpo/margin_mean": 1.627647876739502, "margin_dpo/margin_std": 2.2914857864379883, "step": 565 }, { "epoch": 0.8556311413454271, "fcm_dpo/beta": 0.6040189266204834, "fcm_dpo/delta": 0.15839658677577972, "fcm_dpo/margin": 1.4104197025299072, "fcm_dpo/q_t": 0.3656797707080841, "grad_norm": 174.7579345703125, "learning_rate": 3.1537655732553764e-08, "logits/chosen": 0.1992546021938324, "logits/rejected": 0.1789197027683258, "logps/chosen": -74.33102416992188, "logps/ref_chosen": -71.22373962402344, "logps/ref_rejected": -71.11601257324219, "logps/rejected": -75.63371276855469, "loss": 1.2269, "margin_dpo/margin_mean": 1.4104200601577759, "margin_dpo/margin_std": 2.8447115421295166, "step": 566 }, { "epoch": 0.8571428571428571, "fcm_dpo/beta": 0.579740047454834, "fcm_dpo/delta": -0.17305630445480347, "fcm_dpo/margin": 1.978826880455017, "fcm_dpo/q_t": 0.3130984902381897, "grad_norm": 123.3370132446289, "learning_rate": 3.089785553471233e-08, "logits/chosen": 0.1904851794242859, "logits/rejected": 0.11687320470809937, "logps/chosen": -55.760459899902344, "logps/ref_chosen": -52.669273376464844, "logps/ref_rejected": -74.34785461425781, "logps/rejected": -79.41786193847656, "loss": 0.8807, "margin_dpo/margin_mean": 1.9788269996643066, "margin_dpo/margin_std": 2.6479549407958984, "step": 567 }, { "epoch": 0.8586545729402872, "fcm_dpo/beta": 0.5765559673309326, "fcm_dpo/delta": -0.18778733909130096, "fcm_dpo/margin": 2.0179710388183594, "fcm_dpo/q_t": 0.29743558168411255, "grad_norm": 112.51110076904297, "learning_rate": 3.026418409484513e-08, "logits/chosen": 0.20747870206832886, "logits/rejected": 0.14340844750404358, "logps/chosen": -54.99408721923828, "logps/ref_chosen": -52.178001403808594, "logps/ref_rejected": -85.8277587890625, "logps/rejected": -90.66182708740234, "loss": 0.8043, "margin_dpo/margin_mean": 2.017970561981201, "margin_dpo/margin_std": 2.1505703926086426, "step": 568 }, { "epoch": 0.8601662887377173, "fcm_dpo/beta": 0.5608316659927368, "fcm_dpo/delta": 0.08226889371871948, "fcm_dpo/margin": 1.6459429264068604, "fcm_dpo/q_t": 0.32129180431365967, "grad_norm": 118.24040222167969, "learning_rate": 2.963665913810451e-08, "logits/chosen": 0.10791079699993134, "logits/rejected": 0.08581716567277908, "logps/chosen": -65.56758117675781, "logps/ref_chosen": -62.649261474609375, "logps/ref_rejected": -75.4298324584961, "logps/rejected": -79.99409484863281, "loss": 0.9399, "margin_dpo/margin_mean": 1.6459429264068604, "margin_dpo/margin_std": 2.062713623046875, "step": 569 }, { "epoch": 0.8616780045351474, "fcm_dpo/beta": 0.552453875541687, "fcm_dpo/delta": -0.3056313693523407, "fcm_dpo/margin": 2.29406476020813, "fcm_dpo/q_t": 0.271054744720459, "grad_norm": 103.41246795654297, "learning_rate": 2.9015298217712453e-08, "logits/chosen": 0.12330185621976852, "logits/rejected": 0.05149267241358757, "logps/chosen": -52.63910675048828, "logps/ref_chosen": -50.04179382324219, "logps/ref_rejected": -78.27146911621094, "logps/rejected": -83.162841796875, "loss": 0.7286, "margin_dpo/margin_mean": 2.294064521789551, "margin_dpo/margin_std": 2.158341884613037, "step": 570 }, { "epoch": 0.8631897203325775, "fcm_dpo/beta": 0.5561559200286865, "fcm_dpo/delta": 0.2247203141450882, "fcm_dpo/margin": 1.4249439239501953, "fcm_dpo/q_t": 0.35770949721336365, "grad_norm": 122.92935180664062, "learning_rate": 2.840011871446962e-08, "logits/chosen": 0.1378299742937088, "logits/rejected": 0.10836352407932281, "logps/chosen": -56.67723083496094, "logps/ref_chosen": -53.65681457519531, "logps/ref_rejected": -66.13298034667969, "logps/rejected": -70.57833862304688, "loss": 1.0635, "margin_dpo/margin_mean": 1.4249444007873535, "margin_dpo/margin_std": 2.3282651901245117, "step": 571 }, { "epoch": 0.8647014361300076, "fcm_dpo/beta": 0.5793824195861816, "fcm_dpo/delta": 0.19740980863571167, "fcm_dpo/margin": 1.4121860265731812, "fcm_dpo/q_t": 0.3518349528312683, "grad_norm": 155.68394470214844, "learning_rate": 2.7791137836269158e-08, "logits/chosen": 0.18134906888008118, "logits/rejected": 0.20107844471931458, "logps/chosen": -77.97372436523438, "logps/ref_chosen": -74.81792449951172, "logps/ref_rejected": -65.88681030273438, "logps/rejected": -70.45478820800781, "loss": 1.0227, "margin_dpo/margin_mean": 1.4121863842010498, "margin_dpo/margin_std": 2.1890487670898438, "step": 572 }, { "epoch": 0.8662131519274376, "fcm_dpo/beta": 0.5722247362136841, "fcm_dpo/delta": -0.10332206636667252, "fcm_dpo/margin": 1.9067519903182983, "fcm_dpo/q_t": 0.3288191854953766, "grad_norm": 141.57550048828125, "learning_rate": 2.718837261761528e-08, "logits/chosen": 0.16030460596084595, "logits/rejected": 0.12534594535827637, "logps/chosen": -71.95037078857422, "logps/ref_chosen": -68.72564697265625, "logps/ref_rejected": -88.16201782226562, "logps/rejected": -93.29348754882812, "loss": 1.0156, "margin_dpo/margin_mean": 1.9067527055740356, "margin_dpo/margin_std": 2.9433274269104004, "step": 573 }, { "epoch": 0.8677248677248677, "fcm_dpo/beta": 0.5721937417984009, "fcm_dpo/delta": -0.07496052980422974, "fcm_dpo/margin": 1.8618470430374146, "fcm_dpo/q_t": 0.3078283965587616, "grad_norm": 114.2187271118164, "learning_rate": 2.659183991914696e-08, "logits/chosen": 0.2320360541343689, "logits/rejected": 0.17432260513305664, "logps/chosen": -59.44536209106445, "logps/ref_chosen": -56.31340026855469, "logps/ref_rejected": -83.91553497314453, "logps/rejected": -88.90934753417969, "loss": 0.8193, "margin_dpo/margin_mean": 1.8618476390838623, "margin_dpo/margin_std": 2.011713981628418, "step": 574 }, { "epoch": 0.8692365835222978, "fcm_dpo/beta": 0.5716170072555542, "fcm_dpo/delta": 0.18006719648838043, "fcm_dpo/margin": 1.45901620388031, "fcm_dpo/q_t": 0.3629787862300873, "grad_norm": 139.16409301757812, "learning_rate": 2.600155642716606e-08, "logits/chosen": 0.24198350310325623, "logits/rejected": 0.18358173966407776, "logps/chosen": -67.59683990478516, "logps/ref_chosen": -64.5841293334961, "logps/ref_rejected": -93.47034454345703, "logps/rejected": -97.94207000732422, "loss": 1.1015, "margin_dpo/margin_mean": 1.4590164422988892, "margin_dpo/margin_std": 2.4919562339782715, "step": 575 }, { "epoch": 0.8707482993197279, "fcm_dpo/beta": 0.5859131813049316, "fcm_dpo/delta": 0.05836522579193115, "fcm_dpo/margin": 1.6160252094268799, "fcm_dpo/q_t": 0.34040236473083496, "grad_norm": 173.32106018066406, "learning_rate": 2.5417538653170754e-08, "logits/chosen": 0.18478666245937347, "logits/rejected": 0.1120123565196991, "logps/chosen": -55.915428161621094, "logps/ref_chosen": -53.28052520751953, "logps/ref_rejected": -84.2000503540039, "logps/rejected": -88.45097351074219, "loss": 1.0198, "margin_dpo/margin_mean": 1.6160247325897217, "margin_dpo/margin_std": 2.3769047260284424, "step": 576 }, { "epoch": 0.872260015117158, "fcm_dpo/beta": 0.6100517511367798, "fcm_dpo/delta": 0.2396540343761444, "fcm_dpo/margin": 1.2779208421707153, "fcm_dpo/q_t": 0.36663320660591125, "grad_norm": 141.51870727539062, "learning_rate": 2.4839802933393607e-08, "logits/chosen": 0.15433327853679657, "logits/rejected": 0.13624653220176697, "logps/chosen": -65.23735046386719, "logps/ref_chosen": -62.32468795776367, "logps/ref_rejected": -67.300537109375, "logps/rejected": -71.49111938476562, "loss": 1.1174, "margin_dpo/margin_mean": 1.2779215574264526, "margin_dpo/margin_std": 2.3165364265441895, "step": 577 }, { "epoch": 0.873771730914588, "fcm_dpo/beta": 0.6398091316223145, "fcm_dpo/delta": 0.2764911949634552, "fcm_dpo/margin": 1.1640191078186035, "fcm_dpo/q_t": 0.3820345997810364, "grad_norm": 158.93179321289062, "learning_rate": 2.4268365428344733e-08, "logits/chosen": 0.20606505870819092, "logits/rejected": 0.1820131242275238, "logps/chosen": -59.829124450683594, "logps/ref_chosen": -56.65557861328125, "logps/ref_rejected": -68.21835327148438, "logps/rejected": -72.55591583251953, "loss": 1.1219, "margin_dpo/margin_mean": 1.1640193462371826, "margin_dpo/margin_std": 2.143388271331787, "step": 578 }, { "epoch": 0.8752834467120182, "fcm_dpo/beta": 0.6147720217704773, "fcm_dpo/delta": -0.3748418092727661, "fcm_dpo/margin": 2.1486082077026367, "fcm_dpo/q_t": 0.2804148197174072, "grad_norm": 129.94529724121094, "learning_rate": 2.3703242122359357e-08, "logits/chosen": 0.1448242962360382, "logits/rejected": 0.1141422688961029, "logps/chosen": -59.766502380371094, "logps/ref_chosen": -56.809661865234375, "logps/ref_rejected": -68.09613037109375, "logps/rejected": -73.20157623291016, "loss": 0.779, "margin_dpo/margin_mean": 2.1486077308654785, "margin_dpo/margin_std": 2.415778636932373, "step": 579 }, { "epoch": 0.8767951625094482, "fcm_dpo/beta": 0.6254961490631104, "fcm_dpo/delta": 0.14091333746910095, "fcm_dpo/margin": 1.389689326286316, "fcm_dpo/q_t": 0.3568766117095947, "grad_norm": 135.1801300048828, "learning_rate": 2.3144448823151392e-08, "logits/chosen": 0.1448507308959961, "logits/rejected": 0.10793846845626831, "logps/chosen": -60.63453674316406, "logps/ref_chosen": -57.70011520385742, "logps/ref_rejected": -77.90664672851562, "logps/rejected": -82.23075866699219, "loss": 1.1175, "margin_dpo/margin_mean": 1.3896892070770264, "margin_dpo/margin_std": 2.38545823097229, "step": 580 }, { "epoch": 0.8783068783068783, "fcm_dpo/beta": 0.6277990341186523, "fcm_dpo/delta": 0.06906095892190933, "fcm_dpo/margin": 1.492587924003601, "fcm_dpo/q_t": 0.3425326943397522, "grad_norm": 166.3179473876953, "learning_rate": 2.259200116137039e-08, "logits/chosen": 0.22638601064682007, "logits/rejected": 0.17741230130195618, "logps/chosen": -62.56743240356445, "logps/ref_chosen": -59.332359313964844, "logps/ref_rejected": -83.64482116699219, "logps/rejected": -88.37248229980469, "loss": 1.0185, "margin_dpo/margin_mean": 1.4925878047943115, "margin_dpo/margin_std": 2.3351616859436035, "step": 581 }, { "epoch": 0.8798185941043084, "fcm_dpo/beta": 0.6253724098205566, "fcm_dpo/delta": -0.07393016666173935, "fcm_dpo/margin": 1.7041115760803223, "fcm_dpo/q_t": 0.31156644225120544, "grad_norm": 143.69277954101562, "learning_rate": 2.204591459016525e-08, "logits/chosen": 0.1830708086490631, "logits/rejected": 0.19862952828407288, "logps/chosen": -66.9874267578125, "logps/ref_chosen": -64.16285705566406, "logps/ref_rejected": -58.632896423339844, "logps/rejected": -63.161582946777344, "loss": 0.8907, "margin_dpo/margin_mean": 1.7041112184524536, "margin_dpo/margin_std": 2.1104648113250732, "step": 582 }, { "epoch": 0.8813303099017384, "fcm_dpo/beta": 0.6326082944869995, "fcm_dpo/delta": 0.08823379874229431, "fcm_dpo/margin": 1.4541677236557007, "fcm_dpo/q_t": 0.35390377044677734, "grad_norm": 158.56170654296875, "learning_rate": 2.1506204384751064e-08, "logits/chosen": 0.27797916531562805, "logits/rejected": 0.20349468290805817, "logps/chosen": -54.86329650878906, "logps/ref_chosen": -51.87239456176758, "logps/ref_rejected": -83.86331176757812, "logps/rejected": -88.30838012695312, "loss": 1.1301, "margin_dpo/margin_mean": 1.4541676044464111, "margin_dpo/margin_std": 2.6575491428375244, "step": 583 }, { "epoch": 0.8828420256991686, "fcm_dpo/beta": 0.6200550198554993, "fcm_dpo/delta": -0.1539473831653595, "fcm_dpo/margin": 1.8320786952972412, "fcm_dpo/q_t": 0.32225194573402405, "grad_norm": 134.9540252685547, "learning_rate": 2.09728856419826e-08, "logits/chosen": 0.23410704731941223, "logits/rejected": 0.15715327858924866, "logps/chosen": -49.28221130371094, "logps/ref_chosen": -46.571388244628906, "logps/ref_rejected": -80.67969512939453, "logps/rejected": -85.22259521484375, "loss": 0.9929, "margin_dpo/margin_mean": 1.8320791721343994, "margin_dpo/margin_std": 2.6820805072784424, "step": 584 }, { "epoch": 0.8843537414965986, "fcm_dpo/beta": 0.6162490844726562, "fcm_dpo/delta": 0.03570966795086861, "fcm_dpo/margin": 1.5696362257003784, "fcm_dpo/q_t": 0.3259866535663605, "grad_norm": 138.78614807128906, "learning_rate": 2.044597327993153e-08, "logits/chosen": 0.16352935135364532, "logits/rejected": 0.13112275302410126, "logps/chosen": -60.831146240234375, "logps/ref_chosen": -58.124534606933594, "logps/ref_rejected": -79.00538635253906, "logps/rejected": -83.28163146972656, "loss": 0.9773, "margin_dpo/margin_mean": 1.5696361064910889, "margin_dpo/margin_std": 2.2082860469818115, "step": 585 }, { "epoch": 0.8858654572940288, "fcm_dpo/beta": 0.602626621723175, "fcm_dpo/delta": -0.15734781324863434, "fcm_dpo/margin": 1.8877205848693848, "fcm_dpo/q_t": 0.30060410499572754, "grad_norm": 121.3906021118164, "learning_rate": 1.9925482037469187e-08, "logits/chosen": 0.16395865380764008, "logits/rejected": 0.12222093343734741, "logps/chosen": -57.20876693725586, "logps/ref_chosen": -54.10163879394531, "logps/ref_rejected": -63.72113037109375, "logps/rejected": -68.71598052978516, "loss": 0.8797, "margin_dpo/margin_mean": 1.8877204656600952, "margin_dpo/margin_std": 2.3431496620178223, "step": 586 }, { "epoch": 0.8873771730914588, "fcm_dpo/beta": 0.6007733941078186, "fcm_dpo/delta": 0.015819646418094635, "fcm_dpo/margin": 1.6403778791427612, "fcm_dpo/q_t": 0.3280307352542877, "grad_norm": 161.0563507080078, "learning_rate": 1.9411426473854687e-08, "logits/chosen": 0.20693828165531158, "logits/rejected": 0.18662090599536896, "logps/chosen": -66.15540313720703, "logps/ref_chosen": -63.41719436645508, "logps/ref_rejected": -63.47003936767578, "logps/rejected": -67.84861755371094, "loss": 1.0373, "margin_dpo/margin_mean": 1.6403785943984985, "margin_dpo/margin_std": 2.5440006256103516, "step": 587 }, { "epoch": 0.8888888888888888, "fcm_dpo/beta": 0.5988498330116272, "fcm_dpo/delta": -0.019005782902240753, "fcm_dpo/margin": 1.6973689794540405, "fcm_dpo/q_t": 0.3193732500076294, "grad_norm": 139.44850158691406, "learning_rate": 1.890382096832699e-08, "logits/chosen": 0.2137565314769745, "logits/rejected": 0.17403748631477356, "logps/chosen": -65.03866577148438, "logps/ref_chosen": -62.20103454589844, "logps/ref_rejected": -82.10249328613281, "logps/rejected": -86.63749694824219, "loss": 0.9558, "margin_dpo/margin_mean": 1.697368860244751, "margin_dpo/margin_std": 2.3547918796539307, "step": 588 }, { "epoch": 0.890400604686319, "fcm_dpo/beta": 0.5705356597900391, "fcm_dpo/delta": -0.3349605202674866, "fcm_dpo/margin": 2.2617807388305664, "fcm_dpo/q_t": 0.2769893407821655, "grad_norm": 125.83250427246094, "learning_rate": 1.840267971970344e-08, "logits/chosen": 0.16948330402374268, "logits/rejected": 0.14072127640247345, "logps/chosen": -59.358924865722656, "logps/ref_chosen": -56.71361541748047, "logps/ref_rejected": -76.7366943359375, "logps/rejected": -81.64378356933594, "loss": 0.7735, "margin_dpo/margin_mean": 2.2617812156677246, "margin_dpo/margin_std": 2.493180274963379, "step": 589 }, { "epoch": 0.891912320483749, "fcm_dpo/beta": 0.5450751781463623, "fcm_dpo/delta": -0.2230319082736969, "fcm_dpo/margin": 2.1952528953552246, "fcm_dpo/q_t": 0.2884211540222168, "grad_norm": 120.71878814697266, "learning_rate": 1.7908016745981856e-08, "logits/chosen": 0.12301607429981232, "logits/rejected": 0.09767352789640427, "logps/chosen": -69.55258178710938, "logps/ref_chosen": -66.5138168334961, "logps/ref_rejected": -85.70820617675781, "logps/rejected": -90.94223022460938, "loss": 0.8243, "margin_dpo/margin_mean": 2.1952528953552246, "margin_dpo/margin_std": 2.4201831817626953, "step": 590 }, { "epoch": 0.8934240362811792, "fcm_dpo/beta": 0.5360534191131592, "fcm_dpo/delta": 0.06881964951753616, "fcm_dpo/margin": 1.7440357208251953, "fcm_dpo/q_t": 0.3370535969734192, "grad_norm": 125.32640838623047, "learning_rate": 1.7419845883949098e-08, "logits/chosen": 0.2466006577014923, "logits/rejected": 0.1991560012102127, "logps/chosen": -63.56512451171875, "logps/ref_chosen": -60.697181701660156, "logps/ref_rejected": -86.12278747558594, "logps/rejected": -90.73477172851562, "loss": 0.9852, "margin_dpo/margin_mean": 1.7440353631973267, "margin_dpo/margin_std": 2.4317829608917236, "step": 591 }, { "epoch": 0.8949357520786092, "fcm_dpo/beta": 0.5456318855285645, "fcm_dpo/delta": 0.048897288739681244, "fcm_dpo/margin": 1.750661849975586, "fcm_dpo/q_t": 0.3427332043647766, "grad_norm": 132.4198760986328, "learning_rate": 1.6938180788793556e-08, "logits/chosen": 0.18844714760780334, "logits/rejected": 0.10343804955482483, "logps/chosen": -54.16868591308594, "logps/ref_chosen": -51.237327575683594, "logps/ref_rejected": -81.60242462158203, "logps/rejected": -86.2844467163086, "loss": 0.9699, "margin_dpo/margin_mean": 1.750661849975586, "margin_dpo/margin_std": 2.502837657928467, "step": 592 }, { "epoch": 0.8964474678760394, "fcm_dpo/beta": 0.5585802793502808, "fcm_dpo/delta": 0.09174495190382004, "fcm_dpo/margin": 1.640941858291626, "fcm_dpo/q_t": 0.3366505205631256, "grad_norm": 122.52076721191406, "learning_rate": 1.6463034933723336e-08, "logits/chosen": 0.13056322932243347, "logits/rejected": 0.061800211668014526, "logps/chosen": -44.626976013183594, "logps/ref_chosen": -42.08000183105469, "logps/ref_rejected": -68.47499084472656, "logps/rejected": -72.66290283203125, "loss": 1.0262, "margin_dpo/margin_mean": 1.640941858291626, "margin_dpo/margin_std": 2.4754600524902344, "step": 593 }, { "epoch": 0.8979591836734694, "fcm_dpo/beta": 0.5791330337524414, "fcm_dpo/delta": 0.23276039958000183, "fcm_dpo/margin": 1.3571163415908813, "fcm_dpo/q_t": 0.36690980195999146, "grad_norm": 131.3136749267578, "learning_rate": 1.5994421609589385e-08, "logits/chosen": 0.12303361296653748, "logits/rejected": 0.10529161989688873, "logps/chosen": -66.68218994140625, "logps/ref_chosen": -63.658668518066406, "logps/ref_rejected": -70.35597229003906, "logps/rejected": -74.73660278320312, "loss": 1.0883, "margin_dpo/margin_mean": 1.3571163415908813, "margin_dpo/margin_std": 2.378408193588257, "step": 594 }, { "epoch": 0.8994708994708994, "fcm_dpo/beta": 0.5620189905166626, "fcm_dpo/delta": -0.30020615458488464, "fcm_dpo/margin": 2.245738983154297, "fcm_dpo/q_t": 0.2832089960575104, "grad_norm": 123.73275756835938, "learning_rate": 1.553235392451377e-08, "logits/chosen": 0.2004849910736084, "logits/rejected": 0.12713924050331116, "logps/chosen": -59.19725799560547, "logps/ref_chosen": -56.21875762939453, "logps/ref_rejected": -83.95773315429688, "logps/rejected": -89.18197631835938, "loss": 0.8647, "margin_dpo/margin_mean": 2.245739459991455, "margin_dpo/margin_std": 2.714977741241455, "step": 595 }, { "epoch": 0.9009826152683296, "fcm_dpo/beta": 0.5812000036239624, "fcm_dpo/delta": 0.33480995893478394, "fcm_dpo/margin": 1.184531569480896, "fcm_dpo/q_t": 0.3911210000514984, "grad_norm": 142.58895874023438, "learning_rate": 1.507684480352292e-08, "logits/chosen": 0.11143651604652405, "logits/rejected": 0.11058718711137772, "logps/chosen": -71.51290130615234, "logps/ref_chosen": -68.48088073730469, "logps/ref_rejected": -61.732967376708984, "logps/rejected": -65.94952392578125, "loss": 1.1753, "margin_dpo/margin_mean": 1.1845312118530273, "margin_dpo/margin_std": 2.4391579627990723, "step": 596 }, { "epoch": 0.9024943310657596, "fcm_dpo/beta": 0.5947533249855042, "fcm_dpo/delta": 0.0645606741309166, "fcm_dpo/margin": 1.5828521251678467, "fcm_dpo/q_t": 0.3390669822692871, "grad_norm": 118.60302734375, "learning_rate": 1.4627906988186111e-08, "logits/chosen": 0.1252235770225525, "logits/rejected": 0.09921297430992126, "logps/chosen": -51.60657501220703, "logps/ref_chosen": -48.85750961303711, "logps/ref_rejected": -55.068084716796875, "logps/rejected": -59.400001525878906, "loss": 0.963, "margin_dpo/margin_mean": 1.5828520059585571, "margin_dpo/margin_std": 2.2610087394714355, "step": 597 }, { "epoch": 0.9040060468631897, "fcm_dpo/beta": 0.6207314729690552, "fcm_dpo/delta": 0.23146888613700867, "fcm_dpo/margin": 1.2671819925308228, "fcm_dpo/q_t": 0.38011521100997925, "grad_norm": 151.75205993652344, "learning_rate": 1.4185553036259095e-08, "logits/chosen": 0.1728542447090149, "logits/rejected": 0.11698366701602936, "logps/chosen": -62.22666931152344, "logps/ref_chosen": -58.88715362548828, "logps/ref_rejected": -81.43145751953125, "logps/rejected": -86.03816223144531, "loss": 1.1674, "margin_dpo/margin_mean": 1.2671819925308228, "margin_dpo/margin_std": 2.4872426986694336, "step": 598 }, { "epoch": 0.9055177626606198, "fcm_dpo/beta": 0.6450425386428833, "fcm_dpo/delta": 0.09236402064561844, "fcm_dpo/margin": 1.41303288936615, "fcm_dpo/q_t": 0.3642594814300537, "grad_norm": 166.78338623046875, "learning_rate": 1.3749795321332885e-08, "logits/chosen": 0.280083566904068, "logits/rejected": 0.24155710637569427, "logps/chosen": -60.974185943603516, "logps/ref_chosen": -57.60719299316406, "logps/ref_rejected": -71.80469512939453, "logps/rejected": -76.584716796875, "loss": 1.1191, "margin_dpo/margin_mean": 1.4130332469940186, "margin_dpo/margin_std": 2.563547134399414, "step": 599 }, { "epoch": 0.9070294784580499, "fcm_dpo/beta": 0.6648072004318237, "fcm_dpo/delta": 0.13435859978199005, "fcm_dpo/margin": 1.3058445453643799, "fcm_dpo/q_t": 0.35756915807724, "grad_norm": 156.00218200683594, "learning_rate": 1.3320646032487393e-08, "logits/chosen": 0.2286624014377594, "logits/rejected": 0.19494396448135376, "logps/chosen": -61.69401550292969, "logps/ref_chosen": -58.44231414794922, "logps/ref_rejected": -83.64639282226562, "logps/rejected": -88.20393371582031, "loss": 1.0999, "margin_dpo/margin_mean": 1.3058440685272217, "margin_dpo/margin_std": 2.190995693206787, "step": 600 }, { "epoch": 0.9070294784580499, "eval_fcm_dpo/beta": 0.6568994522094727, "eval_logits/chosen": 0.19621142745018005, "eval_logits/rejected": 0.15854774415493011, "eval_logps/chosen": -77.82337951660156, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -83.91117095947266, "eval_loss": 0.5679579377174377, "eval_margin_dpo/margin_mean": 1.3982763290405273, "eval_margin_dpo/margin_std": 2.4811148643493652, "eval_runtime": 37.9915, "eval_samples_per_second": 60.619, "eval_steps_per_second": 1.895, "step": 600 }, { "epoch": 0.90854119425548, "fcm_dpo/beta": 0.6349242925643921, "fcm_dpo/delta": -0.1545478105545044, "fcm_dpo/margin": 1.7845826148986816, "fcm_dpo/q_t": 0.3199031949043274, "grad_norm": 141.2012176513672, "learning_rate": 1.2898117173950868e-08, "logits/chosen": 0.18726052343845367, "logits/rejected": 0.13292908668518066, "logps/chosen": -58.19509506225586, "logps/ref_chosen": -55.59432601928711, "logps/ref_rejected": -83.68630981445312, "logps/rejected": -88.07167053222656, "loss": 0.9784, "margin_dpo/margin_mean": 1.7845829725265503, "margin_dpo/margin_std": 2.5417721271514893, "step": 601 }, { "epoch": 0.91005291005291, "fcm_dpo/beta": 0.6274067759513855, "fcm_dpo/delta": -0.15826797485351562, "fcm_dpo/margin": 1.8172000646591187, "fcm_dpo/q_t": 0.3086685538291931, "grad_norm": 139.8549346923828, "learning_rate": 1.2482220564763667e-08, "logits/chosen": 0.15303432941436768, "logits/rejected": 0.11989377439022064, "logps/chosen": -58.714149475097656, "logps/ref_chosen": -56.349185943603516, "logps/ref_rejected": -71.9959716796875, "logps/rejected": -76.17813110351562, "loss": 0.887, "margin_dpo/margin_mean": 1.81719970703125, "margin_dpo/margin_std": 2.2619805335998535, "step": 602 }, { "epoch": 0.9115646258503401, "fcm_dpo/beta": 0.6145593523979187, "fcm_dpo/delta": 0.003954831510782242, "fcm_dpo/margin": 1.620744228363037, "fcm_dpo/q_t": 0.32022157311439514, "grad_norm": 121.14891052246094, "learning_rate": 1.2072967838448051e-08, "logits/chosen": 0.13416381180286407, "logits/rejected": 0.08777206391096115, "logps/chosen": -55.47536849975586, "logps/ref_chosen": -53.16838836669922, "logps/ref_rejected": -73.8604736328125, "logps/rejected": -77.78820037841797, "loss": 0.9039, "margin_dpo/margin_mean": 1.620744228363037, "margin_dpo/margin_std": 2.0299124717712402, "step": 603 }, { "epoch": 0.9130763416477702, "fcm_dpo/beta": 0.6099786162376404, "fcm_dpo/delta": -0.06591600924730301, "fcm_dpo/margin": 1.7355551719665527, "fcm_dpo/q_t": 0.3229469954967499, "grad_norm": 138.081298828125, "learning_rate": 1.1670370442682459e-08, "logits/chosen": 0.1237378865480423, "logits/rejected": 0.11915041506290436, "logps/chosen": -75.28321838378906, "logps/ref_chosen": -72.64942169189453, "logps/ref_rejected": -69.8792724609375, "logps/rejected": -74.24861907958984, "loss": 0.9893, "margin_dpo/margin_mean": 1.7355563640594482, "margin_dpo/margin_std": 2.5649847984313965, "step": 604 }, { "epoch": 0.9145880574452003, "fcm_dpo/beta": 0.6022388339042664, "fcm_dpo/delta": -0.05076969414949417, "fcm_dpo/margin": 1.7350056171417236, "fcm_dpo/q_t": 0.3227270543575287, "grad_norm": 143.35333251953125, "learning_rate": 1.1274439638981532e-08, "logits/chosen": 0.22395864129066467, "logits/rejected": 0.17329740524291992, "logps/chosen": -64.78287506103516, "logps/ref_chosen": -61.61284637451172, "logps/ref_rejected": -79.34398651123047, "logps/rejected": -84.24903106689453, "loss": 0.9715, "margin_dpo/margin_mean": 1.7350056171417236, "margin_dpo/margin_std": 2.434452772140503, "step": 605 }, { "epoch": 0.9160997732426304, "fcm_dpo/beta": 0.6224014759063721, "fcm_dpo/delta": 0.12130826711654663, "fcm_dpo/margin": 1.4212950468063354, "fcm_dpo/q_t": 0.35074251890182495, "grad_norm": 151.14439392089844, "learning_rate": 1.0885186502381016e-08, "logits/chosen": 0.19038084149360657, "logits/rejected": 0.14078065752983093, "logps/chosen": -57.25608825683594, "logps/ref_chosen": -54.46424102783203, "logps/ref_rejected": -79.62708282470703, "logps/rejected": -83.84022521972656, "loss": 1.0273, "margin_dpo/margin_mean": 1.421295404434204, "margin_dpo/margin_std": 2.172295331954956, "step": 606 }, { "epoch": 0.9176114890400605, "fcm_dpo/beta": 0.5959450006484985, "fcm_dpo/delta": -0.13134704530239105, "fcm_dpo/margin": 1.8611319065093994, "fcm_dpo/q_t": 0.31175029277801514, "grad_norm": 141.19546508789062, "learning_rate": 1.0502621921127774e-08, "logits/chosen": 0.14506211876869202, "logits/rejected": 0.1156059056520462, "logps/chosen": -65.67162322998047, "logps/ref_chosen": -62.86086654663086, "logps/ref_rejected": -72.5501937866211, "logps/rejected": -77.22207641601562, "loss": 0.9338, "margin_dpo/margin_mean": 1.8611321449279785, "margin_dpo/margin_std": 2.463554859161377, "step": 607 }, { "epoch": 0.9191232048374905, "fcm_dpo/beta": 0.6229555606842041, "fcm_dpo/delta": 0.24546337127685547, "fcm_dpo/margin": 1.2424054145812988, "fcm_dpo/q_t": 0.36555221676826477, "grad_norm": 181.87330627441406, "learning_rate": 1.0126756596375685e-08, "logits/chosen": 0.16855892539024353, "logits/rejected": 0.12452598661184311, "logps/chosen": -66.26856994628906, "logps/ref_chosen": -63.18071746826172, "logps/ref_rejected": -99.15888214111328, "logps/rejected": -103.48914337158203, "loss": 1.0982, "margin_dpo/margin_mean": 1.2424057722091675, "margin_dpo/margin_std": 2.158616542816162, "step": 608 }, { "epoch": 0.9206349206349206, "fcm_dpo/beta": 0.600549578666687, "fcm_dpo/delta": -0.21868403255939484, "fcm_dpo/margin": 1.9679713249206543, "fcm_dpo/q_t": 0.2971384525299072, "grad_norm": 103.79540252685547, "learning_rate": 9.757601041885694e-09, "logits/chosen": 0.21697968244552612, "logits/rejected": 0.18763911724090576, "logps/chosen": -51.205020904541016, "logps/ref_chosen": -48.62322235107422, "logps/ref_rejected": -68.28271484375, "logps/rejected": -72.83248901367188, "loss": 0.8665, "margin_dpo/margin_mean": 1.9679714441299438, "margin_dpo/margin_std": 2.3338661193847656, "step": 609 }, { "epoch": 0.9221466364323507, "fcm_dpo/beta": 0.6042653322219849, "fcm_dpo/delta": -0.03405376523733139, "fcm_dpo/margin": 1.7052839994430542, "fcm_dpo/q_t": 0.3533821403980255, "grad_norm": 194.40943908691406, "learning_rate": 9.395165583732379e-09, "logits/chosen": 0.1579647958278656, "logits/rejected": 0.1604270040988922, "logps/chosen": -75.84585571289062, "logps/ref_chosen": -72.66513061523438, "logps/ref_rejected": -87.15310668945312, "logps/rejected": -92.03912353515625, "loss": 1.1081, "margin_dpo/margin_mean": 1.7052838802337646, "margin_dpo/margin_std": 3.0106663703918457, "step": 610 }, { "epoch": 0.9236583522297808, "fcm_dpo/beta": 0.6121885776519775, "fcm_dpo/delta": 0.25364020466804504, "fcm_dpo/margin": 1.2458240985870361, "fcm_dpo/q_t": 0.36857932806015015, "grad_norm": 133.9265899658203, "learning_rate": 9.03946036001449e-09, "logits/chosen": 0.2268466055393219, "logits/rejected": 0.1877264380455017, "logps/chosen": -51.286346435546875, "logps/ref_chosen": -48.30857849121094, "logps/ref_rejected": -70.6141128540039, "logps/rejected": -74.83771514892578, "loss": 1.0992, "margin_dpo/margin_mean": 1.2458235025405884, "margin_dpo/margin_std": 2.115691661834717, "step": 611 }, { "epoch": 0.9251700680272109, "fcm_dpo/beta": 0.6135225296020508, "fcm_dpo/delta": -0.24519576132297516, "fcm_dpo/margin": 1.9819927215576172, "fcm_dpo/q_t": 0.30337560176849365, "grad_norm": 141.30982971191406, "learning_rate": 8.690495320571839e-09, "logits/chosen": 0.10801693797111511, "logits/rejected": 0.059915874153375626, "logps/chosen": -64.04043579101562, "logps/ref_chosen": -61.23155975341797, "logps/ref_rejected": -94.37979888916016, "logps/rejected": -99.17066955566406, "loss": 0.9409, "margin_dpo/margin_mean": 1.9819923639297485, "margin_dpo/margin_std": 2.702681064605713, "step": 612 }, { "epoch": 0.926681783824641, "fcm_dpo/beta": 0.5761626958847046, "fcm_dpo/delta": -0.3495950698852539, "fcm_dpo/margin": 2.265277862548828, "fcm_dpo/q_t": 0.28574904799461365, "grad_norm": 127.05760955810547, "learning_rate": 8.348280226706722e-09, "logits/chosen": 0.11358965933322906, "logits/rejected": 0.09932907670736313, "logps/chosen": -56.594276428222656, "logps/ref_chosen": -53.98310852050781, "logps/ref_rejected": -58.32208251953125, "logps/rejected": -63.19852828979492, "loss": 0.8489, "margin_dpo/margin_mean": 2.2652783393859863, "margin_dpo/margin_std": 2.772947072982788, "step": 613 }, { "epoch": 0.9281934996220711, "fcm_dpo/beta": 0.569689929485321, "fcm_dpo/delta": 0.031611911952495575, "fcm_dpo/margin": 1.7016217708587646, "fcm_dpo/q_t": 0.3136371970176697, "grad_norm": 121.61019134521484, "learning_rate": 8.012824650910937e-09, "logits/chosen": 0.2070479393005371, "logits/rejected": 0.1917956918478012, "logps/chosen": -63.198665618896484, "logps/ref_chosen": -60.24303436279297, "logps/ref_rejected": -72.26258850097656, "logps/rejected": -76.91984558105469, "loss": 0.8664, "margin_dpo/margin_mean": 1.701621174812317, "margin_dpo/margin_std": 1.9565538167953491, "step": 614 }, { "epoch": 0.9297052154195011, "fcm_dpo/beta": 0.5607982873916626, "fcm_dpo/delta": -0.03319869935512543, "fcm_dpo/margin": 1.835862398147583, "fcm_dpo/q_t": 0.329486608505249, "grad_norm": 150.09515380859375, "learning_rate": 7.684137976598088e-09, "logits/chosen": 0.1885606348514557, "logits/rejected": 0.15545931458473206, "logps/chosen": -75.15232849121094, "logps/ref_chosen": -72.09467315673828, "logps/ref_rejected": -104.02980041503906, "logps/rejected": -108.92332458496094, "loss": 0.9693, "margin_dpo/margin_mean": 1.8358616828918457, "margin_dpo/margin_std": 2.62037992477417, "step": 615 }, { "epoch": 0.9312169312169312, "fcm_dpo/beta": 0.5787394046783447, "fcm_dpo/delta": 0.22798365354537964, "fcm_dpo/margin": 1.3659520149230957, "fcm_dpo/q_t": 0.36905261874198914, "grad_norm": 138.57786560058594, "learning_rate": 7.36222939784098e-09, "logits/chosen": 0.17158043384552002, "logits/rejected": 0.1132676973938942, "logps/chosen": -61.71527862548828, "logps/ref_chosen": -58.530723571777344, "logps/ref_rejected": -75.48025512695312, "logps/rejected": -80.03076171875, "loss": 1.0819, "margin_dpo/margin_mean": 1.365952491760254, "margin_dpo/margin_std": 2.3640403747558594, "step": 616 }, { "epoch": 0.9327286470143613, "fcm_dpo/beta": 0.5880202651023865, "fcm_dpo/delta": 0.12598100304603577, "fcm_dpo/margin": 1.4980671405792236, "fcm_dpo/q_t": 0.35036730766296387, "grad_norm": 136.1953887939453, "learning_rate": 7.047107919114586e-09, "logits/chosen": 0.1694568246603012, "logits/rejected": 0.1344485729932785, "logps/chosen": -60.68329620361328, "logps/ref_chosen": -57.608673095703125, "logps/ref_rejected": -81.22109985351562, "logps/rejected": -85.79379272460938, "loss": 1.0066, "margin_dpo/margin_mean": 1.4980677366256714, "margin_dpo/margin_std": 2.2009902000427246, "step": 617 }, { "epoch": 0.9342403628117913, "fcm_dpo/beta": 0.5951837301254272, "fcm_dpo/delta": 0.013019606471061707, "fcm_dpo/margin": 1.654120922088623, "fcm_dpo/q_t": 0.326251745223999, "grad_norm": 141.9059295654297, "learning_rate": 6.738782355044048e-09, "logits/chosen": 0.1482279747724533, "logits/rejected": 0.07427319884300232, "logps/chosen": -59.329376220703125, "logps/ref_chosen": -56.69594192504883, "logps/ref_rejected": -85.92362976074219, "logps/rejected": -90.21118927001953, "loss": 0.9343, "margin_dpo/margin_mean": 1.6541210412979126, "margin_dpo/margin_std": 2.201620578765869, "step": 618 }, { "epoch": 0.9357520786092215, "fcm_dpo/beta": 0.6088467836380005, "fcm_dpo/delta": 0.07530087977647781, "fcm_dpo/margin": 1.53031587600708, "fcm_dpo/q_t": 0.3448328375816345, "grad_norm": 150.22421264648438, "learning_rate": 6.437261330158206e-09, "logits/chosen": 0.22948047518730164, "logits/rejected": 0.18049129843711853, "logps/chosen": -56.79576873779297, "logps/ref_chosen": -54.05841827392578, "logps/ref_rejected": -83.55493927001953, "logps/rejected": -87.82260131835938, "loss": 1.0339, "margin_dpo/margin_mean": 1.53031587600708, "margin_dpo/margin_std": 2.370837450027466, "step": 619 }, { "epoch": 0.9372637944066515, "fcm_dpo/beta": 0.6322528123855591, "fcm_dpo/delta": 0.09385835379362106, "fcm_dpo/margin": 1.4362270832061768, "fcm_dpo/q_t": 0.3465351164340973, "grad_norm": 173.79412841796875, "learning_rate": 6.142553278648238e-09, "logits/chosen": 0.17980945110321045, "logits/rejected": 0.16840043663978577, "logps/chosen": -66.160888671875, "logps/ref_chosen": -63.36971664428711, "logps/ref_rejected": -65.68269348144531, "logps/rejected": -69.91009521484375, "loss": 1.0082, "margin_dpo/margin_mean": 1.4362270832061768, "margin_dpo/margin_std": 2.130248546600342, "step": 620 }, { "epoch": 0.9387755102040817, "fcm_dpo/beta": 0.6423999071121216, "fcm_dpo/delta": 0.0907469242811203, "fcm_dpo/margin": 1.4205418825149536, "fcm_dpo/q_t": 0.35109221935272217, "grad_norm": 151.4163055419922, "learning_rate": 5.854666444131934e-09, "logits/chosen": 0.19905002415180206, "logits/rejected": 0.12930849194526672, "logps/chosen": -55.218666076660156, "logps/ref_chosen": -52.321224212646484, "logps/ref_rejected": -88.09001159667969, "logps/rejected": -92.40798950195312, "loss": 1.0421, "margin_dpo/margin_mean": 1.420541524887085, "margin_dpo/margin_std": 2.2282516956329346, "step": 621 }, { "epoch": 0.9402872260015117, "fcm_dpo/beta": 0.640907347202301, "fcm_dpo/delta": 0.03911030665040016, "fcm_dpo/margin": 1.5050252676010132, "fcm_dpo/q_t": 0.3320958614349365, "grad_norm": 137.0905303955078, "learning_rate": 5.573608879422875e-09, "logits/chosen": 0.13063614070415497, "logits/rejected": 0.1020088791847229, "logps/chosen": -62.50457763671875, "logps/ref_chosen": -59.86545944213867, "logps/ref_rejected": -81.86668395996094, "logps/rejected": -86.01081848144531, "loss": 0.9325, "margin_dpo/margin_mean": 1.5050253868103027, "margin_dpo/margin_std": 2.0284576416015625, "step": 622 }, { "epoch": 0.9417989417989417, "fcm_dpo/beta": 0.6268334984779358, "fcm_dpo/delta": -0.16526103019714355, "fcm_dpo/margin": 1.8289923667907715, "fcm_dpo/q_t": 0.30245280265808105, "grad_norm": 126.97518920898438, "learning_rate": 5.299388446305342e-09, "logits/chosen": 0.16525626182556152, "logits/rejected": 0.11132755130529404, "logps/chosen": -70.49075317382812, "logps/ref_chosen": -67.36846160888672, "logps/ref_rejected": -82.02733612060547, "logps/rejected": -86.97863006591797, "loss": 0.8618, "margin_dpo/margin_mean": 1.828992247581482, "margin_dpo/margin_std": 2.1596312522888184, "step": 623 }, { "epoch": 0.9433106575963719, "fcm_dpo/beta": 0.6166424751281738, "fcm_dpo/delta": -0.002019442617893219, "fcm_dpo/margin": 1.6235092878341675, "fcm_dpo/q_t": 0.33338773250579834, "grad_norm": 134.533447265625, "learning_rate": 5.03201281531429e-09, "logits/chosen": 0.17911109328269958, "logits/rejected": 0.10127080231904984, "logps/chosen": -53.71633529663086, "logps/ref_chosen": -51.02655029296875, "logps/ref_rejected": -76.49203491210938, "logps/rejected": -80.80532836914062, "loss": 0.9784, "margin_dpo/margin_mean": 1.623509407043457, "margin_dpo/margin_std": 2.3413143157958984, "step": 624 }, { "epoch": 0.9448223733938019, "fcm_dpo/beta": 0.6369531154632568, "fcm_dpo/delta": 0.11569374054670334, "fcm_dpo/margin": 1.400294542312622, "fcm_dpo/q_t": 0.3541494905948639, "grad_norm": 148.3859100341797, "learning_rate": 4.7714894655209174e-09, "logits/chosen": 0.20979532599449158, "logits/rejected": 0.1446528434753418, "logps/chosen": -57.27146530151367, "logps/ref_chosen": -54.20761489868164, "logps/ref_rejected": -84.93669128417969, "logps/rejected": -89.40084075927734, "loss": 1.0516, "margin_dpo/margin_mean": 1.4002941846847534, "margin_dpo/margin_std": 2.27569317817688, "step": 625 }, { "epoch": 0.9463340891912321, "fcm_dpo/beta": 0.6106563806533813, "fcm_dpo/delta": -0.2672712206840515, "fcm_dpo/margin": 2.0213675498962402, "fcm_dpo/q_t": 0.31362709403038025, "grad_norm": 127.49701690673828, "learning_rate": 4.517825684323323e-09, "logits/chosen": 0.24240854382514954, "logits/rejected": 0.1507807821035385, "logps/chosen": -47.64917755126953, "logps/ref_chosen": -45.06201934814453, "logps/ref_rejected": -89.66368103027344, "logps/rejected": -94.27220153808594, "loss": 0.9523, "margin_dpo/margin_mean": 2.0213675498962402, "margin_dpo/margin_std": 2.8801679611206055, "step": 626 }, { "epoch": 0.9478458049886621, "fcm_dpo/beta": 0.5770605802536011, "fcm_dpo/delta": -0.23875784873962402, "fcm_dpo/margin": 2.0926780700683594, "fcm_dpo/q_t": 0.3067499101161957, "grad_norm": 135.4554901123047, "learning_rate": 4.271028567242818e-09, "logits/chosen": 0.12116050720214844, "logits/rejected": 0.04095185548067093, "logps/chosen": -61.70726776123047, "logps/ref_chosen": -58.791053771972656, "logps/ref_rejected": -94.90802001953125, "logps/rejected": -99.91691589355469, "loss": 0.888, "margin_dpo/margin_mean": 2.0926778316497803, "margin_dpo/margin_std": 2.661421298980713, "step": 627 }, { "epoch": 0.9493575207860923, "fcm_dpo/beta": 0.5617523193359375, "fcm_dpo/delta": -0.23493322730064392, "fcm_dpo/margin": 2.1469919681549072, "fcm_dpo/q_t": 0.29745668172836304, "grad_norm": 120.84452056884766, "learning_rate": 4.0311050177251895e-09, "logits/chosen": 0.19451142847537994, "logits/rejected": 0.16487011313438416, "logps/chosen": -55.42369842529297, "logps/ref_chosen": -52.80357360839844, "logps/ref_rejected": -76.49468994140625, "logps/rejected": -81.26181030273438, "loss": 0.9097, "margin_dpo/margin_mean": 2.146991729736328, "margin_dpo/margin_std": 2.7369041442871094, "step": 628 }, { "epoch": 0.9508692365835223, "fcm_dpo/beta": 0.5566290616989136, "fcm_dpo/delta": 0.07311487942934036, "fcm_dpo/margin": 1.6754742860794067, "fcm_dpo/q_t": 0.3302459716796875, "grad_norm": 120.0956802368164, "learning_rate": 3.798061746947995e-09, "logits/chosen": 0.17901673913002014, "logits/rejected": 0.17002899944782257, "logps/chosen": -73.4151382446289, "logps/ref_chosen": -70.71749877929688, "logps/ref_rejected": -78.96273803710938, "logps/rejected": -83.33586120605469, "loss": 0.8944, "margin_dpo/margin_mean": 1.6754742860794067, "margin_dpo/margin_std": 2.052621603012085, "step": 629 }, { "epoch": 0.9523809523809523, "fcm_dpo/beta": 0.5548304915428162, "fcm_dpo/delta": -0.06957443058490753, "fcm_dpo/margin": 1.912638783454895, "fcm_dpo/q_t": 0.32286617159843445, "grad_norm": 108.01695251464844, "learning_rate": 3.5719052736323806e-09, "logits/chosen": 0.1528751254081726, "logits/rejected": 0.11582262814044952, "logps/chosen": -58.81450653076172, "logps/ref_chosen": -56.201412200927734, "logps/ref_rejected": -74.69807434082031, "logps/rejected": -79.22380065917969, "loss": 0.891, "margin_dpo/margin_mean": 1.9126391410827637, "margin_dpo/margin_std": 2.4235522747039795, "step": 630 }, { "epoch": 0.9538926681783825, "fcm_dpo/beta": 0.5212994813919067, "fcm_dpo/delta": -0.20469579100608826, "fcm_dpo/margin": 2.244609832763672, "fcm_dpo/q_t": 0.2998984754085541, "grad_norm": 126.16265869140625, "learning_rate": 3.352641923861144e-09, "logits/chosen": 0.24326127767562866, "logits/rejected": 0.16665881872177124, "logps/chosen": -61.588680267333984, "logps/ref_chosen": -58.82059860229492, "logps/ref_rejected": -96.51437377929688, "logps/rejected": -101.5270767211914, "loss": 0.8734, "margin_dpo/margin_mean": 2.244609832763672, "margin_dpo/margin_std": 2.739107370376587, "step": 631 }, { "epoch": 0.9554043839758125, "fcm_dpo/beta": 0.5139273405075073, "fcm_dpo/delta": -0.14226512610912323, "fcm_dpo/margin": 2.19085693359375, "fcm_dpo/q_t": 0.28850066661834717, "grad_norm": 94.82600402832031, "learning_rate": 3.140277830901428e-09, "logits/chosen": 0.21174263954162598, "logits/rejected": 0.18780021369457245, "logps/chosen": -61.52200698852539, "logps/ref_chosen": -58.786048889160156, "logps/ref_rejected": -67.21923828125, "logps/rejected": -72.14605712890625, "loss": 0.8071, "margin_dpo/margin_mean": 2.19085693359375, "margin_dpo/margin_std": 2.2715227603912354, "step": 632 }, { "epoch": 0.9569160997732427, "fcm_dpo/beta": 0.5195468664169312, "fcm_dpo/delta": 0.10754405707120895, "fcm_dpo/margin": 1.7362931966781616, "fcm_dpo/q_t": 0.3419226408004761, "grad_norm": 99.7951431274414, "learning_rate": 2.9348189350335007e-09, "logits/chosen": 0.1292242705821991, "logits/rejected": 0.08500467240810394, "logps/chosen": -54.51697540283203, "logps/ref_chosen": -52.13019561767578, "logps/ref_rejected": -67.23016357421875, "logps/rejected": -71.35323333740234, "loss": 0.971, "margin_dpo/margin_mean": 1.7362935543060303, "margin_dpo/margin_std": 2.4587457180023193, "step": 633 }, { "epoch": 0.9584278155706727, "fcm_dpo/beta": 0.5588383674621582, "fcm_dpo/delta": 0.42835086584091187, "fcm_dpo/margin": 1.0649152994155884, "fcm_dpo/q_t": 0.4058130383491516, "grad_norm": 154.30062866210938, "learning_rate": 2.736270983384276e-09, "logits/chosen": 0.22938773036003113, "logits/rejected": 0.22998389601707458, "logps/chosen": -63.95368194580078, "logps/ref_chosen": -60.97979736328125, "logps/ref_rejected": -58.50825119018555, "logps/rejected": -62.54705047607422, "loss": 1.2187, "margin_dpo/margin_mean": 1.0649151802062988, "margin_dpo/margin_std": 2.3998215198516846, "step": 634 }, { "epoch": 0.9599395313681028, "fcm_dpo/beta": 0.5766314268112183, "fcm_dpo/delta": 0.07918489724397659, "fcm_dpo/margin": 1.6092689037322998, "fcm_dpo/q_t": 0.3603755235671997, "grad_norm": 152.15786743164062, "learning_rate": 2.5446395297668287e-09, "logits/chosen": 0.10523584485054016, "logits/rejected": 0.06159904971718788, "logps/chosen": -69.22645568847656, "logps/ref_chosen": -65.9730224609375, "logps/ref_rejected": -85.61317443847656, "logps/rejected": -90.47587585449219, "loss": 1.0967, "margin_dpo/margin_mean": 1.6092685461044312, "margin_dpo/margin_std": 2.7506871223449707, "step": 635 }, { "epoch": 0.9614512471655329, "fcm_dpo/beta": 0.5654884576797485, "fcm_dpo/delta": -0.11426550894975662, "fcm_dpo/margin": 1.9465041160583496, "fcm_dpo/q_t": 0.2917436957359314, "grad_norm": 109.79877471923828, "learning_rate": 2.359929934524829e-09, "logits/chosen": 0.1432669758796692, "logits/rejected": 0.07593058794736862, "logps/chosen": -51.8205680847168, "logps/ref_chosen": -49.140167236328125, "logps/ref_rejected": -81.26971435546875, "logps/rejected": -85.89661407470703, "loss": 0.7784, "margin_dpo/margin_mean": 1.9465045928955078, "margin_dpo/margin_std": 1.9250316619873047, "step": 636 }, { "epoch": 0.9629629629629629, "fcm_dpo/beta": 0.5774806141853333, "fcm_dpo/delta": 0.09339653700590134, "fcm_dpo/margin": 1.5811973810195923, "fcm_dpo/q_t": 0.3516439199447632, "grad_norm": 156.06459045410156, "learning_rate": 2.1821473643827137e-09, "logits/chosen": 0.14332839846611023, "logits/rejected": 0.08663806319236755, "logps/chosen": -76.97013854980469, "logps/ref_chosen": -73.69658660888672, "logps/ref_rejected": -83.01487731933594, "logps/rejected": -87.86962890625, "loss": 1.0302, "margin_dpo/margin_mean": 1.5811975002288818, "margin_dpo/margin_std": 2.551753044128418, "step": 637 }, { "epoch": 0.9644746787603931, "fcm_dpo/beta": 0.5492261648178101, "fcm_dpo/delta": -0.27356475591659546, "fcm_dpo/margin": 2.252800941467285, "fcm_dpo/q_t": 0.28849977254867554, "grad_norm": 118.49738311767578, "learning_rate": 2.0112967923011646e-09, "logits/chosen": 0.1485036015510559, "logits/rejected": 0.10658858716487885, "logps/chosen": -65.84213256835938, "logps/ref_chosen": -62.78158187866211, "logps/ref_rejected": -85.40478515625, "logps/rejected": -90.71813201904297, "loss": 0.8175, "margin_dpo/margin_mean": 2.252800464630127, "margin_dpo/margin_std": 2.519404888153076, "step": 638 }, { "epoch": 0.9659863945578231, "fcm_dpo/beta": 0.5499449372291565, "fcm_dpo/delta": 0.053651005029678345, "fcm_dpo/margin": 1.7299044132232666, "fcm_dpo/q_t": 0.33649200201034546, "grad_norm": 118.28497314453125, "learning_rate": 1.847382997337943e-09, "logits/chosen": 0.14743350446224213, "logits/rejected": 0.052376627922058105, "logps/chosen": -56.702606201171875, "logps/ref_chosen": -53.76658630371094, "logps/ref_rejected": -72.30009460449219, "logps/rejected": -76.96601867675781, "loss": 0.9522, "margin_dpo/margin_mean": 1.7299044132232666, "margin_dpo/margin_std": 2.438140392303467, "step": 639 }, { "epoch": 0.9674981103552532, "fcm_dpo/beta": 0.5619925260543823, "fcm_dpo/delta": 0.15041983127593994, "fcm_dpo/margin": 1.5351797342300415, "fcm_dpo/q_t": 0.3423158526420593, "grad_norm": 126.87297821044922, "learning_rate": 1.690410564514244e-09, "logits/chosen": 0.23225465416908264, "logits/rejected": 0.19046524167060852, "logps/chosen": -54.199310302734375, "logps/ref_chosen": -51.41777801513672, "logps/ref_rejected": -77.27879333496094, "logps/rejected": -81.59550476074219, "loss": 1.0331, "margin_dpo/margin_mean": 1.5351800918579102, "margin_dpo/margin_std": 2.3253612518310547, "step": 640 }, { "epoch": 0.9690098261526833, "fcm_dpo/beta": 0.5881354212760925, "fcm_dpo/delta": 0.17225059866905212, "fcm_dpo/margin": 1.4259064197540283, "fcm_dpo/q_t": 0.34505194425582886, "grad_norm": 165.1323699951172, "learning_rate": 1.5403838846864692e-09, "logits/chosen": 0.13332295417785645, "logits/rejected": 0.10575246065855026, "logps/chosen": -74.10830688476562, "logps/ref_chosen": -71.0546646118164, "logps/ref_rejected": -82.2440185546875, "logps/rejected": -86.72355651855469, "loss": 1.0339, "margin_dpo/margin_mean": 1.4259059429168701, "margin_dpo/margin_std": 2.282135486602783, "step": 641 }, { "epoch": 0.9705215419501134, "fcm_dpo/beta": 0.5971714854240417, "fcm_dpo/delta": 0.17347897589206696, "fcm_dpo/margin": 1.4081449508666992, "fcm_dpo/q_t": 0.3561500310897827, "grad_norm": 159.68527221679688, "learning_rate": 1.3973071544233218e-09, "logits/chosen": 0.12671303749084473, "logits/rejected": 0.12140335142612457, "logps/chosen": -72.1884994506836, "logps/ref_chosen": -68.92927551269531, "logps/ref_rejected": -70.85682678222656, "logps/rejected": -75.52420043945312, "loss": 1.078, "margin_dpo/margin_mean": 1.4081450700759888, "margin_dpo/margin_std": 2.3124876022338867, "step": 642 }, { "epoch": 0.9720332577475435, "fcm_dpo/beta": 0.6271607875823975, "fcm_dpo/delta": 0.3203769028186798, "fcm_dpo/margin": 1.1184437274932861, "fcm_dpo/q_t": 0.34422484040260315, "grad_norm": 2246.781982421875, "learning_rate": 1.261184375888541e-09, "logits/chosen": 0.1141979843378067, "logits/rejected": 0.05281982570886612, "logps/chosen": -68.73892974853516, "logps/ref_chosen": -65.30903625488281, "logps/ref_rejected": -83.61613464355469, "logps/rejected": -88.16447448730469, "loss": 1.4846, "margin_dpo/margin_mean": 1.1184438467025757, "margin_dpo/margin_std": 3.4751362800598145, "step": 643 }, { "epoch": 0.9735449735449735, "fcm_dpo/beta": 0.6507506966590881, "fcm_dpo/delta": 0.037187736481428146, "fcm_dpo/margin": 1.4850530624389648, "fcm_dpo/q_t": 0.3464422821998596, "grad_norm": 163.06446838378906, "learning_rate": 1.1320193567288527e-09, "logits/chosen": 0.2439633458852768, "logits/rejected": 0.21551108360290527, "logps/chosen": -53.948020935058594, "logps/ref_chosen": -51.002601623535156, "logps/ref_rejected": -64.46372985839844, "logps/rejected": -68.89420318603516, "loss": 1.0968, "margin_dpo/margin_mean": 1.4850530624389648, "margin_dpo/margin_std": 2.556273937225342, "step": 644 }, { "epoch": 0.9750566893424036, "fcm_dpo/beta": 0.6438242197036743, "fcm_dpo/delta": -0.08629032969474792, "fcm_dpo/margin": 1.6724817752838135, "fcm_dpo/q_t": 0.31522613763809204, "grad_norm": 150.0731964111328, "learning_rate": 1.0098157099674987e-09, "logits/chosen": 0.0987434908747673, "logits/rejected": 0.07496701180934906, "logps/chosen": -64.18575286865234, "logps/ref_chosen": -60.963409423828125, "logps/ref_rejected": -69.73353576660156, "logps/rejected": -74.62835693359375, "loss": 0.8818, "margin_dpo/margin_mean": 1.6724815368652344, "margin_dpo/margin_std": 2.17503023147583, "step": 645 }, { "epoch": 0.9765684051398337, "fcm_dpo/beta": 0.6224489212036133, "fcm_dpo/delta": -0.2178870290517807, "fcm_dpo/margin": 1.9151108264923096, "fcm_dpo/q_t": 0.30455827713012695, "grad_norm": 125.87390899658203, "learning_rate": 8.945768539031783e-10, "logits/chosen": 0.2028486728668213, "logits/rejected": 0.16405922174453735, "logps/chosen": -65.69281005859375, "logps/ref_chosen": -62.290069580078125, "logps/ref_rejected": -85.54812622070312, "logps/rejected": -90.86598205566406, "loss": 0.9126, "margin_dpo/margin_mean": 1.9151103496551514, "margin_dpo/margin_std": 2.5356059074401855, "step": 646 }, { "epoch": 0.9780801209372638, "fcm_dpo/beta": 0.5961904525756836, "fcm_dpo/delta": -0.26335006952285767, "fcm_dpo/margin": 2.065751552581787, "fcm_dpo/q_t": 0.30408942699432373, "grad_norm": 186.32350158691406, "learning_rate": 7.863060120144316e-10, "logits/chosen": 0.20603081583976746, "logits/rejected": 0.13498756289482117, "logps/chosen": -70.44767761230469, "logps/ref_chosen": -67.515869140625, "logps/ref_rejected": -101.50871276855469, "logps/rejected": -106.50627136230469, "loss": 0.9259, "margin_dpo/margin_mean": 2.065751075744629, "margin_dpo/margin_std": 2.7231507301330566, "step": 647 }, { "epoch": 0.9795918367346939, "fcm_dpo/beta": 0.5804147720336914, "fcm_dpo/delta": 0.019788160920143127, "fcm_dpo/margin": 1.6904387474060059, "fcm_dpo/q_t": 0.33119359612464905, "grad_norm": 137.32240295410156, "learning_rate": 6.850062128694045e-10, "logits/chosen": 0.13131186366081238, "logits/rejected": 0.0793139785528183, "logps/chosen": -67.7803726196289, "logps/ref_chosen": -64.59593963623047, "logps/ref_rejected": -83.384033203125, "logps/rejected": -88.25890350341797, "loss": 0.9714, "margin_dpo/margin_mean": 1.6904385089874268, "margin_dpo/margin_std": 2.3690028190612793, "step": 648 }, { "epoch": 0.981103552532124, "fcm_dpo/beta": 0.5746303796768188, "fcm_dpo/delta": -0.07765467464923859, "fcm_dpo/margin": 1.8583083152770996, "fcm_dpo/q_t": 0.3117806911468506, "grad_norm": 126.6104736328125, "learning_rate": 5.906802900412788e-10, "logits/chosen": 0.17333395779132843, "logits/rejected": 0.1271257847547531, "logps/chosen": -52.0538330078125, "logps/ref_chosen": -49.30964660644531, "logps/ref_rejected": -73.73710632324219, "logps/rejected": -78.339599609375, "loss": 0.931, "margin_dpo/margin_mean": 1.8583080768585205, "margin_dpo/margin_std": 2.380646228790283, "step": 649 }, { "epoch": 0.982615268329554, "fcm_dpo/beta": 0.5875668525695801, "fcm_dpo/delta": 0.08222602307796478, "fcm_dpo/margin": 1.5724918842315674, "fcm_dpo/q_t": 0.3484124541282654, "grad_norm": 169.66038513183594, "learning_rate": 5.033308820289184e-10, "logits/chosen": 0.2582772374153137, "logits/rejected": 0.21315817534923553, "logps/chosen": -57.925880432128906, "logps/ref_chosen": -55.06325912475586, "logps/ref_rejected": -77.39610290527344, "logps/rejected": -81.83120727539062, "loss": 1.0498, "margin_dpo/margin_mean": 1.5724914073944092, "margin_dpo/margin_std": 2.477647542953491, "step": 650 }, { "epoch": 0.9841269841269841, "fcm_dpo/beta": 0.600640058517456, "fcm_dpo/delta": 0.055977076292037964, "fcm_dpo/margin": 1.5707225799560547, "fcm_dpo/q_t": 0.3421096205711365, "grad_norm": 148.89381408691406, "learning_rate": 4.2296043218295606e-10, "logits/chosen": 0.24641644954681396, "logits/rejected": 0.1934581995010376, "logps/chosen": -56.82018280029297, "logps/ref_chosen": -54.065162658691406, "logps/ref_rejected": -77.79080200195312, "logps/rejected": -82.1165542602539, "loss": 0.9553, "margin_dpo/margin_mean": 1.5707224607467651, "margin_dpo/margin_std": 2.1566340923309326, "step": 651 }, { "epoch": 0.9856386999244142, "fcm_dpo/beta": 0.5921408534049988, "fcm_dpo/delta": 0.02783304452896118, "fcm_dpo/margin": 1.6457834243774414, "fcm_dpo/q_t": 0.3308331370353699, "grad_norm": 141.38235473632812, "learning_rate": 3.4957118863768176e-10, "logits/chosen": 0.20549950003623962, "logits/rejected": 0.15806761384010315, "logps/chosen": -66.84440612792969, "logps/ref_chosen": -63.64030456542969, "logps/ref_rejected": -78.86882019042969, "logps/rejected": -83.71870422363281, "loss": 0.9701, "margin_dpo/margin_mean": 1.6457829475402832, "margin_dpo/margin_std": 2.2666330337524414, "step": 652 }, { "epoch": 0.9871504157218443, "fcm_dpo/beta": 0.6015132665634155, "fcm_dpo/delta": 0.07994754612445831, "fcm_dpo/margin": 1.541993260383606, "fcm_dpo/q_t": 0.352220356464386, "grad_norm": 151.9169464111328, "learning_rate": 2.831652042480093e-10, "logits/chosen": 0.18092171847820282, "logits/rejected": 0.13986843824386597, "logps/chosen": -64.58534240722656, "logps/ref_chosen": -61.668373107910156, "logps/ref_rejected": -73.83012390136719, "logps/rejected": -78.28909301757812, "loss": 1.0303, "margin_dpo/margin_mean": 1.5419931411743164, "margin_dpo/margin_std": 2.441131591796875, "step": 653 }, { "epoch": 0.9886621315192744, "fcm_dpo/beta": 0.5887470245361328, "fcm_dpo/delta": -0.13443875312805176, "fcm_dpo/margin": 1.8974157571792603, "fcm_dpo/q_t": 0.31610941886901855, "grad_norm": 141.95655822753906, "learning_rate": 2.2374433653205016e-10, "logits/chosen": 0.170884370803833, "logits/rejected": 0.10386738926172256, "logps/chosen": -60.361976623535156, "logps/ref_chosen": -57.568267822265625, "logps/ref_rejected": -87.74789428710938, "logps/rejected": -92.43901824951172, "loss": 0.905, "margin_dpo/margin_mean": 1.8974157571792603, "margin_dpo/margin_std": 2.5552029609680176, "step": 654 }, { "epoch": 0.9901738473167044, "fcm_dpo/beta": 0.5729016661643982, "fcm_dpo/delta": -0.060607388615608215, "fcm_dpo/margin": 1.8254590034484863, "fcm_dpo/q_t": 0.309136301279068, "grad_norm": 104.31941223144531, "learning_rate": 1.7131024761923852e-10, "logits/chosen": 0.1574726104736328, "logits/rejected": 0.0878022313117981, "logps/chosen": -54.848045349121094, "logps/ref_chosen": -52.14714813232422, "logps/ref_rejected": -80.85014343261719, "logps/rejected": -85.37649536132812, "loss": 0.8471, "margin_dpo/margin_mean": 1.8254594802856445, "margin_dpo/margin_std": 2.0447256565093994, "step": 655 }, { "epoch": 0.9916855631141346, "fcm_dpo/beta": 0.555420458316803, "fcm_dpo/delta": -0.24778330326080322, "fcm_dpo/margin": 2.184002161026001, "fcm_dpo/q_t": 0.29599303007125854, "grad_norm": 112.63178253173828, "learning_rate": 1.2586440420372934e-10, "logits/chosen": 0.16302891075611115, "logits/rejected": 0.12122651189565659, "logps/chosen": -76.25328063964844, "logps/ref_chosen": -73.25672912597656, "logps/ref_rejected": -85.35127258300781, "logps/rejected": -90.53182220458984, "loss": 0.82, "margin_dpo/margin_mean": 2.1840016841888428, "margin_dpo/margin_std": 2.538489818572998, "step": 656 }, { "epoch": 0.9931972789115646, "fcm_dpo/beta": 0.5423201322555542, "fcm_dpo/delta": -0.10848333686590195, "fcm_dpo/margin": 2.0198726654052734, "fcm_dpo/q_t": 0.3146105408668518, "grad_norm": 127.41387176513672, "learning_rate": 8.740807750345913e-11, "logits/chosen": 0.26283034682273865, "logits/rejected": 0.19647637009620667, "logps/chosen": -52.75443649291992, "logps/ref_chosen": -49.72339630126953, "logps/ref_rejected": -75.1568603515625, "logps/rejected": -80.20777893066406, "loss": 0.9165, "margin_dpo/margin_mean": 2.0198724269866943, "margin_dpo/margin_std": 2.636084794998169, "step": 657 }, { "epoch": 0.9947089947089947, "fcm_dpo/beta": 0.556152880191803, "fcm_dpo/delta": 0.08222407847642899, "fcm_dpo/margin": 1.6563022136688232, "fcm_dpo/q_t": 0.35666483640670776, "grad_norm": 134.63629150390625, "learning_rate": 5.594234322453539e-11, "logits/chosen": 0.1849232316017151, "logits/rejected": 0.15448370575904846, "logps/chosen": -66.14913940429688, "logps/ref_chosen": -63.04634094238281, "logps/ref_rejected": -83.44963073730469, "logps/rejected": -88.20873260498047, "loss": 1.0747, "margin_dpo/margin_mean": 1.656302809715271, "margin_dpo/margin_std": 2.766972541809082, "step": 658 }, { "epoch": 0.9962207105064248, "fcm_dpo/beta": 0.5667056441307068, "fcm_dpo/delta": 0.22809654474258423, "fcm_dpo/margin": 1.3948873281478882, "fcm_dpo/q_t": 0.36462146043777466, "grad_norm": 134.068603515625, "learning_rate": 3.146808153123293e-11, "logits/chosen": 0.2203397899866104, "logits/rejected": 0.16544032096862793, "logps/chosen": -58.36835479736328, "logps/ref_chosen": -55.0802001953125, "logps/ref_rejected": -71.91049194335938, "logps/rejected": -76.59353637695312, "loss": 1.1224, "margin_dpo/margin_mean": 1.3948872089385986, "margin_dpo/margin_std": 2.498440742492676, "step": 659 }, { "epoch": 0.9977324263038548, "fcm_dpo/beta": 0.5684385895729065, "fcm_dpo/delta": -0.10104553401470184, "fcm_dpo/margin": 1.9175570011138916, "fcm_dpo/q_t": 0.3092523217201233, "grad_norm": 133.181396484375, "learning_rate": 1.3985977021235829e-11, "logits/chosen": 0.2678339183330536, "logits/rejected": 0.2140858918428421, "logps/chosen": -57.32783889770508, "logps/ref_chosen": -54.525917053222656, "logps/ref_rejected": -81.23604583740234, "logps/rejected": -85.95552825927734, "loss": 0.8659, "margin_dpo/margin_mean": 1.9175567626953125, "margin_dpo/margin_std": 2.366135597229004, "step": 660 }, { "epoch": 0.999244142101285, "fcm_dpo/beta": 0.5669894218444824, "fcm_dpo/delta": 0.021186619997024536, "fcm_dpo/margin": 1.7300082445144653, "fcm_dpo/q_t": 0.33315783739089966, "grad_norm": 129.08389282226562, "learning_rate": 3.4965187065971735e-12, "logits/chosen": 0.15239998698234558, "logits/rejected": 0.0887516513466835, "logps/chosen": -63.79823684692383, "logps/ref_chosen": -60.37263870239258, "logps/ref_rejected": -77.42874145507812, "logps/rejected": -82.5843505859375, "loss": 1.0387, "margin_dpo/margin_mean": 1.730008840560913, "margin_dpo/margin_std": 2.6242144107818604, "step": 661 }, { "epoch": 0.999244142101285, "step": 661, "total_flos": 0.0, "train_loss": 1.1173522615757363, "train_runtime": 1752.852, "train_samples_per_second": 24.153, "train_steps_per_second": 0.377 } ], "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 }