{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.999244142101285, "eval_steps": 100, "global_step": 661, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0015117157974300832, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.002980858087539673, "fcm_dpo/q_t": 0.5000747442245483, "grad_norm": 17.89813995361328, "learning_rate": 0.0, "logits/chosen": 1.702779769897461, "logits/rejected": 1.6965749263763428, "logps/chosen": -80.20932006835938, "logps/ref_chosen": -80.27740478515625, "logps/ref_rejected": -83.5943374633789, "logps/rejected": -83.52326965332031, "loss": 1.387, "margin_dpo/margin_mean": -0.0029816031455993652, "margin_dpo/margin_std": 0.3835117816925049, "step": 1 }, { "epoch": 0.0030234315948601664, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.029325395822525024, "fcm_dpo/q_t": 0.4992692470550537, "grad_norm": 21.475933074951172, "learning_rate": 7.462686567164179e-09, "logits/chosen": 1.7006168365478516, "logits/rejected": 1.6698178052902222, "logps/chosen": -74.51097869873047, "logps/ref_chosen": -74.56095886230469, "logps/ref_rejected": -83.53636169433594, "logps/rejected": -83.51570892333984, "loss": 1.3839, "margin_dpo/margin_mean": 0.029325813055038452, "margin_dpo/margin_std": 0.4646317958831787, "step": 2 }, { "epoch": 0.0045351473922902496, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02612045407295227, "fcm_dpo/q_t": 0.5006523132324219, "grad_norm": 19.997602462768555, "learning_rate": 1.4925373134328357e-08, "logits/chosen": 1.626187801361084, "logits/rejected": 1.5350779294967651, "logps/chosen": -82.14555358886719, "logps/ref_chosen": -82.1510009765625, "logps/ref_rejected": -109.82986450195312, "logps/rejected": -109.79829406738281, "loss": 1.3894, "margin_dpo/margin_mean": -0.026119887828826904, "margin_dpo/margin_std": 0.42776572704315186, "step": 3 }, { "epoch": 0.006046863189720333, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.09968307614326477, "fcm_dpo/q_t": 0.5024900436401367, "grad_norm": 19.80933952331543, "learning_rate": 2.2388059701492534e-08, "logits/chosen": 1.7662708759307861, "logits/rejected": 1.754350185394287, "logps/chosen": -92.36078643798828, "logps/ref_chosen": -92.37549591064453, "logps/ref_rejected": -99.59553527832031, "logps/rejected": -99.48114776611328, "loss": 1.3967, "margin_dpo/margin_mean": -0.09968316555023193, "margin_dpo/margin_std": 0.4198913276195526, "step": 4 }, { "epoch": 0.007558578987150416, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.057308733463287354, "fcm_dpo/q_t": 0.501429557800293, "grad_norm": 19.01952362060547, "learning_rate": 2.9850746268656714e-08, "logits/chosen": 1.547282338142395, "logits/rejected": 1.4952952861785889, "logps/chosen": -78.90206909179688, "logps/ref_chosen": -78.84872436523438, "logps/ref_rejected": -97.88040161132812, "logps/rejected": -97.87643432617188, "loss": 1.3927, "margin_dpo/margin_mean": -0.057308852672576904, "margin_dpo/margin_std": 0.4990013837814331, "step": 5 }, { "epoch": 0.009070294784580499, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06428655982017517, "fcm_dpo/q_t": 0.49839404225349426, "grad_norm": 18.07597541809082, "learning_rate": 3.731343283582089e-08, "logits/chosen": 1.5872879028320312, "logits/rejected": 1.4798492193222046, "logps/chosen": -68.31825256347656, "logps/ref_chosen": -68.34607696533203, "logps/ref_rejected": -99.24614715576172, "logps/rejected": -99.28260803222656, "loss": 1.3803, "margin_dpo/margin_mean": 0.06428647041320801, "margin_dpo/margin_std": 0.4039618670940399, "step": 6 }, { "epoch": 0.010582010582010581, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.08962023258209229, "fcm_dpo/q_t": 0.5022399425506592, "grad_norm": 17.446853637695312, "learning_rate": 4.477611940298507e-08, "logits/chosen": 1.458909273147583, "logits/rejected": 1.3968511819839478, "logps/chosen": -69.16606903076172, "logps/ref_chosen": -69.11282348632812, "logps/ref_rejected": -84.01641845703125, "logps/rejected": -83.98003387451172, "loss": 1.3958, "margin_dpo/margin_mean": -0.08962073922157288, "margin_dpo/margin_std": 0.4627250134944916, "step": 7 }, { "epoch": 0.012093726379440665, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.03348463773727417, "fcm_dpo/q_t": 0.5008403062820435, "grad_norm": 18.309078216552734, "learning_rate": 5.223880597014925e-08, "logits/chosen": 1.6482627391815186, "logits/rejected": 1.6344785690307617, "logps/chosen": -78.40745544433594, "logps/ref_chosen": -78.3912353515625, "logps/ref_rejected": -91.06254577636719, "logps/rejected": -91.04528045654297, "loss": 1.3902, "margin_dpo/margin_mean": -0.03348389267921448, "margin_dpo/margin_std": 0.4609626531600952, "step": 8 }, { "epoch": 0.013605442176870748, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06733879446983337, "fcm_dpo/q_t": 0.4983159303665161, "grad_norm": 19.262584686279297, "learning_rate": 5.970149253731343e-08, "logits/chosen": 1.9055404663085938, "logits/rejected": 1.7056143283843994, "logps/chosen": -69.64701843261719, "logps/ref_chosen": -69.67422485351562, "logps/ref_rejected": -105.00473022460938, "logps/rejected": -105.04486846923828, "loss": 1.3801, "margin_dpo/margin_mean": 0.06733927130699158, "margin_dpo/margin_std": 0.44255518913269043, "step": 9 }, { "epoch": 0.015117157974300832, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.12282580137252808, "fcm_dpo/q_t": 0.496931791305542, "grad_norm": 18.874061584472656, "learning_rate": 6.71641791044776e-08, "logits/chosen": 1.5117344856262207, "logits/rejected": 1.4378249645233154, "logps/chosen": -79.62625122070312, "logps/ref_chosen": -79.730712890625, "logps/ref_rejected": -105.50645446777344, "logps/rejected": -105.52481079101562, "loss": 1.3745, "margin_dpo/margin_mean": 0.12282583117485046, "margin_dpo/margin_std": 0.4421079754829407, "step": 10 }, { "epoch": 0.016628873771730914, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0041959285736083984, "fcm_dpo/q_t": 0.49989521503448486, "grad_norm": 17.292160034179688, "learning_rate": 7.462686567164178e-08, "logits/chosen": 1.5478227138519287, "logits/rejected": 1.5047085285186768, "logps/chosen": -85.41871643066406, "logps/ref_chosen": -85.41248321533203, "logps/ref_rejected": -86.50241088867188, "logps/rejected": -86.51283264160156, "loss": 1.3863, "margin_dpo/margin_mean": 0.004195868968963623, "margin_dpo/margin_std": 0.41407087445259094, "step": 11 }, { "epoch": 0.018140589569160998, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.005808025598526001, "fcm_dpo/q_t": 0.500144898891449, "grad_norm": 17.388395309448242, "learning_rate": 8.208955223880596e-08, "logits/chosen": 1.7034389972686768, "logits/rejected": 1.6622624397277832, "logps/chosen": -81.38066101074219, "logps/ref_chosen": -81.38086700439453, "logps/ref_rejected": -89.88151550292969, "logps/rejected": -89.87550354003906, "loss": 1.3872, "margin_dpo/margin_mean": -0.0058082640171051025, "margin_dpo/margin_std": 0.35740458965301514, "step": 12 }, { "epoch": 0.019652305366591082, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.09430631995201111, "fcm_dpo/q_t": 0.4976438283920288, "grad_norm": 17.77792739868164, "learning_rate": 8.955223880597014e-08, "logits/chosen": 1.5007734298706055, "logits/rejected": 1.3290517330169678, "logps/chosen": -63.13867950439453, "logps/ref_chosen": -63.17030715942383, "logps/ref_rejected": -105.61166381835938, "logps/rejected": -105.67434692382812, "loss": 1.3772, "margin_dpo/margin_mean": 0.09430572390556335, "margin_dpo/margin_std": 0.35614386200904846, "step": 13 }, { "epoch": 0.021164021164021163, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.09425640106201172, "fcm_dpo/q_t": 0.5023550987243652, "grad_norm": 19.559247970581055, "learning_rate": 9.701492537313432e-08, "logits/chosen": 1.846583604812622, "logits/rejected": 1.8075042963027954, "logps/chosen": -80.72798156738281, "logps/ref_chosen": -80.71014404296875, "logps/ref_rejected": -89.86041259765625, "logps/rejected": -89.78399658203125, "loss": 1.396, "margin_dpo/margin_mean": -0.0942547619342804, "margin_dpo/margin_std": 0.3276791572570801, "step": 14 }, { "epoch": 0.022675736961451247, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02120572328567505, "fcm_dpo/q_t": 0.5005297660827637, "grad_norm": 20.666595458984375, "learning_rate": 1.044776119402985e-07, "logits/chosen": 1.7599084377288818, "logits/rejected": 1.6542625427246094, "logps/chosen": -82.008056640625, "logps/ref_chosen": -82.00294494628906, "logps/ref_rejected": -106.43550109863281, "logps/rejected": -106.41940307617188, "loss": 1.3889, "margin_dpo/margin_mean": -0.0212060809135437, "margin_dpo/margin_std": 0.42040663957595825, "step": 15 }, { "epoch": 0.02418745275888133, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.055621325969696045, "fcm_dpo/q_t": 0.5013903379440308, "grad_norm": 17.204975128173828, "learning_rate": 1.1194029850746268e-07, "logits/chosen": 1.84678316116333, "logits/rejected": 1.722154140472412, "logps/chosen": -62.32442092895508, "logps/ref_chosen": -62.308345794677734, "logps/ref_rejected": -89.6508560180664, "logps/rejected": -89.61131286621094, "loss": 1.3922, "margin_dpo/margin_mean": -0.05562084913253784, "margin_dpo/margin_std": 0.3397616744041443, "step": 16 }, { "epoch": 0.025699168556311415, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.05070582032203674, "fcm_dpo/q_t": 0.501266360282898, "grad_norm": 18.484811782836914, "learning_rate": 1.1940298507462686e-07, "logits/chosen": 1.670881748199463, "logits/rejected": 1.6353247165679932, "logps/chosen": -85.19845581054688, "logps/ref_chosen": -85.16903686523438, "logps/ref_rejected": -102.57087707519531, "logps/rejected": -102.54957580566406, "loss": 1.3918, "margin_dpo/margin_mean": -0.050705909729003906, "margin_dpo/margin_std": 0.3820461928844452, "step": 17 }, { "epoch": 0.027210884353741496, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10842156410217285, "fcm_dpo/q_t": 0.4972917437553406, "grad_norm": 17.016756057739258, "learning_rate": 1.2686567164179106e-07, "logits/chosen": 1.951217770576477, "logits/rejected": 1.7988412380218506, "logps/chosen": -63.13896560668945, "logps/ref_chosen": -63.17793273925781, "logps/ref_rejected": -86.06461334228516, "logps/rejected": -86.13406372070312, "loss": 1.3759, "margin_dpo/margin_mean": 0.10842183232307434, "margin_dpo/margin_std": 0.40306586027145386, "step": 18 }, { "epoch": 0.02872260015117158, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02614566683769226, "fcm_dpo/q_t": 0.500652551651001, "grad_norm": 19.710174560546875, "learning_rate": 1.343283582089552e-07, "logits/chosen": 1.5407326221466064, "logits/rejected": 1.539353609085083, "logps/chosen": -85.8155746459961, "logps/ref_chosen": -85.82405853271484, "logps/ref_rejected": -100.07136535644531, "logps/rejected": -100.0367431640625, "loss": 1.3893, "margin_dpo/margin_mean": -0.026145905256271362, "margin_dpo/margin_std": 0.36016714572906494, "step": 19 }, { "epoch": 0.030234315948601664, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.03781422972679138, "fcm_dpo/q_t": 0.5009453296661377, "grad_norm": 18.178298950195312, "learning_rate": 1.4179104477611938e-07, "logits/chosen": 2.0051045417785645, "logits/rejected": 1.9216687679290771, "logps/chosen": -73.64117431640625, "logps/ref_chosen": -73.58621215820312, "logps/ref_rejected": -91.21690368652344, "logps/rejected": -91.23403930664062, "loss": 1.3905, "margin_dpo/margin_mean": -0.03781440854072571, "margin_dpo/margin_std": 0.39522528648376465, "step": 20 }, { "epoch": 0.031746031746031744, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07584941387176514, "fcm_dpo/q_t": 0.49810492992401123, "grad_norm": 18.019695281982422, "learning_rate": 1.4925373134328355e-07, "logits/chosen": 1.7651563882827759, "logits/rejected": 1.6562526226043701, "logps/chosen": -81.92829895019531, "logps/ref_chosen": -81.97251892089844, "logps/ref_rejected": -98.05976867675781, "logps/rejected": -98.09140014648438, "loss": 1.3791, "margin_dpo/margin_mean": 0.07584935426712036, "margin_dpo/margin_std": 0.3969747722148895, "step": 21 }, { "epoch": 0.03325774754346183, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.05920052528381348, "fcm_dpo/q_t": 0.5014793872833252, "grad_norm": 18.35191535949707, "learning_rate": 1.5671641791044775e-07, "logits/chosen": 1.5420267581939697, "logits/rejected": 1.507912278175354, "logps/chosen": -76.98601531982422, "logps/ref_chosen": -76.99579620361328, "logps/ref_rejected": -95.76089477539062, "logps/rejected": -95.6919174194336, "loss": 1.3926, "margin_dpo/margin_mean": -0.0592007040977478, "margin_dpo/margin_std": 0.40154093503952026, "step": 22 }, { "epoch": 0.03476946334089191, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03171822428703308, "fcm_dpo/q_t": 0.49920740723609924, "grad_norm": 19.181137084960938, "learning_rate": 1.6417910447761193e-07, "logits/chosen": 1.9651353359222412, "logits/rejected": 1.8704643249511719, "logps/chosen": -84.74580383300781, "logps/ref_chosen": -84.76856994628906, "logps/ref_rejected": -107.28266906738281, "logps/rejected": -107.29161834716797, "loss": 1.3835, "margin_dpo/margin_mean": 0.03171861171722412, "margin_dpo/margin_std": 0.3716067969799042, "step": 23 }, { "epoch": 0.036281179138321996, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03071492910385132, "fcm_dpo/q_t": 0.49923175573349, "grad_norm": 17.107723236083984, "learning_rate": 1.716417910447761e-07, "logits/chosen": 1.72803795337677, "logits/rejected": 1.6687504053115845, "logps/chosen": -69.82955932617188, "logps/ref_chosen": -69.87112426757812, "logps/ref_rejected": -84.02084350585938, "logps/rejected": -84.00999450683594, "loss": 1.3836, "margin_dpo/margin_mean": 0.030714750289916992, "margin_dpo/margin_std": 0.3912420868873596, "step": 24 }, { "epoch": 0.03779289493575208, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.02832847833633423, "fcm_dpo/q_t": 0.49929308891296387, "grad_norm": 19.49996566772461, "learning_rate": 1.7910447761194027e-07, "logits/chosen": 1.7510415315628052, "logits/rejected": 1.6113269329071045, "logps/chosen": -78.23873901367188, "logps/ref_chosen": -78.22694396972656, "logps/ref_rejected": -106.65234375, "logps/rejected": -106.69246673583984, "loss": 1.384, "margin_dpo/margin_mean": 0.028328508138656616, "margin_dpo/margin_std": 0.45287278294563293, "step": 25 }, { "epoch": 0.039304610733182165, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.09531095623970032, "fcm_dpo/q_t": 0.49761906266212463, "grad_norm": 17.824203491210938, "learning_rate": 1.8656716417910447e-07, "logits/chosen": 1.8126094341278076, "logits/rejected": 1.7892074584960938, "logps/chosen": -74.5651626586914, "logps/ref_chosen": -74.59750366210938, "logps/ref_rejected": -93.57858276367188, "logps/rejected": -93.64155578613281, "loss": 1.3772, "margin_dpo/margin_mean": 0.09531095623970032, "margin_dpo/margin_std": 0.4165651202201843, "step": 26 }, { "epoch": 0.04081632653061224, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.008702605962753296, "fcm_dpo/q_t": 0.4997824430465698, "grad_norm": 18.663766860961914, "learning_rate": 1.9402985074626865e-07, "logits/chosen": 1.5955054759979248, "logits/rejected": 1.5449596643447876, "logps/chosen": -78.6395263671875, "logps/ref_chosen": -78.64625549316406, "logps/ref_rejected": -92.33645629882812, "logps/rejected": -92.33843994140625, "loss": 1.3857, "margin_dpo/margin_mean": 0.00870358943939209, "margin_dpo/margin_std": 0.33294573426246643, "step": 27 }, { "epoch": 0.042328042328042326, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.017391502857208252, "fcm_dpo/q_t": 0.49956539273262024, "grad_norm": 17.787866592407227, "learning_rate": 2.0149253731343282e-07, "logits/chosen": 1.8521151542663574, "logits/rejected": 1.7968239784240723, "logps/chosen": -76.90467071533203, "logps/ref_chosen": -76.91271209716797, "logps/ref_rejected": -88.48194885253906, "logps/rejected": -88.49130249023438, "loss": 1.3849, "margin_dpo/margin_mean": 0.017391175031661987, "margin_dpo/margin_std": 0.35291895270347595, "step": 28 }, { "epoch": 0.04383975812547241, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04088890552520752, "fcm_dpo/q_t": 0.4989783763885498, "grad_norm": 21.243146896362305, "learning_rate": 2.08955223880597e-07, "logits/chosen": 1.9115407466888428, "logits/rejected": 1.848691463470459, "logps/chosen": -89.5405502319336, "logps/ref_chosen": -89.62060546875, "logps/ref_rejected": -100.57090759277344, "logps/rejected": -100.53173828125, "loss": 1.3828, "margin_dpo/margin_mean": 0.040889084339141846, "margin_dpo/margin_std": 0.48537182807922363, "step": 29 }, { "epoch": 0.045351473922902494, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06612420082092285, "fcm_dpo/q_t": 0.4983447790145874, "grad_norm": 19.041486740112305, "learning_rate": 2.1641791044776117e-07, "logits/chosen": 1.7888214588165283, "logits/rejected": 1.6230595111846924, "logps/chosen": -68.77651977539062, "logps/ref_chosen": -68.82381439208984, "logps/ref_rejected": -104.7047119140625, "logps/rejected": -104.72354125976562, "loss": 1.3803, "margin_dpo/margin_mean": 0.06612381339073181, "margin_dpo/margin_std": 0.4717687964439392, "step": 30 }, { "epoch": 0.04686318972033258, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0026360750198364258, "fcm_dpo/q_t": 0.4999334514141083, "grad_norm": 20.6657772064209, "learning_rate": 2.2388059701492537e-07, "logits/chosen": 1.8177580833435059, "logits/rejected": 1.69121515750885, "logps/chosen": -86.07583618164062, "logps/ref_chosen": -86.06916809082031, "logps/ref_rejected": -116.66394805908203, "logps/rejected": -116.67324829101562, "loss": 1.3865, "margin_dpo/margin_mean": 0.0026363134384155273, "margin_dpo/margin_std": 0.4249485731124878, "step": 31 }, { "epoch": 0.04837490551776266, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.011246144771575928, "fcm_dpo/q_t": 0.5002791285514832, "grad_norm": 18.716094970703125, "learning_rate": 2.3134328358208954e-07, "logits/chosen": 1.7052407264709473, "logits/rejected": 1.7523813247680664, "logps/chosen": -87.57007598876953, "logps/ref_chosen": -87.59808349609375, "logps/ref_rejected": -100.26905822753906, "logps/rejected": -100.22979736328125, "loss": 1.3879, "margin_dpo/margin_mean": -0.011246174573898315, "margin_dpo/margin_std": 0.4132556915283203, "step": 32 }, { "epoch": 0.049886621315192746, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.009764552116394043, "fcm_dpo/q_t": 0.4997554421424866, "grad_norm": 19.76194190979004, "learning_rate": 2.388059701492537e-07, "logits/chosen": 1.4848226308822632, "logits/rejected": 1.3876771926879883, "logps/chosen": -83.2780990600586, "logps/ref_chosen": -83.29850769042969, "logps/ref_rejected": -94.60990142822266, "logps/rejected": -94.59925842285156, "loss": 1.3857, "margin_dpo/margin_mean": 0.009763926267623901, "margin_dpo/margin_std": 0.39006873965263367, "step": 33 }, { "epoch": 0.05139833711262283, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.010292023420333862, "fcm_dpo/q_t": 0.4997427761554718, "grad_norm": 17.98219108581543, "learning_rate": 2.4626865671641786e-07, "logits/chosen": 1.826859712600708, "logits/rejected": 1.7421438694000244, "logps/chosen": -70.12477111816406, "logps/ref_chosen": -70.15069580078125, "logps/ref_rejected": -84.4693832397461, "logps/rejected": -84.45375061035156, "loss": 1.3857, "margin_dpo/margin_mean": 0.010291993618011475, "margin_dpo/margin_std": 0.4043731391429901, "step": 34 }, { "epoch": 0.05291005291005291, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08489355444908142, "fcm_dpo/q_t": 0.4978798031806946, "grad_norm": 18.01149559020996, "learning_rate": 2.537313432835821e-07, "logits/chosen": 1.5768963098526, "logits/rejected": 1.5215035676956177, "logps/chosen": -78.22184753417969, "logps/ref_chosen": -78.25238037109375, "logps/ref_rejected": -91.06356811523438, "logps/rejected": -91.1179428100586, "loss": 1.3782, "margin_dpo/margin_mean": 0.08489382266998291, "margin_dpo/margin_std": 0.3789323568344116, "step": 35 }, { "epoch": 0.05442176870748299, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.05526149272918701, "fcm_dpo/q_t": 0.5013813972473145, "grad_norm": 18.324071884155273, "learning_rate": 2.611940298507462e-07, "logits/chosen": 2.020540237426758, "logits/rejected": 1.886946678161621, "logps/chosen": -67.08876037597656, "logps/ref_chosen": -67.06676483154297, "logps/ref_rejected": -99.34661865234375, "logps/rejected": -99.31333923339844, "loss": 1.3922, "margin_dpo/margin_mean": -0.05526161193847656, "margin_dpo/margin_std": 0.39409780502319336, "step": 36 }, { "epoch": 0.055933484504913075, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.05224241316318512, "fcm_dpo/q_t": 0.5013055205345154, "grad_norm": 23.652420043945312, "learning_rate": 2.686567164179104e-07, "logits/chosen": 1.912178635597229, "logits/rejected": 1.6196130514144897, "logps/chosen": -75.90277099609375, "logps/ref_chosen": -75.9269790649414, "logps/ref_rejected": -130.34371948242188, "logps/rejected": -130.26727294921875, "loss": 1.3919, "margin_dpo/margin_mean": -0.05224302411079407, "margin_dpo/margin_std": 0.3591179847717285, "step": 37 }, { "epoch": 0.05744520030234316, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0010804533958435059, "fcm_dpo/q_t": 0.4999734163284302, "grad_norm": 18.430307388305664, "learning_rate": 2.761194029850746e-07, "logits/chosen": 1.7531952857971191, "logits/rejected": 1.7149598598480225, "logps/chosen": -83.66578674316406, "logps/ref_chosen": -83.65460205078125, "logps/ref_rejected": -89.15221405029297, "logps/rejected": -89.16448974609375, "loss": 1.3867, "margin_dpo/margin_mean": 0.0010806024074554443, "margin_dpo/margin_std": 0.42552512884140015, "step": 38 }, { "epoch": 0.05895691609977324, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.01400206983089447, "fcm_dpo/q_t": 0.49964994192123413, "grad_norm": 18.991655349731445, "learning_rate": 2.8358208955223876e-07, "logits/chosen": 2.085275888442993, "logits/rejected": 2.0355820655822754, "logps/chosen": -76.13926696777344, "logps/ref_chosen": -76.18706512451172, "logps/ref_rejected": -94.39262390136719, "logps/rejected": -94.35882568359375, "loss": 1.3852, "margin_dpo/margin_mean": 0.014002442359924316, "margin_dpo/margin_std": 0.32028961181640625, "step": 39 }, { "epoch": 0.06046863189720333, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07038983702659607, "fcm_dpo/q_t": 0.49824073910713196, "grad_norm": 17.965362548828125, "learning_rate": 2.9104477611940296e-07, "logits/chosen": 1.860640048980713, "logits/rejected": 1.7459245920181274, "logps/chosen": -77.41645050048828, "logps/ref_chosen": -77.43476867675781, "logps/ref_rejected": -98.58720397949219, "logps/rejected": -98.63927459716797, "loss": 1.3795, "margin_dpo/margin_mean": 0.07039055228233337, "margin_dpo/margin_std": 0.324634850025177, "step": 40 }, { "epoch": 0.06198034769463341, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08602339029312134, "fcm_dpo/q_t": 0.49785032868385315, "grad_norm": 18.264806747436523, "learning_rate": 2.985074626865671e-07, "logits/chosen": 1.8096897602081299, "logits/rejected": 1.7466790676116943, "logps/chosen": -86.822998046875, "logps/ref_chosen": -86.87640380859375, "logps/ref_rejected": -101.0856704711914, "logps/rejected": -101.1182861328125, "loss": 1.378, "margin_dpo/margin_mean": 0.08602365851402283, "margin_dpo/margin_std": 0.35287344455718994, "step": 41 }, { "epoch": 0.06349206349206349, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07593846321105957, "fcm_dpo/q_t": 0.49810317158699036, "grad_norm": 18.090917587280273, "learning_rate": 3.059701492537313e-07, "logits/chosen": 1.737091064453125, "logits/rejected": 1.6874079704284668, "logps/chosen": -79.32017517089844, "logps/ref_chosen": -79.35625457763672, "logps/ref_rejected": -91.54881286621094, "logps/rejected": -91.58865356445312, "loss": 1.3791, "margin_dpo/margin_mean": 0.07593908905982971, "margin_dpo/margin_std": 0.4052484929561615, "step": 42 }, { "epoch": 0.06500377928949358, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.010380953550338745, "fcm_dpo/q_t": 0.4997408986091614, "grad_norm": 19.443647384643555, "learning_rate": 3.134328358208955e-07, "logits/chosen": 1.8869857788085938, "logits/rejected": 1.7890424728393555, "logps/chosen": -90.7708969116211, "logps/ref_chosen": -90.81220245361328, "logps/ref_rejected": -94.16316986083984, "logps/rejected": -94.13224792480469, "loss": 1.3857, "margin_dpo/margin_mean": 0.010380983352661133, "margin_dpo/margin_std": 0.424875408411026, "step": 43 }, { "epoch": 0.06651549508692366, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.007240593433380127, "fcm_dpo/q_t": 0.5001809000968933, "grad_norm": 19.059425354003906, "learning_rate": 3.2089552238805965e-07, "logits/chosen": 1.4611210823059082, "logits/rejected": 1.3848025798797607, "logps/chosen": -88.25433349609375, "logps/ref_chosen": -88.27932739257812, "logps/ref_rejected": -101.14324951171875, "logps/rejected": -101.11101531982422, "loss": 1.3875, "margin_dpo/margin_mean": -0.007240712642669678, "margin_dpo/margin_std": 0.4283526539802551, "step": 44 }, { "epoch": 0.06802721088435375, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.055614978075027466, "fcm_dpo/q_t": 0.49861085414886475, "grad_norm": 19.44415855407715, "learning_rate": 3.2835820895522385e-07, "logits/chosen": 1.6400771141052246, "logits/rejected": 1.5230625867843628, "logps/chosen": -78.38670349121094, "logps/ref_chosen": -78.40264892578125, "logps/ref_rejected": -109.39339447021484, "logps/rejected": -109.43305969238281, "loss": 1.3812, "margin_dpo/margin_mean": 0.05561518669128418, "margin_dpo/margin_std": 0.44193029403686523, "step": 45 }, { "epoch": 0.06953892668178382, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.01664257049560547, "fcm_dpo/q_t": 0.49958425760269165, "grad_norm": 18.43988800048828, "learning_rate": 3.3582089552238805e-07, "logits/chosen": 1.6313531398773193, "logits/rejected": 1.507079839706421, "logps/chosen": -77.99624633789062, "logps/ref_chosen": -78.08491516113281, "logps/ref_rejected": -97.42544555664062, "logps/rejected": -97.35342407226562, "loss": 1.3849, "margin_dpo/margin_mean": 0.01664254069328308, "margin_dpo/margin_std": 0.34713345766067505, "step": 46 }, { "epoch": 0.0710506424792139, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.048463642597198486, "fcm_dpo/q_t": 0.4987889528274536, "grad_norm": 19.143239974975586, "learning_rate": 3.432835820895522e-07, "logits/chosen": 1.6348772048950195, "logits/rejected": 1.5529307126998901, "logps/chosen": -70.75807189941406, "logps/ref_chosen": -70.78988647460938, "logps/ref_rejected": -91.17266845703125, "logps/rejected": -91.18931579589844, "loss": 1.3817, "margin_dpo/margin_mean": 0.0484640896320343, "margin_dpo/margin_std": 0.3223074674606323, "step": 47 }, { "epoch": 0.07256235827664399, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07357487082481384, "fcm_dpo/q_t": 0.49816131591796875, "grad_norm": 17.006410598754883, "learning_rate": 3.507462686567164e-07, "logits/chosen": 1.8455934524536133, "logits/rejected": 1.7849183082580566, "logps/chosen": -66.62384033203125, "logps/ref_chosen": -66.67327880859375, "logps/ref_rejected": -79.28543853759766, "logps/rejected": -79.3095703125, "loss": 1.3792, "margin_dpo/margin_mean": 0.07357525825500488, "margin_dpo/margin_std": 0.30543482303619385, "step": 48 }, { "epoch": 0.07407407407407407, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.005153179168701172, "fcm_dpo/q_t": 0.5001289248466492, "grad_norm": 17.522451400756836, "learning_rate": 3.5820895522388055e-07, "logits/chosen": 1.5053105354309082, "logits/rejected": 1.4593782424926758, "logps/chosen": -75.131103515625, "logps/ref_chosen": -75.17504119873047, "logps/ref_rejected": -80.5369873046875, "logps/rejected": -80.48789978027344, "loss": 1.3871, "margin_dpo/margin_mean": -0.005153149366378784, "margin_dpo/margin_std": 0.34840038418769836, "step": 49 }, { "epoch": 0.07558578987150416, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.008986026048660278, "fcm_dpo/q_t": 0.49977606534957886, "grad_norm": 17.941247940063477, "learning_rate": 3.6567164179104475e-07, "logits/chosen": 1.9036304950714111, "logits/rejected": 1.8325103521347046, "logps/chosen": -71.21244812011719, "logps/ref_chosen": -71.2314224243164, "logps/ref_rejected": -87.59088134765625, "logps/rejected": -87.58090209960938, "loss": 1.3857, "margin_dpo/margin_mean": 0.008985787630081177, "margin_dpo/margin_std": 0.34656456112861633, "step": 50 }, { "epoch": 0.07709750566893424, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.026766493916511536, "fcm_dpo/q_t": 0.5006682872772217, "grad_norm": 18.8486385345459, "learning_rate": 3.7313432835820895e-07, "logits/chosen": 1.8772060871124268, "logits/rejected": 1.823173999786377, "logps/chosen": -78.64732360839844, "logps/ref_chosen": -78.69171142578125, "logps/ref_rejected": -100.78950500488281, "logps/rejected": -100.71833801269531, "loss": 1.3894, "margin_dpo/margin_mean": -0.026766330003738403, "margin_dpo/margin_std": 0.4199393391609192, "step": 51 }, { "epoch": 0.07860922146636433, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.02304530143737793, "fcm_dpo/q_t": 0.4994242787361145, "grad_norm": 20.32939910888672, "learning_rate": 3.805970149253731e-07, "logits/chosen": 1.8706003427505493, "logits/rejected": 1.694793939590454, "logps/chosen": -89.05101776123047, "logps/ref_chosen": -89.09419250488281, "logps/ref_rejected": -116.87469482421875, "logps/rejected": -116.85455322265625, "loss": 1.3845, "margin_dpo/margin_mean": 0.023045867681503296, "margin_dpo/margin_std": 0.4608180522918701, "step": 52 }, { "epoch": 0.0801209372637944, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10077275335788727, "fcm_dpo/q_t": 0.49748221039772034, "grad_norm": 17.24545669555664, "learning_rate": 3.880597014925373e-07, "logits/chosen": 1.8220537900924683, "logits/rejected": 1.7768255472183228, "logps/chosen": -74.13320922851562, "logps/ref_chosen": -74.21418762207031, "logps/ref_rejected": -75.71168518066406, "logps/rejected": -75.73147583007812, "loss": 1.3768, "margin_dpo/margin_mean": 0.1007724404335022, "margin_dpo/margin_std": 0.4598434269428253, "step": 53 }, { "epoch": 0.08163265306122448, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03820416331291199, "fcm_dpo/q_t": 0.4990437626838684, "grad_norm": 16.42124366760254, "learning_rate": 3.9552238805970144e-07, "logits/chosen": 1.7365084886550903, "logits/rejected": 1.7177741527557373, "logps/chosen": -65.5650863647461, "logps/ref_chosen": -65.63475799560547, "logps/ref_rejected": -76.4462890625, "logps/rejected": -76.41482543945312, "loss": 1.383, "margin_dpo/margin_mean": 0.03820416331291199, "margin_dpo/margin_std": 0.41922712326049805, "step": 54 }, { "epoch": 0.08314436885865457, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.20346182584762573, "fcm_dpo/q_t": 0.49491745233535767, "grad_norm": 20.51268768310547, "learning_rate": 4.0298507462686564e-07, "logits/chosen": 1.6703486442565918, "logits/rejected": 1.4564030170440674, "logps/chosen": -68.64900207519531, "logps/ref_chosen": -68.7640380859375, "logps/ref_rejected": -108.80074310302734, "logps/rejected": -108.88917541503906, "loss": 1.3666, "margin_dpo/margin_mean": 0.20346179604530334, "margin_dpo/margin_std": 0.45147860050201416, "step": 55 }, { "epoch": 0.08465608465608465, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08395838737487793, "fcm_dpo/q_t": 0.4979020357131958, "grad_norm": 17.007356643676758, "learning_rate": 4.1044776119402984e-07, "logits/chosen": 1.8042173385620117, "logits/rejected": 1.777266502380371, "logps/chosen": -74.74409484863281, "logps/ref_chosen": -74.7939453125, "logps/ref_rejected": -81.83535766601562, "logps/rejected": -81.86946105957031, "loss": 1.3784, "margin_dpo/margin_mean": 0.08395865559577942, "margin_dpo/margin_std": 0.44492030143737793, "step": 56 }, { "epoch": 0.08616780045351474, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.251179039478302, "fcm_dpo/q_t": 0.49372607469558716, "grad_norm": 18.769027709960938, "learning_rate": 4.17910447761194e-07, "logits/chosen": 1.770645022392273, "logits/rejected": 1.5965385437011719, "logps/chosen": -74.42295837402344, "logps/ref_chosen": -74.5794677734375, "logps/ref_rejected": -105.61981964111328, "logps/rejected": -105.71448516845703, "loss": 1.362, "margin_dpo/margin_mean": 0.25117942690849304, "margin_dpo/margin_std": 0.5226833820343018, "step": 57 }, { "epoch": 0.08767951625094482, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05911412835121155, "fcm_dpo/q_t": 0.4985233545303345, "grad_norm": 19.216299057006836, "learning_rate": 4.253731343283582e-07, "logits/chosen": 1.771589994430542, "logits/rejected": 1.6873626708984375, "logps/chosen": -92.22261047363281, "logps/ref_chosen": -92.24464416503906, "logps/ref_rejected": -103.18975830078125, "logps/rejected": -103.22683715820312, "loss": 1.3809, "margin_dpo/margin_mean": 0.059114038944244385, "margin_dpo/margin_std": 0.44918161630630493, "step": 58 }, { "epoch": 0.08919123204837491, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1401573121547699, "fcm_dpo/q_t": 0.4964984953403473, "grad_norm": 16.799150466918945, "learning_rate": 4.3283582089552234e-07, "logits/chosen": 1.7912099361419678, "logits/rejected": 1.5354535579681396, "logps/chosen": -66.98262786865234, "logps/ref_chosen": -67.12688446044922, "logps/ref_rejected": -91.69569396972656, "logps/rejected": -91.69160461425781, "loss": 1.3729, "margin_dpo/margin_mean": 0.14015719294548035, "margin_dpo/margin_std": 0.4380917549133301, "step": 59 }, { "epoch": 0.09070294784580499, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.030887484550476074, "fcm_dpo/q_t": 0.49922844767570496, "grad_norm": 18.56014060974121, "learning_rate": 4.4029850746268654e-07, "logits/chosen": 1.6983575820922852, "logits/rejected": 1.7200520038604736, "logps/chosen": -79.62914276123047, "logps/ref_chosen": -79.74327087402344, "logps/ref_rejected": -77.89244079589844, "logps/rejected": -77.8092041015625, "loss": 1.3838, "margin_dpo/margin_mean": 0.030887097120285034, "margin_dpo/margin_std": 0.4888887405395508, "step": 60 }, { "epoch": 0.09221466364323508, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03356589376926422, "fcm_dpo/q_t": 0.4991615414619446, "grad_norm": 16.52377700805664, "learning_rate": 4.4776119402985074e-07, "logits/chosen": 1.9934542179107666, "logits/rejected": 1.939957857131958, "logps/chosen": -66.00021362304688, "logps/ref_chosen": -66.08685302734375, "logps/ref_rejected": -88.1458740234375, "logps/rejected": -88.09280395507812, "loss": 1.3835, "margin_dpo/margin_mean": 0.03356604278087616, "margin_dpo/margin_std": 0.47667667269706726, "step": 61 }, { "epoch": 0.09372637944066516, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.13860514760017395, "fcm_dpo/q_t": 0.4965377449989319, "grad_norm": 17.864152908325195, "learning_rate": 4.552238805970149e-07, "logits/chosen": 1.8296890258789062, "logits/rejected": 1.7605338096618652, "logps/chosen": -80.92373657226562, "logps/ref_chosen": -81.0108871459961, "logps/ref_rejected": -95.50444793701172, "logps/rejected": -95.555908203125, "loss": 1.3732, "margin_dpo/margin_mean": 0.13860544562339783, "margin_dpo/margin_std": 0.5384140014648438, "step": 62 }, { "epoch": 0.09523809523809523, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.19448095560073853, "fcm_dpo/q_t": 0.4951436519622803, "grad_norm": 19.299381256103516, "learning_rate": 4.626865671641791e-07, "logits/chosen": 2.365891456604004, "logits/rejected": 2.2761406898498535, "logps/chosen": -78.40909576416016, "logps/ref_chosen": -78.57593536376953, "logps/ref_rejected": -99.71000671386719, "logps/rejected": -99.73765563964844, "loss": 1.3678, "margin_dpo/margin_mean": 0.19448035955429077, "margin_dpo/margin_std": 0.5808597803115845, "step": 63 }, { "epoch": 0.09674981103552532, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.02114969491958618, "fcm_dpo/q_t": 0.49946916103363037, "grad_norm": 16.543537139892578, "learning_rate": 4.701492537313433e-07, "logits/chosen": 1.6816076040267944, "logits/rejected": 1.6188148260116577, "logps/chosen": -69.18231201171875, "logps/ref_chosen": -69.24063110351562, "logps/ref_rejected": -84.14842987060547, "logps/rejected": -84.11125183105469, "loss": 1.3849, "margin_dpo/margin_mean": 0.021149873733520508, "margin_dpo/margin_std": 0.5421885848045349, "step": 64 }, { "epoch": 0.0982615268329554, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.17138445377349854, "fcm_dpo/q_t": 0.4957193434238434, "grad_norm": 18.766584396362305, "learning_rate": 4.776119402985074e-07, "logits/chosen": 2.203514337539673, "logits/rejected": 2.1390552520751953, "logps/chosen": -83.96290588378906, "logps/ref_chosen": -84.0351333618164, "logps/ref_rejected": -96.42926788330078, "logps/rejected": -96.52842712402344, "loss": 1.3701, "margin_dpo/margin_mean": 0.171383798122406, "margin_dpo/margin_std": 0.5873081684112549, "step": 65 }, { "epoch": 0.09977324263038549, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.18604600429534912, "fcm_dpo/q_t": 0.49535736441612244, "grad_norm": 18.556272506713867, "learning_rate": 4.850746268656717e-07, "logits/chosen": 1.7415666580200195, "logits/rejected": 1.6523596048355103, "logps/chosen": -87.71971893310547, "logps/ref_chosen": -87.79238891601562, "logps/ref_rejected": -95.26547241210938, "logps/rejected": -95.37884521484375, "loss": 1.3688, "margin_dpo/margin_mean": 0.18604576587677002, "margin_dpo/margin_std": 0.6094644665718079, "step": 66 }, { "epoch": 0.10128495842781557, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.19521397352218628, "fcm_dpo/q_t": 0.49512505531311035, "grad_norm": 19.030086517333984, "learning_rate": 4.925373134328357e-07, "logits/chosen": 1.739748477935791, "logits/rejected": 1.6200695037841797, "logps/chosen": -77.81809997558594, "logps/ref_chosen": -78.00114440917969, "logps/ref_rejected": -96.03421020507812, "logps/rejected": -96.04637908935547, "loss": 1.3675, "margin_dpo/margin_mean": 0.1952143907546997, "margin_dpo/margin_std": 0.5094437003135681, "step": 67 }, { "epoch": 0.10279667422524566, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05804261565208435, "fcm_dpo/q_t": 0.4985469579696655, "grad_norm": 19.809654235839844, "learning_rate": 5e-07, "logits/chosen": 1.5946201086044312, "logits/rejected": 1.511031985282898, "logps/chosen": -96.04458618164062, "logps/ref_chosen": -96.04267883300781, "logps/ref_rejected": -110.91169738769531, "logps/rejected": -110.97163391113281, "loss": 1.3815, "margin_dpo/margin_mean": 0.05804276466369629, "margin_dpo/margin_std": 0.5992398262023926, "step": 68 }, { "epoch": 0.10430839002267574, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1814037561416626, "fcm_dpo/q_t": 0.4954685568809509, "grad_norm": 20.02865219116211, "learning_rate": 4.999965034812934e-07, "logits/chosen": 1.5734765529632568, "logits/rejected": 1.4613232612609863, "logps/chosen": -84.95499420166016, "logps/ref_chosen": -85.11124420166016, "logps/ref_rejected": -107.57357025146484, "logps/rejected": -107.59872436523438, "loss": 1.3692, "margin_dpo/margin_mean": 0.18140414357185364, "margin_dpo/margin_std": 0.6236604452133179, "step": 69 }, { "epoch": 0.10582010582010581, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2574300467967987, "fcm_dpo/q_t": 0.4935743510723114, "grad_norm": 18.575429916381836, "learning_rate": 4.999860140229787e-07, "logits/chosen": 1.9775105714797974, "logits/rejected": 1.9188661575317383, "logps/chosen": -81.60387420654297, "logps/ref_chosen": -81.87960815429688, "logps/ref_rejected": -92.63243103027344, "logps/rejected": -92.61412048339844, "loss": 1.3619, "margin_dpo/margin_mean": 0.2574295997619629, "margin_dpo/margin_std": 0.6903476715087891, "step": 70 }, { "epoch": 0.1073318216175359, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1851758360862732, "fcm_dpo/q_t": 0.4953737258911133, "grad_norm": 17.575109481811523, "learning_rate": 4.999685319184688e-07, "logits/chosen": 1.6189801692962646, "logits/rejected": 1.613362193107605, "logps/chosen": -79.58382415771484, "logps/ref_chosen": -79.74766540527344, "logps/ref_rejected": -83.39110565185547, "logps/rejected": -83.41243743896484, "loss": 1.3688, "margin_dpo/margin_mean": 0.18517541885375977, "margin_dpo/margin_std": 0.6131936311721802, "step": 71 }, { "epoch": 0.10884353741496598, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.31794312596321106, "fcm_dpo/q_t": 0.4920644462108612, "grad_norm": 18.973140716552734, "learning_rate": 4.999440576567755e-07, "logits/chosen": 1.853528380393982, "logits/rejected": 1.660147786140442, "logps/chosen": -72.76451873779297, "logps/ref_chosen": -73.04458618164062, "logps/ref_rejected": -92.64720153808594, "logps/rejected": -92.6850814819336, "loss": 1.3561, "margin_dpo/margin_mean": 0.3179430365562439, "margin_dpo/margin_std": 0.7343186140060425, "step": 72 }, { "epoch": 0.11035525321239607, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03165990114212036, "fcm_dpo/q_t": 0.4992097020149231, "grad_norm": 19.453399658203125, "learning_rate": 4.999125919224965e-07, "logits/chosen": 1.5204589366912842, "logits/rejected": 1.4590003490447998, "logps/chosen": -87.63735961914062, "logps/ref_chosen": -87.71681213378906, "logps/ref_rejected": -96.93572998046875, "logps/rejected": -96.88794708251953, "loss": 1.3846, "margin_dpo/margin_mean": 0.031659454107284546, "margin_dpo/margin_std": 0.734971821308136, "step": 73 }, { "epoch": 0.11186696900982615, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.31990575790405273, "fcm_dpo/q_t": 0.4920119047164917, "grad_norm": 17.90540313720703, "learning_rate": 4.998741355957963e-07, "logits/chosen": 1.754788875579834, "logits/rejected": 1.5819981098175049, "logps/chosen": -66.66778564453125, "logps/ref_chosen": -67.07321166992188, "logps/ref_rejected": -96.5340347290039, "logps/rejected": -96.44851684570312, "loss": 1.3557, "margin_dpo/margin_mean": 0.31990575790405273, "margin_dpo/margin_std": 0.6406757235527039, "step": 74 }, { "epoch": 0.11337868480725624, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2673693597316742, "fcm_dpo/q_t": 0.49332353472709656, "grad_norm": 17.061166763305664, "learning_rate": 4.998286897523808e-07, "logits/chosen": 1.7332627773284912, "logits/rejected": 1.5775887966156006, "logps/chosen": -61.539031982421875, "logps/ref_chosen": -61.80186462402344, "logps/ref_rejected": -82.37368774414062, "logps/rejected": -82.37822723388672, "loss": 1.361, "margin_dpo/margin_mean": 0.26736900210380554, "margin_dpo/margin_std": 0.7122489213943481, "step": 75 }, { "epoch": 0.11489040060468632, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.22301767766475677, "fcm_dpo/q_t": 0.4944329559803009, "grad_norm": 17.773778915405273, "learning_rate": 4.997762556634679e-07, "logits/chosen": 1.411060094833374, "logits/rejected": 1.2945332527160645, "logps/chosen": -69.60801696777344, "logps/ref_chosen": -69.92233276367188, "logps/ref_rejected": -97.08378601074219, "logps/rejected": -96.99249267578125, "loss": 1.3657, "margin_dpo/margin_mean": 0.22301700711250305, "margin_dpo/margin_std": 0.7884755730628967, "step": 76 }, { "epoch": 0.1164021164021164, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.39924293756484985, "fcm_dpo/q_t": 0.49004054069519043, "grad_norm": 18.30223846435547, "learning_rate": 4.99716834795752e-07, "logits/chosen": 1.6913138628005981, "logits/rejected": 1.5843505859375, "logps/chosen": -70.841552734375, "logps/ref_chosen": -71.206298828125, "logps/ref_rejected": -95.22071075439453, "logps/rejected": -95.25521850585938, "loss": 1.3483, "margin_dpo/margin_mean": 0.3992432653903961, "margin_dpo/margin_std": 0.720382809638977, "step": 77 }, { "epoch": 0.11791383219954649, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.33443766832351685, "fcm_dpo/q_t": 0.4916543960571289, "grad_norm": 17.89702606201172, "learning_rate": 4.996504288113623e-07, "logits/chosen": 1.7041699886322021, "logits/rejected": 1.6885236501693726, "logps/chosen": -83.9932861328125, "logps/ref_chosen": -84.40055847167969, "logps/ref_rejected": -95.41949462890625, "logps/rejected": -95.34666442871094, "loss": 1.3547, "margin_dpo/margin_mean": 0.3344375789165497, "margin_dpo/margin_std": 0.7655009031295776, "step": 78 }, { "epoch": 0.11942554799697656, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4601489007472992, "fcm_dpo/q_t": 0.48854029178619385, "grad_norm": 19.30738639831543, "learning_rate": 4.995770395678171e-07, "logits/chosen": 1.777226209640503, "logits/rejected": 1.5804474353790283, "logps/chosen": -65.54251098632812, "logps/ref_chosen": -65.93923950195312, "logps/ref_rejected": -102.92240905761719, "logps/rejected": -102.98582458496094, "loss": 1.3432, "margin_dpo/margin_mean": 0.46014824509620667, "margin_dpo/margin_std": 0.9754971265792847, "step": 79 }, { "epoch": 0.12093726379440665, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.280136376619339, "fcm_dpo/q_t": 0.4930172264575958, "grad_norm": 17.493688583374023, "learning_rate": 4.994966691179711e-07, "logits/chosen": 1.8364324569702148, "logits/rejected": 1.6422932147979736, "logps/chosen": -78.30735778808594, "logps/ref_chosen": -78.61624908447266, "logps/ref_rejected": -99.9122314453125, "logps/rejected": -99.88346862792969, "loss": 1.3604, "margin_dpo/margin_mean": 0.2801358699798584, "margin_dpo/margin_std": 0.8769626617431641, "step": 80 }, { "epoch": 0.12244897959183673, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.37103793025016785, "fcm_dpo/q_t": 0.4907492399215698, "grad_norm": 17.90415382385254, "learning_rate": 4.994093197099587e-07, "logits/chosen": 1.5202568769454956, "logits/rejected": 1.4248197078704834, "logps/chosen": -79.17378234863281, "logps/ref_chosen": -79.49641418457031, "logps/ref_rejected": -94.52413940429688, "logps/rejected": -94.57255554199219, "loss": 1.3515, "margin_dpo/margin_mean": 0.3710388243198395, "margin_dpo/margin_std": 0.8805793523788452, "step": 81 }, { "epoch": 0.12396069538926682, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6147125959396362, "fcm_dpo/q_t": 0.48470258712768555, "grad_norm": 17.819875717163086, "learning_rate": 4.993149937871306e-07, "logits/chosen": 1.7468092441558838, "logits/rejected": 1.5803828239440918, "logps/chosen": -64.35386657714844, "logps/ref_chosen": -64.97168731689453, "logps/ref_rejected": -86.69085693359375, "logps/rejected": -86.68775939941406, "loss": 1.3285, "margin_dpo/margin_mean": 0.6147127151489258, "margin_dpo/margin_std": 1.030060887336731, "step": 82 }, { "epoch": 0.1254724111866969, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5431385040283203, "fcm_dpo/q_t": 0.4864582121372223, "grad_norm": 18.520225524902344, "learning_rate": 4.992136939879856e-07, "logits/chosen": 1.7743589878082275, "logits/rejected": 1.6339887380599976, "logps/chosen": -72.32333374023438, "logps/ref_chosen": -72.92498779296875, "logps/ref_rejected": -92.27165222167969, "logps/rejected": -92.213134765625, "loss": 1.3347, "margin_dpo/margin_mean": 0.5431385636329651, "margin_dpo/margin_std": 0.8640463948249817, "step": 83 }, { "epoch": 0.12698412698412698, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5130627155303955, "fcm_dpo/q_t": 0.4872024953365326, "grad_norm": 19.25313377380371, "learning_rate": 4.991054231460969e-07, "logits/chosen": 1.8109040260314941, "logits/rejected": 1.6407761573791504, "logps/chosen": -81.33572387695312, "logps/ref_chosen": -81.79109191894531, "logps/ref_rejected": -99.20896911621094, "logps/rejected": -99.26666259765625, "loss": 1.3379, "margin_dpo/margin_mean": 0.5130621194839478, "margin_dpo/margin_std": 0.9440656900405884, "step": 84 }, { "epoch": 0.12849584278155707, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5673836469650269, "fcm_dpo/q_t": 0.4858438968658447, "grad_norm": 17.447330474853516, "learning_rate": 4.989901842900325e-07, "logits/chosen": 1.441693663597107, "logits/rejected": 1.329040765762329, "logps/chosen": -67.29985046386719, "logps/ref_chosen": -67.94147491455078, "logps/ref_rejected": -85.76875305175781, "logps/rejected": -85.69450378417969, "loss": 1.3324, "margin_dpo/margin_mean": 0.5673837661743164, "margin_dpo/margin_std": 0.907517671585083, "step": 85 }, { "epoch": 0.13000755857898716, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.34173232316970825, "fcm_dpo/q_t": 0.4914761185646057, "grad_norm": 17.49716567993164, "learning_rate": 4.988679806432711e-07, "logits/chosen": 1.9244778156280518, "logits/rejected": 1.8607923984527588, "logps/chosen": -78.9068603515625, "logps/ref_chosen": -79.21485900878906, "logps/ref_rejected": -88.69877624511719, "logps/rejected": -88.73251342773438, "loss": 1.3541, "margin_dpo/margin_mean": 0.3417322337627411, "margin_dpo/margin_std": 0.8293349742889404, "step": 86 }, { "epoch": 0.13151927437641722, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7085888385772705, "fcm_dpo/q_t": 0.48235026001930237, "grad_norm": 18.823593139648438, "learning_rate": 4.987388156241114e-07, "logits/chosen": 1.5132489204406738, "logits/rejected": 1.296330213546753, "logps/chosen": -83.90985107421875, "logps/ref_chosen": -84.45362854003906, "logps/ref_rejected": -103.43824005126953, "logps/rejected": -103.60305786132812, "loss": 1.3201, "margin_dpo/margin_mean": 0.7085880041122437, "margin_dpo/margin_std": 1.1765565872192383, "step": 87 }, { "epoch": 0.1330309901738473, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3276621699333191, "fcm_dpo/q_t": 0.4918380081653595, "grad_norm": 18.28816032409668, "learning_rate": 4.986026928455767e-07, "logits/chosen": 1.723911166191101, "logits/rejected": 1.7007009983062744, "logps/chosen": -80.91500091552734, "logps/ref_chosen": -81.27230834960938, "logps/ref_rejected": -89.51646423339844, "logps/rejected": -89.48681640625, "loss": 1.3574, "margin_dpo/margin_mean": 0.3276621699333191, "margin_dpo/margin_std": 1.1452457904815674, "step": 88 }, { "epoch": 0.1345427059712774, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7864416837692261, "fcm_dpo/q_t": 0.4804958701133728, "grad_norm": 18.09223175048828, "learning_rate": 4.984596161153135e-07, "logits/chosen": 1.9219346046447754, "logits/rejected": 1.647355079650879, "logps/chosen": -57.445068359375, "logps/ref_chosen": -58.142333984375, "logps/ref_rejected": -102.53756713867188, "logps/rejected": -102.62675476074219, "loss": 1.3138, "margin_dpo/margin_mean": 0.7864429354667664, "margin_dpo/margin_std": 1.318110466003418, "step": 89 }, { "epoch": 0.1360544217687075, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5627226233482361, "fcm_dpo/q_t": 0.4859766960144043, "grad_norm": 19.66481590270996, "learning_rate": 4.983095894354857e-07, "logits/chosen": 1.6534063816070557, "logits/rejected": 1.4348580837249756, "logps/chosen": -74.80726623535156, "logps/ref_chosen": -75.26505279541016, "logps/ref_rejected": -104.32842254638672, "logps/rejected": -104.43335723876953, "loss": 1.3341, "margin_dpo/margin_mean": 0.5627224445343018, "margin_dpo/margin_std": 1.1481568813323975, "step": 90 }, { "epoch": 0.13756613756613756, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5365311503410339, "fcm_dpo/q_t": 0.48667216300964355, "grad_norm": 17.257240295410156, "learning_rate": 4.98152617002662e-07, "logits/chosen": 1.9292263984680176, "logits/rejected": 1.7573230266571045, "logps/chosen": -68.76639556884766, "logps/ref_chosen": -69.33901977539062, "logps/ref_rejected": -90.31411743164062, "logps/rejected": -90.27801513671875, "loss": 1.338, "margin_dpo/margin_mean": 0.5365312099456787, "margin_dpo/margin_std": 1.3548262119293213, "step": 91 }, { "epoch": 0.13907785336356765, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.637989342212677, "fcm_dpo/q_t": 0.4841547906398773, "grad_norm": 20.7775936126709, "learning_rate": 4.979887032076988e-07, "logits/chosen": 1.601667881011963, "logits/rejected": 1.4548040628433228, "logps/chosen": -71.87997436523438, "logps/ref_chosen": -72.4566650390625, "logps/ref_rejected": -91.6706771850586, "logps/rejected": -91.73197937011719, "loss": 1.3276, "margin_dpo/margin_mean": 0.6379896402359009, "margin_dpo/margin_std": 1.2460086345672607, "step": 92 }, { "epoch": 0.14058956916099774, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4430373013019562, "fcm_dpo/q_t": 0.489046186208725, "grad_norm": 16.134719848632812, "learning_rate": 4.978178526356172e-07, "logits/chosen": 1.73604416847229, "logits/rejected": 1.6414337158203125, "logps/chosen": -63.422340393066406, "logps/ref_chosen": -64.08897399902344, "logps/ref_rejected": -75.09095764160156, "logps/rejected": -74.86735534667969, "loss": 1.3489, "margin_dpo/margin_mean": 0.4430374205112457, "margin_dpo/margin_std": 1.593023419380188, "step": 93 }, { "epoch": 0.1421012849584278, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0405027866363525, "fcm_dpo/q_t": 0.4742341637611389, "grad_norm": 37.81088638305664, "learning_rate": 4.976400700654751e-07, "logits/chosen": 2.0334482192993164, "logits/rejected": 1.8472344875335693, "logps/chosen": -78.81251525878906, "logps/ref_chosen": -79.67372131347656, "logps/ref_rejected": -94.64076232910156, "logps/rejected": -94.82006072998047, "loss": 1.292, "margin_dpo/margin_mean": 1.040502667427063, "margin_dpo/margin_std": 1.6742515563964844, "step": 94 }, { "epoch": 0.1436130007558579, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6995920538902283, "fcm_dpo/q_t": 0.48263174295425415, "grad_norm": 18.631126403808594, "learning_rate": 4.974553604702332e-07, "logits/chosen": 1.4102903604507446, "logits/rejected": 1.2439355850219727, "logps/chosen": -78.23617553710938, "logps/ref_chosen": -78.65760803222656, "logps/ref_rejected": -109.4048080444336, "logps/rejected": -109.68296813964844, "loss": 1.3237, "margin_dpo/margin_mean": 0.6995913982391357, "margin_dpo/margin_std": 1.566126823425293, "step": 95 }, { "epoch": 0.14512471655328799, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.8266540169715881, "fcm_dpo/q_t": 0.4794994592666626, "grad_norm": 18.881208419799805, "learning_rate": 4.972637290166157e-07, "logits/chosen": 1.4069623947143555, "logits/rejected": 1.2911567687988281, "logps/chosen": -77.2410659790039, "logps/ref_chosen": -77.708251953125, "logps/ref_rejected": -104.36044311523438, "logps/rejected": -104.71990203857422, "loss": 1.3116, "margin_dpo/margin_mean": 0.8266537189483643, "margin_dpo/margin_std": 1.5478097200393677, "step": 96 }, { "epoch": 0.14663643235071808, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3047708570957184, "fcm_dpo/q_t": 0.4924178719520569, "grad_norm": 19.201942443847656, "learning_rate": 4.970651810649666e-07, "logits/chosen": 1.6156105995178223, "logits/rejected": 1.5133846998214722, "logps/chosen": -84.23475646972656, "logps/ref_chosen": -84.58917999267578, "logps/ref_rejected": -99.25704956054688, "logps/rejected": -99.2073974609375, "loss": 1.3622, "margin_dpo/margin_mean": 0.3047705590724945, "margin_dpo/margin_std": 1.56210196018219, "step": 97 }, { "epoch": 0.14814814814814814, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4234497845172882, "fcm_dpo/q_t": 0.48944059014320374, "grad_norm": 17.522523880004883, "learning_rate": 4.968597221690985e-07, "logits/chosen": 1.7427449226379395, "logits/rejected": 1.6909432411193848, "logps/chosen": -74.09392547607422, "logps/ref_chosen": -74.42477416992188, "logps/ref_rejected": -88.93840026855469, "logps/rejected": -89.03099822998047, "loss": 1.3494, "margin_dpo/margin_mean": 0.42344966530799866, "margin_dpo/margin_std": 1.4119412899017334, "step": 98 }, { "epoch": 0.14965986394557823, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6521110534667969, "fcm_dpo/q_t": 0.4839605689048767, "grad_norm": 18.090377807617188, "learning_rate": 4.966473580761389e-07, "logits/chosen": 2.0218091011047363, "logits/rejected": 1.9115238189697266, "logps/chosen": -75.04991912841797, "logps/ref_chosen": -75.59742736816406, "logps/ref_rejected": -98.2310791015625, "logps/rejected": -98.33567810058594, "loss": 1.332, "margin_dpo/margin_mean": 0.6521108150482178, "margin_dpo/margin_std": 1.979665994644165, "step": 99 }, { "epoch": 0.15117157974300832, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7829453349113464, "fcm_dpo/q_t": 0.48082974553108215, "grad_norm": 19.518943786621094, "learning_rate": 4.964280947263676e-07, "logits/chosen": 1.8196897506713867, "logits/rejected": 1.7950403690338135, "logps/chosen": -98.03388977050781, "logps/ref_chosen": -98.55859375, "logps/ref_rejected": -106.01295471191406, "logps/rejected": -106.27117919921875, "loss": 1.3228, "margin_dpo/margin_mean": 0.7829455137252808, "margin_dpo/margin_std": 2.2036726474761963, "step": 100 }, { "epoch": 0.15117157974300832, "eval_fcm_dpo/beta": 0.10000000894069672, "eval_logits/chosen": 1.691135048866272, "eval_logits/rejected": 1.577427864074707, "eval_logps/chosen": -86.23526763916016, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -96.92163848876953, "eval_loss": 0.6546333432197571, "eval_margin_dpo/margin_mean": 0.891772449016571, "eval_margin_dpo/margin_std": 1.9760589599609375, "eval_runtime": 42.284, "eval_samples_per_second": 54.465, "eval_steps_per_second": 1.703, "step": 100 }, { "epoch": 0.15268329554043839, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.9559938907623291, "fcm_dpo/q_t": 0.476234495639801, "grad_norm": 16.3298282623291, "learning_rate": 4.96201938253052e-07, "logits/chosen": 1.4772682189941406, "logits/rejected": 1.4215991497039795, "logps/chosen": -68.6669921875, "logps/ref_chosen": -69.45216369628906, "logps/ref_rejected": -88.0458755493164, "logps/rejected": -88.21668243408203, "loss": 1.3001, "margin_dpo/margin_mean": 0.9559941291809082, "margin_dpo/margin_std": 1.6717371940612793, "step": 101 }, { "epoch": 0.15419501133786848, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5958479642868042, "fcm_dpo/q_t": 0.48534175753593445, "grad_norm": 17.342573165893555, "learning_rate": 4.959688949822748e-07, "logits/chosen": 1.3332377672195435, "logits/rejected": 1.264945387840271, "logps/chosen": -79.86666870117188, "logps/ref_chosen": -80.35308837890625, "logps/ref_rejected": -90.61380004882812, "logps/rejected": -90.72323608398438, "loss": 1.3377, "margin_dpo/margin_mean": 0.5958476662635803, "margin_dpo/margin_std": 2.011353015899658, "step": 102 }, { "epoch": 0.15570672713529857, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.210947036743164, "fcm_dpo/q_t": 0.47017818689346313, "grad_norm": 17.412508010864258, "learning_rate": 4.957289714327572e-07, "logits/chosen": 1.8353683948516846, "logits/rejected": 1.7695558071136475, "logps/chosen": -78.50149536132812, "logps/ref_chosen": -79.30392456054688, "logps/ref_rejected": -93.745361328125, "logps/rejected": -94.15388488769531, "loss": 1.2799, "margin_dpo/margin_mean": 1.210945963859558, "margin_dpo/margin_std": 2.1122875213623047, "step": 103 }, { "epoch": 0.15721844293272866, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.138046145439148, "fcm_dpo/q_t": 0.4720809757709503, "grad_norm": 18.750715255737305, "learning_rate": 4.954821743156767e-07, "logits/chosen": 1.8510358333587646, "logits/rejected": 1.610731601715088, "logps/chosen": -73.64946746826172, "logps/ref_chosen": -74.50674438476562, "logps/ref_rejected": -116.09912872314453, "logps/rejected": -116.37989807128906, "loss": 1.2898, "margin_dpo/margin_mean": 1.1380459070205688, "margin_dpo/margin_std": 2.3527626991271973, "step": 104 }, { "epoch": 0.15873015873015872, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.992182731628418, "fcm_dpo/q_t": 0.47577622532844543, "grad_norm": 18.913448333740234, "learning_rate": 4.952285105344791e-07, "logits/chosen": 1.8412593603134155, "logits/rejected": 1.6697262525558472, "logps/chosen": -87.25275421142578, "logps/ref_chosen": -87.76654815673828, "logps/ref_rejected": -108.07927703857422, "logps/rejected": -108.55766296386719, "loss": 1.3093, "margin_dpo/margin_mean": 0.9921829700469971, "margin_dpo/margin_std": 2.8260951042175293, "step": 105 }, { "epoch": 0.1602418745275888, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.8636230230331421, "fcm_dpo/q_t": 0.47864556312561035, "grad_norm": 17.13074493408203, "learning_rate": 4.949679871846857e-07, "logits/chosen": 1.705589771270752, "logits/rejected": 1.6496243476867676, "logps/chosen": -75.5406494140625, "logps/ref_chosen": -76.38548278808594, "logps/ref_rejected": -81.63407897949219, "logps/rejected": -81.65287017822266, "loss": 1.3159, "margin_dpo/margin_mean": 0.8636230826377869, "margin_dpo/margin_std": 2.3666610717773438, "step": 106 }, { "epoch": 0.1617535903250189, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6637716889381409, "fcm_dpo/q_t": 0.4834796190261841, "grad_norm": 19.10794448852539, "learning_rate": 4.947006115536947e-07, "logits/chosen": 1.299161434173584, "logits/rejected": 1.2430386543273926, "logps/chosen": -95.80593872070312, "logps/ref_chosen": -96.14849853515625, "logps/ref_rejected": -107.0481185913086, "logps/rejected": -107.36932373046875, "loss": 1.3336, "margin_dpo/margin_mean": 0.6637719869613647, "margin_dpo/margin_std": 2.2503273487091064, "step": 107 }, { "epoch": 0.16326530612244897, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.853187084197998, "fcm_dpo/q_t": 0.47897130250930786, "grad_norm": 17.166805267333984, "learning_rate": 4.944263911205772e-07, "logits/chosen": 1.4326887130737305, "logits/rejected": 1.309265375137329, "logps/chosen": -84.71664428710938, "logps/ref_chosen": -85.39241027832031, "logps/ref_rejected": -97.79592895507812, "logps/rejected": -97.97334289550781, "loss": 1.3162, "margin_dpo/margin_mean": 0.8531871438026428, "margin_dpo/margin_std": 2.3266098499298096, "step": 108 }, { "epoch": 0.16477702191987906, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.3385164737701416, "fcm_dpo/q_t": 0.46750587224960327, "grad_norm": 18.09962272644043, "learning_rate": 4.941453335558681e-07, "logits/chosen": 1.723799467086792, "logits/rejected": 1.4775876998901367, "logps/chosen": -78.03567504882812, "logps/ref_chosen": -78.99874877929688, "logps/ref_rejected": -100.79278564453125, "logps/rejected": -101.16822814941406, "loss": 1.2754, "margin_dpo/margin_mean": 1.3385167121887207, "margin_dpo/margin_std": 2.743279218673706, "step": 109 }, { "epoch": 0.16628873771730915, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.241614431142807, "fcm_dpo/q_t": 0.4939076602458954, "grad_norm": 20.46181297302246, "learning_rate": 4.938574467213517e-07, "logits/chosen": 1.4239675998687744, "logits/rejected": 1.485081434249878, "logps/chosen": -96.55082702636719, "logps/ref_chosen": -96.95277404785156, "logps/ref_rejected": -91.44450378417969, "logps/rejected": -91.2841567993164, "loss": 1.3772, "margin_dpo/margin_mean": 0.2416144609451294, "margin_dpo/margin_std": 2.4270529747009277, "step": 110 }, { "epoch": 0.16780045351473924, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.9300182461738586, "fcm_dpo/q_t": 0.47719940543174744, "grad_norm": 16.58338737487793, "learning_rate": 4.935627386698418e-07, "logits/chosen": 1.696366548538208, "logits/rejected": 1.5432548522949219, "logps/chosen": -69.25906372070312, "logps/ref_chosen": -70.01641845703125, "logps/ref_rejected": -92.87696838378906, "logps/rejected": -93.04963684082031, "loss": 1.3109, "margin_dpo/margin_mean": 0.9300180077552795, "margin_dpo/margin_std": 2.4844956398010254, "step": 111 }, { "epoch": 0.1693121693121693, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2315406799316406, "fcm_dpo/q_t": 0.4697922468185425, "grad_norm": 19.882429122924805, "learning_rate": 4.932612176449559e-07, "logits/chosen": 1.7238032817840576, "logits/rejected": 1.5111950635910034, "logps/chosen": -76.88069152832031, "logps/ref_chosen": -77.80027770996094, "logps/ref_rejected": -123.10624694824219, "logps/rejected": -123.41820526123047, "loss": 1.2845, "margin_dpo/margin_mean": 1.2315394878387451, "margin_dpo/margin_std": 2.6262001991271973, "step": 112 }, { "epoch": 0.1708238851095994, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.9178091287612915, "fcm_dpo/q_t": 0.47745102643966675, "grad_norm": 16.442792892456055, "learning_rate": 4.929528920808854e-07, "logits/chosen": 1.3930963277816772, "logits/rejected": 1.3205270767211914, "logps/chosen": -69.23207092285156, "logps/ref_chosen": -70.54346466064453, "logps/ref_rejected": -88.79286193847656, "logps/rejected": -88.39927673339844, "loss": 1.3135, "margin_dpo/margin_mean": 0.9178098440170288, "margin_dpo/margin_std": 2.57177472114563, "step": 113 }, { "epoch": 0.17233560090702948, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.3459113836288452, "fcm_dpo/q_t": 0.4675772786140442, "grad_norm": 17.996028900146484, "learning_rate": 4.92637770602159e-07, "logits/chosen": 1.716050386428833, "logits/rejected": 1.5711562633514404, "logps/chosen": -82.80506134033203, "logps/ref_chosen": -83.9239501953125, "logps/ref_rejected": -92.85765838623047, "logps/rejected": -93.08468627929688, "loss": 1.2804, "margin_dpo/margin_mean": 1.3459113836288452, "margin_dpo/margin_std": 3.139867067337036, "step": 114 }, { "epoch": 0.17384731670445955, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2582768201828003, "fcm_dpo/q_t": 0.4693111777305603, "grad_norm": 16.850736618041992, "learning_rate": 4.923158620234019e-07, "logits/chosen": 1.575303316116333, "logits/rejected": 1.4128024578094482, "logps/chosen": -68.53926086425781, "logps/ref_chosen": -69.82767486572266, "logps/ref_rejected": -96.51564025878906, "logps/rejected": -96.48550415039062, "loss": 1.2811, "margin_dpo/margin_mean": 1.2582770586013794, "margin_dpo/margin_std": 2.586811065673828, "step": 115 }, { "epoch": 0.17535903250188964, "fcm_dpo/beta": 0.1008991077542305, "fcm_dpo/delta": 0.08911184966564178, "fcm_dpo/margin": 1.6396100521087646, "fcm_dpo/q_t": 0.4599873721599579, "grad_norm": 18.267988204956055, "learning_rate": 4.91987175349089e-07, "logits/chosen": 1.6626043319702148, "logits/rejected": 1.510635495185852, "logps/chosen": -64.81758880615234, "logps/ref_chosen": -66.19773864746094, "logps/ref_rejected": -90.88304138183594, "logps/rejected": -91.14250183105469, "loss": 1.2453, "margin_dpo/margin_mean": 1.6396100521087646, "margin_dpo/margin_std": 2.6216936111450195, "step": 116 }, { "epoch": 0.17687074829931973, "fcm_dpo/beta": 0.10179821401834488, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.585550308227539, "fcm_dpo/q_t": 0.4604984521865845, "grad_norm": 16.826187133789062, "learning_rate": 4.916517197732933e-07, "logits/chosen": 1.6091898679733276, "logits/rejected": 1.5089998245239258, "logps/chosen": -70.49270629882812, "logps/ref_chosen": -72.15988159179688, "logps/ref_rejected": -85.30296325683594, "logps/rejected": -85.22134399414062, "loss": 1.2524, "margin_dpo/margin_mean": 1.58555006980896, "margin_dpo/margin_std": 2.8378653526306152, "step": 117 }, { "epoch": 0.17838246409674982, "fcm_dpo/beta": 0.10179821401834488, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2571253776550293, "fcm_dpo/q_t": 0.46861183643341064, "grad_norm": 16.774629592895508, "learning_rate": 4.913095046794281e-07, "logits/chosen": 1.6488983631134033, "logits/rejected": 1.5311334133148193, "logps/chosen": -70.03057098388672, "logps/ref_chosen": -71.47773742675781, "logps/ref_rejected": -96.95051574707031, "logps/rejected": -96.76048278808594, "loss": 1.2838, "margin_dpo/margin_mean": 1.2571251392364502, "margin_dpo/margin_std": 2.8928143978118896, "step": 118 }, { "epoch": 0.17989417989417988, "fcm_dpo/beta": 0.10179821401834488, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.358586311340332, "fcm_dpo/q_t": 0.4668753743171692, "grad_norm": 17.19606590270996, "learning_rate": 4.909605396399855e-07, "logits/chosen": 1.7519330978393555, "logits/rejected": 1.6531188488006592, "logps/chosen": -76.84124755859375, "logps/ref_chosen": -78.2727279663086, "logps/ref_rejected": -94.71317291259766, "logps/rejected": -94.64027404785156, "loss": 1.2855, "margin_dpo/margin_mean": 1.3585854768753052, "margin_dpo/margin_std": 3.5642504692077637, "step": 119 }, { "epoch": 0.18140589569160998, "fcm_dpo/beta": 0.10353302955627441, "fcm_dpo/delta": 0.08449066430330276, "fcm_dpo/margin": 1.9712594747543335, "fcm_dpo/q_t": 0.4509131908416748, "grad_norm": 18.62177085876465, "learning_rate": 4.906048344162676e-07, "logits/chosen": 1.8366992473602295, "logits/rejected": 1.6866886615753174, "logps/chosen": -76.55362701416016, "logps/ref_chosen": -78.43109130859375, "logps/ref_rejected": -100.2771987915039, "logps/rejected": -100.37100219726562, "loss": 1.2151, "margin_dpo/margin_mean": 1.9712594747543335, "margin_dpo/margin_std": 2.9160304069519043, "step": 120 }, { "epoch": 0.18291761148904007, "fcm_dpo/beta": 0.10353302955627441, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.4688817262649536, "fcm_dpo/q_t": 0.4635319113731384, "grad_norm": 19.076129913330078, "learning_rate": 4.902423989581143e-07, "logits/chosen": 2.027730941772461, "logits/rejected": 1.7624661922454834, "logps/chosen": -72.38795471191406, "logps/ref_chosen": -74.08768463134766, "logps/ref_rejected": -118.6731948852539, "logps/rejected": -118.44233703613281, "loss": 1.268, "margin_dpo/margin_mean": 1.4688820838928223, "margin_dpo/margin_std": 3.261903762817383, "step": 121 }, { "epoch": 0.18442932728647016, "fcm_dpo/beta": 0.10353302955627441, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.4933545589447021, "fcm_dpo/q_t": 0.46344470977783203, "grad_norm": 18.32591438293457, "learning_rate": 4.898732434036243e-07, "logits/chosen": 1.5426521301269531, "logits/rejected": 1.419804573059082, "logps/chosen": -77.63048553466797, "logps/ref_chosen": -79.36762237548828, "logps/ref_rejected": -92.42371368408203, "logps/rejected": -92.179931640625, "loss": 1.2719, "margin_dpo/margin_mean": 1.493354082107544, "margin_dpo/margin_std": 3.603337287902832, "step": 122 }, { "epoch": 0.18594104308390022, "fcm_dpo/beta": 0.10353302955627441, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.630068063735962, "fcm_dpo/q_t": 0.45924896001815796, "grad_norm": 17.652915954589844, "learning_rate": 4.894973780788722e-07, "logits/chosen": 1.6213600635528564, "logits/rejected": 1.5191831588745117, "logps/chosen": -69.88966369628906, "logps/ref_chosen": -71.91705322265625, "logps/ref_rejected": -96.36418151855469, "logps/rejected": -95.96685028076172, "loss": 1.2503, "margin_dpo/margin_mean": 1.630068063735962, "margin_dpo/margin_std": 3.0862808227539062, "step": 123 }, { "epoch": 0.1874527588813303, "fcm_dpo/beta": 0.10440807044506073, "fcm_dpo/delta": 0.0838112086057663, "fcm_dpo/margin": 1.9508870840072632, "fcm_dpo/q_t": 0.4518246054649353, "grad_norm": 18.525684356689453, "learning_rate": 4.89114813497619e-07, "logits/chosen": 1.9808456897735596, "logits/rejected": 1.8465297222137451, "logps/chosen": -69.44947052001953, "logps/ref_chosen": -71.72529602050781, "logps/ref_rejected": -111.17984771728516, "logps/rejected": -110.85490417480469, "loss": 1.2301, "margin_dpo/margin_mean": 1.9508872032165527, "margin_dpo/margin_std": 3.758726119995117, "step": 124 }, { "epoch": 0.1889644746787604, "fcm_dpo/beta": 0.1063346192240715, "fcm_dpo/delta": 0.04968992993235588, "fcm_dpo/margin": 2.074455976486206, "fcm_dpo/q_t": 0.44727903604507446, "grad_norm": 19.388065338134766, "learning_rate": 4.887255603610184e-07, "logits/chosen": 1.8083943128585815, "logits/rejected": 1.618847131729126, "logps/chosen": -79.10482025146484, "logps/ref_chosen": -81.55532836914062, "logps/ref_rejected": -110.9144287109375, "logps/rejected": -110.53838348388672, "loss": 1.2137, "margin_dpo/margin_mean": 2.074455499649048, "margin_dpo/margin_std": 3.5274128913879395, "step": 125 }, { "epoch": 0.19047619047619047, "fcm_dpo/beta": 0.10724936425685883, "fcm_dpo/delta": 0.08529400080442429, "fcm_dpo/margin": 1.6670329570770264, "fcm_dpo/q_t": 0.4588611423969269, "grad_norm": 19.64308738708496, "learning_rate": 4.883296295573176e-07, "logits/chosen": 1.1320900917053223, "logits/rejected": 1.1597490310668945, "logps/chosen": -83.641357421875, "logps/ref_chosen": -87.07349395751953, "logps/ref_rejected": -85.05271911621094, "logps/rejected": -83.28761291503906, "loss": 1.2669, "margin_dpo/margin_mean": 1.6670331954956055, "margin_dpo/margin_std": 4.1900434494018555, "step": 126 }, { "epoch": 0.19198790627362056, "fcm_dpo/beta": 0.10914330929517746, "fcm_dpo/delta": 0.04506043344736099, "fcm_dpo/margin": 2.2443766593933105, "fcm_dpo/q_t": 0.44131654500961304, "grad_norm": 18.411846160888672, "learning_rate": 4.87927032161552e-07, "logits/chosen": 1.8658032417297363, "logits/rejected": 1.8027232885360718, "logps/chosen": -77.10999298095703, "logps/ref_chosen": -80.4578857421875, "logps/ref_rejected": -90.50740051269531, "logps/rejected": -89.40388488769531, "loss": 1.1873, "margin_dpo/margin_mean": 2.2443761825561523, "margin_dpo/margin_std": 3.211836338043213, "step": 127 }, { "epoch": 0.19349962207105065, "fcm_dpo/beta": 0.10914330929517746, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.470666527748108, "fcm_dpo/q_t": 0.4621870517730713, "grad_norm": 20.818655014038086, "learning_rate": 4.875177794352363e-07, "logits/chosen": 1.4957571029663086, "logits/rejected": 1.3097018003463745, "logps/chosen": -82.75028991699219, "logps/ref_chosen": -85.77519226074219, "logps/ref_rejected": -112.63516998291016, "logps/rejected": -111.08094024658203, "loss": 1.2966, "margin_dpo/margin_mean": 1.4706671237945557, "margin_dpo/margin_std": 4.722267150878906, "step": 128 }, { "epoch": 0.19501133786848074, "fcm_dpo/beta": 0.10914330929517746, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6215567588806152, "fcm_dpo/q_t": 0.45845484733581543, "grad_norm": 20.263038635253906, "learning_rate": 4.871018828260491e-07, "logits/chosen": 1.5004316568374634, "logits/rejected": 1.5068594217300415, "logps/chosen": -81.99288940429688, "logps/ref_chosen": -84.94615173339844, "logps/ref_rejected": -85.36473846435547, "logps/rejected": -84.03303527832031, "loss": 1.2797, "margin_dpo/margin_mean": 1.6215569972991943, "margin_dpo/margin_std": 4.648078918457031, "step": 129 }, { "epoch": 0.1965230536659108, "fcm_dpo/beta": 0.11057177186012268, "fcm_dpo/delta": 0.06501518934965134, "fcm_dpo/margin": 2.0028374195098877, "fcm_dpo/q_t": 0.44799530506134033, "grad_norm": 20.011611938476562, "learning_rate": 4.866793539675126e-07, "logits/chosen": 1.4784972667694092, "logits/rejected": 1.3632197380065918, "logps/chosen": -75.56727600097656, "logps/ref_chosen": -79.0184555053711, "logps/ref_rejected": -97.63998413085938, "logps/rejected": -96.19164276123047, "loss": 1.2255, "margin_dpo/margin_mean": 2.0028371810913086, "margin_dpo/margin_std": 3.9271349906921387, "step": 130 }, { "epoch": 0.1980347694633409, "fcm_dpo/beta": 0.11228152364492416, "fcm_dpo/delta": 0.11718001961708069, "fcm_dpo/margin": 2.5501041412353516, "fcm_dpo/q_t": 0.4346443712711334, "grad_norm": 18.658187866210938, "learning_rate": 4.86250204678667e-07, "logits/chosen": 1.426423192024231, "logits/rejected": 1.1902827024459839, "logps/chosen": -64.61331939697266, "logps/ref_chosen": -68.24565887451172, "logps/ref_rejected": -97.99555969238281, "logps/rejected": -96.91332244873047, "loss": 1.1839, "margin_dpo/margin_mean": 2.5501043796539307, "margin_dpo/margin_std": 4.5746660232543945, "step": 131 }, { "epoch": 0.19954648526077098, "fcm_dpo/beta": 0.11424092948436737, "fcm_dpo/delta": 0.09166815131902695, "fcm_dpo/margin": 1.7558990716934204, "fcm_dpo/q_t": 0.45261579751968384, "grad_norm": 19.862043380737305, "learning_rate": 4.858144469637408e-07, "logits/chosen": 1.655261516571045, "logits/rejected": 1.5392924547195435, "logps/chosen": -78.36050415039062, "logps/ref_chosen": -82.06532287597656, "logps/ref_rejected": -89.47691345214844, "logps/rejected": -87.52798461914062, "loss": 1.2439, "margin_dpo/margin_mean": 1.7558988332748413, "margin_dpo/margin_std": 3.9155571460723877, "step": 132 }, { "epoch": 0.20105820105820105, "fcm_dpo/beta": 0.11807942390441895, "fcm_dpo/delta": 0.16631799936294556, "fcm_dpo/margin": 2.0136890411376953, "fcm_dpo/q_t": 0.44534194469451904, "grad_norm": 20.467275619506836, "learning_rate": 4.853720930118138e-07, "logits/chosen": 1.4204668998718262, "logits/rejected": 1.4540302753448486, "logps/chosen": -79.44143676757812, "logps/ref_chosen": -83.70661163330078, "logps/ref_rejected": -89.3868179321289, "logps/rejected": -87.13533020019531, "loss": 1.2249, "margin_dpo/margin_mean": 2.0136892795562744, "margin_dpo/margin_std": 4.2390546798706055, "step": 133 }, { "epoch": 0.20256991685563114, "fcm_dpo/beta": 0.12024421989917755, "fcm_dpo/delta": 0.0528922863304615, "fcm_dpo/margin": 2.901494026184082, "fcm_dpo/q_t": 0.4216766357421875, "grad_norm": 18.669694900512695, "learning_rate": 4.849231551964771e-07, "logits/chosen": 1.9196783304214478, "logits/rejected": 1.7756422758102417, "logps/chosen": -66.67295837402344, "logps/ref_chosen": -71.57601928710938, "logps/ref_rejected": -92.34259033203125, "logps/rejected": -90.3410415649414, "loss": 1.1417, "margin_dpo/margin_mean": 2.901494026184082, "margin_dpo/margin_std": 4.747570991516113, "step": 134 }, { "epoch": 0.20408163265306123, "fcm_dpo/beta": 0.12111984193325043, "fcm_dpo/delta": 0.054889243096113205, "fcm_dpo/margin": 2.007038116455078, "fcm_dpo/q_t": 0.4430903196334839, "grad_norm": 19.303375244140625, "learning_rate": 4.844676460754862e-07, "logits/chosen": 1.465507984161377, "logits/rejected": 1.408734917640686, "logps/chosen": -61.307891845703125, "logps/ref_chosen": -66.39884948730469, "logps/ref_rejected": -81.38636779785156, "logps/rejected": -78.30242919921875, "loss": 1.2202, "margin_dpo/margin_mean": 2.007038116455078, "margin_dpo/margin_std": 4.225521087646484, "step": 135 }, { "epoch": 0.20559334845049132, "fcm_dpo/beta": 0.12096191942691803, "fcm_dpo/delta": -0.0338929109275341, "fcm_dpo/margin": 2.4650301933288574, "fcm_dpo/q_t": 0.4351881146430969, "grad_norm": 22.06288719177246, "learning_rate": 4.840055783904106e-07, "logits/chosen": 1.5917037725448608, "logits/rejected": 1.3086708784103394, "logps/chosen": -82.8524169921875, "logps/ref_chosen": -86.75381469726562, "logps/ref_rejected": -113.35548400878906, "logps/rejected": -111.91911315917969, "loss": 1.232, "margin_dpo/margin_mean": 2.4650299549102783, "margin_dpo/margin_std": 5.841936111450195, "step": 136 }, { "epoch": 0.20710506424792138, "fcm_dpo/beta": 0.12060153484344482, "fcm_dpo/delta": -0.029881920665502548, "fcm_dpo/margin": 2.5589451789855957, "fcm_dpo/q_t": 0.4295978248119354, "grad_norm": 18.40294075012207, "learning_rate": 4.835369650662767e-07, "logits/chosen": 1.843554973602295, "logits/rejected": 1.72433340549469, "logps/chosen": -67.306396484375, "logps/ref_chosen": -72.21119689941406, "logps/ref_rejected": -88.30802917480469, "logps/rejected": -85.96217346191406, "loss": 1.1748, "margin_dpo/margin_mean": 2.5589451789855957, "margin_dpo/margin_std": 4.473442077636719, "step": 137 }, { "epoch": 0.20861678004535147, "fcm_dpo/beta": 0.12128353118896484, "fcm_dpo/delta": 0.08594699203968048, "fcm_dpo/margin": 1.882028579711914, "fcm_dpo/q_t": 0.45027002692222595, "grad_norm": 20.4240779876709, "learning_rate": 4.830618192112065e-07, "logits/chosen": 1.2977592945098877, "logits/rejected": 1.201406717300415, "logps/chosen": -70.62525939941406, "logps/ref_chosen": -74.54273223876953, "logps/ref_rejected": -84.63615417480469, "logps/rejected": -82.60069274902344, "loss": 1.2605, "margin_dpo/margin_mean": 1.8820289373397827, "margin_dpo/margin_std": 5.117735385894775, "step": 138 }, { "epoch": 0.21012849584278157, "fcm_dpo/beta": 0.12293344736099243, "fcm_dpo/delta": 0.049421682953834534, "fcm_dpo/margin": 1.9987208843231201, "fcm_dpo/q_t": 0.4412422180175781, "grad_norm": 24.668521881103516, "learning_rate": 4.825801541160509e-07, "logits/chosen": 1.370278000831604, "logits/rejected": 1.3114217519760132, "logps/chosen": -84.10897827148438, "logps/ref_chosen": -87.63740539550781, "logps/ref_rejected": -101.3896484375, "logps/rejected": -99.85993957519531, "loss": 1.2438, "margin_dpo/margin_mean": 1.9987210035324097, "margin_dpo/margin_std": 4.9212646484375, "step": 139 }, { "epoch": 0.21164021164021163, "fcm_dpo/beta": 0.12314370274543762, "fcm_dpo/delta": 0.0037242621183395386, "fcm_dpo/margin": 3.2170228958129883, "fcm_dpo/q_t": 0.4121108651161194, "grad_norm": 24.398794174194336, "learning_rate": 4.820919832540181e-07, "logits/chosen": 1.258446216583252, "logits/rejected": 1.1674163341522217, "logps/chosen": -76.97390747070312, "logps/ref_chosen": -81.32339477539062, "logps/ref_rejected": -99.7275619506836, "logps/rejected": -98.59510040283203, "loss": 1.1392, "margin_dpo/margin_mean": 3.2170233726501465, "margin_dpo/margin_std": 5.567892074584961, "step": 140 }, { "epoch": 0.21315192743764172, "fcm_dpo/beta": 0.12346719205379486, "fcm_dpo/delta": 0.01799967885017395, "fcm_dpo/margin": 3.0974936485290527, "fcm_dpo/q_t": 0.4152432382106781, "grad_norm": 18.94412612915039, "learning_rate": 4.815973202802966e-07, "logits/chosen": 1.5887515544891357, "logits/rejected": 1.4874029159545898, "logps/chosen": -74.20223236083984, "logps/ref_chosen": -78.08534240722656, "logps/ref_rejected": -101.70516967773438, "logps/rejected": -100.9195556640625, "loss": 1.1532, "margin_dpo/margin_mean": 3.0974936485290527, "margin_dpo/margin_std": 5.584391117095947, "step": 141 }, { "epoch": 0.2146636432350718, "fcm_dpo/beta": 0.12660640478134155, "fcm_dpo/delta": 0.12214577943086624, "fcm_dpo/margin": 2.2193617820739746, "fcm_dpo/q_t": 0.4376070201396942, "grad_norm": 20.792604446411133, "learning_rate": 4.810961790316729e-07, "logits/chosen": 1.5094895362854004, "logits/rejected": 1.4517810344696045, "logps/chosen": -79.07592010498047, "logps/ref_chosen": -82.84616088867188, "logps/ref_rejected": -95.14714050292969, "logps/rejected": -93.59625244140625, "loss": 1.2179, "margin_dpo/margin_mean": 2.2193617820739746, "margin_dpo/margin_std": 4.92032527923584, "step": 142 }, { "epoch": 0.2161753590325019, "fcm_dpo/beta": 0.12763959169387817, "fcm_dpo/delta": 0.03856246545910835, "fcm_dpo/margin": 2.0553064346313477, "fcm_dpo/q_t": 0.442947119474411, "grad_norm": 29.368139266967773, "learning_rate": 4.805885735261454e-07, "logits/chosen": 1.50538170337677, "logits/rejected": 1.4741663932800293, "logps/chosen": -76.84846496582031, "logps/ref_chosen": -80.29791259765625, "logps/ref_rejected": -87.44291687011719, "logps/rejected": -86.04878234863281, "loss": 1.2779, "margin_dpo/margin_mean": 2.055307388305664, "margin_dpo/margin_std": 6.01151180267334, "step": 143 }, { "epoch": 0.21768707482993196, "fcm_dpo/beta": 0.129078671336174, "fcm_dpo/delta": 0.03681322559714317, "fcm_dpo/margin": 1.5377318859100342, "fcm_dpo/q_t": 0.4620184898376465, "grad_norm": 30.103649139404297, "learning_rate": 4.800745179625307e-07, "logits/chosen": 1.5828511714935303, "logits/rejected": 1.5261180400848389, "logps/chosen": -76.47662353515625, "logps/ref_chosen": -79.09429168701172, "logps/ref_rejected": -92.42912292480469, "logps/rejected": -91.34918212890625, "loss": 1.3799, "margin_dpo/margin_mean": 1.5377315282821655, "margin_dpo/margin_std": 6.86276912689209, "step": 144 }, { "epoch": 0.21919879062736206, "fcm_dpo/beta": 0.129638671875, "fcm_dpo/delta": 0.03883426636457443, "fcm_dpo/margin": 2.796900510787964, "fcm_dpo/q_t": 0.41991177201271057, "grad_norm": 27.09717559814453, "learning_rate": 4.795540267200686e-07, "logits/chosen": 1.4565389156341553, "logits/rejected": 1.4920694828033447, "logps/chosen": -94.13655090332031, "logps/ref_chosen": -97.7087173461914, "logps/ref_rejected": -97.63011169433594, "logps/rejected": -96.85485076904297, "loss": 1.2269, "margin_dpo/margin_mean": 2.7969002723693848, "margin_dpo/margin_std": 6.733441352844238, "step": 145 }, { "epoch": 0.22071050642479215, "fcm_dpo/beta": 0.13282790780067444, "fcm_dpo/delta": 0.12798386812210083, "fcm_dpo/margin": 2.072317123413086, "fcm_dpo/q_t": 0.44036737084388733, "grad_norm": 25.196712493896484, "learning_rate": 4.790271143580173e-07, "logits/chosen": 1.359689712524414, "logits/rejected": 1.3533302545547485, "logps/chosen": -72.72005462646484, "logps/ref_chosen": -76.56294250488281, "logps/ref_rejected": -83.78160095214844, "logps/rejected": -82.01103210449219, "loss": 1.261, "margin_dpo/margin_mean": 2.072317361831665, "margin_dpo/margin_std": 5.668929100036621, "step": 146 }, { "epoch": 0.2222222222222222, "fcm_dpo/beta": 0.1359969526529312, "fcm_dpo/delta": 0.132144033908844, "fcm_dpo/margin": 1.9967001676559448, "fcm_dpo/q_t": 0.44437921047210693, "grad_norm": 28.042757034301758, "learning_rate": 4.784937956152489e-07, "logits/chosen": 1.5116376876831055, "logits/rejected": 1.4199936389923096, "logps/chosen": -80.23736572265625, "logps/ref_chosen": -83.24113464355469, "logps/ref_rejected": -97.50960540771484, "logps/rejected": -96.50253295898438, "loss": 1.285, "margin_dpo/margin_mean": 1.9967000484466553, "margin_dpo/margin_std": 5.979328155517578, "step": 147 }, { "epoch": 0.2237339380196523, "fcm_dpo/beta": 0.13668569922447205, "fcm_dpo/delta": -0.06743814051151276, "fcm_dpo/margin": 3.392505407333374, "fcm_dpo/q_t": 0.4045356810092926, "grad_norm": 19.915895462036133, "learning_rate": 4.779540854098347e-07, "logits/chosen": 1.6008939743041992, "logits/rejected": 1.3894622325897217, "logps/chosen": -62.7821159362793, "logps/ref_chosen": -66.36277770996094, "logps/ref_rejected": -87.66487121582031, "logps/rejected": -87.47671508789062, "loss": 1.166, "margin_dpo/margin_mean": 3.392505407333374, "margin_dpo/margin_std": 6.748303413391113, "step": 148 }, { "epoch": 0.2252456538170824, "fcm_dpo/beta": 0.13758441805839539, "fcm_dpo/delta": 0.12122346460819244, "fcm_dpo/margin": 2.0515592098236084, "fcm_dpo/q_t": 0.43767067790031433, "grad_norm": 24.13621711730957, "learning_rate": 4.774079988386296e-07, "logits/chosen": 1.406876802444458, "logits/rejected": 1.2928733825683594, "logps/chosen": -70.20016479492188, "logps/ref_chosen": -72.0576171875, "logps/ref_rejected": -83.94097900390625, "logps/rejected": -84.13508605957031, "loss": 1.2556, "margin_dpo/margin_mean": 2.0515592098236084, "margin_dpo/margin_std": 5.487130641937256, "step": 149 }, { "epoch": 0.22675736961451248, "fcm_dpo/beta": 0.13595245778560638, "fcm_dpo/delta": -0.11543023586273193, "fcm_dpo/margin": 3.746461868286133, "fcm_dpo/q_t": 0.39566880464553833, "grad_norm": 24.9544734954834, "learning_rate": 4.768555511768486e-07, "logits/chosen": 1.7028197050094604, "logits/rejected": 1.6237821578979492, "logps/chosen": -83.20237731933594, "logps/ref_chosen": -85.52684783935547, "logps/ref_rejected": -108.37449645996094, "logps/rejected": -109.79650115966797, "loss": 1.1504, "margin_dpo/margin_mean": 3.746461868286133, "margin_dpo/margin_std": 7.140506267547607, "step": 150 }, { "epoch": 0.22826908541194255, "fcm_dpo/beta": 0.13261011242866516, "fcm_dpo/delta": -0.1349739134311676, "fcm_dpo/margin": 3.980025291442871, "fcm_dpo/q_t": 0.38910791277885437, "grad_norm": 20.386749267578125, "learning_rate": 4.762967578776406e-07, "logits/chosen": 1.5325286388397217, "logits/rejected": 1.386871337890625, "logps/chosen": -65.10581970214844, "logps/ref_chosen": -69.160888671875, "logps/ref_rejected": -91.42207336425781, "logps/rejected": -91.34703063964844, "loss": 1.0712, "margin_dpo/margin_mean": 3.980025291442871, "margin_dpo/margin_std": 6.253316879272461, "step": 151 }, { "epoch": 0.22978080120937264, "fcm_dpo/beta": 0.13304655253887177, "fcm_dpo/delta": 0.022187475115060806, "fcm_dpo/margin": 2.838752508163452, "fcm_dpo/q_t": 0.4203815460205078, "grad_norm": 24.348512649536133, "learning_rate": 4.757316345716553e-07, "logits/chosen": 1.8199957609176636, "logits/rejected": 1.6512593030929565, "logps/chosen": -69.91305541992188, "logps/ref_chosen": -72.48135375976562, "logps/ref_rejected": -94.44818878173828, "logps/rejected": -94.71864318847656, "loss": 1.215, "margin_dpo/margin_mean": 2.838752508163452, "margin_dpo/margin_std": 6.552783489227295, "step": 152 }, { "epoch": 0.23129251700680273, "fcm_dpo/beta": 0.13313767313957214, "fcm_dpo/delta": -0.022764816880226135, "fcm_dpo/margin": 3.1567814350128174, "fcm_dpo/q_t": 0.4093555212020874, "grad_norm": 22.365690231323242, "learning_rate": 4.751601970666064e-07, "logits/chosen": 1.4918171167373657, "logits/rejected": 1.4286854267120361, "logps/chosen": -87.7032470703125, "logps/ref_chosen": -89.6655044555664, "logps/ref_rejected": -90.67737579345703, "logps/rejected": -91.87190246582031, "loss": 1.1532, "margin_dpo/margin_mean": 3.1567811965942383, "margin_dpo/margin_std": 5.790462970733643, "step": 153 }, { "epoch": 0.2328042328042328, "fcm_dpo/beta": 0.13354626297950745, "fcm_dpo/delta": 0.10732235014438629, "fcm_dpo/margin": 2.216388702392578, "fcm_dpo/q_t": 0.4366697072982788, "grad_norm": 25.555784225463867, "learning_rate": 4.745824613468292e-07, "logits/chosen": 1.8585283756256104, "logits/rejected": 1.8039864301681519, "logps/chosen": -73.84849548339844, "logps/ref_chosen": -76.58096313476562, "logps/ref_rejected": -78.18669891357422, "logps/rejected": -77.67062377929688, "loss": 1.2893, "margin_dpo/margin_mean": 2.216388702392578, "margin_dpo/margin_std": 6.572654724121094, "step": 154 }, { "epoch": 0.23431594860166288, "fcm_dpo/beta": 0.1331787407398224, "fcm_dpo/delta": -0.039101965725421906, "fcm_dpo/margin": 3.2796497344970703, "fcm_dpo/q_t": 0.4062679409980774, "grad_norm": 27.18946075439453, "learning_rate": 4.7399844357283393e-07, "logits/chosen": 1.571397066116333, "logits/rejected": 1.5497605800628662, "logps/chosen": -80.94525146484375, "logps/ref_chosen": -82.65617370605469, "logps/ref_rejected": -95.52484130859375, "logps/rejected": -97.09357452392578, "loss": 1.1833, "margin_dpo/margin_mean": 3.2796502113342285, "margin_dpo/margin_std": 6.847566604614258, "step": 155 }, { "epoch": 0.23582766439909297, "fcm_dpo/beta": 0.1313672512769699, "fcm_dpo/delta": -0.10709728300571442, "fcm_dpo/margin": 3.818819999694824, "fcm_dpo/q_t": 0.39108410477638245, "grad_norm": 24.178529739379883, "learning_rate": 4.7340816008085305e-07, "logits/chosen": 1.4785492420196533, "logits/rejected": 1.3886404037475586, "logps/chosen": -85.74685668945312, "logps/ref_chosen": -87.66494750976562, "logps/ref_rejected": -108.2437744140625, "logps/rejected": -110.14449310302734, "loss": 1.0881, "margin_dpo/margin_mean": 3.818819522857666, "margin_dpo/margin_std": 6.0646257400512695, "step": 156 }, { "epoch": 0.23733938019652306, "fcm_dpo/beta": 0.12807613611221313, "fcm_dpo/delta": -0.10187535732984543, "fcm_dpo/margin": 2.6955342292785645, "fcm_dpo/q_t": 0.42544877529144287, "grad_norm": 20.409114837646484, "learning_rate": 4.728116273823847e-07, "logits/chosen": 1.3515766859054565, "logits/rejected": 1.3558907508850098, "logps/chosen": -67.93927001953125, "logps/ref_chosen": -70.77095794677734, "logps/ref_rejected": -78.78271484375, "logps/rejected": -78.64656066894531, "loss": 1.2191, "margin_dpo/margin_mean": 2.6955342292785645, "margin_dpo/margin_std": 6.043845176696777, "step": 157 }, { "epoch": 0.23885109599395313, "fcm_dpo/beta": 0.1278986632823944, "fcm_dpo/delta": -0.0014535868540406227, "fcm_dpo/margin": 3.138162136077881, "fcm_dpo/q_t": 0.4160592555999756, "grad_norm": 23.344018936157227, "learning_rate": 4.7220886216373085e-07, "logits/chosen": 1.5847368240356445, "logits/rejected": 1.5203303098678589, "logps/chosen": -78.30911254882812, "logps/ref_chosen": -81.21516418457031, "logps/ref_rejected": -97.8381118774414, "logps/rejected": -98.07023620605469, "loss": 1.1989, "margin_dpo/margin_mean": 3.138162136077881, "margin_dpo/margin_std": 6.922576904296875, "step": 158 }, { "epoch": 0.24036281179138322, "fcm_dpo/beta": 0.1267024725675583, "fcm_dpo/delta": -0.06470802426338196, "fcm_dpo/margin": 3.643596649169922, "fcm_dpo/q_t": 0.39887940883636475, "grad_norm": 20.22292709350586, "learning_rate": 4.715998812855304e-07, "logits/chosen": 1.5062403678894043, "logits/rejected": 1.4325311183929443, "logps/chosen": -69.44245910644531, "logps/ref_chosen": -72.33412170410156, "logps/ref_rejected": -89.49591064453125, "logps/rejected": -90.24784851074219, "loss": 1.123, "margin_dpo/margin_mean": 3.6435976028442383, "margin_dpo/margin_std": 6.381722450256348, "step": 159 }, { "epoch": 0.2418745275888133, "fcm_dpo/beta": 0.12704569101333618, "fcm_dpo/delta": 0.028272677212953568, "fcm_dpo/margin": 2.9338912963867188, "fcm_dpo/q_t": 0.4206444323062897, "grad_norm": 20.26133155822754, "learning_rate": 4.7098470178228755e-07, "logits/chosen": 1.3379184007644653, "logits/rejected": 1.1803221702575684, "logps/chosen": -60.8831901550293, "logps/ref_chosen": -63.26386260986328, "logps/ref_rejected": -82.27867126464844, "logps/rejected": -82.83187866210938, "loss": 1.1803, "margin_dpo/margin_mean": 2.933891773223877, "margin_dpo/margin_std": 6.030610084533691, "step": 160 }, { "epoch": 0.24338624338624337, "fcm_dpo/beta": 0.12744706869125366, "fcm_dpo/delta": -0.0032501202076673508, "fcm_dpo/margin": 3.161600112915039, "fcm_dpo/q_t": 0.4082825183868408, "grad_norm": 19.699172973632812, "learning_rate": 4.703633408618955e-07, "logits/chosen": 1.5183749198913574, "logits/rejected": 1.43423330783844, "logps/chosen": -67.91627502441406, "logps/ref_chosen": -70.69304656982422, "logps/ref_rejected": -82.73606872558594, "logps/rejected": -83.12089538574219, "loss": 1.1607, "margin_dpo/margin_mean": 3.161600112915039, "margin_dpo/margin_std": 6.068305015563965, "step": 161 }, { "epoch": 0.24489795918367346, "fcm_dpo/beta": 0.12381276488304138, "fcm_dpo/delta": -0.1807851493358612, "fcm_dpo/margin": 4.608911037445068, "fcm_dpo/q_t": 0.37070053815841675, "grad_norm": 21.9332275390625, "learning_rate": 4.697358159051549e-07, "logits/chosen": 1.7209584712982178, "logits/rejected": 1.6164560317993164, "logps/chosen": -86.98110961914062, "logps/ref_chosen": -89.3046646118164, "logps/ref_rejected": -114.05778503417969, "logps/rejected": -116.34314727783203, "loss": 1.0072, "margin_dpo/margin_mean": 4.608911514282227, "margin_dpo/margin_std": 5.768190383911133, "step": 162 }, { "epoch": 0.24640967498110355, "fcm_dpo/beta": 0.12266433238983154, "fcm_dpo/delta": -0.04616904258728027, "fcm_dpo/margin": 3.6132428646087646, "fcm_dpo/q_t": 0.40427160263061523, "grad_norm": 20.070598602294922, "learning_rate": 4.691021444652876e-07, "logits/chosen": 1.5364042520523071, "logits/rejected": 1.4361138343811035, "logps/chosen": -65.273193359375, "logps/ref_chosen": -68.61222076416016, "logps/ref_rejected": -89.03155517578125, "logps/rejected": -89.3057861328125, "loss": 1.1298, "margin_dpo/margin_mean": 3.6132428646087646, "margin_dpo/margin_std": 6.241079807281494, "step": 163 }, { "epoch": 0.24792139077853365, "fcm_dpo/beta": 0.1191563755273819, "fcm_dpo/delta": -0.14316755533218384, "fcm_dpo/margin": 4.495945930480957, "fcm_dpo/q_t": 0.3841249942779541, "grad_norm": 19.52849578857422, "learning_rate": 4.6846234426744624e-07, "logits/chosen": 1.4956767559051514, "logits/rejected": 1.3199671506881714, "logps/chosen": -70.1712646484375, "logps/ref_chosen": -73.55902862548828, "logps/ref_rejected": -94.16201782226562, "logps/rejected": -95.27018737792969, "loss": 1.0776, "margin_dpo/margin_mean": 4.495945930480957, "margin_dpo/margin_std": 6.860191345214844, "step": 164 }, { "epoch": 0.2494331065759637, "fcm_dpo/beta": 0.11708627641201019, "fcm_dpo/delta": -0.10044585913419724, "fcm_dpo/margin": 4.228605270385742, "fcm_dpo/q_t": 0.39115095138549805, "grad_norm": 18.38561248779297, "learning_rate": 4.678164332082175e-07, "logits/chosen": 1.5222417116165161, "logits/rejected": 1.3784247636795044, "logps/chosen": -65.81343078613281, "logps/ref_chosen": -68.67132568359375, "logps/ref_rejected": -85.95689392089844, "logps/rejected": -87.32759857177734, "loss": 1.0836, "margin_dpo/margin_mean": 4.2286057472229, "margin_dpo/margin_std": 6.33597469329834, "step": 165 }, { "epoch": 0.2509448223733938, "fcm_dpo/beta": 0.1159178614616394, "fcm_dpo/delta": 0.04237668961286545, "fcm_dpo/margin": 3.097219944000244, "fcm_dpo/q_t": 0.42350584268569946, "grad_norm": 21.20147705078125, "learning_rate": 4.6716442935512214e-07, "logits/chosen": 1.5135124921798706, "logits/rejected": 1.2978200912475586, "logps/chosen": -78.65182495117188, "logps/ref_chosen": -80.89755249023438, "logps/ref_rejected": -111.91075134277344, "logps/rejected": -112.76225280761719, "loss": 1.1855, "margin_dpo/margin_mean": 3.097219467163086, "margin_dpo/margin_std": 6.383951187133789, "step": 166 }, { "epoch": 0.25245653817082386, "fcm_dpo/beta": 0.11525711417198181, "fcm_dpo/delta": -0.03476811572909355, "fcm_dpo/margin": 3.750250816345215, "fcm_dpo/q_t": 0.4018850326538086, "grad_norm": 18.29037094116211, "learning_rate": 4.6650635094610966e-07, "logits/chosen": 1.3140567541122437, "logits/rejected": 1.2441596984863281, "logps/chosen": -73.19277954101562, "logps/ref_chosen": -76.73136138916016, "logps/ref_rejected": -92.57389068603516, "logps/rejected": -92.78555297851562, "loss": 1.1115, "margin_dpo/margin_mean": 3.750251293182373, "margin_dpo/margin_std": 5.980134963989258, "step": 167 }, { "epoch": 0.25396825396825395, "fcm_dpo/beta": 0.11641091853380203, "fcm_dpo/delta": 0.05485351383686066, "fcm_dpo/margin": 2.9811627864837646, "fcm_dpo/q_t": 0.42268872261047363, "grad_norm": 19.12883949279785, "learning_rate": 4.6584221638904767e-07, "logits/chosen": 1.411829948425293, "logits/rejected": 1.2877988815307617, "logps/chosen": -79.95015716552734, "logps/ref_chosen": -82.63671112060547, "logps/ref_rejected": -96.72691345214844, "logps/rejected": -97.02153015136719, "loss": 1.1565, "margin_dpo/margin_mean": 2.981163501739502, "margin_dpo/margin_std": 5.256748676300049, "step": 168 }, { "epoch": 0.25547996976568405, "fcm_dpo/beta": 0.11612222343683243, "fcm_dpo/delta": -0.07028884440660477, "fcm_dpo/margin": 4.022111892700195, "fcm_dpo/q_t": 0.39903855323791504, "grad_norm": 20.48174476623535, "learning_rate": 4.651720442612075e-07, "logits/chosen": 1.6285090446472168, "logits/rejected": 1.6098928451538086, "logps/chosen": -74.69467163085938, "logps/ref_chosen": -78.87673950195312, "logps/ref_rejected": -94.18919372558594, "logps/rejected": -94.02923583984375, "loss": 1.1083, "margin_dpo/margin_mean": 4.022111892700195, "margin_dpo/margin_std": 6.698416709899902, "step": 169 }, { "epoch": 0.25699168556311414, "fcm_dpo/beta": 0.1153804212808609, "fcm_dpo/delta": -0.004839559551328421, "fcm_dpo/margin": 2.7229254245758057, "fcm_dpo/q_t": 0.4343235194683075, "grad_norm": 20.735963821411133, "learning_rate": 4.6449585330874425e-07, "logits/chosen": 1.4833619594573975, "logits/rejected": 1.5050253868103027, "logps/chosen": -69.8090591430664, "logps/ref_chosen": -73.35820007324219, "logps/ref_rejected": -76.85077667236328, "logps/rejected": -76.02455139160156, "loss": 1.3037, "margin_dpo/margin_mean": 2.7229256629943848, "margin_dpo/margin_std": 8.234253883361816, "step": 170 }, { "epoch": 0.2585034013605442, "fcm_dpo/beta": 0.11291978508234024, "fcm_dpo/delta": -0.09370312094688416, "fcm_dpo/margin": 4.317270278930664, "fcm_dpo/q_t": 0.3929103910923004, "grad_norm": 18.698780059814453, "learning_rate": 4.6381366244617224e-07, "logits/chosen": 1.7400933504104614, "logits/rejected": 1.5900516510009766, "logps/chosen": -77.08047485351562, "logps/ref_chosen": -80.4322738647461, "logps/ref_rejected": -96.99999237060547, "logps/rejected": -97.96546936035156, "loss": 1.1195, "margin_dpo/margin_mean": 4.3172712326049805, "margin_dpo/margin_std": 7.4338483810424805, "step": 171 }, { "epoch": 0.2600151171579743, "fcm_dpo/beta": 0.11300242692232132, "fcm_dpo/delta": -0.033088989555835724, "fcm_dpo/margin": 3.8189542293548584, "fcm_dpo/q_t": 0.4061174988746643, "grad_norm": 18.41037940979004, "learning_rate": 4.631254907558365e-07, "logits/chosen": 1.6603069305419922, "logits/rejected": 1.5482029914855957, "logps/chosen": -67.79273223876953, "logps/ref_chosen": -70.45406341552734, "logps/ref_rejected": -99.85603332519531, "logps/rejected": -101.01365661621094, "loss": 1.1392, "margin_dpo/margin_mean": 3.8189544677734375, "margin_dpo/margin_std": 6.91829252243042, "step": 172 }, { "epoch": 0.2615268329554044, "fcm_dpo/beta": 0.10994724184274673, "fcm_dpo/delta": -0.11195747554302216, "fcm_dpo/margin": 3.7884159088134766, "fcm_dpo/q_t": 0.41372668743133545, "grad_norm": 21.74614715576172, "learning_rate": 4.624313574873786e-07, "logits/chosen": 1.496457815170288, "logits/rejected": 1.2725635766983032, "logps/chosen": -69.65293884277344, "logps/ref_chosen": -72.15026092529297, "logps/ref_rejected": -94.10212707519531, "logps/rejected": -95.39321899414062, "loss": 1.1928, "margin_dpo/margin_mean": 3.7884161472320557, "margin_dpo/margin_std": 7.9902544021606445, "step": 173 }, { "epoch": 0.26303854875283444, "fcm_dpo/beta": 0.10757105052471161, "fcm_dpo/delta": -0.1165534034371376, "fcm_dpo/margin": 4.738400459289551, "fcm_dpo/q_t": 0.3892877399921417, "grad_norm": 18.56683349609375, "learning_rate": 4.61731282057198e-07, "logits/chosen": 1.6890777349472046, "logits/rejected": 1.4859000444412231, "logps/chosen": -73.22291564941406, "logps/ref_chosen": -75.99629211425781, "logps/ref_rejected": -106.2359619140625, "logps/rejected": -108.20098876953125, "loss": 1.0893, "margin_dpo/margin_mean": 4.738400459289551, "margin_dpo/margin_std": 7.599691390991211, "step": 174 }, { "epoch": 0.26455026455026454, "fcm_dpo/beta": 0.10623383522033691, "fcm_dpo/delta": -0.05162263289093971, "fcm_dpo/margin": 4.22575569152832, "fcm_dpo/q_t": 0.40296652913093567, "grad_norm": 19.128814697265625, "learning_rate": 4.6102528404790965e-07, "logits/chosen": 1.7655320167541504, "logits/rejected": 1.6901226043701172, "logps/chosen": -81.8990249633789, "logps/ref_chosen": -84.51177978515625, "logps/ref_rejected": -104.46299743652344, "logps/rejected": -106.07599639892578, "loss": 1.1556, "margin_dpo/margin_mean": 4.225754737854004, "margin_dpo/margin_std": 8.150654792785645, "step": 175 }, { "epoch": 0.2660619803476946, "fcm_dpo/beta": 0.10651630163192749, "fcm_dpo/delta": 0.009663296863436699, "fcm_dpo/margin": 2.732745409011841, "fcm_dpo/q_t": 0.4402855634689331, "grad_norm": 21.021997451782227, "learning_rate": 4.603133832077953e-07, "logits/chosen": 1.352994441986084, "logits/rejected": 1.2789125442504883, "logps/chosen": -97.29902648925781, "logps/ref_chosen": -98.2034912109375, "logps/ref_rejected": -103.2023696899414, "logps/rejected": -105.03064727783203, "loss": 1.2824, "margin_dpo/margin_mean": 2.732745409011841, "margin_dpo/margin_std": 8.030749320983887, "step": 176 }, { "epoch": 0.2675736961451247, "fcm_dpo/beta": 0.10103031247854233, "fcm_dpo/delta": -0.35967448353767395, "fcm_dpo/margin": 7.258904457092285, "fcm_dpo/q_t": 0.3416978716850281, "grad_norm": 20.529727935791016, "learning_rate": 4.5959559945025183e-07, "logits/chosen": 2.0382192134857178, "logits/rejected": 1.7976688146591187, "logps/chosen": -73.77230834960938, "logps/ref_chosen": -78.029541015625, "logps/ref_rejected": -112.57099914550781, "logps/rejected": -115.57266235351562, "loss": 0.9238, "margin_dpo/margin_mean": 7.25890588760376, "margin_dpo/margin_std": 8.024457931518555, "step": 177 }, { "epoch": 0.2690854119425548, "fcm_dpo/beta": 0.09761218726634979, "fcm_dpo/delta": -0.05611484497785568, "fcm_dpo/margin": 4.6330156326293945, "fcm_dpo/q_t": 0.3984643220901489, "grad_norm": 16.481779098510742, "learning_rate": 4.588719528532341e-07, "logits/chosen": 1.4155144691467285, "logits/rejected": 1.2903285026550293, "logps/chosen": -76.18199157714844, "logps/ref_chosen": -79.48869323730469, "logps/ref_rejected": -96.62449645996094, "logps/rejected": -97.9508056640625, "loss": 1.0842, "margin_dpo/margin_mean": 4.633016109466553, "margin_dpo/margin_std": 6.5637617111206055, "step": 178 }, { "epoch": 0.2705971277399849, "fcm_dpo/beta": 0.09835878014564514, "fcm_dpo/delta": 0.011413969099521637, "fcm_dpo/margin": 3.954383134841919, "fcm_dpo/q_t": 0.4162752032279968, "grad_norm": 17.731225967407227, "learning_rate": 4.581424636586928e-07, "logits/chosen": 1.7793190479278564, "logits/rejected": 1.718389630317688, "logps/chosen": -80.90934753417969, "logps/ref_chosen": -84.5088119506836, "logps/ref_rejected": -93.07945251464844, "logps/rejected": -93.43437194824219, "loss": 1.1847, "margin_dpo/margin_mean": 3.954383134841919, "margin_dpo/margin_std": 8.3207426071167, "step": 179 }, { "epoch": 0.272108843537415, "fcm_dpo/beta": 0.09892720729112625, "fcm_dpo/delta": 0.015113111585378647, "fcm_dpo/margin": 3.8914146423339844, "fcm_dpo/q_t": 0.4185601472854614, "grad_norm": 16.331005096435547, "learning_rate": 4.5740715227200897e-07, "logits/chosen": 1.004435658454895, "logits/rejected": 0.9405149221420288, "logps/chosen": -70.37954711914062, "logps/ref_chosen": -74.5645523071289, "logps/ref_rejected": -81.02266693115234, "logps/rejected": -80.72906494140625, "loss": 1.1612, "margin_dpo/margin_mean": 3.8914148807525635, "margin_dpo/margin_std": 7.424806118011475, "step": 180 }, { "epoch": 0.273620559334845, "fcm_dpo/beta": 0.0971488356590271, "fcm_dpo/delta": -0.08372343331575394, "fcm_dpo/margin": 4.935826301574707, "fcm_dpo/q_t": 0.39608198404312134, "grad_norm": 19.247838973999023, "learning_rate": 4.566660392614228e-07, "logits/chosen": 1.4877427816390991, "logits/rejected": 1.3681514263153076, "logps/chosen": -73.86396789550781, "logps/ref_chosen": -78.77166748046875, "logps/ref_rejected": -98.29750061035156, "logps/rejected": -98.32562255859375, "loss": 1.0703, "margin_dpo/margin_mean": 4.935826301574707, "margin_dpo/margin_std": 7.191739082336426, "step": 181 }, { "epoch": 0.2751322751322751, "fcm_dpo/beta": 0.09410342574119568, "fcm_dpo/delta": -0.2044612020254135, "fcm_dpo/margin": 6.296449661254883, "fcm_dpo/q_t": 0.37253063917160034, "grad_norm": 17.99175453186035, "learning_rate": 4.5591914535745817e-07, "logits/chosen": 2.003591537475586, "logits/rejected": 1.784796953201294, "logps/chosen": -72.50923156738281, "logps/ref_chosen": -75.67765045166016, "logps/ref_rejected": -107.47894287109375, "logps/rejected": -110.60696411132812, "loss": 1.0184, "margin_dpo/margin_mean": 6.296449661254883, "margin_dpo/margin_std": 8.511575698852539, "step": 182 }, { "epoch": 0.2766439909297052, "fcm_dpo/beta": 0.09400541335344315, "fcm_dpo/delta": 0.05271903797984123, "fcm_dpo/margin": 2.27756667137146, "fcm_dpo/q_t": 0.4522141218185425, "grad_norm": 18.433788299560547, "learning_rate": 4.551664914523433e-07, "logits/chosen": 1.6212289333343506, "logits/rejected": 1.5513460636138916, "logps/chosen": -80.0240707397461, "logps/ref_chosen": -79.99969482421875, "logps/ref_rejected": -89.35220336914062, "logps/rejected": -91.65414428710938, "loss": 1.2929, "margin_dpo/margin_mean": 2.277566909790039, "margin_dpo/margin_std": 7.199431419372559, "step": 183 }, { "epoch": 0.2781557067271353, "fcm_dpo/beta": 0.09329868853092194, "fcm_dpo/delta": -0.021785538643598557, "fcm_dpo/margin": 4.506135940551758, "fcm_dpo/q_t": 0.4057055115699768, "grad_norm": 14.286376953125, "learning_rate": 4.544080985994258e-07, "logits/chosen": 1.6556332111358643, "logits/rejected": 1.5036529302597046, "logps/chosen": -59.074684143066406, "logps/ref_chosen": -62.133941650390625, "logps/ref_rejected": -84.44404602050781, "logps/rejected": -85.89093017578125, "loss": 1.0965, "margin_dpo/margin_mean": 4.506134986877441, "margin_dpo/margin_std": 6.4260478019714355, "step": 184 }, { "epoch": 0.2796674225245654, "fcm_dpo/beta": 0.09228262305259705, "fcm_dpo/delta": -0.06251490116119385, "fcm_dpo/margin": 4.972772598266602, "fcm_dpo/q_t": 0.4024953842163086, "grad_norm": 14.063694953918457, "learning_rate": 4.5364398801258394e-07, "logits/chosen": 1.575112223625183, "logits/rejected": 1.4495570659637451, "logps/chosen": -66.57908630371094, "logps/ref_chosen": -67.93174743652344, "logps/ref_rejected": -83.76744079589844, "logps/rejected": -87.38755798339844, "loss": 1.1587, "margin_dpo/margin_mean": 4.97277307510376, "margin_dpo/margin_std": 9.61676025390625, "step": 185 }, { "epoch": 0.2811791383219955, "fcm_dpo/beta": 0.09120054543018341, "fcm_dpo/delta": -0.07828307151794434, "fcm_dpo/margin": 5.2014384269714355, "fcm_dpo/q_t": 0.40300649404525757, "grad_norm": 16.713613510131836, "learning_rate": 4.5287418106563354e-07, "logits/chosen": 1.347496509552002, "logits/rejected": 1.2185564041137695, "logps/chosen": -84.60653686523438, "logps/ref_chosen": -86.22174072265625, "logps/ref_rejected": -100.42019653320312, "logps/rejected": -104.00642395019531, "loss": 1.1461, "margin_dpo/margin_mean": 5.201438903808594, "margin_dpo/margin_std": 9.859893798828125, "step": 186 }, { "epoch": 0.28269085411942557, "fcm_dpo/beta": 0.09041302651166916, "fcm_dpo/delta": -0.028823696076869965, "fcm_dpo/margin": 4.727651119232178, "fcm_dpo/q_t": 0.4054437577724457, "grad_norm": 17.96294593811035, "learning_rate": 4.520986992917297e-07, "logits/chosen": 1.6717660427093506, "logits/rejected": 1.5524613857269287, "logps/chosen": -93.56632995605469, "logps/ref_chosen": -92.81202697753906, "logps/ref_rejected": -117.28926086425781, "logps/rejected": -122.77122497558594, "loss": 1.1282, "margin_dpo/margin_mean": 4.7276506423950195, "margin_dpo/margin_std": 8.11026668548584, "step": 187 }, { "epoch": 0.2842025699168556, "fcm_dpo/beta": 0.09080306440591812, "fcm_dpo/delta": -0.020409882068634033, "fcm_dpo/margin": 4.612438201904297, "fcm_dpo/q_t": 0.4097273349761963, "grad_norm": 16.726558685302734, "learning_rate": 4.5131756438276466e-07, "logits/chosen": 1.7009717226028442, "logits/rejected": 1.5912370681762695, "logps/chosen": -88.00601196289062, "logps/ref_chosen": -87.85247802734375, "logps/ref_rejected": -94.58252716064453, "logps/rejected": -99.34849548339844, "loss": 1.1286, "margin_dpo/margin_mean": 4.612437725067139, "margin_dpo/margin_std": 7.809675216674805, "step": 188 }, { "epoch": 0.2857142857142857, "fcm_dpo/beta": 0.08787811547517776, "fcm_dpo/delta": -0.1251910775899887, "fcm_dpo/margin": 3.8941988945007324, "fcm_dpo/q_t": 0.42143514752388, "grad_norm": 19.706933975219727, "learning_rate": 4.5053079818876096e-07, "logits/chosen": 1.5907940864562988, "logits/rejected": 1.5847728252410889, "logps/chosen": -94.68544006347656, "logps/ref_chosen": -95.00414276123047, "logps/ref_rejected": -90.50090789794922, "logps/rejected": -94.07640075683594, "loss": 1.2243, "margin_dpo/margin_mean": 3.8941988945007324, "margin_dpo/margin_std": 8.557470321655273, "step": 189 }, { "epoch": 0.2872260015117158, "fcm_dpo/beta": 0.08545570075511932, "fcm_dpo/delta": -0.16769738495349884, "fcm_dpo/margin": 6.525961399078369, "fcm_dpo/q_t": 0.37398993968963623, "grad_norm": 19.378028869628906, "learning_rate": 4.4973842271726024e-07, "logits/chosen": 1.5073564052581787, "logits/rejected": 1.1648344993591309, "logps/chosen": -69.2900390625, "logps/ref_chosen": -70.79264831542969, "logps/ref_rejected": -122.56155395507812, "logps/rejected": -127.58491516113281, "loss": 1.0211, "margin_dpo/margin_mean": 6.525960922241211, "margin_dpo/margin_std": 8.354592323303223, "step": 190 }, { "epoch": 0.2887377173091459, "fcm_dpo/beta": 0.08557368814945221, "fcm_dpo/delta": 0.06597628444433212, "fcm_dpo/margin": 3.9295034408569336, "fcm_dpo/q_t": 0.4254741966724396, "grad_norm": 17.88901138305664, "learning_rate": 4.48940460132708e-07, "logits/chosen": 1.7644071578979492, "logits/rejected": 1.663835883140564, "logps/chosen": -94.00773620605469, "logps/ref_chosen": -92.15048217773438, "logps/ref_rejected": -106.4153060913086, "logps/rejected": -112.20206451416016, "loss": 1.1986, "margin_dpo/margin_mean": 3.9295034408569336, "margin_dpo/margin_std": 8.408856391906738, "step": 191 }, { "epoch": 0.29024943310657597, "fcm_dpo/beta": 0.08797520399093628, "fcm_dpo/delta": 0.1374768316745758, "fcm_dpo/margin": 3.0240793228149414, "fcm_dpo/q_t": 0.43987464904785156, "grad_norm": 13.99423885345459, "learning_rate": 4.481369327558329e-07, "logits/chosen": 1.7628107070922852, "logits/rejected": 1.6950645446777344, "logps/chosen": -70.67575073242188, "logps/ref_chosen": -69.51527404785156, "logps/ref_rejected": -80.15898132324219, "logps/rejected": -84.34353637695312, "loss": 1.2409, "margin_dpo/margin_mean": 3.0240793228149414, "margin_dpo/margin_std": 7.654026508331299, "step": 192 }, { "epoch": 0.29176114890400606, "fcm_dpo/beta": 0.0865698754787445, "fcm_dpo/delta": -0.14727595448493958, "fcm_dpo/margin": 6.229554176330566, "fcm_dpo/q_t": 0.38533905148506165, "grad_norm": 14.288257598876953, "learning_rate": 4.47327863063023e-07, "logits/chosen": 1.495797038078308, "logits/rejected": 1.507264256477356, "logps/chosen": -72.861083984375, "logps/ref_chosen": -73.43276977539062, "logps/ref_rejected": -77.81238555908203, "logps/rejected": -83.47024536132812, "loss": 1.0529, "margin_dpo/margin_mean": 6.229554176330566, "margin_dpo/margin_std": 9.042031288146973, "step": 193 }, { "epoch": 0.29327286470143615, "fcm_dpo/beta": 0.08675388991832733, "fcm_dpo/delta": 0.04714217036962509, "fcm_dpo/margin": 2.9702444076538086, "fcm_dpo/q_t": 0.4436585009098053, "grad_norm": 17.544597625732422, "learning_rate": 4.4651327368569684e-07, "logits/chosen": 1.4156489372253418, "logits/rejected": 1.3410913944244385, "logps/chosen": -78.87250518798828, "logps/ref_chosen": -76.63236999511719, "logps/ref_rejected": -85.67449188232422, "logps/rejected": -90.88487243652344, "loss": 1.3064, "margin_dpo/margin_mean": 2.9702446460723877, "margin_dpo/margin_std": 9.60417366027832, "step": 194 }, { "epoch": 0.2947845804988662, "fcm_dpo/beta": 0.08641483634710312, "fcm_dpo/delta": -0.002406056970357895, "fcm_dpo/margin": 4.652583599090576, "fcm_dpo/q_t": 0.4115176796913147, "grad_norm": 16.650192260742188, "learning_rate": 4.4569318740967043e-07, "logits/chosen": 1.422691822052002, "logits/rejected": 1.44659423828125, "logps/chosen": -92.10873413085938, "logps/ref_chosen": -89.43354797363281, "logps/ref_rejected": -91.25908660888672, "logps/rejected": -98.58685302734375, "loss": 1.1326, "margin_dpo/margin_mean": 4.652583122253418, "margin_dpo/margin_std": 7.990112781524658, "step": 195 }, { "epoch": 0.2962962962962963, "fcm_dpo/beta": 0.08771341294050217, "fcm_dpo/delta": 0.04121620953083038, "fcm_dpo/margin": 4.0985236167907715, "fcm_dpo/q_t": 0.4218651056289673, "grad_norm": 15.421774864196777, "learning_rate": 4.448676271745197e-07, "logits/chosen": 1.4758151769638062, "logits/rejected": 1.361769199371338, "logps/chosen": -76.75479888916016, "logps/ref_chosen": -75.47528839111328, "logps/ref_rejected": -99.37582397460938, "logps/rejected": -104.75386047363281, "loss": 1.1852, "margin_dpo/margin_mean": 4.098524570465088, "margin_dpo/margin_std": 8.5865478515625, "step": 196 }, { "epoch": 0.29780801209372637, "fcm_dpo/beta": 0.0871208906173706, "fcm_dpo/delta": -0.0760483369231224, "fcm_dpo/margin": 5.416136264801025, "fcm_dpo/q_t": 0.39775335788726807, "grad_norm": 18.421781539916992, "learning_rate": 4.440366160729392e-07, "logits/chosen": 1.9163366556167603, "logits/rejected": 1.760741114616394, "logps/chosen": -67.76097106933594, "logps/ref_chosen": -67.57392883300781, "logps/ref_rejected": -89.97993469238281, "logps/rejected": -95.58311462402344, "loss": 1.1732, "margin_dpo/margin_mean": 5.416136264801025, "margin_dpo/margin_std": 10.692497253417969, "step": 197 }, { "epoch": 0.29931972789115646, "fcm_dpo/beta": 0.08542082458734512, "fcm_dpo/delta": -0.08968013525009155, "fcm_dpo/margin": 5.680561542510986, "fcm_dpo/q_t": 0.3916228413581848, "grad_norm": 15.027694702148438, "learning_rate": 4.432001773500957e-07, "logits/chosen": 1.7204902172088623, "logits/rejected": 1.608296513557434, "logps/chosen": -78.00040435791016, "logps/ref_chosen": -77.36013793945312, "logps/ref_rejected": -90.55670166015625, "logps/rejected": -96.87753295898438, "loss": 1.0841, "margin_dpo/margin_mean": 5.6805620193481445, "margin_dpo/margin_std": 8.565601348876953, "step": 198 }, { "epoch": 0.30083144368858655, "fcm_dpo/beta": 0.08444124460220337, "fcm_dpo/delta": -0.018757252022624016, "fcm_dpo/margin": 4.949807167053223, "fcm_dpo/q_t": 0.4094560742378235, "grad_norm": 16.759029388427734, "learning_rate": 4.4235833440297856e-07, "logits/chosen": 1.376772165298462, "logits/rejected": 1.1592652797698975, "logps/chosen": -74.80636596679688, "logps/ref_chosen": -73.05004119873047, "logps/ref_rejected": -95.21923065185547, "logps/rejected": -101.92535400390625, "loss": 1.194, "margin_dpo/margin_mean": 4.949808120727539, "margin_dpo/margin_std": 10.570771217346191, "step": 199 }, { "epoch": 0.30234315948601664, "fcm_dpo/beta": 0.08244869858026505, "fcm_dpo/delta": -0.09225839376449585, "fcm_dpo/margin": 5.892239570617676, "fcm_dpo/q_t": 0.3961694836616516, "grad_norm": 15.525456428527832, "learning_rate": 4.415111107797445e-07, "logits/chosen": 1.6795740127563477, "logits/rejected": 1.4383859634399414, "logps/chosen": -73.5711441040039, "logps/ref_chosen": -73.75833129882812, "logps/ref_rejected": -105.00157165527344, "logps/rejected": -110.70662689208984, "loss": 1.1508, "margin_dpo/margin_mean": 5.892240524291992, "margin_dpo/margin_std": 11.194169998168945, "step": 200 }, { "epoch": 0.30234315948601664, "eval_fcm_dpo/beta": 0.0827222391963005, "eval_logits/chosen": 1.640813946723938, "eval_logits/rejected": 1.512371301651001, "eval_logps/chosen": -86.83106994628906, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -102.12706756591797, "eval_loss": 0.5663687586784363, "eval_margin_dpo/margin_mean": 5.50140905380249, "eval_margin_dpo/margin_std": 9.768028259277344, "eval_runtime": 42.2671, "eval_samples_per_second": 54.487, "eval_steps_per_second": 1.703, "step": 200 }, { "epoch": 0.30385487528344673, "fcm_dpo/beta": 0.08134329319000244, "fcm_dpo/delta": -0.10130226612091064, "fcm_dpo/margin": 6.098217964172363, "fcm_dpo/q_t": 0.39153581857681274, "grad_norm": 16.230323791503906, "learning_rate": 4.4065853017905953e-07, "logits/chosen": 1.7866268157958984, "logits/rejected": 1.635867714881897, "logps/chosen": -79.84793853759766, "logps/ref_chosen": -79.4841079711914, "logps/ref_rejected": -100.94435119628906, "logps/rejected": -107.40640258789062, "loss": 1.0776, "margin_dpo/margin_mean": 6.098217964172363, "margin_dpo/margin_std": 9.222325325012207, "step": 201 }, { "epoch": 0.30536659108087677, "fcm_dpo/beta": 0.08060562610626221, "fcm_dpo/delta": -0.08434365689754486, "fcm_dpo/margin": 5.954107284545898, "fcm_dpo/q_t": 0.395026832818985, "grad_norm": 18.101303100585938, "learning_rate": 4.3980061644943575e-07, "logits/chosen": 1.5323119163513184, "logits/rejected": 1.3420054912567139, "logps/chosen": -65.55935668945312, "logps/ref_chosen": -66.83952331542969, "logps/ref_rejected": -93.05116271972656, "logps/rejected": -97.72509765625, "loss": 1.0971, "margin_dpo/margin_mean": 5.954107284545898, "margin_dpo/margin_std": 9.408431053161621, "step": 202 }, { "epoch": 0.30687830687830686, "fcm_dpo/beta": 0.07931406050920486, "fcm_dpo/delta": -0.04864976555109024, "fcm_dpo/margin": 5.629288196563721, "fcm_dpo/q_t": 0.4027126431465149, "grad_norm": 15.56915283203125, "learning_rate": 4.3893739358856455e-07, "logits/chosen": 1.7592835426330566, "logits/rejected": 1.4586081504821777, "logps/chosen": -80.4451904296875, "logps/ref_chosen": -80.32998657226562, "logps/ref_rejected": -113.52803039550781, "logps/rejected": -119.27251434326172, "loss": 1.1022, "margin_dpo/margin_mean": 5.6292877197265625, "margin_dpo/margin_std": 8.988699913024902, "step": 203 }, { "epoch": 0.30839002267573695, "fcm_dpo/beta": 0.07697254419326782, "fcm_dpo/delta": -0.09150978177785873, "fcm_dpo/margin": 6.280875205993652, "fcm_dpo/q_t": 0.39389508962631226, "grad_norm": 15.502331733703613, "learning_rate": 4.380688857426449e-07, "logits/chosen": 1.646850347518921, "logits/rejected": 1.469982624053955, "logps/chosen": -65.03179168701172, "logps/ref_chosen": -66.68875885009766, "logps/ref_rejected": -85.07585906982422, "logps/rejected": -89.69976806640625, "loss": 1.0762, "margin_dpo/margin_mean": 6.280874252319336, "margin_dpo/margin_std": 8.902972221374512, "step": 204 }, { "epoch": 0.30990173847316704, "fcm_dpo/beta": 0.07790710031986237, "fcm_dpo/delta": 0.00653502345085144, "fcm_dpo/margin": 5.048940658569336, "fcm_dpo/q_t": 0.41906195878982544, "grad_norm": 16.413204193115234, "learning_rate": 4.3719511720570814e-07, "logits/chosen": 1.6623713970184326, "logits/rejected": 1.5246727466583252, "logps/chosen": -86.12635040283203, "logps/ref_chosen": -86.51950073242188, "logps/ref_rejected": -112.55376434326172, "logps/rejected": -117.20954895019531, "loss": 1.1914, "margin_dpo/margin_mean": 5.048940658569336, "margin_dpo/margin_std": 10.76982593536377, "step": 205 }, { "epoch": 0.31141345427059713, "fcm_dpo/beta": 0.07731396704912186, "fcm_dpo/delta": -0.03766930103302002, "fcm_dpo/margin": 3.6308658123016357, "fcm_dpo/q_t": 0.4388701319694519, "grad_norm": 15.725306510925293, "learning_rate": 4.363161124189387e-07, "logits/chosen": 1.8234915733337402, "logits/rejected": 1.7651959657669067, "logps/chosen": -87.84285736083984, "logps/ref_chosen": -88.68557739257812, "logps/ref_rejected": -97.75945281982422, "logps/rejected": -100.54759979248047, "loss": 1.2596, "margin_dpo/margin_mean": 3.6308655738830566, "margin_dpo/margin_std": 9.579620361328125, "step": 206 }, { "epoch": 0.3129251700680272, "fcm_dpo/beta": 0.07598337531089783, "fcm_dpo/delta": -0.0768546462059021, "fcm_dpo/margin": 6.224672317504883, "fcm_dpo/q_t": 0.4005652666091919, "grad_norm": 15.966635704040527, "learning_rate": 4.3543189596998986e-07, "logits/chosen": 1.6207809448242188, "logits/rejected": 1.3825271129608154, "logps/chosen": -85.65934753417969, "logps/ref_chosen": -85.12134552001953, "logps/ref_rejected": -103.34955596923828, "logps/rejected": -110.11223602294922, "loss": 1.0937, "margin_dpo/margin_mean": 6.224671840667725, "margin_dpo/margin_std": 10.07467269897461, "step": 207 }, { "epoch": 0.3144368858654573, "fcm_dpo/beta": 0.07625681906938553, "fcm_dpo/delta": 0.05362796038389206, "fcm_dpo/margin": 3.379854917526245, "fcm_dpo/q_t": 0.4436368942260742, "grad_norm": 15.4779691696167, "learning_rate": 4.3454249259229664e-07, "logits/chosen": 1.4077962636947632, "logits/rejected": 1.3891024589538574, "logps/chosen": -76.3375473022461, "logps/ref_chosen": -78.84121704101562, "logps/ref_rejected": -89.82504272460938, "logps/rejected": -90.70123291015625, "loss": 1.2865, "margin_dpo/margin_mean": 3.379855155944824, "margin_dpo/margin_std": 10.172846794128418, "step": 208 }, { "epoch": 0.31594860166288735, "fcm_dpo/beta": 0.074808269739151, "fcm_dpo/delta": -0.16914093494415283, "fcm_dpo/margin": 7.482547760009766, "fcm_dpo/q_t": 0.3817582130432129, "grad_norm": 14.531926155090332, "learning_rate": 4.336479271643833e-07, "logits/chosen": 1.150145411491394, "logits/rejected": 1.0794119834899902, "logps/chosen": -83.18844604492188, "logps/ref_chosen": -85.98588562011719, "logps/ref_rejected": -107.1638412475586, "logps/rejected": -111.84893798828125, "loss": 1.0456, "margin_dpo/margin_mean": 7.482547760009766, "margin_dpo/margin_std": 10.86307430267334, "step": 209 }, { "epoch": 0.31746031746031744, "fcm_dpo/beta": 0.07286547869443893, "fcm_dpo/delta": -0.15477873384952545, "fcm_dpo/margin": 7.500222206115723, "fcm_dpo/q_t": 0.37924668192863464, "grad_norm": 15.154267311096191, "learning_rate": 4.327482247091679e-07, "logits/chosen": 1.9730315208435059, "logits/rejected": 1.707854986190796, "logps/chosen": -69.033203125, "logps/ref_chosen": -71.75653076171875, "logps/ref_rejected": -102.47966003417969, "logps/rejected": -107.25654602050781, "loss": 1.0331, "margin_dpo/margin_mean": 7.5002217292785645, "margin_dpo/margin_std": 9.832921981811523, "step": 210 }, { "epoch": 0.31897203325774753, "fcm_dpo/beta": 0.07181885093450546, "fcm_dpo/delta": -0.03351927548646927, "fcm_dpo/margin": 6.012731075286865, "fcm_dpo/q_t": 0.40568870306015015, "grad_norm": 13.956618309020996, "learning_rate": 4.3184341039326217e-07, "logits/chosen": 1.502807378768921, "logits/rejected": 1.2806059122085571, "logps/chosen": -67.71188354492188, "logps/ref_chosen": -70.95170593261719, "logps/ref_rejected": -108.51902770996094, "logps/rejected": -111.29193115234375, "loss": 1.1199, "margin_dpo/margin_mean": 6.012731552124023, "margin_dpo/margin_std": 10.054574966430664, "step": 211 }, { "epoch": 0.3204837490551776, "fcm_dpo/beta": 0.06993983685970306, "fcm_dpo/delta": -0.14037087559700012, "fcm_dpo/margin": 7.620143890380859, "fcm_dpo/q_t": 0.3832334280014038, "grad_norm": 15.770182609558105, "learning_rate": 4.309335095262675e-07, "logits/chosen": 1.540845513343811, "logits/rejected": 1.3966429233551025, "logps/chosen": -70.31551361083984, "logps/ref_chosen": -74.34010314941406, "logps/ref_rejected": -97.58259582519531, "logps/rejected": -101.17815399169922, "loss": 1.0673, "margin_dpo/margin_mean": 7.620143890380859, "margin_dpo/margin_std": 11.508644104003906, "step": 212 }, { "epoch": 0.3219954648526077, "fcm_dpo/beta": 0.06910738348960876, "fcm_dpo/delta": -0.01402133610099554, "fcm_dpo/margin": 5.980879306793213, "fcm_dpo/q_t": 0.41256964206695557, "grad_norm": 13.321818351745605, "learning_rate": 4.3001854756006724e-07, "logits/chosen": 1.6784493923187256, "logits/rejected": 1.6832369565963745, "logps/chosen": -75.61842346191406, "logps/ref_chosen": -80.2526626586914, "logps/ref_rejected": -94.76947021484375, "logps/rejected": -96.11610412597656, "loss": 1.1607, "margin_dpo/margin_mean": 5.980880260467529, "margin_dpo/margin_std": 11.716890335083008, "step": 213 }, { "epoch": 0.3235071806500378, "fcm_dpo/beta": 0.06942355632781982, "fcm_dpo/delta": 0.0006352830678224564, "fcm_dpo/margin": 5.750314712524414, "fcm_dpo/q_t": 0.4148133099079132, "grad_norm": 22.03853416442871, "learning_rate": 4.290985500881143e-07, "logits/chosen": 1.4918372631072998, "logits/rejected": 1.4288208484649658, "logps/chosen": -73.66145324707031, "logps/ref_chosen": -77.9675064086914, "logps/ref_rejected": -84.0354232788086, "logps/rejected": -85.47969055175781, "loss": 1.1484, "margin_dpo/margin_mean": 5.750315189361572, "margin_dpo/margin_std": 10.650880813598633, "step": 214 }, { "epoch": 0.3250188964474679, "fcm_dpo/beta": 0.06811805069446564, "fcm_dpo/delta": -0.17968696355819702, "fcm_dpo/margin": 8.354536056518555, "fcm_dpo/q_t": 0.37755051255226135, "grad_norm": 12.484495162963867, "learning_rate": 4.281735428447157e-07, "logits/chosen": 1.399054765701294, "logits/rejected": 1.164167046546936, "logps/chosen": -78.92556762695312, "logps/ref_chosen": -81.2047348022461, "logps/ref_rejected": -116.18414306640625, "logps/rejected": -122.259521484375, "loss": 1.0274, "margin_dpo/margin_mean": 8.354536056518555, "margin_dpo/margin_std": 10.667181015014648, "step": 215 }, { "epoch": 0.32653061224489793, "fcm_dpo/beta": 0.06630684435367584, "fcm_dpo/delta": -0.036182302981615067, "fcm_dpo/margin": 6.55294132232666, "fcm_dpo/q_t": 0.403317928314209, "grad_norm": 14.035476684570312, "learning_rate": 4.2724355170431247e-07, "logits/chosen": 1.8719550371170044, "logits/rejected": 1.6385178565979004, "logps/chosen": -81.72055053710938, "logps/ref_chosen": -83.57113647460938, "logps/ref_rejected": -112.51902770996094, "logps/rejected": -117.22138977050781, "loss": 1.1132, "margin_dpo/margin_mean": 6.552940368652344, "margin_dpo/margin_std": 10.67953109741211, "step": 216 }, { "epoch": 0.328042328042328, "fcm_dpo/beta": 0.0655495673418045, "fcm_dpo/delta": -0.072674959897995, "fcm_dpo/margin": 7.158425331115723, "fcm_dpo/q_t": 0.40195178985595703, "grad_norm": 12.58739185333252, "learning_rate": 4.26308602680756e-07, "logits/chosen": 1.6943777799606323, "logits/rejected": 1.4496433734893799, "logps/chosen": -76.8424301147461, "logps/ref_chosen": -77.01390075683594, "logps/ref_rejected": -105.28099822998047, "logps/rejected": -112.26795196533203, "loss": 1.1127, "margin_dpo/margin_mean": 7.158425331115723, "margin_dpo/margin_std": 12.355752944946289, "step": 217 }, { "epoch": 0.3295540438397581, "fcm_dpo/beta": 0.06492602825164795, "fcm_dpo/delta": -0.06278467178344727, "fcm_dpo/margin": 4.715144634246826, "fcm_dpo/q_t": 0.4337427616119385, "grad_norm": 13.146940231323242, "learning_rate": 4.253687219265803e-07, "logits/chosen": 1.2216265201568604, "logits/rejected": 1.214383602142334, "logps/chosen": -92.66165924072266, "logps/ref_chosen": -92.47299194335938, "logps/ref_rejected": -92.80751037597656, "logps/rejected": -97.71131896972656, "loss": 1.2578, "margin_dpo/margin_mean": 4.715145111083984, "margin_dpo/margin_std": 12.045462608337402, "step": 218 }, { "epoch": 0.3310657596371882, "fcm_dpo/beta": 0.06420271098613739, "fcm_dpo/delta": -0.02842680737376213, "fcm_dpo/margin": 6.653696060180664, "fcm_dpo/q_t": 0.40601614117622375, "grad_norm": 11.388204574584961, "learning_rate": 4.2442393573227043e-07, "logits/chosen": 1.2830934524536133, "logits/rejected": 1.1858066320419312, "logps/chosen": -76.35826110839844, "logps/ref_chosen": -77.10382080078125, "logps/ref_rejected": -92.3438949584961, "logps/rejected": -98.25202941894531, "loss": 1.1086, "margin_dpo/margin_mean": 6.653696060180664, "margin_dpo/margin_std": 10.547959327697754, "step": 219 }, { "epoch": 0.3325774754346183, "fcm_dpo/beta": 0.06442397832870483, "fcm_dpo/delta": 0.024653874337673187, "fcm_dpo/margin": 5.8402934074401855, "fcm_dpo/q_t": 0.41860347986221313, "grad_norm": 12.554850578308105, "learning_rate": 4.234742705255272e-07, "logits/chosen": 1.8427441120147705, "logits/rejected": 1.6540197134017944, "logps/chosen": -60.972564697265625, "logps/ref_chosen": -62.48021697998047, "logps/ref_rejected": -86.93276977539062, "logps/rejected": -91.26541137695312, "loss": 1.1688, "margin_dpo/margin_mean": 5.840293884277344, "margin_dpo/margin_std": 11.364240646362305, "step": 220 }, { "epoch": 0.3340891912320484, "fcm_dpo/beta": 0.06460246443748474, "fcm_dpo/delta": -0.025578338652849197, "fcm_dpo/margin": 6.562741279602051, "fcm_dpo/q_t": 0.41032537817955017, "grad_norm": 14.2130126953125, "learning_rate": 4.22519752870528e-07, "logits/chosen": 1.5389156341552734, "logits/rejected": 1.3322972059249878, "logps/chosen": -77.62300109863281, "logps/ref_chosen": -78.35491943359375, "logps/ref_rejected": -108.17631530761719, "logps/rejected": -114.00713348388672, "loss": 1.1522, "margin_dpo/margin_mean": 6.562740325927734, "margin_dpo/margin_std": 12.393266677856445, "step": 221 }, { "epoch": 0.3356009070294785, "fcm_dpo/beta": 0.06298163533210754, "fcm_dpo/delta": -0.14094004034996033, "fcm_dpo/margin": 8.472888946533203, "fcm_dpo/q_t": 0.3822260797023773, "grad_norm": 13.83692741394043, "learning_rate": 4.2156040946718343e-07, "logits/chosen": 1.9284300804138184, "logits/rejected": 1.653240442276001, "logps/chosen": -77.398193359375, "logps/ref_chosen": -77.2734375, "logps/ref_rejected": -126.41007995605469, "logps/rejected": -135.00772094726562, "loss": 1.0311, "margin_dpo/margin_mean": 8.47288990020752, "margin_dpo/margin_std": 10.996661186218262, "step": 222 }, { "epoch": 0.3371126228269085, "fcm_dpo/beta": 0.06107974052429199, "fcm_dpo/delta": -0.14540138840675354, "fcm_dpo/margin": 8.803044319152832, "fcm_dpo/q_t": 0.38139283657073975, "grad_norm": 11.530991554260254, "learning_rate": 4.2059626715039065e-07, "logits/chosen": 1.380382776260376, "logits/rejected": 1.242692232131958, "logps/chosen": -79.19699096679688, "logps/ref_chosen": -78.4210205078125, "logps/ref_rejected": -101.38420867919922, "logps/rejected": -110.96322631835938, "loss": 1.036, "margin_dpo/margin_mean": 8.803045272827148, "margin_dpo/margin_std": 11.874062538146973, "step": 223 }, { "epoch": 0.3386243386243386, "fcm_dpo/beta": 0.06099741533398628, "fcm_dpo/delta": 0.058463674038648605, "fcm_dpo/margin": 5.63276481628418, "fcm_dpo/q_t": 0.4232059121131897, "grad_norm": 15.79139232635498, "learning_rate": 4.1962735288928304e-07, "logits/chosen": 1.628061056137085, "logits/rejected": 1.5595409870147705, "logps/chosen": -81.38064575195312, "logps/ref_chosen": -79.36337280273438, "logps/ref_rejected": -89.99789428710938, "logps/rejected": -97.64793395996094, "loss": 1.1645, "margin_dpo/margin_mean": 5.63276481628418, "margin_dpo/margin_std": 10.35816764831543, "step": 224 }, { "epoch": 0.3401360544217687, "fcm_dpo/beta": 0.05994480103254318, "fcm_dpo/delta": -0.10221153497695923, "fcm_dpo/margin": 8.266857147216797, "fcm_dpo/q_t": 0.3963213264942169, "grad_norm": 13.923603057861328, "learning_rate": 4.186536937864752e-07, "logits/chosen": 1.6372480392456055, "logits/rejected": 1.35588800907135, "logps/chosen": -91.33637237548828, "logps/ref_chosen": -88.99606323242188, "logps/ref_rejected": -127.55032348632812, "logps/rejected": -138.15748596191406, "loss": 1.121, "margin_dpo/margin_mean": 8.266858100891113, "margin_dpo/margin_std": 14.600061416625977, "step": 225 }, { "epoch": 0.3416477702191988, "fcm_dpo/beta": 0.059474602341651917, "fcm_dpo/delta": -0.0711555927991867, "fcm_dpo/margin": 7.866827964782715, "fcm_dpo/q_t": 0.39824751019477844, "grad_norm": 10.33656120300293, "learning_rate": 4.176753170773052e-07, "logits/chosen": 1.4881277084350586, "logits/rejected": 1.3798588514328003, "logps/chosen": -69.00434112548828, "logps/ref_chosen": -68.68444061279297, "logps/ref_rejected": -85.81898498535156, "logps/rejected": -94.0057144165039, "loss": 1.1187, "margin_dpo/margin_mean": 7.866827011108398, "margin_dpo/margin_std": 13.508081436157227, "step": 226 }, { "epoch": 0.3431594860166289, "fcm_dpo/beta": 0.05821537971496582, "fcm_dpo/delta": -0.07453415542840958, "fcm_dpo/margin": 8.073860168457031, "fcm_dpo/q_t": 0.4008241593837738, "grad_norm": 11.849678993225098, "learning_rate": 4.166922501290729e-07, "logits/chosen": 1.849320411682129, "logits/rejected": 1.7355461120605469, "logps/chosen": -74.02799224853516, "logps/ref_chosen": -72.52029418945312, "logps/ref_rejected": -90.7720718383789, "logps/rejected": -100.35362243652344, "loss": 1.1423, "margin_dpo/margin_mean": 8.073861122131348, "margin_dpo/margin_std": 14.962955474853516, "step": 227 }, { "epoch": 0.34467120181405897, "fcm_dpo/beta": 0.05832822620868683, "fcm_dpo/delta": -0.017807405441999435, "fcm_dpo/margin": 7.148486137390137, "fcm_dpo/q_t": 0.410169780254364, "grad_norm": 11.562895774841309, "learning_rate": 4.1570452044027405e-07, "logits/chosen": 1.5992982387542725, "logits/rejected": 1.4247077703475952, "logps/chosen": -73.79998779296875, "logps/ref_chosen": -72.23167419433594, "logps/ref_rejected": -95.45873260498047, "logps/rejected": -104.175537109375, "loss": 1.1584, "margin_dpo/margin_mean": 7.148487567901611, "margin_dpo/margin_std": 13.711039543151855, "step": 228 }, { "epoch": 0.34618291761148906, "fcm_dpo/beta": 0.05742615461349487, "fcm_dpo/delta": -0.07237652689218521, "fcm_dpo/margin": 8.164986610412598, "fcm_dpo/q_t": 0.3964681625366211, "grad_norm": 11.010472297668457, "learning_rate": 4.147121556398312e-07, "logits/chosen": 1.90861177444458, "logits/rejected": 1.6754601001739502, "logps/chosen": -64.9649429321289, "logps/ref_chosen": -66.88822174072266, "logps/ref_rejected": -92.27890014648438, "logps/rejected": -98.52061462402344, "loss": 1.0953, "margin_dpo/margin_mean": 8.164986610412598, "margin_dpo/margin_std": 12.750991821289062, "step": 229 }, { "epoch": 0.3476946334089191, "fcm_dpo/beta": 0.0579565167427063, "fcm_dpo/delta": 0.03130243346095085, "fcm_dpo/margin": 6.360637664794922, "fcm_dpo/q_t": 0.4185139238834381, "grad_norm": 13.974564552307129, "learning_rate": 4.137151834863213e-07, "logits/chosen": 1.3691972494125366, "logits/rejected": 1.3622510433197021, "logps/chosen": -76.63909912109375, "logps/ref_chosen": -76.12332153320312, "logps/ref_rejected": -78.19171905517578, "logps/rejected": -85.06814575195312, "loss": 1.194, "margin_dpo/margin_mean": 6.360638618469238, "margin_dpo/margin_std": 13.329750061035156, "step": 230 }, { "epoch": 0.3492063492063492, "fcm_dpo/beta": 0.056014273315668106, "fcm_dpo/delta": -0.1440962851047516, "fcm_dpo/margin": 9.541065216064453, "fcm_dpo/q_t": 0.3836032450199127, "grad_norm": 13.153295516967773, "learning_rate": 4.1271363186719835e-07, "logits/chosen": 1.2926297187805176, "logits/rejected": 1.297825813293457, "logps/chosen": -93.82791137695312, "logps/ref_chosen": -92.45181274414062, "logps/ref_rejected": -100.89735412597656, "logps/rejected": -111.81452178955078, "loss": 1.0597, "margin_dpo/margin_mean": 9.541065216064453, "margin_dpo/margin_std": 13.813907623291016, "step": 231 }, { "epoch": 0.3507180650037793, "fcm_dpo/beta": 0.05636624991893768, "fcm_dpo/delta": 0.038746319711208344, "fcm_dpo/margin": 6.433258533477783, "fcm_dpo/q_t": 0.42595604062080383, "grad_norm": 52.55567932128906, "learning_rate": 4.1170752879801436e-07, "logits/chosen": 1.5060627460479736, "logits/rejected": 1.4304571151733398, "logps/chosen": -87.07777404785156, "logps/ref_chosen": -86.75383758544922, "logps/ref_rejected": -98.16909790039062, "logps/rejected": -104.9262924194336, "loss": 1.2126, "margin_dpo/margin_mean": 6.433259010314941, "margin_dpo/margin_std": 14.871390342712402, "step": 232 }, { "epoch": 0.35222978080120937, "fcm_dpo/beta": 0.055456142872571945, "fcm_dpo/delta": -0.08742889761924744, "fcm_dpo/margin": 6.654336452484131, "fcm_dpo/q_t": 0.4225269556045532, "grad_norm": 10.734969139099121, "learning_rate": 4.106969024216348e-07, "logits/chosen": 1.5932424068450928, "logits/rejected": 1.4768648147583008, "logps/chosen": -72.19903564453125, "logps/ref_chosen": -72.87556457519531, "logps/ref_rejected": -85.22943115234375, "logps/rejected": -91.20724487304688, "loss": 1.1889, "margin_dpo/margin_mean": 6.654335975646973, "margin_dpo/margin_std": 13.669539451599121, "step": 233 }, { "epoch": 0.35374149659863946, "fcm_dpo/beta": 0.05513335019350052, "fcm_dpo/delta": -0.058548182249069214, "fcm_dpo/margin": 6.796961784362793, "fcm_dpo/q_t": 0.41953760385513306, "grad_norm": 11.108580589294434, "learning_rate": 4.09681781007452e-07, "logits/chosen": 1.1654977798461914, "logits/rejected": 1.123396635055542, "logps/chosen": -68.56211853027344, "logps/ref_chosen": -70.05477905273438, "logps/ref_rejected": -68.7240982055664, "logps/rejected": -74.02839660644531, "loss": 1.1745, "margin_dpo/margin_mean": 6.796961784362793, "margin_dpo/margin_std": 13.235331535339355, "step": 234 }, { "epoch": 0.35525321239606955, "fcm_dpo/beta": 0.05425561964511871, "fcm_dpo/delta": -0.07811406254768372, "fcm_dpo/margin": 8.745163917541504, "fcm_dpo/q_t": 0.39355021715164185, "grad_norm": 14.021645545959473, "learning_rate": 4.08662192950594e-07, "logits/chosen": 1.5307154655456543, "logits/rejected": 1.5013929605484009, "logps/chosen": -82.95243835449219, "logps/ref_chosen": -85.86051940917969, "logps/ref_rejected": -96.14663696289062, "logps/rejected": -101.98373413085938, "loss": 1.0673, "margin_dpo/margin_mean": 8.745163917541504, "margin_dpo/margin_std": 12.195037841796875, "step": 235 }, { "epoch": 0.35676492819349964, "fcm_dpo/beta": 0.05361638218164444, "fcm_dpo/delta": -0.029304729774594307, "fcm_dpo/margin": 7.980991363525391, "fcm_dpo/q_t": 0.40694406628608704, "grad_norm": 11.097721099853516, "learning_rate": 4.076381667711306e-07, "logits/chosen": 1.3134417533874512, "logits/rejected": 1.3008582592010498, "logps/chosen": -89.44915771484375, "logps/ref_chosen": -89.75252532958984, "logps/ref_rejected": -99.28534698486328, "logps/rejected": -106.96296691894531, "loss": 1.168, "margin_dpo/margin_mean": 7.980992317199707, "margin_dpo/margin_std": 15.801910400390625, "step": 236 }, { "epoch": 0.35827664399092973, "fcm_dpo/beta": 0.0544375479221344, "fcm_dpo/delta": 0.109842948615551, "fcm_dpo/margin": 5.391643524169922, "fcm_dpo/q_t": 0.4364463984966278, "grad_norm": 12.905435562133789, "learning_rate": 4.066097311132753e-07, "logits/chosen": 1.6906208992004395, "logits/rejected": 1.6837265491485596, "logps/chosen": -92.12542724609375, "logps/ref_chosen": -92.59001922607422, "logps/ref_rejected": -101.45584869384766, "logps/rejected": -106.38289642333984, "loss": 1.2223, "margin_dpo/margin_mean": 5.391643524169922, "margin_dpo/margin_std": 12.385063171386719, "step": 237 }, { "epoch": 0.35978835978835977, "fcm_dpo/beta": 0.054263561964035034, "fcm_dpo/delta": -0.04640674218535423, "fcm_dpo/margin": 8.181036949157715, "fcm_dpo/q_t": 0.4023571014404297, "grad_norm": 10.64908218383789, "learning_rate": 4.0557691474458414e-07, "logits/chosen": 1.474362850189209, "logits/rejected": 1.4516832828521729, "logps/chosen": -80.7723388671875, "logps/ref_chosen": -82.2470474243164, "logps/ref_rejected": -92.59944152832031, "logps/rejected": -99.30577850341797, "loss": 1.1047, "margin_dpo/margin_mean": 8.181037902832031, "margin_dpo/margin_std": 12.993268013000488, "step": 238 }, { "epoch": 0.36130007558578986, "fcm_dpo/beta": 0.05412614718079567, "fcm_dpo/delta": -0.05782606452703476, "fcm_dpo/margin": 8.406447410583496, "fcm_dpo/q_t": 0.4042254686355591, "grad_norm": 12.209634780883789, "learning_rate": 4.045397465551513e-07, "logits/chosen": 1.825083613395691, "logits/rejected": 1.4843999147415161, "logps/chosen": -75.3488998413086, "logps/ref_chosen": -75.30878448486328, "logps/ref_rejected": -131.2318115234375, "logps/rejected": -139.67837524414062, "loss": 1.1293, "margin_dpo/margin_mean": 8.40644645690918, "margin_dpo/margin_std": 14.963945388793945, "step": 239 }, { "epoch": 0.36281179138321995, "fcm_dpo/beta": 0.05183475464582443, "fcm_dpo/delta": -0.2632848024368286, "fcm_dpo/margin": 12.479537010192871, "fcm_dpo/q_t": 0.3565759062767029, "grad_norm": 16.51203727722168, "learning_rate": 4.0349825555680045e-07, "logits/chosen": 1.3655710220336914, "logits/rejected": 1.1419577598571777, "logps/chosen": -68.95610809326172, "logps/ref_chosen": -70.81785583496094, "logps/ref_rejected": -98.53778076171875, "logps/rejected": -109.15557861328125, "loss": 0.9813, "margin_dpo/margin_mean": 12.479536056518555, "margin_dpo/margin_std": 15.227754592895508, "step": 240 }, { "epoch": 0.36432350718065004, "fcm_dpo/beta": 0.05159341171383858, "fcm_dpo/delta": 0.07422710210084915, "fcm_dpo/margin": 6.358600616455078, "fcm_dpo/q_t": 0.4284597635269165, "grad_norm": 16.396221160888672, "learning_rate": 4.0245247088227377e-07, "logits/chosen": 1.2168505191802979, "logits/rejected": 1.1646525859832764, "logps/chosen": -88.27828979492188, "logps/ref_chosen": -88.60260772705078, "logps/ref_rejected": -101.42214965820312, "logps/rejected": -107.4564208984375, "loss": 1.1956, "margin_dpo/margin_mean": 6.35860013961792, "margin_dpo/margin_std": 13.391414642333984, "step": 241 }, { "epoch": 0.36583522297808013, "fcm_dpo/beta": 0.050503283739089966, "fcm_dpo/delta": -0.1315668523311615, "fcm_dpo/margin": 10.372234344482422, "fcm_dpo/q_t": 0.3853452801704407, "grad_norm": 13.189273834228516, "learning_rate": 4.0140242178441665e-07, "logits/chosen": 1.3361291885375977, "logits/rejected": 1.249570608139038, "logps/chosen": -75.59468841552734, "logps/ref_chosen": -77.34110260009766, "logps/ref_rejected": -84.76332092285156, "logps/rejected": -93.38914489746094, "loss": 1.0559, "margin_dpo/margin_mean": 10.372234344482422, "margin_dpo/margin_std": 14.9345703125, "step": 242 }, { "epoch": 0.3673469387755102, "fcm_dpo/beta": 0.05032705143094063, "fcm_dpo/delta": 0.002032870426774025, "fcm_dpo/margin": 7.908749580383301, "fcm_dpo/q_t": 0.41314005851745605, "grad_norm": 13.510807991027832, "learning_rate": 4.003481376353596e-07, "logits/chosen": 1.4432936906814575, "logits/rejected": 1.4497876167297363, "logps/chosen": -95.52180480957031, "logps/ref_chosen": -93.55897521972656, "logps/ref_rejected": -89.33551025390625, "logps/rejected": -99.20707702636719, "loss": 1.1565, "margin_dpo/margin_mean": 7.908749580383301, "margin_dpo/margin_std": 14.790718078613281, "step": 243 }, { "epoch": 0.3688586545729403, "fcm_dpo/beta": 0.0483069010078907, "fcm_dpo/delta": -0.277275025844574, "fcm_dpo/margin": 13.635225296020508, "fcm_dpo/q_t": 0.3540143668651581, "grad_norm": 9.192960739135742, "learning_rate": 3.9928964792569654e-07, "logits/chosen": 1.7651071548461914, "logits/rejected": 1.618671178817749, "logps/chosen": -69.32542419433594, "logps/ref_chosen": -69.82603454589844, "logps/ref_rejected": -92.4764175415039, "logps/rejected": -105.61103057861328, "loss": 0.9311, "margin_dpo/margin_mean": 13.635225296020508, "margin_dpo/margin_std": 14.008907318115234, "step": 244 }, { "epoch": 0.37037037037037035, "fcm_dpo/beta": 0.04629525542259216, "fcm_dpo/delta": -0.18837378919124603, "fcm_dpo/margin": 12.46827220916748, "fcm_dpo/q_t": 0.3724544048309326, "grad_norm": 10.381243705749512, "learning_rate": 3.982269822636601e-07, "logits/chosen": 1.560800552368164, "logits/rejected": 1.4880516529083252, "logps/chosen": -89.27384948730469, "logps/ref_chosen": -85.68216705322266, "logps/ref_rejected": -93.8754653930664, "logps/rejected": -109.9354248046875, "loss": 1.0069, "margin_dpo/margin_mean": 12.46827220916748, "margin_dpo/margin_std": 15.679994583129883, "step": 245 }, { "epoch": 0.37188208616780044, "fcm_dpo/beta": 0.045213352888822556, "fcm_dpo/delta": -0.11620943248271942, "fcm_dpo/margin": 11.289131164550781, "fcm_dpo/q_t": 0.38982075452804565, "grad_norm": 11.111617088317871, "learning_rate": 3.971601703742932e-07, "logits/chosen": 1.5159287452697754, "logits/rejected": 1.3694238662719727, "logps/chosen": -97.58953857421875, "logps/ref_chosen": -90.05093383789062, "logps/ref_rejected": -112.77645874023438, "logps/rejected": -131.60418701171875, "loss": 1.0648, "margin_dpo/margin_mean": 11.289131164550781, "margin_dpo/margin_std": 16.619611740112305, "step": 246 }, { "epoch": 0.37339380196523053, "fcm_dpo/beta": 0.045018598437309265, "fcm_dpo/delta": 0.03637855127453804, "fcm_dpo/margin": 5.280495643615723, "fcm_dpo/q_t": 0.44779178500175476, "grad_norm": 14.453449249267578, "learning_rate": 3.960892420986177e-07, "logits/chosen": 1.7113162279129028, "logits/rejected": 1.6601393222808838, "logps/chosen": -112.38603973388672, "logps/ref_chosen": -103.23979187011719, "logps/ref_rejected": -105.26278686523438, "logps/rejected": -119.68952941894531, "loss": 1.2806, "margin_dpo/margin_mean": 5.280494689941406, "margin_dpo/margin_std": 15.77202033996582, "step": 247 }, { "epoch": 0.3749055177626606, "fcm_dpo/beta": 0.045170750468969345, "fcm_dpo/delta": -0.044324249029159546, "fcm_dpo/margin": 9.78189468383789, "fcm_dpo/q_t": 0.40565359592437744, "grad_norm": 12.155468940734863, "learning_rate": 3.9501422739279953e-07, "logits/chosen": 1.5762226581573486, "logits/rejected": 1.7077572345733643, "logps/chosen": -95.51654052734375, "logps/ref_chosen": -88.16007995605469, "logps/ref_rejected": -75.11514282226562, "logps/rejected": -92.25349426269531, "loss": 1.1298, "margin_dpo/margin_mean": 9.78189468383789, "margin_dpo/margin_std": 17.028648376464844, "step": 248 }, { "epoch": 0.3764172335600907, "fcm_dpo/beta": 0.04518796131014824, "fcm_dpo/delta": 0.08948609232902527, "fcm_dpo/margin": 3.5041308403015137, "fcm_dpo/q_t": 0.46222302317619324, "grad_norm": 11.555813789367676, "learning_rate": 3.9393515632731094e-07, "logits/chosen": 1.6086914539337158, "logits/rejected": 1.6677453517913818, "logps/chosen": -102.29704284667969, "logps/ref_chosen": -91.01773071289062, "logps/ref_rejected": -80.51113891601562, "logps/rejected": -95.29458618164062, "loss": 1.3803, "margin_dpo/margin_mean": 3.5041308403015137, "margin_dpo/margin_std": 17.459102630615234, "step": 249 }, { "epoch": 0.3779289493575208, "fcm_dpo/beta": 0.04499515891075134, "fcm_dpo/delta": -0.1342799812555313, "fcm_dpo/margin": 11.716409683227539, "fcm_dpo/q_t": 0.3838357627391815, "grad_norm": 14.925415992736816, "learning_rate": 3.9285205908608934e-07, "logits/chosen": 1.6468373537063599, "logits/rejected": 1.5891878604888916, "logps/chosen": -91.09884643554688, "logps/ref_chosen": -80.5888671875, "logps/ref_rejected": -90.15093994140625, "logps/rejected": -112.37733459472656, "loss": 1.0709, "margin_dpo/margin_mean": 11.716409683227539, "margin_dpo/margin_std": 17.74170684814453, "step": 250 }, { "epoch": 0.3794406651549509, "fcm_dpo/beta": 0.04469338804483414, "fcm_dpo/delta": 0.06509894132614136, "fcm_dpo/margin": 7.542912483215332, "fcm_dpo/q_t": 0.42707180976867676, "grad_norm": 10.588440895080566, "learning_rate": 3.9176496596569265e-07, "logits/chosen": 1.8731662034988403, "logits/rejected": 1.7417778968811035, "logps/chosen": -93.5906982421875, "logps/ref_chosen": -82.70405578613281, "logps/ref_rejected": -98.94266510009766, "logps/rejected": -117.37222290039062, "loss": 1.195, "margin_dpo/margin_mean": 7.542912483215332, "margin_dpo/margin_std": 15.997113227844238, "step": 251 }, { "epoch": 0.38095238095238093, "fcm_dpo/beta": 0.04474301263689995, "fcm_dpo/delta": -0.04980228468775749, "fcm_dpo/margin": 7.735275745391846, "fcm_dpo/q_t": 0.42199695110321045, "grad_norm": 10.110880851745605, "learning_rate": 3.9067390737445254e-07, "logits/chosen": 1.2721498012542725, "logits/rejected": 1.1554539203643799, "logps/chosen": -81.79570007324219, "logps/ref_chosen": -73.10369110107422, "logps/ref_rejected": -94.90235900878906, "logps/rejected": -111.32964324951172, "loss": 1.2083, "margin_dpo/margin_mean": 7.735275745391846, "margin_dpo/margin_std": 17.073593139648438, "step": 252 }, { "epoch": 0.382464096749811, "fcm_dpo/beta": 0.044133760035037994, "fcm_dpo/delta": -0.0435882993042469, "fcm_dpo/margin": 6.770211696624756, "fcm_dpo/q_t": 0.43785423040390015, "grad_norm": 13.272875785827637, "learning_rate": 3.8957891383162304e-07, "logits/chosen": 1.7309565544128418, "logits/rejected": 1.6086697578430176, "logps/chosen": -79.59036254882812, "logps/ref_chosen": -68.7789535522461, "logps/ref_rejected": -75.98162078857422, "logps/rejected": -93.563232421875, "loss": 1.2324, "margin_dpo/margin_mean": 6.770212173461914, "margin_dpo/margin_std": 16.197599411010742, "step": 253 }, { "epoch": 0.3839758125472411, "fcm_dpo/beta": 0.043851204216480255, "fcm_dpo/delta": -0.030714038759469986, "fcm_dpo/margin": 9.789338111877441, "fcm_dpo/q_t": 0.40718260407447815, "grad_norm": 12.212136268615723, "learning_rate": 3.884800159665276e-07, "logits/chosen": 1.1488699913024902, "logits/rejected": 1.0198073387145996, "logps/chosen": -92.68016815185547, "logps/ref_chosen": -81.49362182617188, "logps/ref_rejected": -101.43672943115234, "logps/rejected": -122.4126205444336, "loss": 1.1105, "margin_dpo/margin_mean": 9.789339065551758, "margin_dpo/margin_std": 15.794754028320312, "step": 254 }, { "epoch": 0.3854875283446712, "fcm_dpo/beta": 0.04367159307003021, "fcm_dpo/delta": 0.004656985402107239, "fcm_dpo/margin": 9.045727729797363, "fcm_dpo/q_t": 0.41058290004730225, "grad_norm": 15.931432723999023, "learning_rate": 3.873772445177015e-07, "logits/chosen": 1.6107978820800781, "logits/rejected": 1.5556628704071045, "logps/chosen": -101.8770751953125, "logps/ref_chosen": -90.46351623535156, "logps/ref_rejected": -105.32445526123047, "logps/rejected": -125.78373718261719, "loss": 1.1632, "margin_dpo/margin_mean": 9.045727729797363, "margin_dpo/margin_std": 17.29879379272461, "step": 255 }, { "epoch": 0.3869992441421013, "fcm_dpo/beta": 0.043372705578804016, "fcm_dpo/delta": -0.06455010920763016, "fcm_dpo/margin": 10.635354042053223, "fcm_dpo/q_t": 0.401132732629776, "grad_norm": 10.090229034423828, "learning_rate": 3.862706303320329e-07, "logits/chosen": 1.5428683757781982, "logits/rejected": 1.3247222900390625, "logps/chosen": -94.11489868164062, "logps/ref_chosen": -81.56578063964844, "logps/ref_rejected": -108.58460998535156, "logps/rejected": -131.76907348632812, "loss": 1.1149, "margin_dpo/margin_mean": 10.635353088378906, "margin_dpo/margin_std": 17.988861083984375, "step": 256 }, { "epoch": 0.3885109599395314, "fcm_dpo/beta": 0.04340306669473648, "fcm_dpo/delta": 0.01081980112940073, "fcm_dpo/margin": 8.97639274597168, "fcm_dpo/q_t": 0.41701555252075195, "grad_norm": 14.49475383758545, "learning_rate": 3.851602043638994e-07, "logits/chosen": 1.6534913778305054, "logits/rejected": 1.538425326347351, "logps/chosen": -103.74288940429688, "logps/ref_chosen": -89.57557678222656, "logps/ref_rejected": -123.74462127685547, "logps/rejected": -146.88832092285156, "loss": 1.1878, "margin_dpo/margin_mean": 8.97639274597168, "margin_dpo/margin_std": 19.01647186279297, "step": 257 }, { "epoch": 0.3900226757369615, "fcm_dpo/beta": 0.04306865483522415, "fcm_dpo/delta": -0.06294693052768707, "fcm_dpo/margin": 10.682682037353516, "fcm_dpo/q_t": 0.3954845368862152, "grad_norm": 15.105984687805176, "learning_rate": 3.840459976743023e-07, "logits/chosen": 1.445413589477539, "logits/rejected": 1.294599175453186, "logps/chosen": -91.37678527832031, "logps/ref_chosen": -77.34173583984375, "logps/ref_rejected": -99.5709228515625, "logps/rejected": -124.28864288330078, "loss": 1.0614, "margin_dpo/margin_mean": 10.682682991027832, "margin_dpo/margin_std": 13.988033294677734, "step": 258 }, { "epoch": 0.3915343915343915, "fcm_dpo/beta": 0.041154563426971436, "fcm_dpo/delta": -0.2557143568992615, "fcm_dpo/margin": 15.510650634765625, "fcm_dpo/q_t": 0.3617561161518097, "grad_norm": 9.235639572143555, "learning_rate": 3.8292804142999796e-07, "logits/chosen": 1.6183881759643555, "logits/rejected": 1.4107770919799805, "logps/chosen": -90.75386047363281, "logps/ref_chosen": -82.39556121826172, "logps/ref_rejected": -113.73309326171875, "logps/rejected": -137.60205078125, "loss": 0.9892, "margin_dpo/margin_mean": 15.510651588439941, "margin_dpo/margin_std": 19.06281280517578, "step": 259 }, { "epoch": 0.3930461073318216, "fcm_dpo/beta": 0.04060753434896469, "fcm_dpo/delta": -0.0027508698403835297, "fcm_dpo/margin": 9.910329818725586, "fcm_dpo/q_t": 0.41329747438430786, "grad_norm": 16.31837272644043, "learning_rate": 3.818063669026256e-07, "logits/chosen": 1.426304578781128, "logits/rejected": 1.2057493925094604, "logps/chosen": -78.43536376953125, "logps/ref_chosen": -65.98947143554688, "logps/ref_rejected": -94.59706115722656, "logps/rejected": -116.95327758789062, "loss": 1.2009, "margin_dpo/margin_mean": 9.910329818725586, "margin_dpo/margin_std": 21.698278427124023, "step": 260 }, { "epoch": 0.3945578231292517, "fcm_dpo/beta": 0.0410933792591095, "fcm_dpo/delta": 0.07607575505971909, "fcm_dpo/margin": 7.944557189941406, "fcm_dpo/q_t": 0.4305505156517029, "grad_norm": 12.40672779083252, "learning_rate": 3.806810054678331e-07, "logits/chosen": 1.302069902420044, "logits/rejected": 1.3645875453948975, "logps/chosen": -99.55830383300781, "logps/ref_chosen": -88.87684631347656, "logps/ref_rejected": -82.34838104248047, "logps/rejected": -100.97440338134766, "loss": 1.228, "margin_dpo/margin_mean": 7.944557189941406, "margin_dpo/margin_std": 18.96116828918457, "step": 261 }, { "epoch": 0.3960695389266818, "fcm_dpo/beta": 0.04135086387395859, "fcm_dpo/delta": -0.03751703351736069, "fcm_dpo/margin": 10.531759262084961, "fcm_dpo/q_t": 0.40397828817367554, "grad_norm": 9.584962844848633, "learning_rate": 3.7955198860439887e-07, "logits/chosen": 1.743009090423584, "logits/rejected": 1.5424811840057373, "logps/chosen": -93.39926147460938, "logps/ref_chosen": -85.81719970703125, "logps/ref_rejected": -105.49027252197266, "logps/rejected": -123.60408782958984, "loss": 1.102, "margin_dpo/margin_mean": 10.531759262084961, "margin_dpo/margin_std": 16.00366973876953, "step": 262 }, { "epoch": 0.3975812547241119, "fcm_dpo/beta": 0.040634773671627045, "fcm_dpo/delta": -0.06253870576620102, "fcm_dpo/margin": 11.31068229675293, "fcm_dpo/q_t": 0.4001634120941162, "grad_norm": 10.176654815673828, "learning_rate": 3.784193478933516e-07, "logits/chosen": 1.6961557865142822, "logits/rejected": 1.3818693161010742, "logps/chosen": -81.1441421508789, "logps/ref_chosen": -73.61693572998047, "logps/ref_rejected": -102.39161682128906, "logps/rejected": -121.22950744628906, "loss": 1.1071, "margin_dpo/margin_mean": 11.31068229675293, "margin_dpo/margin_std": 18.635833740234375, "step": 263 }, { "epoch": 0.39909297052154197, "fcm_dpo/beta": 0.04004526510834694, "fcm_dpo/delta": -0.07845936715602875, "fcm_dpo/margin": 11.853878021240234, "fcm_dpo/q_t": 0.39685431122779846, "grad_norm": 9.623083114624023, "learning_rate": 3.7728311501708674e-07, "logits/chosen": 1.3879857063293457, "logits/rejected": 1.2453668117523193, "logps/chosen": -110.89766693115234, "logps/ref_chosen": -101.57856750488281, "logps/ref_rejected": -111.65735626220703, "logps/rejected": -132.83033752441406, "loss": 1.0743, "margin_dpo/margin_mean": 11.853878021240234, "margin_dpo/margin_std": 17.387245178222656, "step": 264 }, { "epoch": 0.40060468631897206, "fcm_dpo/beta": 0.038992609828710556, "fcm_dpo/delta": -0.1944998949766159, "fcm_dpo/margin": 14.966062545776367, "fcm_dpo/q_t": 0.3741992712020874, "grad_norm": 9.02346420288086, "learning_rate": 3.7614332175848027e-07, "logits/chosen": 1.6460204124450684, "logits/rejected": 1.5074365139007568, "logps/chosen": -70.97564697265625, "logps/ref_chosen": -65.76426696777344, "logps/ref_rejected": -85.19627380371094, "logps/rejected": -105.37371826171875, "loss": 1.047, "margin_dpo/margin_mean": 14.966060638427734, "margin_dpo/margin_std": 21.565500259399414, "step": 265 }, { "epoch": 0.4021164021164021, "fcm_dpo/beta": 0.03780415654182434, "fcm_dpo/delta": -0.0783822163939476, "fcm_dpo/margin": 12.542582511901855, "fcm_dpo/q_t": 0.3959764838218689, "grad_norm": 12.275056838989258, "learning_rate": 3.75e-07, "logits/chosen": 1.502774715423584, "logits/rejected": 1.3121366500854492, "logps/chosen": -79.05199432373047, "logps/ref_chosen": -75.05682373046875, "logps/ref_rejected": -97.52758026123047, "logps/rejected": -114.06533813476562, "loss": 1.0869, "margin_dpo/margin_mean": 12.542583465576172, "margin_dpo/margin_std": 19.404693603515625, "step": 266 }, { "epoch": 0.4036281179138322, "fcm_dpo/beta": 0.038025856018066406, "fcm_dpo/delta": -0.019957251846790314, "fcm_dpo/margin": 10.98253059387207, "fcm_dpo/q_t": 0.40621858835220337, "grad_norm": 7.956915378570557, "learning_rate": 3.738531817228131e-07, "logits/chosen": 1.4474161863327026, "logits/rejected": 1.3807919025421143, "logps/chosen": -73.45161437988281, "logps/ref_chosen": -71.13494110107422, "logps/ref_rejected": -81.14566040039062, "logps/rejected": -94.44486999511719, "loss": 1.124, "margin_dpo/margin_mean": 10.98253059387207, "margin_dpo/margin_std": 17.64596176147461, "step": 267 }, { "epoch": 0.4051398337112623, "fcm_dpo/beta": 0.037608757615089417, "fcm_dpo/delta": -0.006927022244781256, "fcm_dpo/margin": 7.760471820831299, "fcm_dpo/q_t": 0.4383552074432373, "grad_norm": 8.453049659729004, "learning_rate": 3.7270289900589204e-07, "logits/chosen": 1.5731756687164307, "logits/rejected": 1.5207586288452148, "logps/chosen": -83.85981750488281, "logps/ref_chosen": -80.06082153320312, "logps/ref_rejected": -87.43035888671875, "logps/rejected": -98.98983001708984, "loss": 1.2233, "margin_dpo/margin_mean": 7.760472297668457, "margin_dpo/margin_std": 18.012435913085938, "step": 268 }, { "epoch": 0.40665154950869237, "fcm_dpo/beta": 0.03780782222747803, "fcm_dpo/delta": -0.01912636309862137, "fcm_dpo/margin": 11.029001235961914, "fcm_dpo/q_t": 0.4066724181175232, "grad_norm": 10.276360511779785, "learning_rate": 3.7154918402511714e-07, "logits/chosen": 1.6505131721496582, "logits/rejected": 1.5838286876678467, "logps/chosen": -89.04998016357422, "logps/ref_chosen": -83.36944580078125, "logps/ref_rejected": -100.66839599609375, "logps/rejected": -117.37794494628906, "loss": 1.1236, "margin_dpo/margin_mean": 11.029001235961914, "margin_dpo/margin_std": 17.85173797607422, "step": 269 }, { "epoch": 0.40816326530612246, "fcm_dpo/beta": 0.03762676566839218, "fcm_dpo/delta": 0.02844144031405449, "fcm_dpo/margin": 9.90216064453125, "fcm_dpo/q_t": 0.41595685482025146, "grad_norm": 10.71731185913086, "learning_rate": 3.7039206905237656e-07, "logits/chosen": 1.6075557470321655, "logits/rejected": 1.3942244052886963, "logps/chosen": -90.50005340576172, "logps/ref_chosen": -85.35945129394531, "logps/ref_rejected": -104.47489929199219, "logps/rejected": -119.51766967773438, "loss": 1.1518, "margin_dpo/margin_mean": 9.902159690856934, "margin_dpo/margin_std": 17.91962242126465, "step": 270 }, { "epoch": 0.40967498110355255, "fcm_dpo/beta": 0.03780830278992653, "fcm_dpo/delta": 0.020608875900506973, "fcm_dpo/margin": 7.860053539276123, "fcm_dpo/q_t": 0.4410387873649597, "grad_norm": 10.222976684570312, "learning_rate": 3.692315864546635e-07, "logits/chosen": 1.4401030540466309, "logits/rejected": 1.307854175567627, "logps/chosen": -92.08882141113281, "logps/ref_chosen": -86.01373291015625, "logps/ref_rejected": -109.99561309814453, "logps/rejected": -123.93075561523438, "loss": 1.2552, "margin_dpo/margin_mean": 7.860054016113281, "margin_dpo/margin_std": 21.145322799682617, "step": 271 }, { "epoch": 0.41118669690098264, "fcm_dpo/beta": 0.03660859167575836, "fcm_dpo/delta": -0.21598142385482788, "fcm_dpo/margin": 16.469074249267578, "fcm_dpo/q_t": 0.3663064241409302, "grad_norm": 11.490429878234863, "learning_rate": 3.6806776869317067e-07, "logits/chosen": 1.662217378616333, "logits/rejected": 1.6814091205596924, "logps/chosen": -88.33952331542969, "logps/ref_chosen": -86.37013244628906, "logps/ref_rejected": -85.74638366699219, "logps/rejected": -104.1848373413086, "loss": 0.9742, "margin_dpo/margin_mean": 16.469074249267578, "margin_dpo/margin_std": 18.660140991210938, "step": 272 }, { "epoch": 0.4126984126984127, "fcm_dpo/beta": 0.03605865687131882, "fcm_dpo/delta": -0.018457308411598206, "fcm_dpo/margin": 11.581703186035156, "fcm_dpo/q_t": 0.4078387916088104, "grad_norm": 10.316041946411133, "learning_rate": 3.669006483223828e-07, "logits/chosen": 1.6665234565734863, "logits/rejected": 1.5125619173049927, "logps/chosen": -82.37163543701172, "logps/ref_chosen": -75.51087951660156, "logps/ref_rejected": -101.60345458984375, "logps/rejected": -120.04591369628906, "loss": 1.1738, "margin_dpo/margin_mean": 11.581703186035156, "margin_dpo/margin_std": 23.04998016357422, "step": 273 }, { "epoch": 0.41421012849584277, "fcm_dpo/beta": 0.0356617271900177, "fcm_dpo/delta": -0.07967354357242584, "fcm_dpo/margin": 13.34488582611084, "fcm_dpo/q_t": 0.3960227370262146, "grad_norm": 7.991804599761963, "learning_rate": 3.657302579891656e-07, "logits/chosen": 1.3249926567077637, "logits/rejected": 1.2660455703735352, "logps/chosen": -85.18216705322266, "logps/ref_chosen": -79.040283203125, "logps/ref_rejected": -86.31329345703125, "logps/rejected": -105.80005645751953, "loss": 1.081, "margin_dpo/margin_mean": 13.34488582611084, "margin_dpo/margin_std": 19.965038299560547, "step": 274 }, { "epoch": 0.41572184429327286, "fcm_dpo/beta": 0.03469148278236389, "fcm_dpo/delta": -0.2013135701417923, "fcm_dpo/margin": 17.002704620361328, "fcm_dpo/q_t": 0.37113866209983826, "grad_norm": 9.240900039672852, "learning_rate": 3.645566304318526e-07, "logits/chosen": 1.4302536249160767, "logits/rejected": 1.2116038799285889, "logps/chosen": -74.8931884765625, "logps/ref_chosen": -71.82034301757812, "logps/ref_rejected": -94.29946899414062, "logps/rejected": -114.37501525878906, "loss": 0.996, "margin_dpo/margin_mean": 17.002704620361328, "margin_dpo/margin_std": 20.238014221191406, "step": 275 }, { "epoch": 0.41723356009070295, "fcm_dpo/beta": 0.03372751921415329, "fcm_dpo/delta": -0.07244005799293518, "fcm_dpo/margin": 13.906018257141113, "fcm_dpo/q_t": 0.3970295786857605, "grad_norm": 13.785542488098145, "learning_rate": 3.633797984793294e-07, "logits/chosen": 1.2479619979858398, "logits/rejected": 1.1853089332580566, "logps/chosen": -73.82440185546875, "logps/ref_chosen": -69.54020690917969, "logps/ref_rejected": -78.59674072265625, "logps/rejected": -96.78695678710938, "loss": 1.0941, "margin_dpo/margin_mean": 13.906018257141113, "margin_dpo/margin_std": 21.78231430053711, "step": 276 }, { "epoch": 0.41874527588813304, "fcm_dpo/beta": 0.03437428921461105, "fcm_dpo/delta": 0.14951635897159576, "fcm_dpo/margin": 7.397580146789551, "fcm_dpo/q_t": 0.4457745850086212, "grad_norm": 10.012701034545898, "learning_rate": 3.6219979505011555e-07, "logits/chosen": 1.7152776718139648, "logits/rejected": 1.794743537902832, "logps/chosen": -104.70123291015625, "logps/ref_chosen": -94.4896240234375, "logps/ref_rejected": -85.45901489257812, "logps/rejected": -103.06820678710938, "loss": 1.281, "margin_dpo/margin_mean": 7.397579669952393, "margin_dpo/margin_std": 22.03266716003418, "step": 277 }, { "epoch": 0.42025699168556313, "fcm_dpo/beta": 0.034356892108917236, "fcm_dpo/delta": -0.08088916540145874, "fcm_dpo/margin": 10.672147750854492, "fcm_dpo/q_t": 0.4208662509918213, "grad_norm": 11.455450057983398, "learning_rate": 3.6101665315144353e-07, "logits/chosen": 1.6356345415115356, "logits/rejected": 1.4460411071777344, "logps/chosen": -97.5278549194336, "logps/ref_chosen": -87.42613220214844, "logps/ref_rejected": -105.44854736328125, "logps/rejected": -126.222412109375, "loss": 1.1979, "margin_dpo/margin_mean": 10.672147750854492, "margin_dpo/margin_std": 22.455184936523438, "step": 278 }, { "epoch": 0.4217687074829932, "fcm_dpo/beta": 0.03291717916727066, "fcm_dpo/delta": -0.25346609950065613, "fcm_dpo/margin": 19.376266479492188, "fcm_dpo/q_t": 0.3584246337413788, "grad_norm": 10.21809196472168, "learning_rate": 3.5983040587833563e-07, "logits/chosen": 1.4293241500854492, "logits/rejected": 1.3572511672973633, "logps/chosen": -71.79463195800781, "logps/ref_chosen": -70.516845703125, "logps/ref_rejected": -86.04249572753906, "logps/rejected": -106.69654846191406, "loss": 0.9563, "margin_dpo/margin_mean": 19.376266479492188, "margin_dpo/margin_std": 21.07408905029297, "step": 279 }, { "epoch": 0.42328042328042326, "fcm_dpo/beta": 0.031365811824798584, "fcm_dpo/delta": -0.20678894221782684, "fcm_dpo/margin": 18.94228744506836, "fcm_dpo/q_t": 0.36702173948287964, "grad_norm": 10.397953987121582, "learning_rate": 3.586410864126781e-07, "logits/chosen": 1.5499365329742432, "logits/rejected": 1.4086581468582153, "logps/chosen": -82.18923950195312, "logps/ref_chosen": -76.5021743774414, "logps/ref_rejected": -94.2752685546875, "logps/rejected": -118.90461730957031, "loss": 0.9775, "margin_dpo/margin_mean": 18.94228744506836, "margin_dpo/margin_std": 21.44241714477539, "step": 280 }, { "epoch": 0.42479213907785335, "fcm_dpo/beta": 0.030857127159833908, "fcm_dpo/delta": -0.04540977627038956, "fcm_dpo/margin": 14.368467330932617, "fcm_dpo/q_t": 0.4017066955566406, "grad_norm": 8.291491508483887, "learning_rate": 3.574487280222929e-07, "logits/chosen": 1.611108422279358, "logits/rejected": 1.6434032917022705, "logps/chosen": -85.33029174804688, "logps/ref_chosen": -77.50468444824219, "logps/ref_rejected": -79.05717468261719, "logps/rejected": -101.25123596191406, "loss": 1.1022, "margin_dpo/margin_mean": 14.368467330932617, "margin_dpo/margin_std": 22.488018035888672, "step": 281 }, { "epoch": 0.42630385487528344, "fcm_dpo/beta": 0.03047410398721695, "fcm_dpo/delta": -0.10886166244745255, "fcm_dpo/margin": 16.517818450927734, "fcm_dpo/q_t": 0.3921484351158142, "grad_norm": 9.714167594909668, "learning_rate": 3.562533640600075e-07, "logits/chosen": 1.3984555006027222, "logits/rejected": 1.2696608304977417, "logps/chosen": -91.43336486816406, "logps/ref_chosen": -80.31298065185547, "logps/ref_rejected": -83.72120666503906, "logps/rejected": -111.35940551757812, "loss": 1.0718, "margin_dpo/margin_mean": 16.517820358276367, "margin_dpo/margin_std": 24.46967887878418, "step": 282 }, { "epoch": 0.42781557067271353, "fcm_dpo/beta": 0.03024178370833397, "fcm_dpo/delta": -0.0018365830183029175, "fcm_dpo/margin": 13.276296615600586, "fcm_dpo/q_t": 0.41008254885673523, "grad_norm": 11.770947456359863, "learning_rate": 3.550550279627215e-07, "logits/chosen": 1.4370884895324707, "logits/rejected": 1.1001770496368408, "logps/chosen": -96.20101928710938, "logps/ref_chosen": -80.72602844238281, "logps/ref_rejected": -115.68379211425781, "logps/rejected": -144.43507385253906, "loss": 1.15, "margin_dpo/margin_mean": 13.276298522949219, "margin_dpo/margin_std": 24.19223976135254, "step": 283 }, { "epoch": 0.4293272864701436, "fcm_dpo/beta": 0.02937573380768299, "fcm_dpo/delta": -0.15650880336761475, "fcm_dpo/margin": 18.640047073364258, "fcm_dpo/q_t": 0.3789505362510681, "grad_norm": 8.9287748336792, "learning_rate": 3.5385375325047163e-07, "logits/chosen": 1.5543755292892456, "logits/rejected": 1.3925883769989014, "logps/chosen": -87.98505401611328, "logps/ref_chosen": -77.5223388671875, "logps/ref_rejected": -104.1847152709961, "logps/rejected": -133.2874755859375, "loss": 1.0198, "margin_dpo/margin_mean": 18.640047073364258, "margin_dpo/margin_std": 23.840347290039062, "step": 284 }, { "epoch": 0.4308390022675737, "fcm_dpo/beta": 0.029584050178527832, "fcm_dpo/delta": 0.06740415096282959, "fcm_dpo/margin": 11.305548667907715, "fcm_dpo/q_t": 0.42707347869873047, "grad_norm": 11.452542304992676, "learning_rate": 3.5264957352549375e-07, "logits/chosen": 1.760129690170288, "logits/rejected": 1.6619482040405273, "logps/chosen": -107.24560546875, "logps/ref_chosen": -85.79348754882812, "logps/ref_rejected": -96.46463775634766, "logps/rejected": -129.22230529785156, "loss": 1.2082, "margin_dpo/margin_mean": 11.305547714233398, "margin_dpo/margin_std": 25.100751876831055, "step": 285 }, { "epoch": 0.4323507180650038, "fcm_dpo/beta": 0.028857434168457985, "fcm_dpo/delta": -0.15750229358673096, "fcm_dpo/margin": 19.008071899414062, "fcm_dpo/q_t": 0.3800804316997528, "grad_norm": 8.827457427978516, "learning_rate": 3.514425224712835e-07, "logits/chosen": 1.429862141609192, "logits/rejected": 1.1608591079711914, "logps/chosen": -95.72119140625, "logps/ref_chosen": -77.86268615722656, "logps/ref_rejected": -110.77134704589844, "logps/rejected": -147.637939453125, "loss": 1.0384, "margin_dpo/margin_mean": 19.008071899414062, "margin_dpo/margin_std": 26.186504364013672, "step": 286 }, { "epoch": 0.43386243386243384, "fcm_dpo/beta": 0.027673378586769104, "fcm_dpo/delta": -0.29322800040245056, "fcm_dpo/margin": 24.37060546875, "fcm_dpo/q_t": 0.3525708317756653, "grad_norm": 11.625649452209473, "learning_rate": 3.502326338516534e-07, "logits/chosen": 1.506529688835144, "logits/rejected": 1.4740724563598633, "logps/chosen": -74.51770782470703, "logps/ref_chosen": -62.552825927734375, "logps/ref_rejected": -77.7650146484375, "logps/rejected": -114.10049438476562, "loss": 0.9449, "margin_dpo/margin_mean": 24.37060546875, "margin_dpo/margin_std": 26.406269073486328, "step": 287 }, { "epoch": 0.43537414965986393, "fcm_dpo/beta": 0.027158301323652267, "fcm_dpo/delta": 0.03195127844810486, "fcm_dpo/margin": 13.59581470489502, "fcm_dpo/q_t": 0.41956979036331177, "grad_norm": 10.045726776123047, "learning_rate": 3.490199415097892e-07, "logits/chosen": 1.1744755506515503, "logits/rejected": 1.060915470123291, "logps/chosen": -106.11993408203125, "logps/ref_chosen": -83.74117279052734, "logps/ref_rejected": -106.93913269042969, "logps/rejected": -142.91371154785156, "loss": 1.1573, "margin_dpo/margin_mean": 13.595813751220703, "margin_dpo/margin_std": 25.147212982177734, "step": 288 }, { "epoch": 0.436885865457294, "fcm_dpo/beta": 0.026928097009658813, "fcm_dpo/delta": -0.080781489610672, "fcm_dpo/margin": 17.713703155517578, "fcm_dpo/q_t": 0.3959569036960602, "grad_norm": 8.819540023803711, "learning_rate": 3.4780447936730247e-07, "logits/chosen": 1.7476454973220825, "logits/rejected": 1.671339750289917, "logps/chosen": -94.48479461669922, "logps/ref_chosen": -73.04204559326172, "logps/ref_rejected": -88.07904052734375, "logps/rejected": -127.23548889160156, "loss": 1.0789, "margin_dpo/margin_mean": 17.713703155517578, "margin_dpo/margin_std": 26.483057022094727, "step": 289 }, { "epoch": 0.4383975812547241, "fcm_dpo/beta": 0.02662849798798561, "fcm_dpo/delta": 0.0020782388746738434, "fcm_dpo/margin": 14.92031192779541, "fcm_dpo/q_t": 0.4109000563621521, "grad_norm": 9.255086898803711, "learning_rate": 3.465862814232821e-07, "logits/chosen": 1.6757187843322754, "logits/rejected": 1.5849800109863281, "logps/chosen": -106.42593383789062, "logps/ref_chosen": -78.60614013671875, "logps/ref_rejected": -108.50082397460938, "logps/rejected": -151.24093627929688, "loss": 1.1414, "margin_dpo/margin_mean": 14.920310974121094, "margin_dpo/margin_std": 25.820228576660156, "step": 290 }, { "epoch": 0.4399092970521542, "fcm_dpo/beta": 0.02671034447848797, "fcm_dpo/delta": -0.08461640775203705, "fcm_dpo/margin": 17.954689025878906, "fcm_dpo/q_t": 0.39686498045921326, "grad_norm": 10.577985763549805, "learning_rate": 3.4536538175334343e-07, "logits/chosen": 1.9808932542800903, "logits/rejected": 1.711446762084961, "logps/chosen": -90.13938903808594, "logps/ref_chosen": -66.71226501464844, "logps/ref_rejected": -96.14029693603516, "logps/rejected": -137.5220947265625, "loss": 1.1239, "margin_dpo/margin_mean": 17.954689025878906, "margin_dpo/margin_std": 30.840965270996094, "step": 291 }, { "epoch": 0.4414210128495843, "fcm_dpo/beta": 0.026129107922315598, "fcm_dpo/delta": -0.04495128244161606, "fcm_dpo/margin": 16.948894500732422, "fcm_dpo/q_t": 0.40278592705726624, "grad_norm": 10.55582046508789, "learning_rate": 3.4414181450867465e-07, "logits/chosen": 1.4002811908721924, "logits/rejected": 1.318420171737671, "logps/chosen": -103.06094360351562, "logps/ref_chosen": -80.3355484008789, "logps/ref_rejected": -90.44906616210938, "logps/rejected": -130.12335205078125, "loss": 1.1202, "margin_dpo/margin_mean": 16.948894500732422, "margin_dpo/margin_std": 28.7812557220459, "step": 292 }, { "epoch": 0.4429327286470144, "fcm_dpo/beta": 0.025567151606082916, "fcm_dpo/delta": -0.1617434173822403, "fcm_dpo/margin": 21.632625579833984, "fcm_dpo/q_t": 0.3831241726875305, "grad_norm": 11.195418357849121, "learning_rate": 3.4291561391508185e-07, "logits/chosen": 1.8973731994628906, "logits/rejected": 1.7341361045837402, "logps/chosen": -95.07200622558594, "logps/ref_chosen": -71.69970703125, "logps/ref_rejected": -102.13948059082031, "logps/rejected": -147.1444091796875, "loss": 1.0974, "margin_dpo/margin_mean": 21.632625579833984, "margin_dpo/margin_std": 36.16749572753906, "step": 293 }, { "epoch": 0.4444444444444444, "fcm_dpo/beta": 0.024921298027038574, "fcm_dpo/delta": -0.05894307792186737, "fcm_dpo/margin": 18.27361488342285, "fcm_dpo/q_t": 0.39900296926498413, "grad_norm": 12.699747085571289, "learning_rate": 3.4168681427203153e-07, "logits/chosen": 1.6706271171569824, "logits/rejected": 1.5605731010437012, "logps/chosen": -93.49272155761719, "logps/ref_chosen": -70.73458862304688, "logps/ref_rejected": -86.68821716308594, "logps/rejected": -127.719970703125, "loss": 1.0928, "margin_dpo/margin_mean": 18.273616790771484, "margin_dpo/margin_std": 27.647445678710938, "step": 294 }, { "epoch": 0.4459561602418745, "fcm_dpo/beta": 0.025058383122086525, "fcm_dpo/delta": 0.03231852129101753, "fcm_dpo/margin": 14.720477104187012, "fcm_dpo/q_t": 0.4185822606086731, "grad_norm": 13.354880332946777, "learning_rate": 3.4045544995169125e-07, "logits/chosen": 1.2770599126815796, "logits/rejected": 1.0146013498306274, "logps/chosen": -91.09881591796875, "logps/ref_chosen": -66.42644500732422, "logps/ref_rejected": -99.58766174316406, "logps/rejected": -138.9805145263672, "loss": 1.1745, "margin_dpo/margin_mean": 14.720477104187012, "margin_dpo/margin_std": 29.390125274658203, "step": 295 }, { "epoch": 0.4474678760393046, "fcm_dpo/beta": 0.02432844787836075, "fcm_dpo/delta": -0.17003798484802246, "fcm_dpo/margin": 22.94456672668457, "fcm_dpo/q_t": 0.37980806827545166, "grad_norm": 9.647819519042969, "learning_rate": 3.392215553979679e-07, "logits/chosen": 1.331782579421997, "logits/rejected": 1.188205599784851, "logps/chosen": -109.94386291503906, "logps/ref_chosen": -87.47459411621094, "logps/ref_rejected": -103.96894836425781, "logps/rejected": -149.38278198242188, "loss": 1.0534, "margin_dpo/margin_mean": 22.94456672668457, "margin_dpo/margin_std": 33.13970947265625, "step": 296 }, { "epoch": 0.4489795918367347, "fcm_dpo/beta": 0.023840460926294327, "fcm_dpo/delta": -0.18414409458637238, "fcm_dpo/margin": 24.06473159790039, "fcm_dpo/q_t": 0.3722948133945465, "grad_norm": 12.918481826782227, "learning_rate": 3.3798516512554485e-07, "logits/chosen": 1.4045250415802002, "logits/rejected": 1.276968002319336, "logps/chosen": -97.88027954101562, "logps/ref_chosen": -73.46731567382812, "logps/ref_rejected": -88.22674560546875, "logps/rejected": -136.70443725585938, "loss": 0.9989, "margin_dpo/margin_mean": 24.06473159790039, "margin_dpo/margin_std": 27.818138122558594, "step": 297 }, { "epoch": 0.4504913076341648, "fcm_dpo/beta": 0.023241370916366577, "fcm_dpo/delta": -0.06495536863803864, "fcm_dpo/margin": 19.87840461730957, "fcm_dpo/q_t": 0.4035850465297699, "grad_norm": 12.550525665283203, "learning_rate": 3.367463137189156e-07, "logits/chosen": 1.7822816371917725, "logits/rejected": 1.6975232362747192, "logps/chosen": -94.45632934570312, "logps/ref_chosen": -73.21676635742188, "logps/ref_rejected": -84.9563217163086, "logps/rejected": -126.07428741455078, "loss": 1.14, "margin_dpo/margin_mean": 19.878402709960938, "margin_dpo/margin_std": 36.64874267578125, "step": 298 }, { "epoch": 0.4520030234315949, "fcm_dpo/beta": 0.023189421743154526, "fcm_dpo/delta": 0.02762114629149437, "fcm_dpo/margin": 16.102848052978516, "fcm_dpo/q_t": 0.42042386531829834, "grad_norm": 9.376786231994629, "learning_rate": 3.355050358314172e-07, "logits/chosen": 1.2836685180664062, "logits/rejected": 1.1858904361724854, "logps/chosen": -100.31362915039062, "logps/ref_chosen": -76.9534912109375, "logps/ref_rejected": -87.53433227539062, "logps/rejected": -126.997314453125, "loss": 1.19, "margin_dpo/margin_mean": 16.10284996032715, "margin_dpo/margin_std": 34.105308532714844, "step": 299 }, { "epoch": 0.45351473922902497, "fcm_dpo/beta": 0.02318103238940239, "fcm_dpo/delta": -0.0181889571249485, "fcm_dpo/margin": 18.007240295410156, "fcm_dpo/q_t": 0.40901291370391846, "grad_norm": 7.5748467445373535, "learning_rate": 3.3426136618426043e-07, "logits/chosen": 1.4052281379699707, "logits/rejected": 1.2209758758544922, "logps/chosen": -100.68650817871094, "logps/ref_chosen": -78.36398315429688, "logps/ref_rejected": -97.03912353515625, "logps/rejected": -137.368896484375, "loss": 1.1662, "margin_dpo/margin_mean": 18.007240295410156, "margin_dpo/margin_std": 35.32292175292969, "step": 300 }, { "epoch": 0.45351473922902497, "eval_fcm_dpo/beta": 0.02316315658390522, "eval_logits/chosen": 1.3809725046157837, "eval_logits/rejected": 1.253163456916809, "eval_logps/chosen": -106.03602600097656, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -135.76068115234375, "eval_loss": 0.5600526928901672, "eval_margin_dpo/margin_mean": 19.930051803588867, "eval_margin_dpo/margin_std": 33.923973083496094, "eval_runtime": 42.33, "eval_samples_per_second": 54.406, "eval_steps_per_second": 1.701, "step": 300 }, { "epoch": 0.455026455026455, "fcm_dpo/beta": 0.022925032302737236, "fcm_dpo/delta": -0.031985800713300705, "fcm_dpo/margin": 18.749425888061523, "fcm_dpo/q_t": 0.4092825651168823, "grad_norm": 7.92974328994751, "learning_rate": 3.3301533956555885e-07, "logits/chosen": 1.3435275554656982, "logits/rejected": 1.2831223011016846, "logps/chosen": -91.15614318847656, "logps/ref_chosen": -70.6719741821289, "logps/ref_rejected": -87.11650085449219, "logps/rejected": -126.35010528564453, "loss": 1.1464, "margin_dpo/margin_mean": 18.749425888061523, "margin_dpo/margin_std": 34.64903259277344, "step": 301 }, { "epoch": 0.4565381708238851, "fcm_dpo/beta": 0.023184455931186676, "fcm_dpo/delta": 0.0728965699672699, "fcm_dpo/margin": 9.814539909362793, "fcm_dpo/q_t": 0.4503589868545532, "grad_norm": 9.430741310119629, "learning_rate": 3.317669908293554e-07, "logits/chosen": 1.1618998050689697, "logits/rejected": 1.0537515878677368, "logps/chosen": -111.93255615234375, "logps/ref_chosen": -85.29096221923828, "logps/ref_rejected": -106.22589874267578, "logps/rejected": -142.68203735351562, "loss": 1.2936, "margin_dpo/margin_mean": 9.814538955688477, "margin_dpo/margin_std": 31.3475341796875, "step": 302 }, { "epoch": 0.4580498866213152, "fcm_dpo/beta": 0.022597171366214752, "fcm_dpo/delta": -0.17091017961502075, "fcm_dpo/margin": 24.739337921142578, "fcm_dpo/q_t": 0.3804760277271271, "grad_norm": 7.812971115112305, "learning_rate": 3.3051635489464793e-07, "logits/chosen": 1.7512423992156982, "logits/rejected": 1.6072579622268677, "logps/chosen": -104.96895599365234, "logps/ref_chosen": -83.90059661865234, "logps/ref_rejected": -104.7340087890625, "logps/rejected": -150.54171752929688, "loss": 1.0798, "margin_dpo/margin_mean": 24.739337921142578, "margin_dpo/margin_std": 39.02869415283203, "step": 303 }, { "epoch": 0.4595616024187453, "fcm_dpo/beta": 0.022171439602971077, "fcm_dpo/delta": -0.11509604007005692, "fcm_dpo/margin": 22.964481353759766, "fcm_dpo/q_t": 0.38752371072769165, "grad_norm": 9.252838134765625, "learning_rate": 3.292634667444117e-07, "logits/chosen": 1.3123148679733276, "logits/rejected": 1.203786015510559, "logps/chosen": -93.95158386230469, "logps/ref_chosen": -77.39997100830078, "logps/ref_rejected": -94.21647644042969, "logps/rejected": -133.73257446289062, "loss": 1.0487, "margin_dpo/margin_mean": 22.9644832611084, "margin_dpo/margin_std": 31.3437442779541, "step": 304 }, { "epoch": 0.46107331821617537, "fcm_dpo/beta": 0.02172023430466652, "fcm_dpo/delta": -0.04386585205793381, "fcm_dpo/margin": 20.260169982910156, "fcm_dpo/q_t": 0.4058588147163391, "grad_norm": 8.574239730834961, "learning_rate": 3.280083614246217e-07, "logits/chosen": 1.5030393600463867, "logits/rejected": 1.5878260135650635, "logps/chosen": -113.70249938964844, "logps/ref_chosen": -90.90805053710938, "logps/ref_rejected": -85.84992980957031, "logps/rejected": -128.90455627441406, "loss": 1.1498, "margin_dpo/margin_mean": 20.260169982910156, "margin_dpo/margin_std": 37.48213577270508, "step": 305 }, { "epoch": 0.46258503401360546, "fcm_dpo/beta": 0.021479588001966476, "fcm_dpo/delta": -0.06405539065599442, "fcm_dpo/margin": 21.383968353271484, "fcm_dpo/q_t": 0.39899182319641113, "grad_norm": 10.03992748260498, "learning_rate": 3.267510740432719e-07, "logits/chosen": 1.418445348739624, "logits/rejected": 1.1668379306793213, "logps/chosen": -90.37960815429688, "logps/ref_chosen": -71.7261962890625, "logps/ref_rejected": -97.70491027832031, "logps/rejected": -137.74229431152344, "loss": 1.0879, "margin_dpo/margin_mean": 21.383968353271484, "margin_dpo/margin_std": 31.563940048217773, "step": 306 }, { "epoch": 0.46409674981103555, "fcm_dpo/beta": 0.02162957563996315, "fcm_dpo/delta": 0.02099861018359661, "fcm_dpo/margin": 13.357643127441406, "fcm_dpo/q_t": 0.4386698305606842, "grad_norm": 12.694254875183105, "learning_rate": 3.2549163976939285e-07, "logits/chosen": 1.2605947256088257, "logits/rejected": 1.1243689060211182, "logps/chosen": -88.7791976928711, "logps/ref_chosen": -74.38668823242188, "logps/ref_rejected": -84.16001892089844, "logps/rejected": -111.91016387939453, "loss": 1.2375, "margin_dpo/margin_mean": 13.357643127441406, "margin_dpo/margin_std": 33.23987579345703, "step": 307 }, { "epoch": 0.4656084656084656, "fcm_dpo/beta": 0.021374888718128204, "fcm_dpo/delta": -0.07016818970441818, "fcm_dpo/margin": 21.82101821899414, "fcm_dpo/q_t": 0.3989279270172119, "grad_norm": 7.945812225341797, "learning_rate": 3.2423009383206874e-07, "logits/chosen": 1.2982113361358643, "logits/rejected": 1.3503541946411133, "logps/chosen": -103.78570556640625, "logps/ref_chosen": -87.50894165039062, "logps/ref_rejected": -94.80848693847656, "logps/rejected": -132.90626525878906, "loss": 1.1067, "margin_dpo/margin_mean": 21.821016311645508, "margin_dpo/margin_std": 35.72882080078125, "step": 308 }, { "epoch": 0.4671201814058957, "fcm_dpo/beta": 0.02099529653787613, "fcm_dpo/delta": -0.06646728515625, "fcm_dpo/margin": 21.988813400268555, "fcm_dpo/q_t": 0.3982385993003845, "grad_norm": 8.893543243408203, "learning_rate": 3.229664715194511e-07, "logits/chosen": 1.3217096328735352, "logits/rejected": 1.2139837741851807, "logps/chosen": -104.08708190917969, "logps/ref_chosen": -82.15191650390625, "logps/ref_rejected": -95.03496551513672, "logps/rejected": -138.9589385986328, "loss": 1.075, "margin_dpo/margin_mean": 21.988813400268555, "margin_dpo/margin_std": 30.44322967529297, "step": 309 }, { "epoch": 0.46863189720332576, "fcm_dpo/beta": 0.02120891772210598, "fcm_dpo/delta": 0.02794772759079933, "fcm_dpo/margin": 10.018726348876953, "fcm_dpo/q_t": 0.455955445766449, "grad_norm": 11.014235496520996, "learning_rate": 3.2170080817777257e-07, "logits/chosen": 1.458807110786438, "logits/rejected": 1.4317870140075684, "logps/chosen": -115.71739196777344, "logps/ref_chosen": -93.7555160522461, "logps/ref_rejected": -96.93236541748047, "logps/rejected": -128.9129638671875, "loss": 1.3344, "margin_dpo/margin_mean": 10.018726348876953, "margin_dpo/margin_std": 37.276611328125, "step": 310 }, { "epoch": 0.47014361300075586, "fcm_dpo/beta": 0.02111116424202919, "fcm_dpo/delta": 0.006699252873659134, "fcm_dpo/margin": 18.61467742919922, "fcm_dpo/q_t": 0.41649070382118225, "grad_norm": 9.210880279541016, "learning_rate": 3.204331392103574e-07, "logits/chosen": 1.4640614986419678, "logits/rejected": 1.1159043312072754, "logps/chosen": -89.65441131591797, "logps/ref_chosen": -76.20762634277344, "logps/ref_rejected": -110.48141479492188, "logps/rejected": -142.54287719726562, "loss": 1.149, "margin_dpo/margin_mean": 18.61467933654785, "margin_dpo/margin_std": 34.18421936035156, "step": 311 }, { "epoch": 0.47165532879818595, "fcm_dpo/beta": 0.02078743278980255, "fcm_dpo/delta": -0.15529538691043854, "fcm_dpo/margin": 26.313379287719727, "fcm_dpo/q_t": 0.38429784774780273, "grad_norm": 7.8850884437561035, "learning_rate": 3.1916350007663176e-07, "logits/chosen": 1.4752094745635986, "logits/rejected": 1.304732084274292, "logps/chosen": -89.44306182861328, "logps/ref_chosen": -69.08878326416016, "logps/ref_rejected": -91.84494018554688, "logps/rejected": -138.51260375976562, "loss": 1.0482, "margin_dpo/margin_mean": 26.31338119506836, "margin_dpo/margin_std": 37.86506271362305, "step": 312 }, { "epoch": 0.47316704459561604, "fcm_dpo/beta": 0.020855270326137543, "fcm_dpo/delta": 0.07202167809009552, "fcm_dpo/margin": 15.828784942626953, "fcm_dpo/q_t": 0.4334028959274292, "grad_norm": 8.256758689880371, "learning_rate": 3.178919262911314e-07, "logits/chosen": 1.533963680267334, "logits/rejected": 1.4955235719680786, "logps/chosen": -92.3877944946289, "logps/ref_chosen": -78.20826721191406, "logps/ref_rejected": -86.90351867675781, "logps/rejected": -116.9118423461914, "loss": 1.2266, "margin_dpo/margin_mean": 15.828783988952637, "margin_dpo/margin_std": 38.44497299194336, "step": 313 }, { "epoch": 0.47467876039304613, "fcm_dpo/beta": 0.02034802734851837, "fcm_dpo/delta": -0.1469862312078476, "fcm_dpo/margin": 26.44600486755371, "fcm_dpo/q_t": 0.38667362928390503, "grad_norm": 11.084165573120117, "learning_rate": 3.166184534225087e-07, "logits/chosen": 1.629220724105835, "logits/rejected": 1.6717283725738525, "logps/chosen": -105.98693084716797, "logps/ref_chosen": -90.41890716552734, "logps/ref_rejected": -84.33525848388672, "logps/rejected": -126.34929656982422, "loss": 1.0556, "margin_dpo/margin_mean": 26.44600486755371, "margin_dpo/margin_std": 38.6795654296875, "step": 314 }, { "epoch": 0.47619047619047616, "fcm_dpo/beta": 0.020289920270442963, "fcm_dpo/delta": -0.03734702616930008, "fcm_dpo/margin": 21.451175689697266, "fcm_dpo/q_t": 0.4053088426589966, "grad_norm": 8.479659080505371, "learning_rate": 3.1534311709253723e-07, "logits/chosen": 1.165109395980835, "logits/rejected": 1.0913877487182617, "logps/chosen": -103.09860229492188, "logps/ref_chosen": -87.32842254638672, "logps/ref_rejected": -93.71661376953125, "logps/rejected": -130.93795776367188, "loss": 1.0996, "margin_dpo/margin_mean": 21.451175689697266, "margin_dpo/margin_std": 32.51549530029297, "step": 315 }, { "epoch": 0.47770219198790626, "fcm_dpo/beta": 0.020012233406305313, "fcm_dpo/delta": -0.13435091078281403, "fcm_dpo/margin": 26.249847412109375, "fcm_dpo/q_t": 0.38359588384628296, "grad_norm": 9.866154670715332, "learning_rate": 3.1406595297511564e-07, "logits/chosen": 1.4812779426574707, "logits/rejected": 1.1655080318450928, "logps/chosen": -93.73405456542969, "logps/ref_chosen": -73.898681640625, "logps/ref_rejected": -115.42668151855469, "logps/rejected": -161.5118865966797, "loss": 1.0329, "margin_dpo/margin_mean": 26.249849319458008, "margin_dpo/margin_std": 30.74091339111328, "step": 316 }, { "epoch": 0.47921390778533635, "fcm_dpo/beta": 0.019283965229988098, "fcm_dpo/delta": -0.07981756329536438, "fcm_dpo/margin": 24.6502628326416, "fcm_dpo/q_t": 0.3955802917480469, "grad_norm": 14.283778190612793, "learning_rate": 3.1278699679526975e-07, "logits/chosen": 1.5710258483886719, "logits/rejected": 1.4736764430999756, "logps/chosen": -92.30467224121094, "logps/ref_chosen": -75.42947387695312, "logps/ref_rejected": -90.60166931152344, "logps/rejected": -132.12713623046875, "loss": 1.1094, "margin_dpo/margin_mean": 24.650264739990234, "margin_dpo/margin_std": 41.600746154785156, "step": 317 }, { "epoch": 0.48072562358276644, "fcm_dpo/beta": 0.019192587584257126, "fcm_dpo/delta": -0.0218157060444355, "fcm_dpo/margin": 21.925718307495117, "fcm_dpo/q_t": 0.4134993255138397, "grad_norm": 8.415660858154297, "learning_rate": 3.1150628432815336e-07, "logits/chosen": 1.4273430109024048, "logits/rejected": 1.282718300819397, "logps/chosen": -89.98394012451172, "logps/ref_chosen": -70.38318634033203, "logps/ref_rejected": -98.19901275634766, "logps/rejected": -139.72549438476562, "loss": 1.166, "margin_dpo/margin_mean": 21.925718307495117, "margin_dpo/margin_std": 43.43846893310547, "step": 318 }, { "epoch": 0.48223733938019653, "fcm_dpo/beta": 0.018868155777454376, "fcm_dpo/delta": -0.10744694620370865, "fcm_dpo/margin": 26.597312927246094, "fcm_dpo/q_t": 0.39157384634017944, "grad_norm": 7.358407020568848, "learning_rate": 3.1022385139804707e-07, "logits/chosen": 1.4171019792556763, "logits/rejected": 1.3558650016784668, "logps/chosen": -106.76776885986328, "logps/ref_chosen": -83.40225982666016, "logps/ref_rejected": -95.40069580078125, "logps/rejected": -145.36351013183594, "loss": 1.0818, "margin_dpo/margin_mean": 26.597312927246094, "margin_dpo/margin_std": 41.52876281738281, "step": 319 }, { "epoch": 0.4837490551776266, "fcm_dpo/beta": 0.018746916204690933, "fcm_dpo/delta": -0.025277845561504364, "fcm_dpo/margin": 14.949155807495117, "fcm_dpo/q_t": 0.4393833875656128, "grad_norm": 8.69232177734375, "learning_rate": 3.0893973387735683e-07, "logits/chosen": 1.186877727508545, "logits/rejected": 1.0729196071624756, "logps/chosen": -90.08442687988281, "logps/ref_chosen": -68.70979309082031, "logps/ref_rejected": -87.00540924072266, "logps/rejected": -123.32920837402344, "loss": 1.2802, "margin_dpo/margin_mean": 14.949155807495117, "margin_dpo/margin_std": 43.0057487487793, "step": 320 }, { "epoch": 0.4852607709750567, "fcm_dpo/beta": 0.018360208719968796, "fcm_dpo/delta": -0.054547958076000214, "fcm_dpo/margin": 24.483469009399414, "fcm_dpo/q_t": 0.4054255485534668, "grad_norm": 18.02200698852539, "learning_rate": 3.0765396768561004e-07, "logits/chosen": 1.4728999137878418, "logits/rejected": 1.4148732423782349, "logps/chosen": -88.64881896972656, "logps/ref_chosen": -66.48135375976562, "logps/ref_rejected": -71.84545135498047, "logps/rejected": -118.49638366699219, "loss": 1.1465, "margin_dpo/margin_mean": 24.483470916748047, "margin_dpo/margin_std": 44.83539581298828, "step": 321 }, { "epoch": 0.48677248677248675, "fcm_dpo/beta": 0.017961975187063217, "fcm_dpo/delta": -0.21638159453868866, "fcm_dpo/margin": 33.61131286621094, "fcm_dpo/q_t": 0.36547616124153137, "grad_norm": 15.814164161682129, "learning_rate": 3.063665887884511e-07, "logits/chosen": 1.8528566360473633, "logits/rejected": 1.6500463485717773, "logps/chosen": -94.36614990234375, "logps/ref_chosen": -65.94654846191406, "logps/ref_rejected": -94.26603698730469, "logps/rejected": -156.29696655273438, "loss": 0.9901, "margin_dpo/margin_mean": 33.61131286621094, "margin_dpo/margin_std": 40.53282928466797, "step": 322 }, { "epoch": 0.48828420256991684, "fcm_dpo/beta": 0.017774982377886772, "fcm_dpo/delta": -0.011496573686599731, "fcm_dpo/margin": 23.097457885742188, "fcm_dpo/q_t": 0.4170437753200531, "grad_norm": 13.843807220458984, "learning_rate": 3.0507763319663517e-07, "logits/chosen": 1.5236470699310303, "logits/rejected": 1.389400601387024, "logps/chosen": -119.60983276367188, "logps/ref_chosen": -86.5498046875, "logps/ref_rejected": -110.39498901367188, "logps/rejected": -166.55245971679688, "loss": 1.2227, "margin_dpo/margin_mean": 23.097457885742188, "margin_dpo/margin_std": 54.00891876220703, "step": 323 }, { "epoch": 0.4897959183673469, "fcm_dpo/beta": 0.01737690530717373, "fcm_dpo/delta": -0.08369007706642151, "fcm_dpo/margin": 27.562347412109375, "fcm_dpo/q_t": 0.39454299211502075, "grad_norm": 15.58385944366455, "learning_rate": 3.0378713696502097e-07, "logits/chosen": 1.4405279159545898, "logits/rejected": 1.3061044216156006, "logps/chosen": -102.10205841064453, "logps/ref_chosen": -74.44218444824219, "logps/ref_rejected": -85.7646484375, "logps/rejected": -140.98687744140625, "loss": 1.1105, "margin_dpo/margin_mean": 27.562347412109375, "margin_dpo/margin_std": 46.17185592651367, "step": 324 }, { "epoch": 0.491307634164777, "fcm_dpo/beta": 0.016980335116386414, "fcm_dpo/delta": -0.09777640551328659, "fcm_dpo/margin": 28.89820098876953, "fcm_dpo/q_t": 0.394008994102478, "grad_norm": 13.200928688049316, "learning_rate": 3.0249513619156206e-07, "logits/chosen": 1.6928132772445679, "logits/rejected": 1.5178282260894775, "logps/chosen": -123.48445892333984, "logps/ref_chosen": -81.43812561035156, "logps/ref_rejected": -97.04302978515625, "logps/rejected": -167.987548828125, "loss": 1.1019, "margin_dpo/margin_mean": 28.898202896118164, "margin_dpo/margin_std": 46.94188690185547, "step": 325 }, { "epoch": 0.4928193499622071, "fcm_dpo/beta": 0.01719042658805847, "fcm_dpo/delta": 0.0699775367975235, "fcm_dpo/margin": 19.33672332763672, "fcm_dpo/q_t": 0.42833542823791504, "grad_norm": 8.703057289123535, "learning_rate": 3.012016670162977e-07, "logits/chosen": 1.2717374563217163, "logits/rejected": 1.2912553548812866, "logps/chosen": -141.61001586914062, "logps/ref_chosen": -91.65318298339844, "logps/ref_rejected": -90.64222717285156, "logps/rejected": -159.935791015625, "loss": 1.2212, "margin_dpo/margin_mean": 19.33672332763672, "margin_dpo/margin_std": 45.26724624633789, "step": 326 }, { "epoch": 0.4943310657596372, "fcm_dpo/beta": 0.017470695078372955, "fcm_dpo/delta": 0.037908885627985, "fcm_dpo/margin": 20.763715744018555, "fcm_dpo/q_t": 0.42454928159713745, "grad_norm": 28.22562026977539, "learning_rate": 2.99906765620341e-07, "logits/chosen": 1.1321663856506348, "logits/rejected": 1.0784250497817993, "logps/chosen": -141.81544494628906, "logps/ref_chosen": -89.97216796875, "logps/ref_rejected": -97.54869079589844, "logps/rejected": -170.15567016601562, "loss": 1.2403, "margin_dpo/margin_mean": 20.763715744018555, "margin_dpo/margin_std": 51.22842025756836, "step": 327 }, { "epoch": 0.4958427815570673, "fcm_dpo/beta": 0.01727999374270439, "fcm_dpo/delta": -0.0346628800034523, "fcm_dpo/margin": 25.054723739624023, "fcm_dpo/q_t": 0.4068809151649475, "grad_norm": 8.079117774963379, "learning_rate": 2.9861046822486766e-07, "logits/chosen": 1.3686432838439941, "logits/rejected": 1.2675422430038452, "logps/chosen": -128.63833618164062, "logps/ref_chosen": -80.27335357666016, "logps/ref_rejected": -99.04093170166016, "logps/rejected": -172.46063232421875, "loss": 1.1454, "margin_dpo/margin_mean": 25.054725646972656, "margin_dpo/margin_std": 45.77302551269531, "step": 328 }, { "epoch": 0.4973544973544973, "fcm_dpo/beta": 0.016885770484805107, "fcm_dpo/delta": -0.12400930374860764, "fcm_dpo/margin": 30.581024169921875, "fcm_dpo/q_t": 0.39069896936416626, "grad_norm": 9.308547973632812, "learning_rate": 2.9731281109010253e-07, "logits/chosen": 1.4687421321868896, "logits/rejected": 1.3343288898468018, "logps/chosen": -128.83963012695312, "logps/ref_chosen": -79.75892639160156, "logps/ref_rejected": -102.06265258789062, "logps/rejected": -181.72439575195312, "loss": 1.0682, "margin_dpo/margin_mean": 30.581024169921875, "margin_dpo/margin_std": 45.86301040649414, "step": 329 }, { "epoch": 0.4988662131519274, "fcm_dpo/beta": 0.016510095447301865, "fcm_dpo/delta": -0.20744779706001282, "fcm_dpo/margin": 36.0556526184082, "fcm_dpo/q_t": 0.37239986658096313, "grad_norm": 10.562420845031738, "learning_rate": 2.9601383051430505e-07, "logits/chosen": 1.605103850364685, "logits/rejected": 1.4531974792480469, "logps/chosen": -114.76742553710938, "logps/ref_chosen": -70.55734252929688, "logps/ref_rejected": -94.53077697753906, "logps/rejected": -174.7965087890625, "loss": 1.0623, "margin_dpo/margin_mean": 36.05565643310547, "margin_dpo/margin_std": 53.4388427734375, "step": 330 }, { "epoch": 0.5003779289493575, "fcm_dpo/beta": 0.015694059431552887, "fcm_dpo/delta": -0.1913868486881256, "fcm_dpo/margin": 36.94060516357422, "fcm_dpo/q_t": 0.3752760589122772, "grad_norm": 6.33163595199585, "learning_rate": 2.947135628327544e-07, "logits/chosen": 2.0764613151550293, "logits/rejected": 2.02164888381958, "logps/chosen": -129.00782775878906, "logps/ref_chosen": -75.46063232421875, "logps/ref_rejected": -84.78495788574219, "logps/rejected": -175.27273559570312, "loss": 1.0412, "margin_dpo/margin_mean": 36.94060516357422, "margin_dpo/margin_std": 53.34477996826172, "step": 331 }, { "epoch": 0.5018896447467877, "fcm_dpo/beta": 0.015376402996480465, "fcm_dpo/delta": -0.1497417539358139, "fcm_dpo/margin": 35.20355224609375, "fcm_dpo/q_t": 0.3825414180755615, "grad_norm": 9.974367141723633, "learning_rate": 2.934120444167326e-07, "logits/chosen": 1.1533145904541016, "logits/rejected": 1.0612804889678955, "logps/chosen": -136.8844757080078, "logps/ref_chosen": -84.32807922363281, "logps/ref_rejected": -95.63302612304688, "logps/rejected": -183.39297485351562, "loss": 1.0329, "margin_dpo/margin_mean": 35.20355224609375, "margin_dpo/margin_std": 45.5770263671875, "step": 332 }, { "epoch": 0.5034013605442177, "fcm_dpo/beta": 0.014777831733226776, "fcm_dpo/delta": -0.17367474734783173, "fcm_dpo/margin": 38.178375244140625, "fcm_dpo/q_t": 0.37468940019607544, "grad_norm": 8.577049255371094, "learning_rate": 2.921093116725076e-07, "logits/chosen": 1.7649075984954834, "logits/rejected": 1.598151683807373, "logps/chosen": -137.31753540039062, "logps/ref_chosen": -78.2132339477539, "logps/ref_rejected": -103.82716369628906, "logps/rejected": -201.10983276367188, "loss": 0.9973, "margin_dpo/margin_mean": 38.178375244140625, "margin_dpo/margin_std": 44.62627410888672, "step": 333 }, { "epoch": 0.5049130763416477, "fcm_dpo/beta": 0.014649095013737679, "fcm_dpo/delta": 0.020269624888896942, "fcm_dpo/margin": 25.97400665283203, "fcm_dpo/q_t": 0.4182738661766052, "grad_norm": 9.703882217407227, "learning_rate": 2.9080540104031484e-07, "logits/chosen": 1.5542503595352173, "logits/rejected": 1.3984897136688232, "logps/chosen": -144.12342834472656, "logps/ref_chosen": -85.0171127319336, "logps/ref_rejected": -106.79039764404297, "logps/rejected": -191.8707275390625, "loss": 1.2143, "margin_dpo/margin_mean": 25.9740047454834, "margin_dpo/margin_std": 59.42298889160156, "step": 334 }, { "epoch": 0.5064247921390779, "fcm_dpo/beta": 0.014591803774237633, "fcm_dpo/delta": -0.04717499762773514, "fcm_dpo/margin": 25.18863296508789, "fcm_dpo/q_t": 0.42526859045028687, "grad_norm": 14.016016960144043, "learning_rate": 2.895003489933375e-07, "logits/chosen": 1.5459372997283936, "logits/rejected": 1.444729208946228, "logps/chosen": -143.58782958984375, "logps/ref_chosen": -78.56513214111328, "logps/ref_rejected": -92.68515014648438, "logps/rejected": -182.89646911621094, "loss": 1.2477, "margin_dpo/margin_mean": 25.188629150390625, "margin_dpo/margin_std": 63.48894500732422, "step": 335 }, { "epoch": 0.5079365079365079, "fcm_dpo/beta": 0.014425481669604778, "fcm_dpo/delta": -0.014608364552259445, "fcm_dpo/margin": 28.666763305664062, "fcm_dpo/q_t": 0.40841758251190186, "grad_norm": 15.602229118347168, "learning_rate": 2.8819419203668675e-07, "logits/chosen": 1.321707010269165, "logits/rejected": 1.2782776355743408, "logps/chosen": -163.07656860351562, "logps/ref_chosen": -88.63243103027344, "logps/ref_rejected": -107.89385986328125, "logps/rejected": -211.0047607421875, "loss": 1.1316, "margin_dpo/margin_mean": 28.666763305664062, "margin_dpo/margin_std": 49.40132141113281, "step": 336 }, { "epoch": 0.509448223733938, "fcm_dpo/beta": 0.01457132212817669, "fcm_dpo/delta": 0.03601706027984619, "fcm_dpo/margin": 25.066566467285156, "fcm_dpo/q_t": 0.4219823479652405, "grad_norm": 11.307106971740723, "learning_rate": 2.8688696670638053e-07, "logits/chosen": 1.351456642150879, "logits/rejected": 1.244319200515747, "logps/chosen": -164.80145263671875, "logps/ref_chosen": -93.25018310546875, "logps/ref_rejected": -103.8592529296875, "logps/rejected": -200.4770965576172, "loss": 1.1949, "margin_dpo/margin_mean": 25.066566467285156, "margin_dpo/margin_std": 53.922340393066406, "step": 337 }, { "epoch": 0.5109599395313681, "fcm_dpo/beta": 0.014669684693217278, "fcm_dpo/delta": 0.039764419198036194, "fcm_dpo/margin": 24.654888153076172, "fcm_dpo/q_t": 0.42528173327445984, "grad_norm": 9.578775405883789, "learning_rate": 2.8557870956832133e-07, "logits/chosen": 1.6919002532958984, "logits/rejected": 1.6292130947113037, "logps/chosen": -154.7214813232422, "logps/ref_chosen": -81.79462432861328, "logps/ref_rejected": -90.98942565917969, "logps/rejected": -188.5711669921875, "loss": 1.1853, "margin_dpo/margin_mean": 24.654888153076172, "margin_dpo/margin_std": 51.62914276123047, "step": 338 }, { "epoch": 0.5124716553287982, "fcm_dpo/beta": 0.014321266673505306, "fcm_dpo/delta": -0.1600484699010849, "fcm_dpo/margin": 38.45646286010742, "fcm_dpo/q_t": 0.3788467347621918, "grad_norm": 8.566285133361816, "learning_rate": 2.842694572172736e-07, "logits/chosen": 1.7731349468231201, "logits/rejected": 1.559950590133667, "logps/chosen": -119.42196655273438, "logps/ref_chosen": -61.80355453491211, "logps/ref_rejected": -85.16979217529297, "logps/rejected": -181.24465942382812, "loss": 1.0265, "margin_dpo/margin_mean": 38.45646667480469, "margin_dpo/margin_std": 50.06471252441406, "step": 339 }, { "epoch": 0.5139833711262283, "fcm_dpo/beta": 0.014187190681695938, "fcm_dpo/delta": -0.06460127234458923, "fcm_dpo/margin": 32.51493453979492, "fcm_dpo/q_t": 0.40349990129470825, "grad_norm": 11.992464065551758, "learning_rate": 2.8295924627584004e-07, "logits/chosen": 1.505007028579712, "logits/rejected": 1.4854414463043213, "logps/chosen": -140.9080810546875, "logps/ref_chosen": -72.486083984375, "logps/ref_rejected": -79.86129760742188, "logps/rejected": -180.79823303222656, "loss": 1.1434, "margin_dpo/margin_mean": 32.51493453979492, "margin_dpo/margin_std": 60.210792541503906, "step": 340 }, { "epoch": 0.5154950869236583, "fcm_dpo/beta": 0.013437741436064243, "fcm_dpo/delta": -0.22484445571899414, "fcm_dpo/margin": 37.80937576293945, "fcm_dpo/q_t": 0.3881411552429199, "grad_norm": 16.43123435974121, "learning_rate": 2.816481133934373e-07, "logits/chosen": 1.8760827779769897, "logits/rejected": 1.7105937004089355, "logps/chosen": -145.66722106933594, "logps/ref_chosen": -77.36830139160156, "logps/ref_rejected": -94.64933013916016, "logps/rejected": -200.75762939453125, "loss": 1.0886, "margin_dpo/margin_mean": 37.80937576293945, "margin_dpo/margin_std": 55.31840133666992, "step": 341 }, { "epoch": 0.5170068027210885, "fcm_dpo/beta": 0.013311228714883327, "fcm_dpo/delta": -0.059115879237651825, "fcm_dpo/margin": 34.28473663330078, "fcm_dpo/q_t": 0.40271270275115967, "grad_norm": 7.672739028930664, "learning_rate": 2.8033609524527046e-07, "logits/chosen": 1.5947625637054443, "logits/rejected": 1.496432900428772, "logps/chosen": -133.24288940429688, "logps/ref_chosen": -71.00831604003906, "logps/ref_rejected": -84.22953796386719, "logps/rejected": -180.74884033203125, "loss": 1.1126, "margin_dpo/margin_mean": 34.284732818603516, "margin_dpo/margin_std": 57.40011978149414, "step": 342 }, { "epoch": 0.5185185185185185, "fcm_dpo/beta": 0.01319624949246645, "fcm_dpo/delta": -0.03155739977955818, "fcm_dpo/margin": 21.631832122802734, "fcm_dpo/q_t": 0.4371912479400635, "grad_norm": 8.965015411376953, "learning_rate": 2.7902322853130753e-07, "logits/chosen": 1.3000271320343018, "logits/rejected": 1.2619600296020508, "logps/chosen": -152.49551391601562, "logps/ref_chosen": -91.44624328613281, "logps/ref_rejected": -99.06044006347656, "logps/rejected": -181.74154663085938, "loss": 1.245, "margin_dpo/margin_mean": 21.631834030151367, "margin_dpo/margin_std": 53.96006774902344, "step": 343 }, { "epoch": 0.5200302343159486, "fcm_dpo/beta": 0.013017518445849419, "fcm_dpo/delta": -0.04310518503189087, "fcm_dpo/margin": 33.79201126098633, "fcm_dpo/q_t": 0.4027559459209442, "grad_norm": 8.789457321166992, "learning_rate": 2.7770954997525274e-07, "logits/chosen": 1.9475184679031372, "logits/rejected": 1.7549506425857544, "logps/chosen": -145.70272827148438, "logps/ref_chosen": -73.43608093261719, "logps/ref_rejected": -100.76569366455078, "logps/rejected": -206.82437133789062, "loss": 1.1134, "margin_dpo/margin_mean": 33.79201126098633, "margin_dpo/margin_std": 54.33518600463867, "step": 344 }, { "epoch": 0.5215419501133787, "fcm_dpo/beta": 0.013107789680361748, "fcm_dpo/delta": 0.009255402721464634, "fcm_dpo/margin": 29.836973190307617, "fcm_dpo/q_t": 0.4151036739349365, "grad_norm": 8.21016788482666, "learning_rate": 2.7639509632351927e-07, "logits/chosen": 1.7126402854919434, "logits/rejected": 1.5933332443237305, "logps/chosen": -130.25466918945312, "logps/ref_chosen": -75.79296875, "logps/ref_rejected": -94.34156799316406, "logps/rejected": -178.64024353027344, "loss": 1.1678, "margin_dpo/margin_mean": 29.83697509765625, "margin_dpo/margin_std": 58.75901794433594, "step": 345 }, { "epoch": 0.5230536659108088, "fcm_dpo/beta": 0.013089219108223915, "fcm_dpo/delta": -0.04255472123622894, "fcm_dpo/margin": 33.64637756347656, "fcm_dpo/q_t": 0.4048069715499878, "grad_norm": 9.588007926940918, "learning_rate": 2.7507990434420123e-07, "logits/chosen": 1.4854192733764648, "logits/rejected": 1.303803563117981, "logps/chosen": -130.6438446044922, "logps/ref_chosen": -72.26289367675781, "logps/ref_rejected": -106.36925506591797, "logps/rejected": -198.39659118652344, "loss": 1.1534, "margin_dpo/margin_mean": 33.6463737487793, "margin_dpo/margin_std": 63.119842529296875, "step": 346 }, { "epoch": 0.5245653817082389, "fcm_dpo/beta": 0.013037774711847305, "fcm_dpo/delta": 0.005431188270449638, "fcm_dpo/margin": 30.270421981811523, "fcm_dpo/q_t": 0.41649848222732544, "grad_norm": 10.595651626586914, "learning_rate": 2.737640108260456e-07, "logits/chosen": 1.8637452125549316, "logits/rejected": 1.750800609588623, "logps/chosen": -138.73822021484375, "logps/ref_chosen": -71.19871520996094, "logps/ref_rejected": -91.543212890625, "logps/rejected": -189.35311889648438, "loss": 1.1634, "margin_dpo/margin_mean": 30.270421981811523, "margin_dpo/margin_std": 58.804443359375, "step": 347 }, { "epoch": 0.5260770975056689, "fcm_dpo/beta": 0.012842783704400063, "fcm_dpo/delta": -0.06106501445174217, "fcm_dpo/margin": 35.64274597167969, "fcm_dpo/q_t": 0.4025927186012268, "grad_norm": 8.013747215270996, "learning_rate": 2.724474525774229e-07, "logits/chosen": 1.8773822784423828, "logits/rejected": 1.792330265045166, "logps/chosen": -130.76272583007812, "logps/ref_chosen": -69.95603942871094, "logps/ref_rejected": -83.64309692382812, "logps/rejected": -180.092529296875, "loss": 1.1057, "margin_dpo/margin_mean": 35.64274215698242, "margin_dpo/margin_std": 58.040496826171875, "step": 348 }, { "epoch": 0.527588813303099, "fcm_dpo/beta": 0.012860096991062164, "fcm_dpo/delta": -0.034538384526968, "fcm_dpo/margin": 33.63842010498047, "fcm_dpo/q_t": 0.4073810279369354, "grad_norm": 8.36850643157959, "learning_rate": 2.711302664252973e-07, "logits/chosen": 1.7388920783996582, "logits/rejected": 1.5637342929840088, "logps/chosen": -130.26393127441406, "logps/ref_chosen": -70.71857452392578, "logps/ref_rejected": -99.93263244628906, "logps/rejected": -193.11642456054688, "loss": 1.1455, "margin_dpo/margin_mean": 33.63842010498047, "margin_dpo/margin_std": 61.37939453125, "step": 349 }, { "epoch": 0.5291005291005291, "fcm_dpo/beta": 0.012324045412242413, "fcm_dpo/delta": -0.20469069480895996, "fcm_dpo/margin": 47.97602081298828, "fcm_dpo/q_t": 0.3688136339187622, "grad_norm": 8.611337661743164, "learning_rate": 2.698124892141971e-07, "logits/chosen": 1.7097694873809814, "logits/rejected": 1.5507686138153076, "logps/chosen": -139.9197540283203, "logps/ref_chosen": -78.16873168945312, "logps/ref_rejected": -104.84308624267578, "logps/rejected": -214.57012939453125, "loss": 0.9812, "margin_dpo/margin_mean": 47.976016998291016, "margin_dpo/margin_std": 54.40205383300781, "step": 350 }, { "epoch": 0.5306122448979592, "fcm_dpo/beta": 0.012206001207232475, "fcm_dpo/delta": -0.014462406747043133, "fcm_dpo/margin": 33.90068817138672, "fcm_dpo/q_t": 0.40708887577056885, "grad_norm": 9.900226593017578, "learning_rate": 2.6849415780518357e-07, "logits/chosen": 1.5461342334747314, "logits/rejected": 1.3549315929412842, "logps/chosen": -138.26388549804688, "logps/ref_chosen": -71.79151916503906, "logps/ref_rejected": -97.04634094238281, "logps/rejected": -197.41940307617188, "loss": 1.1573, "margin_dpo/margin_mean": 33.90068817138672, "margin_dpo/margin_std": 63.30729293823242, "step": 351 }, { "epoch": 0.5321239606953893, "fcm_dpo/beta": 0.012208282947540283, "fcm_dpo/delta": 0.02938525378704071, "fcm_dpo/margin": 30.417400360107422, "fcm_dpo/q_t": 0.4205033779144287, "grad_norm": 8.994142532348633, "learning_rate": 2.6717530907482024e-07, "logits/chosen": 1.763619065284729, "logits/rejected": 1.622790813446045, "logps/chosen": -151.24343872070312, "logps/ref_chosen": -80.86544799804688, "logps/ref_rejected": -102.02129364013672, "logps/rejected": -202.81666564941406, "loss": 1.1736, "margin_dpo/margin_mean": 30.417404174804688, "margin_dpo/margin_std": 60.021575927734375, "step": 352 }, { "epoch": 0.5336356764928194, "fcm_dpo/beta": 0.012203685939311981, "fcm_dpo/delta": -0.062111545354127884, "fcm_dpo/margin": 37.6354866027832, "fcm_dpo/q_t": 0.39819955825805664, "grad_norm": 8.61074447631836, "learning_rate": 2.658559799141411e-07, "logits/chosen": 1.6032516956329346, "logits/rejected": 1.6155762672424316, "logps/chosen": -138.62901306152344, "logps/ref_chosen": -84.77235412597656, "logps/ref_rejected": -86.77130889892578, "logps/rejected": -178.26345825195312, "loss": 1.1042, "margin_dpo/margin_mean": 37.6354866027832, "margin_dpo/margin_std": 60.413963317871094, "step": 353 }, { "epoch": 0.5351473922902494, "fcm_dpo/beta": 0.01194113027304411, "fcm_dpo/delta": -0.08657968789339066, "fcm_dpo/margin": 40.35693359375, "fcm_dpo/q_t": 0.39217013120651245, "grad_norm": 7.400528907775879, "learning_rate": 2.6453620722761895e-07, "logits/chosen": 1.9535037279129028, "logits/rejected": 1.5858149528503418, "logps/chosen": -109.94905090332031, "logps/ref_chosen": -54.33562088012695, "logps/ref_rejected": -92.4120101928711, "logps/rejected": -188.38235473632812, "loss": 1.0929, "margin_dpo/margin_mean": 40.35693359375, "margin_dpo/margin_std": 63.149314880371094, "step": 354 }, { "epoch": 0.5366591080876795, "fcm_dpo/beta": 0.011824669316411018, "fcm_dpo/delta": -0.07782451808452606, "fcm_dpo/margin": 40.10158920288086, "fcm_dpo/q_t": 0.3945278525352478, "grad_norm": 6.816810131072998, "learning_rate": 2.632160279321328e-07, "logits/chosen": 1.802927017211914, "logits/rejected": 1.4767255783081055, "logps/chosen": -124.52835845947266, "logps/ref_chosen": -61.8388671875, "logps/ref_rejected": -98.65571594238281, "logps/rejected": -201.44680786132812, "loss": 1.0844, "margin_dpo/margin_mean": 40.10158920288086, "margin_dpo/margin_std": 59.98944854736328, "step": 355 }, { "epoch": 0.5381708238851096, "fcm_dpo/beta": 0.011599002406001091, "fcm_dpo/delta": -0.050096526741981506, "fcm_dpo/margin": 38.54914474487305, "fcm_dpo/q_t": 0.4025239646434784, "grad_norm": 9.990870475769043, "learning_rate": 2.618954789559356e-07, "logits/chosen": 1.7717761993408203, "logits/rejected": 1.5829424858093262, "logps/chosen": -113.30570983886719, "logps/ref_chosen": -63.92546463012695, "logps/ref_rejected": -89.682861328125, "logps/rejected": -177.61224365234375, "loss": 1.15, "margin_dpo/margin_mean": 38.54914474487305, "margin_dpo/margin_std": 72.21054077148438, "step": 356 }, { "epoch": 0.5396825396825397, "fcm_dpo/beta": 0.011380214244127274, "fcm_dpo/delta": -0.1073249951004982, "fcm_dpo/margin": 33.18156051635742, "fcm_dpo/q_t": 0.4151083827018738, "grad_norm": 7.118601322174072, "learning_rate": 2.6057459723762076e-07, "logits/chosen": 1.7753883600234985, "logits/rejected": 1.6539323329925537, "logps/chosen": -143.6602783203125, "logps/ref_chosen": -81.07588958740234, "logps/ref_rejected": -85.06967163085938, "logps/rejected": -180.83560180664062, "loss": 1.1537, "margin_dpo/margin_mean": 33.181556701660156, "margin_dpo/margin_std": 57.43933868408203, "step": 357 }, { "epoch": 0.5411942554799698, "fcm_dpo/beta": 0.011224227026104927, "fcm_dpo/delta": -0.10705891996622086, "fcm_dpo/margin": 44.70811462402344, "fcm_dpo/q_t": 0.3878515660762787, "grad_norm": 8.951573371887207, "learning_rate": 2.5925341972508954e-07, "logits/chosen": 1.4298584461212158, "logits/rejected": 1.4541399478912354, "logps/chosen": -139.0441436767578, "logps/ref_chosen": -84.09109497070312, "logps/ref_rejected": -85.07244873046875, "logps/rejected": -184.73361206054688, "loss": 1.0554, "margin_dpo/margin_mean": 44.70811462402344, "margin_dpo/margin_std": 61.42544937133789, "step": 358 }, { "epoch": 0.5427059712773998, "fcm_dpo/beta": 0.011128641664981842, "fcm_dpo/delta": -0.009427734650671482, "fcm_dpo/margin": 25.97496795654297, "fcm_dpo/q_t": 0.43627357482910156, "grad_norm": 7.67764949798584, "learning_rate": 2.579319833745169e-07, "logits/chosen": 1.1958760023117065, "logits/rejected": 1.1496171951293945, "logps/chosen": -139.93809509277344, "logps/ref_chosen": -80.7490234375, "logps/ref_rejected": -94.92911529541016, "logps/rejected": -180.09317016601562, "loss": 1.2185, "margin_dpo/margin_mean": 25.97496795654297, "margin_dpo/margin_std": 57.3719596862793, "step": 359 }, { "epoch": 0.54421768707483, "fcm_dpo/beta": 0.011084508150815964, "fcm_dpo/delta": -0.0067586712539196014, "fcm_dpo/margin": 36.662925720214844, "fcm_dpo/q_t": 0.41188788414001465, "grad_norm": 9.444145202636719, "learning_rate": 2.5661032514931834e-07, "logits/chosen": 1.6899638175964355, "logits/rejected": 1.4466266632080078, "logps/chosen": -139.62124633789062, "logps/ref_chosen": -78.38681030273438, "logps/ref_rejected": -109.68933868408203, "logps/rejected": -207.58673095703125, "loss": 1.1254, "margin_dpo/margin_mean": 36.662925720214844, "margin_dpo/margin_std": 61.32177734375, "step": 360 }, { "epoch": 0.54572940287226, "fcm_dpo/beta": 0.010927680879831314, "fcm_dpo/delta": -0.11040109395980835, "fcm_dpo/margin": 46.201229095458984, "fcm_dpo/q_t": 0.3868357539176941, "grad_norm": 10.33779525756836, "learning_rate": 2.552884820191154e-07, "logits/chosen": 1.3621007204055786, "logits/rejected": 1.2309937477111816, "logps/chosen": -130.75283813476562, "logps/ref_chosen": -73.9055404663086, "logps/ref_rejected": -89.8489990234375, "logps/rejected": -192.89752197265625, "loss": 1.0453, "margin_dpo/margin_mean": 46.201229095458984, "margin_dpo/margin_std": 61.137569427490234, "step": 361 }, { "epoch": 0.54724111866969, "fcm_dpo/beta": 0.010783391073346138, "fcm_dpo/delta": -0.07165468484163284, "fcm_dpo/margin": 32.059444427490234, "fcm_dpo/q_t": 0.4229075610637665, "grad_norm": 7.126445293426514, "learning_rate": 2.53966490958702e-07, "logits/chosen": 1.6053223609924316, "logits/rejected": 1.2717453241348267, "logps/chosen": -150.54673767089844, "logps/ref_chosen": -82.32565307617188, "logps/ref_rejected": -123.14100646972656, "logps/rejected": -223.42153930664062, "loss": 1.1834, "margin_dpo/margin_mean": 32.059444427490234, "margin_dpo/margin_std": 62.51380920410156, "step": 362 }, { "epoch": 0.5487528344671202, "fcm_dpo/beta": 0.010588840581476688, "fcm_dpo/delta": -0.07053353637456894, "fcm_dpo/margin": 44.12169647216797, "fcm_dpo/q_t": 0.39325785636901855, "grad_norm": 9.556055068969727, "learning_rate": 2.526443889470099e-07, "logits/chosen": 1.9458461999893188, "logits/rejected": 1.5623431205749512, "logps/chosen": -133.33316040039062, "logps/ref_chosen": -66.05493927001953, "logps/ref_rejected": -106.79598999023438, "logps/rejected": -218.1959228515625, "loss": 1.0552, "margin_dpo/margin_mean": 44.12169647216797, "margin_dpo/margin_std": 56.65771484375, "step": 363 }, { "epoch": 0.5502645502645502, "fcm_dpo/beta": 0.010463178157806396, "fcm_dpo/delta": -0.044942498207092285, "fcm_dpo/margin": 42.31687545776367, "fcm_dpo/q_t": 0.40370985865592957, "grad_norm": 6.737151622772217, "learning_rate": 2.513222129660744e-07, "logits/chosen": 1.4676833152770996, "logits/rejected": 1.313372254371643, "logps/chosen": -137.15481567382812, "logps/ref_chosen": -76.38365173339844, "logps/ref_rejected": -100.22221374511719, "logps/rejected": -203.31024169921875, "loss": 1.126, "margin_dpo/margin_mean": 42.31687927246094, "margin_dpo/margin_std": 73.18367004394531, "step": 364 }, { "epoch": 0.5517762660619804, "fcm_dpo/beta": 0.010423636063933372, "fcm_dpo/delta": -0.021412841975688934, "fcm_dpo/margin": 40.340911865234375, "fcm_dpo/q_t": 0.403866708278656, "grad_norm": 9.5303955078125, "learning_rate": 2.5e-07, "logits/chosen": 1.8161044120788574, "logits/rejected": 1.81839120388031, "logps/chosen": -129.7283172607422, "logps/ref_chosen": -81.83399963378906, "logps/ref_rejected": -89.06932830810547, "logps/rejected": -177.3045654296875, "loss": 1.0826, "margin_dpo/margin_mean": 40.340911865234375, "margin_dpo/margin_std": 52.74284744262695, "step": 365 }, { "epoch": 0.5532879818594104, "fcm_dpo/beta": 0.010475091636180878, "fcm_dpo/delta": 0.034233205020427704, "fcm_dpo/margin": 35.035865783691406, "fcm_dpo/q_t": 0.41999638080596924, "grad_norm": 7.062388896942139, "learning_rate": 2.486777870339255e-07, "logits/chosen": 1.525472640991211, "logits/rejected": 1.503098964691162, "logps/chosen": -128.90716552734375, "logps/ref_chosen": -72.03398895263672, "logps/ref_rejected": -83.65354919433594, "logps/rejected": -175.56259155273438, "loss": 1.1736, "margin_dpo/margin_mean": 35.035865783691406, "margin_dpo/margin_std": 69.0810546875, "step": 366 }, { "epoch": 0.5547996976568406, "fcm_dpo/beta": 0.010566307231783867, "fcm_dpo/delta": 0.06605655699968338, "fcm_dpo/margin": 31.818260192871094, "fcm_dpo/q_t": 0.4226842224597931, "grad_norm": 7.502590656280518, "learning_rate": 2.4735561105299014e-07, "logits/chosen": 1.4546136856079102, "logits/rejected": 1.2036893367767334, "logps/chosen": -138.4822540283203, "logps/ref_chosen": -72.39828491210938, "logps/ref_rejected": -95.58364868164062, "logps/rejected": -193.4858856201172, "loss": 1.1717, "margin_dpo/margin_mean": 31.818260192871094, "margin_dpo/margin_std": 59.4080810546875, "step": 367 }, { "epoch": 0.5563114134542706, "fcm_dpo/beta": 0.010582388378679752, "fcm_dpo/delta": -0.018604513257741928, "fcm_dpo/margin": 39.4763298034668, "fcm_dpo/q_t": 0.40766745805740356, "grad_norm": 7.9668803215026855, "learning_rate": 2.46033509041298e-07, "logits/chosen": 1.2730112075805664, "logits/rejected": 1.2791517972946167, "logps/chosen": -163.7975311279297, "logps/ref_chosen": -90.12812042236328, "logps/ref_rejected": -91.6636962890625, "logps/rejected": -204.8094482421875, "loss": 1.1212, "margin_dpo/margin_mean": 39.47633361816406, "margin_dpo/margin_std": 65.0750732421875, "step": 368 }, { "epoch": 0.5578231292517006, "fcm_dpo/beta": 0.010618243366479874, "fcm_dpo/delta": 0.027943773195147514, "fcm_dpo/margin": 26.496131896972656, "fcm_dpo/q_t": 0.4377315640449524, "grad_norm": 9.919585227966309, "learning_rate": 2.447115179808846e-07, "logits/chosen": 1.5244429111480713, "logits/rejected": 1.434938669204712, "logps/chosen": -138.58261108398438, "logps/ref_chosen": -71.29417419433594, "logps/ref_rejected": -99.03875732421875, "logps/rejected": -192.82333374023438, "loss": 1.2296, "margin_dpo/margin_mean": 26.49612808227539, "margin_dpo/margin_std": 62.91883850097656, "step": 369 }, { "epoch": 0.5593348450491308, "fcm_dpo/beta": 0.010387836024165154, "fcm_dpo/delta": -0.14537188410758972, "fcm_dpo/margin": 51.689735412597656, "fcm_dpo/q_t": 0.3806115984916687, "grad_norm": 8.526147842407227, "learning_rate": 2.4338967485068164e-07, "logits/chosen": 1.9894087314605713, "logits/rejected": 1.8496686220169067, "logps/chosen": -132.82601928710938, "logps/ref_chosen": -69.14627075195312, "logps/ref_rejected": -93.58651733398438, "logps/rejected": -208.95599365234375, "loss": 1.0493, "margin_dpo/margin_mean": 51.689735412597656, "margin_dpo/margin_std": 72.51553344726562, "step": 370 }, { "epoch": 0.5608465608465608, "fcm_dpo/beta": 0.010324855335056782, "fcm_dpo/delta": -0.04348411411046982, "fcm_dpo/margin": 42.744075775146484, "fcm_dpo/q_t": 0.40137621760368347, "grad_norm": 15.825779914855957, "learning_rate": 2.420680166254831e-07, "logits/chosen": 2.0278491973876953, "logits/rejected": 1.9837315082550049, "logps/chosen": -133.10084533691406, "logps/ref_chosen": -65.76728820800781, "logps/ref_rejected": -79.9320068359375, "logps/rejected": -190.0096435546875, "loss": 1.115, "margin_dpo/margin_mean": 42.74407958984375, "margin_dpo/margin_std": 69.81803894042969, "step": 371 }, { "epoch": 0.562358276643991, "fcm_dpo/beta": 0.010167205706238747, "fcm_dpo/delta": -0.08461183309555054, "fcm_dpo/margin": 34.201377868652344, "fcm_dpo/q_t": 0.4236924350261688, "grad_norm": 11.217028617858887, "learning_rate": 2.4074658027491044e-07, "logits/chosen": 1.6819504499435425, "logits/rejected": 1.464050054550171, "logps/chosen": -131.26275634765625, "logps/ref_chosen": -69.97252655029297, "logps/ref_rejected": -92.38316345214844, "logps/rejected": -187.87478637695312, "loss": 1.2092, "margin_dpo/margin_mean": 34.201377868652344, "margin_dpo/margin_std": 73.22529602050781, "step": 372 }, { "epoch": 0.563869992441421, "fcm_dpo/beta": 0.01011381484568119, "fcm_dpo/delta": 0.01528643537312746, "fcm_dpo/margin": 38.09349822998047, "fcm_dpo/q_t": 0.41596657037734985, "grad_norm": 6.146734237670898, "learning_rate": 2.394254027623792e-07, "logits/chosen": 1.6164660453796387, "logits/rejected": 1.3731865882873535, "logps/chosen": -157.2783203125, "logps/ref_chosen": -79.34700012207031, "logps/ref_rejected": -95.69737243652344, "logps/rejected": -211.72218322753906, "loss": 1.1663, "margin_dpo/margin_mean": 38.093502044677734, "margin_dpo/margin_std": 73.8922119140625, "step": 373 }, { "epoch": 0.5653817082388511, "fcm_dpo/beta": 0.0098224813118577, "fcm_dpo/delta": -0.1968994438648224, "fcm_dpo/margin": 59.59770965576172, "fcm_dpo/q_t": 0.36753708124160767, "grad_norm": 10.473079681396484, "learning_rate": 2.381045210440644e-07, "logits/chosen": 1.605952262878418, "logits/rejected": 1.594057559967041, "logps/chosen": -160.69447326660156, "logps/ref_chosen": -93.45108032226562, "logps/ref_rejected": -93.575927734375, "logps/rejected": -220.41702270507812, "loss": 0.9881, "margin_dpo/margin_mean": 59.59770965576172, "margin_dpo/margin_std": 68.90078735351562, "step": 374 }, { "epoch": 0.5668934240362812, "fcm_dpo/beta": 0.009714547544717789, "fcm_dpo/delta": 0.006051559001207352, "fcm_dpo/margin": 40.572608947753906, "fcm_dpo/q_t": 0.4105721712112427, "grad_norm": 8.553228378295898, "learning_rate": 2.3678397206786715e-07, "logits/chosen": 1.466545581817627, "logits/rejected": 1.3870232105255127, "logps/chosen": -130.58074951171875, "logps/ref_chosen": -77.37177276611328, "logps/ref_rejected": -98.59054565429688, "logps/rejected": -192.37213134765625, "loss": 1.1341, "margin_dpo/margin_mean": 40.572608947753906, "margin_dpo/margin_std": 68.16059112548828, "step": 375 }, { "epoch": 0.5684051398337112, "fcm_dpo/beta": 0.009592314250767231, "fcm_dpo/delta": -0.0968877449631691, "fcm_dpo/margin": 51.301605224609375, "fcm_dpo/q_t": 0.39151814579963684, "grad_norm": 6.652123928070068, "learning_rate": 2.3546379277238103e-07, "logits/chosen": 1.765944480895996, "logits/rejected": 1.5501282215118408, "logps/chosen": -136.59339904785156, "logps/ref_chosen": -68.99790954589844, "logps/ref_rejected": -90.37117004394531, "logps/rejected": -209.2682647705078, "loss": 1.0789, "margin_dpo/margin_mean": 51.301605224609375, "margin_dpo/margin_std": 77.24393463134766, "step": 376 }, { "epoch": 0.5699168556311414, "fcm_dpo/beta": 0.009752610698342323, "fcm_dpo/delta": 0.1289438158273697, "fcm_dpo/margin": 28.12000846862793, "fcm_dpo/q_t": 0.439037561416626, "grad_norm": 8.08517074584961, "learning_rate": 2.3414402008585886e-07, "logits/chosen": 1.6637239456176758, "logits/rejected": 1.628348469734192, "logps/chosen": -143.10202026367188, "logps/ref_chosen": -64.22705841064453, "logps/ref_rejected": -73.10292053222656, "logps/rejected": -180.09786987304688, "loss": 1.2252, "margin_dpo/margin_mean": 28.12000846862793, "margin_dpo/margin_std": 64.73475646972656, "step": 377 }, { "epoch": 0.5714285714285714, "fcm_dpo/beta": 0.009980445727705956, "fcm_dpo/delta": 0.10566462576389313, "fcm_dpo/margin": 29.759580612182617, "fcm_dpo/q_t": 0.43272238969802856, "grad_norm": 11.20825481414795, "learning_rate": 2.3282469092517977e-07, "logits/chosen": 1.4938591718673706, "logits/rejected": 1.3901243209838867, "logps/chosen": -152.7470703125, "logps/ref_chosen": -76.90864562988281, "logps/ref_rejected": -90.53460693359375, "logps/rejected": -196.13259887695312, "loss": 1.2127, "margin_dpo/margin_mean": 29.759580612182617, "margin_dpo/margin_std": 65.59465789794922, "step": 378 }, { "epoch": 0.5729402872260015, "fcm_dpo/beta": 0.00992667581886053, "fcm_dpo/delta": -0.06411861628293991, "fcm_dpo/margin": 46.461585998535156, "fcm_dpo/q_t": 0.3969395160675049, "grad_norm": 10.400646209716797, "learning_rate": 2.3150584219481643e-07, "logits/chosen": 1.529576063156128, "logits/rejected": 1.3459718227386475, "logps/chosen": -159.71408081054688, "logps/ref_chosen": -91.2371597290039, "logps/ref_rejected": -120.1969985961914, "logps/rejected": -235.135498046875, "loss": 1.0854, "margin_dpo/margin_mean": 46.461585998535156, "margin_dpo/margin_std": 68.9682388305664, "step": 379 }, { "epoch": 0.5744520030234316, "fcm_dpo/beta": 0.009684507735073566, "fcm_dpo/delta": -0.12210749089717865, "fcm_dpo/margin": 53.237144470214844, "fcm_dpo/q_t": 0.38306939601898193, "grad_norm": 6.27390718460083, "learning_rate": 2.3018751078580283e-07, "logits/chosen": 1.5858521461486816, "logits/rejected": 1.4949119091033936, "logps/chosen": -135.59014892578125, "logps/ref_chosen": -77.78315734863281, "logps/ref_rejected": -92.56083679199219, "logps/rejected": -203.60496520996094, "loss": 1.0686, "margin_dpo/margin_mean": 53.237144470214844, "margin_dpo/margin_std": 78.18292236328125, "step": 380 }, { "epoch": 0.5759637188208617, "fcm_dpo/beta": 0.009657653979957104, "fcm_dpo/delta": 0.018920322880148888, "fcm_dpo/margin": 24.543724060058594, "fcm_dpo/q_t": 0.4460574686527252, "grad_norm": 9.84516429901123, "learning_rate": 2.288697335747027e-07, "logits/chosen": 1.6163301467895508, "logits/rejected": 1.5747904777526855, "logps/chosen": -155.70779418945312, "logps/ref_chosen": -75.28189086914062, "logps/ref_rejected": -81.1995849609375, "logps/rejected": -186.16920471191406, "loss": 1.2755, "margin_dpo/margin_mean": 24.543724060058594, "margin_dpo/margin_std": 70.27555847167969, "step": 381 }, { "epoch": 0.5774754346182918, "fcm_dpo/beta": 0.009639125317335129, "fcm_dpo/delta": -0.03817974030971527, "fcm_dpo/margin": 33.828857421875, "fcm_dpo/q_t": 0.42681363224983215, "grad_norm": 12.585222244262695, "learning_rate": 2.2755254742257706e-07, "logits/chosen": 1.5320096015930176, "logits/rejected": 1.3989049196243286, "logps/chosen": -159.08447265625, "logps/ref_chosen": -78.74870300292969, "logps/ref_rejected": -99.77484130859375, "logps/rejected": -213.93946838378906, "loss": 1.1735, "margin_dpo/margin_mean": 33.828857421875, "margin_dpo/margin_std": 61.3028450012207, "step": 382 }, { "epoch": 0.5789871504157218, "fcm_dpo/beta": 0.009590652771294117, "fcm_dpo/delta": -0.00014057103544473648, "fcm_dpo/margin": 41.718406677246094, "fcm_dpo/q_t": 0.41220152378082275, "grad_norm": 6.8291192054748535, "learning_rate": 2.2623598917395436e-07, "logits/chosen": 1.277235984802246, "logits/rejected": 1.402155876159668, "logps/chosen": -168.30319213867188, "logps/ref_chosen": -95.92772674560547, "logps/ref_rejected": -92.13604736328125, "logps/rejected": -206.22991943359375, "loss": 1.1432, "margin_dpo/margin_mean": 41.718406677246094, "margin_dpo/margin_std": 74.20654296875, "step": 383 }, { "epoch": 0.5804988662131519, "fcm_dpo/beta": 0.00961991772055626, "fcm_dpo/delta": 0.013582942076027393, "fcm_dpo/margin": 40.22323989868164, "fcm_dpo/q_t": 0.41252365708351135, "grad_norm": 8.104625701904297, "learning_rate": 2.2492009565579875e-07, "logits/chosen": 1.8011313676834106, "logits/rejected": 1.7203788757324219, "logps/chosen": -156.7716064453125, "logps/ref_chosen": -80.208984375, "logps/ref_rejected": -94.39380645751953, "logps/rejected": -211.17965698242188, "loss": 1.1359, "margin_dpo/margin_mean": 40.22323989868164, "margin_dpo/margin_std": 68.09140014648438, "step": 384 }, { "epoch": 0.582010582010582, "fcm_dpo/beta": 0.009620234370231628, "fcm_dpo/delta": -0.018791217356920242, "fcm_dpo/margin": 43.446475982666016, "fcm_dpo/q_t": 0.4067288041114807, "grad_norm": 13.1287260055542, "learning_rate": 2.2360490367648084e-07, "logits/chosen": 1.3702280521392822, "logits/rejected": 1.2520372867584229, "logps/chosen": -159.83053588867188, "logps/ref_chosen": -85.26632690429688, "logps/ref_rejected": -102.1983413696289, "logps/rejected": -220.20901489257812, "loss": 1.1143, "margin_dpo/margin_mean": 43.44647979736328, "margin_dpo/margin_std": 69.08692932128906, "step": 385 }, { "epoch": 0.5835222978080121, "fcm_dpo/beta": 0.009717528708279133, "fcm_dpo/delta": 0.0649806335568428, "fcm_dpo/margin": 21.092330932617188, "fcm_dpo/q_t": 0.45253363251686096, "grad_norm": 12.614625930786133, "learning_rate": 2.2229045002474724e-07, "logits/chosen": 1.76641845703125, "logits/rejected": 1.5830888748168945, "logps/chosen": -182.44427490234375, "logps/ref_chosen": -93.19975280761719, "logps/ref_rejected": -112.98831176757812, "logps/rejected": -223.32516479492188, "loss": 1.296, "margin_dpo/margin_mean": 21.092330932617188, "margin_dpo/margin_std": 68.20382690429688, "step": 386 }, { "epoch": 0.5850340136054422, "fcm_dpo/beta": 0.00968782976269722, "fcm_dpo/delta": -0.054171886295080185, "fcm_dpo/margin": 46.604862213134766, "fcm_dpo/q_t": 0.40007221698760986, "grad_norm": 7.663426399230957, "learning_rate": 2.209767714686924e-07, "logits/chosen": 1.5394854545593262, "logits/rejected": 1.3090683221817017, "logps/chosen": -142.0372772216797, "logps/ref_chosen": -66.32861328125, "logps/ref_rejected": -100.56486511230469, "logps/rejected": -222.87838745117188, "loss": 1.0958, "margin_dpo/margin_mean": 46.604862213134766, "margin_dpo/margin_std": 70.9585189819336, "step": 387 }, { "epoch": 0.5865457294028723, "fcm_dpo/beta": 0.009683318436145782, "fcm_dpo/delta": 0.03653840348124504, "fcm_dpo/margin": 26.106887817382812, "fcm_dpo/q_t": 0.4437592029571533, "grad_norm": 13.81177043914795, "learning_rate": 2.1966390475472954e-07, "logits/chosen": 1.5186080932617188, "logits/rejected": 1.5088022947311401, "logps/chosen": -176.24365234375, "logps/ref_chosen": -92.95967864990234, "logps/ref_rejected": -97.9437255859375, "logps/rejected": -207.33456420898438, "loss": 1.2804, "margin_dpo/margin_mean": 26.106887817382812, "margin_dpo/margin_std": 76.65711975097656, "step": 388 }, { "epoch": 0.5880574452003023, "fcm_dpo/beta": 0.009643949568271637, "fcm_dpo/delta": -0.03215987980365753, "fcm_dpo/margin": 44.669090270996094, "fcm_dpo/q_t": 0.40315672755241394, "grad_norm": 9.2018404006958, "learning_rate": 2.1835188660656265e-07, "logits/chosen": 1.5738117694854736, "logits/rejected": 1.4677269458770752, "logps/chosen": -144.96087646484375, "logps/ref_chosen": -76.89031982421875, "logps/ref_rejected": -93.79212951660156, "logps/rejected": -206.53176879882812, "loss": 1.1067, "margin_dpo/margin_mean": 44.669090270996094, "margin_dpo/margin_std": 69.94828796386719, "step": 389 }, { "epoch": 0.5895691609977324, "fcm_dpo/beta": 0.009584802202880383, "fcm_dpo/delta": -0.008903810754418373, "fcm_dpo/margin": 42.60907745361328, "fcm_dpo/q_t": 0.40758416056632996, "grad_norm": 7.528814792633057, "learning_rate": 2.170407537241599e-07, "logits/chosen": 1.746974229812622, "logits/rejected": 1.6056780815124512, "logps/chosen": -115.68708038330078, "logps/ref_chosen": -61.058815002441406, "logps/ref_rejected": -79.55152893066406, "logps/rejected": -176.7888641357422, "loss": 1.11, "margin_dpo/margin_mean": 42.60907745361328, "margin_dpo/margin_std": 65.27470397949219, "step": 390 }, { "epoch": 0.5910808767951625, "fcm_dpo/beta": 0.009371999651193619, "fcm_dpo/delta": -0.15121221542358398, "fcm_dpo/margin": 57.90020751953125, "fcm_dpo/q_t": 0.3787153959274292, "grad_norm": 7.222849369049072, "learning_rate": 2.1573054278272636e-07, "logits/chosen": 1.3327152729034424, "logits/rejected": 1.2173633575439453, "logps/chosen": -140.46063232421875, "logps/ref_chosen": -78.60820770263672, "logps/ref_rejected": -103.3367691040039, "logps/rejected": -223.08938598632812, "loss": 1.0448, "margin_dpo/margin_mean": 57.900211334228516, "margin_dpo/margin_std": 80.29869842529297, "step": 391 }, { "epoch": 0.5925925925925926, "fcm_dpo/beta": 0.009207520633935928, "fcm_dpo/delta": -0.12280896306037903, "fcm_dpo/margin": 56.0882568359375, "fcm_dpo/q_t": 0.3837066888809204, "grad_norm": 8.607046127319336, "learning_rate": 2.1442129043167873e-07, "logits/chosen": 1.732191801071167, "logits/rejected": 1.638712763786316, "logps/chosen": -139.175537109375, "logps/ref_chosen": -86.99468994140625, "logps/ref_rejected": -112.73616027832031, "logps/rejected": -221.00526428222656, "loss": 1.0457, "margin_dpo/margin_mean": 56.088260650634766, "margin_dpo/margin_std": 74.01065063476562, "step": 392 }, { "epoch": 0.5941043083900227, "fcm_dpo/beta": 0.009078966453671455, "fcm_dpo/delta": 0.0228070467710495, "fcm_dpo/margin": 41.60166931152344, "fcm_dpo/q_t": 0.41527247428894043, "grad_norm": 5.811827182769775, "learning_rate": 2.131130332936195e-07, "logits/chosen": 1.8474013805389404, "logits/rejected": 1.7108526229858398, "logps/chosen": -136.55609130859375, "logps/ref_chosen": -71.26398468017578, "logps/ref_rejected": -88.99722290039062, "logps/rejected": -195.89100646972656, "loss": 1.1339, "margin_dpo/margin_mean": 41.60166931152344, "margin_dpo/margin_std": 67.7447509765625, "step": 393 }, { "epoch": 0.5956160241874527, "fcm_dpo/beta": 0.00911475345492363, "fcm_dpo/delta": -0.02656731568276882, "fcm_dpo/margin": 46.67322540283203, "fcm_dpo/q_t": 0.4027498960494995, "grad_norm": 7.995791435241699, "learning_rate": 2.1180580796331323e-07, "logits/chosen": 1.8606326580047607, "logits/rejected": 1.759063720703125, "logps/chosen": -133.31561279296875, "logps/ref_chosen": -78.70564270019531, "logps/ref_rejected": -87.01431274414062, "logps/rejected": -188.29750061035156, "loss": 1.0838, "margin_dpo/margin_mean": 46.6732292175293, "margin_dpo/margin_std": 63.45305633544922, "step": 394 }, { "epoch": 0.5971277399848829, "fcm_dpo/beta": 0.009070505388081074, "fcm_dpo/delta": -0.01608949899673462, "fcm_dpo/margin": 32.70347213745117, "fcm_dpo/q_t": 0.43361908197402954, "grad_norm": 7.361412048339844, "learning_rate": 2.104996510066625e-07, "logits/chosen": 1.6011724472045898, "logits/rejected": 1.3548979759216309, "logps/chosen": -120.16991424560547, "logps/ref_chosen": -65.30274963378906, "logps/ref_rejected": -93.22492218017578, "logps/rejected": -180.79556274414062, "loss": 1.2047, "margin_dpo/margin_mean": 32.70347213745117, "margin_dpo/margin_std": 67.76264953613281, "step": 395 }, { "epoch": 0.5986394557823129, "fcm_dpo/beta": 0.009096162393689156, "fcm_dpo/delta": 0.06893099844455719, "fcm_dpo/margin": 36.61708068847656, "fcm_dpo/q_t": 0.42313504219055176, "grad_norm": 8.793717384338379, "learning_rate": 2.0919459895968517e-07, "logits/chosen": 1.633918046951294, "logits/rejected": 1.4138092994689941, "logps/chosen": -121.96156311035156, "logps/ref_chosen": -67.33502197265625, "logps/ref_rejected": -98.8193359375, "logps/rejected": -190.06295776367188, "loss": 1.1485, "margin_dpo/margin_mean": 36.61708068847656, "margin_dpo/margin_std": 57.59046173095703, "step": 396 }, { "epoch": 0.600151171579743, "fcm_dpo/beta": 0.009252132847905159, "fcm_dpo/delta": 0.07621178776025772, "fcm_dpo/margin": 24.135467529296875, "fcm_dpo/q_t": 0.4461718797683716, "grad_norm": 8.252684593200684, "learning_rate": 2.078906883274924e-07, "logits/chosen": 1.580981731414795, "logits/rejected": 1.4645161628723145, "logps/chosen": -145.71463012695312, "logps/ref_chosen": -89.6042251586914, "logps/ref_rejected": -104.9779052734375, "logps/rejected": -185.22377014160156, "loss": 1.2607, "margin_dpo/margin_mean": 24.135465621948242, "margin_dpo/margin_std": 64.16143798828125, "step": 397 }, { "epoch": 0.6016628873771731, "fcm_dpo/beta": 0.009271949529647827, "fcm_dpo/delta": -0.04590045288205147, "fcm_dpo/margin": 47.874114990234375, "fcm_dpo/q_t": 0.40098968148231506, "grad_norm": 5.616165637969971, "learning_rate": 2.065879555832674e-07, "logits/chosen": 1.6307227611541748, "logits/rejected": 1.4722830057144165, "logps/chosen": -110.94779205322266, "logps/ref_chosen": -66.43465423583984, "logps/ref_rejected": -90.90376281738281, "logps/rejected": -183.291015625, "loss": 1.0816, "margin_dpo/margin_mean": 47.874114990234375, "margin_dpo/margin_std": 67.85723876953125, "step": 398 }, { "epoch": 0.6031746031746031, "fcm_dpo/beta": 0.009118804708123207, "fcm_dpo/delta": -0.05795277655124664, "fcm_dpo/margin": 49.865936279296875, "fcm_dpo/q_t": 0.39776691794395447, "grad_norm": 14.43397045135498, "learning_rate": 2.052864371672457e-07, "logits/chosen": 1.5677590370178223, "logits/rejected": 1.220775842666626, "logps/chosen": -153.78182983398438, "logps/ref_chosen": -87.22315979003906, "logps/ref_rejected": -136.32411193847656, "logps/rejected": -252.7487030029297, "loss": 1.0823, "margin_dpo/margin_mean": 49.865936279296875, "margin_dpo/margin_std": 71.72140502929688, "step": 399 }, { "epoch": 0.6046863189720333, "fcm_dpo/beta": 0.009189041331410408, "fcm_dpo/delta": 0.031679678708314896, "fcm_dpo/margin": 27.590740203857422, "fcm_dpo/q_t": 0.44303256273269653, "grad_norm": 11.517251014709473, "learning_rate": 2.0398616948569493e-07, "logits/chosen": 1.9134583473205566, "logits/rejected": 1.700062870979309, "logps/chosen": -163.27114868164062, "logps/ref_chosen": -91.1212158203125, "logps/ref_rejected": -108.19235229492188, "logps/rejected": -207.93301391601562, "loss": 1.2381, "margin_dpo/margin_mean": 27.590740203857422, "margin_dpo/margin_std": 66.34977722167969, "step": 400 }, { "epoch": 0.6046863189720333, "eval_fcm_dpo/beta": 0.009189040400087833, "eval_logits/chosen": 1.7609648704528809, "eval_logits/rejected": 1.6208163499832153, "eval_logps/chosen": -144.77313232421875, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -194.42681884765625, "eval_loss": 0.5745885968208313, "eval_margin_dpo/margin_mean": 39.85907745361328, "eval_margin_dpo/margin_std": 68.78260040283203, "eval_runtime": 42.2968, "eval_samples_per_second": 54.449, "eval_steps_per_second": 1.702, "step": 400 }, { "epoch": 0.6061980347694633, "fcm_dpo/beta": 0.009086466394364834, "fcm_dpo/delta": -0.08671090006828308, "fcm_dpo/margin": 53.11298370361328, "fcm_dpo/q_t": 0.39035192131996155, "grad_norm": 6.894371509552002, "learning_rate": 2.0268718890989752e-07, "logits/chosen": 1.9491140842437744, "logits/rejected": 1.7437255382537842, "logps/chosen": -116.19009399414062, "logps/ref_chosen": -67.54151153564453, "logps/ref_rejected": -98.06488800048828, "logps/rejected": -199.82644653320312, "loss": 1.0545, "margin_dpo/margin_mean": 53.11298370361328, "margin_dpo/margin_std": 69.13746643066406, "step": 401 }, { "epoch": 0.6077097505668935, "fcm_dpo/beta": 0.008998140692710876, "fcm_dpo/delta": -0.00021855533123016357, "fcm_dpo/margin": 44.44822692871094, "fcm_dpo/q_t": 0.40850377082824707, "grad_norm": 15.925997734069824, "learning_rate": 2.013895317751323e-07, "logits/chosen": 1.445326328277588, "logits/rejected": 1.4380943775177002, "logps/chosen": -130.43907165527344, "logps/ref_chosen": -77.44487762451172, "logps/ref_rejected": -83.1333236694336, "logps/rejected": -180.57574462890625, "loss": 1.0932, "margin_dpo/margin_mean": 44.44822692871094, "margin_dpo/margin_std": 59.27261734008789, "step": 402 }, { "epoch": 0.6092214663643235, "fcm_dpo/beta": 0.008991338312625885, "fcm_dpo/delta": -0.061279378831386566, "fcm_dpo/margin": 50.972496032714844, "fcm_dpo/q_t": 0.39812153577804565, "grad_norm": 6.25055456161499, "learning_rate": 2.0009323437965898e-07, "logits/chosen": 1.9196542501449585, "logits/rejected": 1.7053799629211426, "logps/chosen": -129.6885528564453, "logps/ref_chosen": -68.8230972290039, "logps/ref_rejected": -99.82356262207031, "logps/rejected": -211.66152954101562, "loss": 1.0887, "margin_dpo/margin_mean": 50.972496032714844, "margin_dpo/margin_std": 75.49934387207031, "step": 403 }, { "epoch": 0.6107331821617535, "fcm_dpo/beta": 0.008772274479269981, "fcm_dpo/delta": -0.05370837450027466, "fcm_dpo/margin": 51.221717834472656, "fcm_dpo/q_t": 0.3986782431602478, "grad_norm": 6.705301761627197, "learning_rate": 1.9879833298370237e-07, "logits/chosen": 1.60830557346344, "logits/rejected": 1.4156162738800049, "logps/chosen": -134.65676879882812, "logps/ref_chosen": -80.26783752441406, "logps/ref_rejected": -111.60258483886719, "logps/rejected": -217.2132110595703, "loss": 1.0884, "margin_dpo/margin_mean": 51.221717834472656, "margin_dpo/margin_std": 72.74002075195312, "step": 404 }, { "epoch": 0.6122448979591837, "fcm_dpo/beta": 0.008734223432838917, "fcm_dpo/delta": -0.05189571902155876, "fcm_dpo/margin": 34.009098052978516, "fcm_dpo/q_t": 0.4333046078681946, "grad_norm": 8.25037956237793, "learning_rate": 1.975048638084379e-07, "logits/chosen": 1.7744567394256592, "logits/rejected": 1.6257699728012085, "logps/chosen": -130.64573669433594, "logps/ref_chosen": -68.31065368652344, "logps/ref_rejected": -81.56044006347656, "logps/rejected": -177.9046173095703, "loss": 1.1885, "margin_dpo/margin_mean": 34.009098052978516, "margin_dpo/margin_std": 61.043663024902344, "step": 405 }, { "epoch": 0.6137566137566137, "fcm_dpo/beta": 0.008658639155328274, "fcm_dpo/delta": -0.0498778373003006, "fcm_dpo/margin": 51.688690185546875, "fcm_dpo/q_t": 0.3979532718658447, "grad_norm": 7.853333950042725, "learning_rate": 1.9621286303497914e-07, "logits/chosen": 1.5959994792938232, "logits/rejected": 1.2231396436691284, "logps/chosen": -121.9815444946289, "logps/ref_chosen": -64.86714935302734, "logps/ref_rejected": -110.06051635742188, "logps/rejected": -218.8636016845703, "loss": 1.0798, "margin_dpo/margin_mean": 51.688690185546875, "margin_dpo/margin_std": 71.60235595703125, "step": 406 }, { "epoch": 0.6152683295540439, "fcm_dpo/beta": 0.008745891973376274, "fcm_dpo/delta": 0.054477281868457794, "fcm_dpo/margin": 39.68103790283203, "fcm_dpo/q_t": 0.42161452770233154, "grad_norm": 6.449524402618408, "learning_rate": 1.9492236680336483e-07, "logits/chosen": 1.6160037517547607, "logits/rejected": 1.3780653476715088, "logps/chosen": -187.60446166992188, "logps/ref_chosen": -102.01712799072266, "logps/ref_rejected": -121.53548431396484, "logps/rejected": -246.80386352539062, "loss": 1.14, "margin_dpo/margin_mean": 39.68103790283203, "margin_dpo/margin_std": 62.7738037109375, "step": 407 }, { "epoch": 0.6167800453514739, "fcm_dpo/beta": 0.008701864629983902, "fcm_dpo/delta": -0.06911702454090118, "fcm_dpo/margin": 53.51190948486328, "fcm_dpo/q_t": 0.39388060569763184, "grad_norm": 6.081846714019775, "learning_rate": 1.9363341121154895e-07, "logits/chosen": 1.58525812625885, "logits/rejected": 1.3879069089889526, "logps/chosen": -135.20721435546875, "logps/ref_chosen": -72.77989959716797, "logps/ref_rejected": -92.01815795898438, "logps/rejected": -207.9573974609375, "loss": 1.0482, "margin_dpo/margin_mean": 53.51191329956055, "margin_dpo/margin_std": 63.87721252441406, "step": 408 }, { "epoch": 0.618291761148904, "fcm_dpo/beta": 0.00872439332306385, "fcm_dpo/delta": 0.05888619273900986, "fcm_dpo/margin": 28.853759765625, "fcm_dpo/q_t": 0.4449183940887451, "grad_norm": 7.567356109619141, "learning_rate": 1.9234603231438994e-07, "logits/chosen": 1.5403639078140259, "logits/rejected": 1.5558340549468994, "logps/chosen": -158.68392944335938, "logps/ref_chosen": -77.7901611328125, "logps/ref_rejected": -79.2997055053711, "logps/rejected": -189.0472412109375, "loss": 1.2434, "margin_dpo/margin_mean": 28.853759765625, "margin_dpo/margin_std": 71.60784149169922, "step": 409 }, { "epoch": 0.6198034769463341, "fcm_dpo/beta": 0.00859205611050129, "fcm_dpo/delta": -0.07421442866325378, "fcm_dpo/margin": 54.710350036621094, "fcm_dpo/q_t": 0.39280834794044495, "grad_norm": 6.605342864990234, "learning_rate": 1.9106026612264315e-07, "logits/chosen": 1.5201159715652466, "logits/rejected": 1.449505090713501, "logps/chosen": -152.5746612548828, "logps/ref_chosen": -80.35844421386719, "logps/ref_rejected": -92.19056701660156, "logps/rejected": -219.11712646484375, "loss": 1.0505, "margin_dpo/margin_mean": 54.710350036621094, "margin_dpo/margin_std": 67.6512451171875, "step": 410 }, { "epoch": 0.6213151927437641, "fcm_dpo/beta": 0.008500929921865463, "fcm_dpo/delta": -0.08138732612133026, "fcm_dpo/margin": 56.17717361450195, "fcm_dpo/q_t": 0.3915950059890747, "grad_norm": 8.85523509979248, "learning_rate": 1.8977614860195296e-07, "logits/chosen": 1.5671194791793823, "logits/rejected": 1.4427812099456787, "logps/chosen": -139.56689453125, "logps/ref_chosen": -70.72857666015625, "logps/ref_rejected": -93.19204711914062, "logps/rejected": -218.20755004882812, "loss": 1.042, "margin_dpo/margin_mean": 56.17717742919922, "margin_dpo/margin_std": 68.0445556640625, "step": 411 }, { "epoch": 0.6228269085411943, "fcm_dpo/beta": 0.008425667881965637, "fcm_dpo/delta": 0.007511887699365616, "fcm_dpo/margin": 46.5667724609375, "fcm_dpo/q_t": 0.41029736399650574, "grad_norm": 9.423548698425293, "learning_rate": 1.8849371567184662e-07, "logits/chosen": 1.6571754217147827, "logits/rejected": 1.5339932441711426, "logps/chosen": -159.95870971679688, "logps/ref_chosen": -72.87568664550781, "logps/ref_rejected": -88.21068572998047, "logps/rejected": -221.8604736328125, "loss": 1.1057, "margin_dpo/margin_mean": 46.5667724609375, "margin_dpo/margin_std": 65.58912658691406, "step": 412 }, { "epoch": 0.6243386243386243, "fcm_dpo/beta": 0.008588114753365517, "fcm_dpo/delta": 0.09323213249444962, "fcm_dpo/margin": 36.05671691894531, "fcm_dpo/q_t": 0.43061697483062744, "grad_norm": 7.6309967041015625, "learning_rate": 1.872130032047302e-07, "logits/chosen": 1.1301251649856567, "logits/rejected": 1.026894211769104, "logps/chosen": -175.1704864501953, "logps/ref_chosen": -84.70051574707031, "logps/ref_rejected": -92.06742095947266, "logps/rejected": -218.5941162109375, "loss": 1.2021, "margin_dpo/margin_mean": 36.05672073364258, "margin_dpo/margin_std": 76.68885803222656, "step": 413 }, { "epoch": 0.6258503401360545, "fcm_dpo/beta": 0.008647723123431206, "fcm_dpo/delta": 0.007389996200799942, "fcm_dpo/margin": 45.430633544921875, "fcm_dpo/q_t": 0.41087737679481506, "grad_norm": 7.641894817352295, "learning_rate": 1.8593404702488436e-07, "logits/chosen": 1.8904352188110352, "logits/rejected": 1.7254257202148438, "logps/chosen": -158.95175170898438, "logps/ref_chosen": -70.97660827636719, "logps/ref_rejected": -92.90523529052734, "logps/rejected": -226.31101989746094, "loss": 1.1192, "margin_dpo/margin_mean": 45.430633544921875, "margin_dpo/margin_std": 70.77737426757812, "step": 414 }, { "epoch": 0.6273620559334845, "fcm_dpo/beta": 0.008703294210135937, "fcm_dpo/delta": 0.05299728363752365, "fcm_dpo/margin": 40.08605194091797, "fcm_dpo/q_t": 0.42089951038360596, "grad_norm": 9.653564453125, "learning_rate": 1.846568829074628e-07, "logits/chosen": 1.6395957469940186, "logits/rejected": 1.5583677291870117, "logps/chosen": -162.5501708984375, "logps/ref_chosen": -71.7189712524414, "logps/ref_rejected": -74.54219818115234, "logps/rejected": -205.45945739746094, "loss": 1.1706, "margin_dpo/margin_mean": 40.086055755615234, "margin_dpo/margin_std": 76.08244323730469, "step": 415 }, { "epoch": 0.6288737717309146, "fcm_dpo/beta": 0.008706950582563877, "fcm_dpo/delta": -0.03111879713833332, "fcm_dpo/margin": 32.41926956176758, "fcm_dpo/q_t": 0.436288446187973, "grad_norm": 7.236203670501709, "learning_rate": 1.8338154657749128e-07, "logits/chosen": 1.453350305557251, "logits/rejected": 1.3089187145233154, "logps/chosen": -167.54823303222656, "logps/ref_chosen": -72.88249206542969, "logps/ref_rejected": -85.30693054199219, "logps/rejected": -212.39193725585938, "loss": 1.2269, "margin_dpo/margin_mean": 32.419273376464844, "margin_dpo/margin_std": 73.30819702148438, "step": 416 }, { "epoch": 0.6303854875283447, "fcm_dpo/beta": 0.008633752353489399, "fcm_dpo/delta": -0.05498684570193291, "fcm_dpo/margin": 52.40943145751953, "fcm_dpo/q_t": 0.3974682092666626, "grad_norm": 7.425167083740234, "learning_rate": 1.8210807370886849e-07, "logits/chosen": 1.7691454887390137, "logits/rejected": 1.5810437202453613, "logps/chosen": -164.9538116455078, "logps/ref_chosen": -72.49703216552734, "logps/ref_rejected": -89.38966369628906, "logps/rejected": -234.25587463378906, "loss": 1.0932, "margin_dpo/margin_mean": 52.40943145751953, "margin_dpo/margin_std": 79.10160064697266, "step": 417 }, { "epoch": 0.6318972033257747, "fcm_dpo/beta": 0.008647737093269825, "fcm_dpo/delta": 0.03645090013742447, "fcm_dpo/margin": 26.66490936279297, "fcm_dpo/q_t": 0.44797760248184204, "grad_norm": 8.460302352905273, "learning_rate": 1.8083649992336825e-07, "logits/chosen": 1.6079257726669312, "logits/rejected": 1.5933177471160889, "logps/chosen": -192.292724609375, "logps/ref_chosen": -89.70926666259766, "logps/ref_rejected": -90.98756408691406, "logps/rejected": -220.23593139648438, "loss": 1.2555, "margin_dpo/margin_mean": 26.664907455444336, "margin_dpo/margin_std": 68.90656280517578, "step": 418 }, { "epoch": 0.6334089191232048, "fcm_dpo/beta": 0.008394574746489525, "fcm_dpo/delta": -0.1813567876815796, "fcm_dpo/margin": 67.94660186767578, "fcm_dpo/q_t": 0.3701520562171936, "grad_norm": 6.600857257843018, "learning_rate": 1.7956686078964255e-07, "logits/chosen": 1.4430088996887207, "logits/rejected": 1.2721772193908691, "logps/chosen": -146.96456909179688, "logps/ref_chosen": -75.652099609375, "logps/ref_rejected": -91.0013427734375, "logps/rejected": -230.2604217529297, "loss": 0.9805, "margin_dpo/margin_mean": 67.94660949707031, "margin_dpo/margin_std": 72.4099349975586, "step": 419 }, { "epoch": 0.6349206349206349, "fcm_dpo/beta": 0.00842717569321394, "fcm_dpo/delta": 0.052176523953676224, "fcm_dpo/margin": 30.08838653564453, "fcm_dpo/q_t": 0.4437430500984192, "grad_norm": 6.708927631378174, "learning_rate": 1.782991918222275e-07, "logits/chosen": 1.5903661251068115, "logits/rejected": 1.4636292457580566, "logps/chosen": -171.9695587158203, "logps/ref_chosen": -72.58027648925781, "logps/ref_rejected": -79.90303802490234, "logps/rejected": -209.38070678710938, "loss": 1.2683, "margin_dpo/margin_mean": 30.08838653564453, "margin_dpo/margin_std": 84.14425659179688, "step": 420 }, { "epoch": 0.636432350718065, "fcm_dpo/beta": 0.008517824113368988, "fcm_dpo/delta": 0.08660048246383667, "fcm_dpo/margin": 37.125335693359375, "fcm_dpo/q_t": 0.42935702204704285, "grad_norm": 7.109913349151611, "learning_rate": 1.7703352848054887e-07, "logits/chosen": 1.5397415161132812, "logits/rejected": 1.3150533437728882, "logps/chosen": -173.07357788085938, "logps/ref_chosen": -78.71546936035156, "logps/ref_rejected": -90.82321166992188, "logps/rejected": -222.30667114257812, "loss": 1.2212, "margin_dpo/margin_mean": 37.12533950805664, "margin_dpo/margin_std": 85.766357421875, "step": 421 }, { "epoch": 0.6379440665154951, "fcm_dpo/beta": 0.008670301176607609, "fcm_dpo/delta": 0.06840848177671432, "fcm_dpo/margin": 38.49900817871094, "fcm_dpo/q_t": 0.42432913184165955, "grad_norm": 9.032937049865723, "learning_rate": 1.7576990616793137e-07, "logits/chosen": 1.5093660354614258, "logits/rejected": 1.4550291299819946, "logps/chosen": -174.02197265625, "logps/ref_chosen": -86.74519348144531, "logps/ref_rejected": -94.02015686035156, "logps/rejected": -219.79595947265625, "loss": 1.1566, "margin_dpo/margin_mean": 38.49900436401367, "margin_dpo/margin_std": 65.20552825927734, "step": 422 }, { "epoch": 0.6394557823129252, "fcm_dpo/beta": 0.008607940748333931, "fcm_dpo/delta": -0.05018071457743645, "fcm_dpo/margin": 52.0123405456543, "fcm_dpo/q_t": 0.39915522933006287, "grad_norm": 6.428966522216797, "learning_rate": 1.745083602306071e-07, "logits/chosen": 1.4873400926589966, "logits/rejected": 1.2859231233596802, "logps/chosen": -158.43319702148438, "logps/ref_chosen": -72.02232360839844, "logps/ref_rejected": -93.26976776123047, "logps/rejected": -231.69296264648438, "loss": 1.0761, "margin_dpo/margin_mean": 52.0123405456543, "margin_dpo/margin_std": 71.97749328613281, "step": 423 }, { "epoch": 0.6409674981103552, "fcm_dpo/beta": 0.008534763008356094, "fcm_dpo/delta": -0.04528824985027313, "fcm_dpo/margin": 51.92509460449219, "fcm_dpo/q_t": 0.3988720178604126, "grad_norm": 10.321234703063965, "learning_rate": 1.7324892595672804e-07, "logits/chosen": 1.3067110776901245, "logits/rejected": 1.2278244495391846, "logps/chosen": -161.60940551757812, "logps/ref_chosen": -68.22148132324219, "logps/ref_rejected": -94.12411499023438, "logps/rejected": -239.4371337890625, "loss": 1.0754, "margin_dpo/margin_mean": 51.92509460449219, "margin_dpo/margin_std": 70.35984802246094, "step": 424 }, { "epoch": 0.6424792139077853, "fcm_dpo/beta": 0.008516119793057442, "fcm_dpo/delta": -0.034915171563625336, "fcm_dpo/margin": 50.874088287353516, "fcm_dpo/q_t": 0.4017060101032257, "grad_norm": 6.930230140686035, "learning_rate": 1.7199163857537824e-07, "logits/chosen": 1.5249474048614502, "logits/rejected": 1.4606943130493164, "logps/chosen": -156.6333770751953, "logps/ref_chosen": -75.90104675292969, "logps/ref_rejected": -86.08673095703125, "logps/rejected": -217.69314575195312, "loss": 1.0782, "margin_dpo/margin_mean": 50.87408447265625, "margin_dpo/margin_std": 67.64917755126953, "step": 425 }, { "epoch": 0.6439909297052154, "fcm_dpo/beta": 0.008540986105799675, "fcm_dpo/delta": 0.08454061299562454, "fcm_dpo/margin": 26.161170959472656, "fcm_dpo/q_t": 0.45000118017196655, "grad_norm": 8.317493438720703, "learning_rate": 1.7073653325558828e-07, "logits/chosen": 1.5708837509155273, "logits/rejected": 1.5647852420806885, "logps/chosen": -192.17919921875, "logps/ref_chosen": -89.93118286132812, "logps/ref_rejected": -91.04658508300781, "logps/rejected": -219.45579528808594, "loss": 1.2791, "margin_dpo/margin_mean": 26.161174774169922, "margin_dpo/margin_std": 78.16825866699219, "step": 426 }, { "epoch": 0.6455026455026455, "fcm_dpo/beta": 0.008664184249937534, "fcm_dpo/delta": 0.034002143889665604, "fcm_dpo/margin": 42.38199996948242, "fcm_dpo/q_t": 0.41778403520584106, "grad_norm": 7.2056684494018555, "learning_rate": 1.6948364510535218e-07, "logits/chosen": 1.706808090209961, "logits/rejected": 1.5242950916290283, "logps/chosen": -175.19021606445312, "logps/ref_chosen": -77.83393859863281, "logps/ref_rejected": -98.69864654541016, "logps/rejected": -238.43692016601562, "loss": 1.1425, "margin_dpo/margin_mean": 42.38199996948242, "margin_dpo/margin_std": 72.0445327758789, "step": 427 }, { "epoch": 0.6470143613000756, "fcm_dpo/beta": 0.00863520335406065, "fcm_dpo/delta": -0.045657265931367874, "fcm_dpo/margin": 51.373043060302734, "fcm_dpo/q_t": 0.4012737274169922, "grad_norm": 7.712218284606934, "learning_rate": 1.6823300917064458e-07, "logits/chosen": 1.3941190242767334, "logits/rejected": 1.3461658954620361, "logps/chosen": -178.5933837890625, "logps/ref_chosen": -90.3450927734375, "logps/ref_rejected": -100.24185180664062, "logps/rejected": -239.86318969726562, "loss": 1.0844, "margin_dpo/margin_mean": 51.373046875, "margin_dpo/margin_std": 73.79509735107422, "step": 428 }, { "epoch": 0.6485260770975056, "fcm_dpo/beta": 0.008655520156025887, "fcm_dpo/delta": 0.015263576060533524, "fcm_dpo/margin": 44.45375442504883, "fcm_dpo/q_t": 0.41261976957321167, "grad_norm": 7.072147846221924, "learning_rate": 1.669846604344412e-07, "logits/chosen": 1.2022795677185059, "logits/rejected": 1.2440556287765503, "logps/chosen": -170.97415161132812, "logps/ref_chosen": -78.24811553955078, "logps/ref_rejected": -75.24495697021484, "logps/rejected": -212.42477416992188, "loss": 1.1794, "margin_dpo/margin_mean": 44.453758239746094, "margin_dpo/margin_std": 88.38716125488281, "step": 429 }, { "epoch": 0.6500377928949358, "fcm_dpo/beta": 0.00848034955561161, "fcm_dpo/delta": -0.09685202687978745, "fcm_dpo/margin": 57.9986686706543, "fcm_dpo/q_t": 0.3889318108558655, "grad_norm": 6.905314922332764, "learning_rate": 1.6573863381573954e-07, "logits/chosen": 1.3700964450836182, "logits/rejected": 1.4100453853607178, "logps/chosen": -157.59194946289062, "logps/ref_chosen": -76.08027648925781, "logps/ref_rejected": -84.09554290771484, "logps/rejected": -223.60589599609375, "loss": 1.0623, "margin_dpo/margin_mean": 57.99867248535156, "margin_dpo/margin_std": 80.27232360839844, "step": 430 }, { "epoch": 0.6515495086923658, "fcm_dpo/beta": 0.008486424572765827, "fcm_dpo/delta": 0.014092888683080673, "fcm_dpo/margin": 45.52803039550781, "fcm_dpo/q_t": 0.4137864112854004, "grad_norm": 6.048389911651611, "learning_rate": 1.6449496416858282e-07, "logits/chosen": 1.37843656539917, "logits/rejected": 1.2628583908081055, "logps/chosen": -143.2403106689453, "logps/ref_chosen": -66.88581085205078, "logps/ref_rejected": -89.56040954589844, "logps/rejected": -211.44293212890625, "loss": 1.1442, "margin_dpo/margin_mean": 45.52803039550781, "margin_dpo/margin_std": 80.02131652832031, "step": 431 }, { "epoch": 0.6530612244897959, "fcm_dpo/beta": 0.008482584729790688, "fcm_dpo/delta": 0.0026997558306902647, "fcm_dpo/margin": 46.84972381591797, "fcm_dpo/q_t": 0.412258505821228, "grad_norm": 6.3566412925720215, "learning_rate": 1.632536862810844e-07, "logits/chosen": 1.5703511238098145, "logits/rejected": 1.4383772611618042, "logps/chosen": -160.373291015625, "logps/ref_chosen": -79.65066528320312, "logps/ref_rejected": -103.92634582519531, "logps/rejected": -231.4987030029297, "loss": 1.155, "margin_dpo/margin_mean": 46.84972381591797, "margin_dpo/margin_std": 87.462646484375, "step": 432 }, { "epoch": 0.654572940287226, "fcm_dpo/beta": 0.008361434563994408, "fcm_dpo/delta": -0.14136160910129547, "fcm_dpo/margin": 63.84764862060547, "fcm_dpo/q_t": 0.38163742423057556, "grad_norm": 6.963474750518799, "learning_rate": 1.6201483487445515e-07, "logits/chosen": 1.7979011535644531, "logits/rejected": 1.780426263809204, "logps/chosen": -154.82102966308594, "logps/ref_chosen": -77.30774688720703, "logps/ref_rejected": -81.65180206298828, "logps/rejected": -223.0127410888672, "loss": 1.057, "margin_dpo/margin_mean": 63.84764862060547, "margin_dpo/margin_std": 90.18070220947266, "step": 433 }, { "epoch": 0.656084656084656, "fcm_dpo/beta": 0.008094357326626778, "fcm_dpo/delta": -0.08986343443393707, "fcm_dpo/margin": 59.861175537109375, "fcm_dpo/q_t": 0.3910486698150635, "grad_norm": 6.0232343673706055, "learning_rate": 1.6077844460203204e-07, "logits/chosen": 1.4988017082214355, "logits/rejected": 1.337533950805664, "logps/chosen": -124.65926361083984, "logps/ref_chosen": -63.31850051879883, "logps/ref_rejected": -89.15093994140625, "logps/rejected": -210.35289001464844, "loss": 1.0736, "margin_dpo/margin_mean": 59.861175537109375, "margin_dpo/margin_std": 86.07001495361328, "step": 434 }, { "epoch": 0.6575963718820862, "fcm_dpo/beta": 0.008202778175473213, "fcm_dpo/delta": 0.058817461133003235, "fcm_dpo/margin": 41.77471923828125, "fcm_dpo/q_t": 0.4230271577835083, "grad_norm": 7.037782192230225, "learning_rate": 1.5954455004830878e-07, "logits/chosen": 1.81736421585083, "logits/rejected": 1.734781265258789, "logps/chosen": -157.01971435546875, "logps/ref_chosen": -71.1719741821289, "logps/ref_rejected": -86.42095184326172, "logps/rejected": -214.0434112548828, "loss": 1.176, "margin_dpo/margin_mean": 41.77471923828125, "margin_dpo/margin_std": 80.3061294555664, "step": 435 }, { "epoch": 0.6591080876795162, "fcm_dpo/beta": 0.008254708722233772, "fcm_dpo/delta": 0.020179305225610733, "fcm_dpo/margin": 46.05265808105469, "fcm_dpo/q_t": 0.41473814845085144, "grad_norm": 7.05510139465332, "learning_rate": 1.5831318572796847e-07, "logits/chosen": 1.3741307258605957, "logits/rejected": 1.2321834564208984, "logps/chosen": -149.34500122070312, "logps/ref_chosen": -74.45087432861328, "logps/ref_rejected": -86.01708984375, "logps/rejected": -206.9638671875, "loss": 1.1532, "margin_dpo/margin_mean": 46.05265808105469, "margin_dpo/margin_std": 83.09732055664062, "step": 436 }, { "epoch": 0.6606198034769464, "fcm_dpo/beta": 0.008225423283874989, "fcm_dpo/delta": -0.0016473679570481181, "fcm_dpo/margin": 33.15676498413086, "fcm_dpo/q_t": 0.43936091661453247, "grad_norm": 8.62120246887207, "learning_rate": 1.5708438608491815e-07, "logits/chosen": 1.6454570293426514, "logits/rejected": 1.309108018875122, "logps/chosen": -168.96128845214844, "logps/ref_chosen": -72.38907623291016, "logps/ref_rejected": -111.03279876708984, "logps/rejected": -240.76177978515625, "loss": 1.2447, "margin_dpo/margin_mean": 33.15676498413086, "margin_dpo/margin_std": 82.86212158203125, "step": 437 }, { "epoch": 0.6621315192743764, "fcm_dpo/beta": 0.00817845668643713, "fcm_dpo/delta": -0.08092987537384033, "fcm_dpo/margin": 58.298133850097656, "fcm_dpo/q_t": 0.39217159152030945, "grad_norm": 7.8519816398620605, "learning_rate": 1.558581854913253e-07, "logits/chosen": 1.588195562362671, "logits/rejected": 1.4413819313049316, "logps/chosen": -132.78721618652344, "logps/ref_chosen": -57.27682876586914, "logps/ref_rejected": -83.07940673828125, "logps/rejected": -216.887939453125, "loss": 1.07, "margin_dpo/margin_mean": 58.298133850097656, "margin_dpo/margin_std": 79.39116668701172, "step": 438 }, { "epoch": 0.6636432350718064, "fcm_dpo/beta": 0.00810486450791359, "fcm_dpo/delta": -0.027155064046382904, "fcm_dpo/margin": 52.506919860839844, "fcm_dpo/q_t": 0.4055066704750061, "grad_norm": 5.769200801849365, "learning_rate": 1.5463461824665658e-07, "logits/chosen": 1.5664254426956177, "logits/rejected": 1.445422887802124, "logps/chosen": -185.7174530029297, "logps/ref_chosen": -98.35890197753906, "logps/ref_rejected": -112.69817352294922, "logps/rejected": -252.5636444091797, "loss": 1.0962, "margin_dpo/margin_mean": 52.506927490234375, "margin_dpo/margin_std": 76.33085632324219, "step": 439 }, { "epoch": 0.6651549508692366, "fcm_dpo/beta": 0.007925990968942642, "fcm_dpo/delta": -0.11724238097667694, "fcm_dpo/margin": 64.5205307006836, "fcm_dpo/q_t": 0.38388943672180176, "grad_norm": 9.103875160217285, "learning_rate": 1.534137185767178e-07, "logits/chosen": 1.3807024955749512, "logits/rejected": 1.1057807207107544, "logps/chosen": -127.14051818847656, "logps/ref_chosen": -61.662452697753906, "logps/ref_rejected": -86.81646728515625, "logps/rejected": -216.8150634765625, "loss": 1.0246, "margin_dpo/margin_mean": 64.5205307006836, "margin_dpo/margin_std": 76.79633331298828, "step": 440 }, { "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.007821722887456417, "fcm_dpo/delta": -0.01750565692782402, "fcm_dpo/margin": 53.261146545410156, "fcm_dpo/q_t": 0.40477800369262695, "grad_norm": 11.901995658874512, "learning_rate": 1.521955206326976e-07, "logits/chosen": 1.5941314697265625, "logits/rejected": 1.301910400390625, "logps/chosen": -140.73521423339844, "logps/ref_chosen": -74.33235168457031, "logps/ref_rejected": -99.654541015625, "logps/rejected": -219.3185577392578, "loss": 1.0951, "margin_dpo/margin_mean": 53.26115417480469, "margin_dpo/margin_std": 74.70690155029297, "step": 441 }, { "epoch": 0.6681783824640968, "fcm_dpo/beta": 0.007868031039834023, "fcm_dpo/delta": 0.006100071594119072, "fcm_dpo/margin": 50.065460205078125, "fcm_dpo/q_t": 0.41083312034606934, "grad_norm": 9.3712797164917, "learning_rate": 1.5098005849021078e-07, "logits/chosen": 1.5173122882843018, "logits/rejected": 1.4266064167022705, "logps/chosen": -173.49835205078125, "logps/ref_chosen": -82.42591857910156, "logps/ref_rejected": -106.71090698242188, "logps/rejected": -247.8488006591797, "loss": 1.1319, "margin_dpo/margin_mean": 50.065460205078125, "margin_dpo/margin_std": 83.03895568847656, "step": 442 }, { "epoch": 0.6696900982615268, "fcm_dpo/beta": 0.007733378559350967, "fcm_dpo/delta": -0.07331807166337967, "fcm_dpo/margin": 60.69889450073242, "fcm_dpo/q_t": 0.3952465057373047, "grad_norm": 9.156957626342773, "learning_rate": 1.4976736614834662e-07, "logits/chosen": 1.7997548580169678, "logits/rejected": 1.6043787002563477, "logps/chosen": -152.74725341796875, "logps/ref_chosen": -72.87019348144531, "logps/ref_rejected": -94.48143005371094, "logps/rejected": -235.05740356445312, "loss": 1.0812, "margin_dpo/margin_mean": 60.69889450073242, "margin_dpo/margin_std": 89.32925415039062, "step": 443 }, { "epoch": 0.671201814058957, "fcm_dpo/beta": 0.00782048236578703, "fcm_dpo/delta": 0.0576334185898304, "fcm_dpo/margin": 21.587419509887695, "fcm_dpo/q_t": 0.46234452724456787, "grad_norm": 10.07677936553955, "learning_rate": 1.4855747752871654e-07, "logits/chosen": 1.544628620147705, "logits/rejected": 1.298480749130249, "logps/chosen": -171.25473022460938, "logps/ref_chosen": -74.650390625, "logps/ref_rejected": -106.89204406738281, "logps/rejected": -225.08380126953125, "loss": 1.3148, "margin_dpo/margin_mean": 21.587419509887695, "margin_dpo/margin_std": 77.45458221435547, "step": 444 }, { "epoch": 0.672713529856387, "fcm_dpo/beta": 0.007657904177904129, "fcm_dpo/delta": -0.1426275670528412, "fcm_dpo/margin": 69.86483764648438, "fcm_dpo/q_t": 0.37939491868019104, "grad_norm": 8.217625617980957, "learning_rate": 1.473504264745062e-07, "logits/chosen": 1.5077362060546875, "logits/rejected": 1.4950257539749146, "logps/chosen": -158.41409301757812, "logps/ref_chosen": -76.26957702636719, "logps/ref_rejected": -89.84994506835938, "logps/rejected": -241.8592987060547, "loss": 1.0212, "margin_dpo/margin_mean": 69.86483001708984, "margin_dpo/margin_std": 85.99278259277344, "step": 445 }, { "epoch": 0.674225245653817, "fcm_dpo/beta": 0.007591135799884796, "fcm_dpo/delta": -0.09901002794504166, "fcm_dpo/margin": 64.8037338256836, "fcm_dpo/q_t": 0.38721728324890137, "grad_norm": 5.767418384552002, "learning_rate": 1.461462467495284e-07, "logits/chosen": 1.5965964794158936, "logits/rejected": 1.43882417678833, "logps/chosen": -137.1719207763672, "logps/ref_chosen": -62.74647903442383, "logps/ref_rejected": -86.395751953125, "logps/rejected": -225.62493896484375, "loss": 1.0341, "margin_dpo/margin_mean": 64.80374145507812, "margin_dpo/margin_std": 68.17994689941406, "step": 446 }, { "epoch": 0.6757369614512472, "fcm_dpo/beta": 0.007440866902470589, "fcm_dpo/delta": -0.015056482516229153, "fcm_dpo/margin": 55.69664764404297, "fcm_dpo/q_t": 0.4054851233959198, "grad_norm": 9.256994247436523, "learning_rate": 1.4494497203727843e-07, "logits/chosen": 1.2242474555969238, "logits/rejected": 0.9210813045501709, "logps/chosen": -139.55810546875, "logps/ref_chosen": -71.06666564941406, "logps/ref_rejected": -103.57111358642578, "logps/rejected": -227.7592010498047, "loss": 1.1148, "margin_dpo/margin_mean": 55.69664764404297, "margin_dpo/margin_std": 87.89811706542969, "step": 447 }, { "epoch": 0.6772486772486772, "fcm_dpo/beta": 0.007497244980186224, "fcm_dpo/delta": 0.03677457943558693, "fcm_dpo/margin": 48.574466705322266, "fcm_dpo/q_t": 0.417564332485199, "grad_norm": 7.66340446472168, "learning_rate": 1.4374663593999256e-07, "logits/chosen": 1.652498483657837, "logits/rejected": 1.538625955581665, "logps/chosen": -159.7042236328125, "logps/ref_chosen": -73.400146484375, "logps/ref_rejected": -96.34330749511719, "logps/rejected": -231.22186279296875, "loss": 1.1446, "margin_dpo/margin_mean": 48.574466705322266, "margin_dpo/margin_std": 81.99598693847656, "step": 448 }, { "epoch": 0.6787603930461074, "fcm_dpo/beta": 0.007546110078692436, "fcm_dpo/delta": 0.04134930670261383, "fcm_dpo/margin": 28.31925392150879, "fcm_dpo/q_t": 0.45164692401885986, "grad_norm": 15.793785095214844, "learning_rate": 1.4255127197770707e-07, "logits/chosen": 0.9623701572418213, "logits/rejected": 0.9727632403373718, "logps/chosen": -188.20831298828125, "logps/ref_chosen": -93.66099548339844, "logps/ref_rejected": -102.53019714355469, "logps/rejected": -225.39675903320312, "loss": 1.2846, "margin_dpo/margin_mean": 28.319255828857422, "margin_dpo/margin_std": 84.99736022949219, "step": 449 }, { "epoch": 0.6802721088435374, "fcm_dpo/beta": 0.007580921053886414, "fcm_dpo/delta": 0.04489194229245186, "fcm_dpo/margin": 47.054134368896484, "fcm_dpo/q_t": 0.41971227526664734, "grad_norm": 6.454341888427734, "learning_rate": 1.4135891358732205e-07, "logits/chosen": 1.755326509475708, "logits/rejected": 1.4095392227172852, "logps/chosen": -133.84730529785156, "logps/ref_chosen": -62.52460479736328, "logps/ref_rejected": -94.04986572265625, "logps/rejected": -212.42669677734375, "loss": 1.1614, "margin_dpo/margin_mean": 47.054134368896484, "margin_dpo/margin_std": 86.08123779296875, "step": 450 }, { "epoch": 0.6817838246409675, "fcm_dpo/beta": 0.007728881202638149, "fcm_dpo/delta": 0.0906517505645752, "fcm_dpo/margin": 40.36960983276367, "fcm_dpo/q_t": 0.43043777346611023, "grad_norm": 7.109012126922607, "learning_rate": 1.4016959412166437e-07, "logits/chosen": 1.3793981075286865, "logits/rejected": 1.2791502475738525, "logps/chosen": -155.21142578125, "logps/ref_chosen": -79.14009094238281, "logps/ref_rejected": -93.23919677734375, "logps/rejected": -209.68016052246094, "loss": 1.1896, "margin_dpo/margin_mean": 40.36961364746094, "margin_dpo/margin_std": 80.68801879882812, "step": 451 }, { "epoch": 0.6832955404383976, "fcm_dpo/beta": 0.00771725969389081, "fcm_dpo/delta": -0.02904406189918518, "fcm_dpo/margin": 55.432777404785156, "fcm_dpo/q_t": 0.40317466855049133, "grad_norm": 7.301806926727295, "learning_rate": 1.3898334684855645e-07, "logits/chosen": 1.551164150238037, "logits/rejected": 1.3258707523345947, "logps/chosen": -147.094482421875, "logps/ref_chosen": -70.38827514648438, "logps/ref_rejected": -95.47691345214844, "logps/rejected": -227.61587524414062, "loss": 1.1117, "margin_dpo/margin_mean": 55.432777404785156, "margin_dpo/margin_std": 87.90678405761719, "step": 452 }, { "epoch": 0.6848072562358276, "fcm_dpo/beta": 0.0077826473861932755, "fcm_dpo/delta": 0.03363284468650818, "fcm_dpo/margin": 47.162410736083984, "fcm_dpo/q_t": 0.4179924726486206, "grad_norm": 7.004802703857422, "learning_rate": 1.3780020494988445e-07, "logits/chosen": 1.5915346145629883, "logits/rejected": 1.474679946899414, "logps/chosen": -158.24111938476562, "logps/ref_chosen": -79.9207763671875, "logps/ref_rejected": -90.20779418945312, "logps/rejected": -215.6905517578125, "loss": 1.1627, "margin_dpo/margin_mean": 47.162410736083984, "margin_dpo/margin_std": 87.50480651855469, "step": 453 }, { "epoch": 0.6863189720332578, "fcm_dpo/beta": 0.007651199586689472, "fcm_dpo/delta": -0.10082674026489258, "fcm_dpo/margin": 64.81204223632812, "fcm_dpo/q_t": 0.38581979274749756, "grad_norm": 7.975934028625488, "learning_rate": 1.366202015206706e-07, "logits/chosen": 1.653881311416626, "logits/rejected": 1.5468456745147705, "logps/chosen": -131.710693359375, "logps/ref_chosen": -69.71887969970703, "logps/ref_rejected": -82.86952209472656, "logps/rejected": -209.67337036132812, "loss": 1.0413, "margin_dpo/margin_mean": 64.81204223632812, "margin_dpo/margin_std": 79.79120635986328, "step": 454 }, { "epoch": 0.6878306878306878, "fcm_dpo/beta": 0.007537747733294964, "fcm_dpo/delta": -0.06928986310958862, "fcm_dpo/margin": 61.83723449707031, "fcm_dpo/q_t": 0.393891841173172, "grad_norm": 5.723858833312988, "learning_rate": 1.354433695681474e-07, "logits/chosen": 1.210571527481079, "logits/rejected": 1.1374115943908691, "logps/chosen": -170.25729370117188, "logps/ref_chosen": -89.51481628417969, "logps/ref_rejected": -97.93235778808594, "logps/rejected": -240.51206970214844, "loss": 1.0549, "margin_dpo/margin_mean": 61.83723449707031, "margin_dpo/margin_std": 78.60615539550781, "step": 455 }, { "epoch": 0.6893424036281179, "fcm_dpo/beta": 0.0074795568361878395, "fcm_dpo/delta": 0.006430737674236298, "fcm_dpo/margin": 52.61790466308594, "fcm_dpo/q_t": 0.4104902744293213, "grad_norm": 9.476668357849121, "learning_rate": 1.3426974201083439e-07, "logits/chosen": 1.48141610622406, "logits/rejected": 1.3234397172927856, "logps/chosen": -151.86122131347656, "logps/ref_chosen": -74.60527038574219, "logps/ref_rejected": -97.98377227783203, "logps/rejected": -227.85763549804688, "loss": 1.1121, "margin_dpo/margin_mean": 52.61790466308594, "margin_dpo/margin_std": 77.86461639404297, "step": 456 }, { "epoch": 0.690854119425548, "fcm_dpo/beta": 0.007550334092229605, "fcm_dpo/delta": 0.06581398844718933, "fcm_dpo/margin": 44.53820037841797, "fcm_dpo/q_t": 0.4231857359409332, "grad_norm": 7.015773773193359, "learning_rate": 1.3309935167761717e-07, "logits/chosen": 1.758169174194336, "logits/rejected": 1.5407148599624634, "logps/chosen": -152.5609893798828, "logps/ref_chosen": -63.927032470703125, "logps/ref_rejected": -83.15243530273438, "logps/rejected": -216.3245849609375, "loss": 1.151, "margin_dpo/margin_mean": 44.53820037841797, "margin_dpo/margin_std": 72.46507263183594, "step": 457 }, { "epoch": 0.6923658352229781, "fcm_dpo/beta": 0.0076367976143956184, "fcm_dpo/delta": -0.017959222197532654, "fcm_dpo/margin": 54.54873275756836, "fcm_dpo/q_t": 0.4054056406021118, "grad_norm": 8.418825149536133, "learning_rate": 1.3193223130682936e-07, "logits/chosen": 1.6931997537612915, "logits/rejected": 1.3649234771728516, "logps/chosen": -142.6044464111328, "logps/ref_chosen": -67.68869018554688, "logps/ref_rejected": -104.40899658203125, "logps/rejected": -233.8734893798828, "loss": 1.1064, "margin_dpo/margin_mean": 54.548736572265625, "margin_dpo/margin_std": 81.37754821777344, "step": 458 }, { "epoch": 0.6938775510204082, "fcm_dpo/beta": 0.007586327847093344, "fcm_dpo/delta": -0.08097509294748306, "fcm_dpo/margin": 62.6253547668457, "fcm_dpo/q_t": 0.3917158246040344, "grad_norm": 6.951050758361816, "learning_rate": 1.3076841354533658e-07, "logits/chosen": 1.7440357208251953, "logits/rejected": 1.6339316368103027, "logps/chosen": -152.58859252929688, "logps/ref_chosen": -83.82363891601562, "logps/ref_rejected": -103.75938415527344, "logps/rejected": -235.14968872070312, "loss": 1.0538, "margin_dpo/margin_mean": 62.6253547668457, "margin_dpo/margin_std": 73.25015258789062, "step": 459 }, { "epoch": 0.6953892668178382, "fcm_dpo/beta": 0.007416483946144581, "fcm_dpo/delta": -0.0269007571041584, "fcm_dpo/margin": 57.389373779296875, "fcm_dpo/q_t": 0.4039474427700043, "grad_norm": 7.818911075592041, "learning_rate": 1.2960793094762345e-07, "logits/chosen": 1.9773120880126953, "logits/rejected": 1.5847508907318115, "logps/chosen": -170.81787109375, "logps/ref_chosen": -79.4836654663086, "logps/ref_rejected": -112.31745910644531, "logps/rejected": -261.041015625, "loss": 1.0852, "margin_dpo/margin_mean": 57.389373779296875, "margin_dpo/margin_std": 78.99481964111328, "step": 460 }, { "epoch": 0.6969009826152683, "fcm_dpo/beta": 0.007295292802155018, "fcm_dpo/delta": -0.07289466261863708, "fcm_dpo/margin": 64.19013977050781, "fcm_dpo/q_t": 0.39158502221107483, "grad_norm": 7.286057949066162, "learning_rate": 1.2845081597488286e-07, "logits/chosen": 1.7993457317352295, "logits/rejected": 1.565769910812378, "logps/chosen": -129.4446563720703, "logps/ref_chosen": -64.28482055664062, "logps/ref_rejected": -93.73818969726562, "logps/rejected": -223.08816528320312, "loss": 1.0383, "margin_dpo/margin_mean": 64.19013977050781, "margin_dpo/margin_std": 71.33786010742188, "step": 461 }, { "epoch": 0.6984126984126984, "fcm_dpo/beta": 0.0071984813548624516, "fcm_dpo/delta": -0.08874664455652237, "fcm_dpo/margin": 67.24839782714844, "fcm_dpo/q_t": 0.3891562223434448, "grad_norm": 9.859999656677246, "learning_rate": 1.27297100994108e-07, "logits/chosen": 1.5112301111221313, "logits/rejected": 1.3689056634902954, "logps/chosen": -158.24203491210938, "logps/ref_chosen": -77.15335083007812, "logps/ref_rejected": -91.12923431396484, "logps/rejected": -239.46630859375, "loss": 1.0579, "margin_dpo/margin_mean": 67.24839782714844, "margin_dpo/margin_std": 89.81712341308594, "step": 462 }, { "epoch": 0.6999244142101285, "fcm_dpo/beta": 0.007196832448244095, "fcm_dpo/delta": 0.008259914815425873, "fcm_dpo/margin": 40.18242645263672, "fcm_dpo/q_t": 0.43439820408821106, "grad_norm": 11.131797790527344, "learning_rate": 1.2614681827718695e-07, "logits/chosen": 1.4589455127716064, "logits/rejected": 1.4636832475662231, "logps/chosen": -173.68218994140625, "logps/ref_chosen": -87.58760070800781, "logps/ref_rejected": -87.97022247314453, "logps/rejected": -214.24722290039062, "loss": 1.198, "margin_dpo/margin_mean": 40.18242645263672, "margin_dpo/margin_std": 79.24378967285156, "step": 463 }, { "epoch": 0.7014361300075586, "fcm_dpo/beta": 0.007220131810754538, "fcm_dpo/delta": -0.008086636662483215, "fcm_dpo/margin": 56.42665100097656, "fcm_dpo/q_t": 0.40838852524757385, "grad_norm": 8.530649185180664, "learning_rate": 1.2500000000000005e-07, "logits/chosen": 1.1444811820983887, "logits/rejected": 1.046148419380188, "logps/chosen": -166.62535095214844, "logps/ref_chosen": -75.83175659179688, "logps/ref_rejected": -84.4811019897461, "logps/rejected": -231.70135498046875, "loss": 1.1332, "margin_dpo/margin_mean": 56.42665100097656, "margin_dpo/margin_std": 94.67403411865234, "step": 464 }, { "epoch": 0.7029478458049887, "fcm_dpo/beta": 0.007176822982728481, "fcm_dpo/delta": 0.00978757068514824, "fcm_dpo/margin": 54.408992767333984, "fcm_dpo/q_t": 0.411876916885376, "grad_norm": 13.895620346069336, "learning_rate": 1.238566782415197e-07, "logits/chosen": 1.7335069179534912, "logits/rejected": 1.5657398700714111, "logps/chosen": -169.32749938964844, "logps/ref_chosen": -77.057861328125, "logps/ref_rejected": -102.75727844238281, "logps/rejected": -249.4359130859375, "loss": 1.1214, "margin_dpo/margin_mean": 54.40899658203125, "margin_dpo/margin_std": 85.08840942382812, "step": 465 }, { "epoch": 0.7044595616024187, "fcm_dpo/beta": 0.0072374227456748486, "fcm_dpo/delta": 0.026420170441269875, "fcm_dpo/margin": 36.631690979003906, "fcm_dpo/q_t": 0.43956026434898376, "grad_norm": 11.89194393157959, "learning_rate": 1.2271688498291334e-07, "logits/chosen": 1.5139234066009521, "logits/rejected": 1.562455415725708, "logps/chosen": -190.8651885986328, "logps/ref_chosen": -91.7751693725586, "logps/ref_rejected": -90.2679443359375, "logps/rejected": -225.98965454101562, "loss": 1.2149, "margin_dpo/margin_mean": 36.63169479370117, "margin_dpo/margin_std": 77.32295989990234, "step": 466 }, { "epoch": 0.7059712773998488, "fcm_dpo/beta": 0.007246088702231646, "fcm_dpo/delta": -0.0002483353018760681, "fcm_dpo/margin": 55.23102569580078, "fcm_dpo/q_t": 0.4089811146259308, "grad_norm": 7.205003261566162, "learning_rate": 1.2158065210664848e-07, "logits/chosen": 1.6517345905303955, "logits/rejected": 1.2684385776519775, "logps/chosen": -141.467041015625, "logps/ref_chosen": -64.77557373046875, "logps/ref_rejected": -102.58863830566406, "logps/rejected": -234.51113891601562, "loss": 1.0982, "margin_dpo/margin_mean": 55.23102569580078, "margin_dpo/margin_std": 77.0743408203125, "step": 467 }, { "epoch": 0.7074829931972789, "fcm_dpo/beta": 0.007257150486111641, "fcm_dpo/delta": -0.006373733282089233, "fcm_dpo/margin": 55.92658233642578, "fcm_dpo/q_t": 0.4085814356803894, "grad_norm": 9.413773536682129, "learning_rate": 1.204480113956011e-07, "logits/chosen": 1.6140985488891602, "logits/rejected": 1.6181747913360596, "logps/chosen": -170.14039611816406, "logps/ref_chosen": -82.22445678710938, "logps/ref_rejected": -92.99041748046875, "logps/rejected": -236.83291625976562, "loss": 1.1214, "margin_dpo/margin_mean": 55.92658233642578, "margin_dpo/margin_std": 89.77207946777344, "step": 468 }, { "epoch": 0.708994708994709, "fcm_dpo/beta": 0.007210570853203535, "fcm_dpo/delta": 0.03220885246992111, "fcm_dpo/margin": 51.060340881347656, "fcm_dpo/q_t": 0.41571545600891113, "grad_norm": 17.996444702148438, "learning_rate": 1.1931899453216697e-07, "logits/chosen": 1.7104898691177368, "logits/rejected": 1.6818902492523193, "logps/chosen": -161.86837768554688, "logps/ref_chosen": -75.93031311035156, "logps/ref_rejected": -92.26559448242188, "logps/rejected": -229.2639923095703, "loss": 1.1197, "margin_dpo/margin_mean": 51.060340881347656, "margin_dpo/margin_std": 71.57931518554688, "step": 469 }, { "epoch": 0.7105064247921391, "fcm_dpo/beta": 0.0072533609345555305, "fcm_dpo/delta": -0.014294777996838093, "fcm_dpo/margin": 57.032936096191406, "fcm_dpo/q_t": 0.40608060359954834, "grad_norm": 6.177210330963135, "learning_rate": 1.1819363309737438e-07, "logits/chosen": 1.6594834327697754, "logits/rejected": 1.4640038013458252, "logps/chosen": -148.8472900390625, "logps/ref_chosen": -65.86345672607422, "logps/ref_rejected": -85.89832305908203, "logps/rejected": -225.91510009765625, "loss": 1.1063, "margin_dpo/margin_mean": 57.032936096191406, "margin_dpo/margin_std": 86.32967376708984, "step": 470 }, { "epoch": 0.7120181405895691, "fcm_dpo/beta": 0.007146833464503288, "fcm_dpo/delta": -0.09884218871593475, "fcm_dpo/margin": 69.11367797851562, "fcm_dpo/q_t": 0.3864142894744873, "grad_norm": 7.910089015960693, "learning_rate": 1.1707195857000215e-07, "logits/chosen": 1.5441298484802246, "logits/rejected": 1.4003360271453857, "logps/chosen": -151.22935485839844, "logps/ref_chosen": -74.3460922241211, "logps/ref_rejected": -93.43672943115234, "logps/rejected": -239.4336700439453, "loss": 1.0383, "margin_dpo/margin_mean": 69.11367797851562, "margin_dpo/margin_std": 84.86093139648438, "step": 471 }, { "epoch": 0.7135298563869993, "fcm_dpo/beta": 0.007175215519964695, "fcm_dpo/delta": 0.021523576229810715, "fcm_dpo/margin": 52.75737380981445, "fcm_dpo/q_t": 0.41504496335983276, "grad_norm": 7.508319854736328, "learning_rate": 1.1595400232569768e-07, "logits/chosen": 1.6507457494735718, "logits/rejected": 1.5339549779891968, "logps/chosen": -153.89007568359375, "logps/ref_chosen": -74.75674438476562, "logps/ref_rejected": -95.18183135986328, "logps/rejected": -227.07254028320312, "loss": 1.1488, "margin_dpo/margin_mean": 52.75737380981445, "margin_dpo/margin_std": 92.27010345458984, "step": 472 }, { "epoch": 0.7150415721844293, "fcm_dpo/beta": 0.00716636935248971, "fcm_dpo/delta": 0.0196013655513525, "fcm_dpo/margin": 53.183753967285156, "fcm_dpo/q_t": 0.41339337825775146, "grad_norm": 13.5819091796875, "learning_rate": 1.1483979563610069e-07, "logits/chosen": 1.4470967054367065, "logits/rejected": 1.1765843629837036, "logps/chosen": -144.18069458007812, "logps/ref_chosen": -71.65933227539062, "logps/ref_rejected": -109.99200439453125, "logps/rejected": -235.69711303710938, "loss": 1.1334, "margin_dpo/margin_mean": 53.183753967285156, "margin_dpo/margin_std": 87.66374206542969, "step": 473 }, { "epoch": 0.7165532879818595, "fcm_dpo/beta": 0.0072361379861831665, "fcm_dpo/delta": 0.07070795446634293, "fcm_dpo/margin": 45.83869934082031, "fcm_dpo/q_t": 0.42501258850097656, "grad_norm": 10.908045768737793, "learning_rate": 1.1372936966796709e-07, "logits/chosen": 1.7690794467926025, "logits/rejected": 1.540644884109497, "logps/chosen": -154.1907958984375, "logps/ref_chosen": -65.91990661621094, "logps/ref_rejected": -89.09432983398438, "logps/rejected": -223.20391845703125, "loss": 1.1675, "margin_dpo/margin_mean": 45.83869934082031, "margin_dpo/margin_std": 83.50460815429688, "step": 474 }, { "epoch": 0.7180650037792895, "fcm_dpo/beta": 0.007119235582649708, "fcm_dpo/delta": -0.13208799064159393, "fcm_dpo/margin": 73.72787475585938, "fcm_dpo/q_t": 0.37819790840148926, "grad_norm": 6.977424621582031, "learning_rate": 1.126227554822985e-07, "logits/chosen": 1.305135726928711, "logits/rejected": 1.2358264923095703, "logps/chosen": -163.15887451171875, "logps/ref_chosen": -79.02459716796875, "logps/ref_rejected": -107.33058166503906, "logps/rejected": -265.1927185058594, "loss": 0.9995, "margin_dpo/margin_mean": 73.72787475585938, "margin_dpo/margin_std": 77.44041442871094, "step": 475 }, { "epoch": 0.7195767195767195, "fcm_dpo/beta": 0.0071355425752699375, "fcm_dpo/delta": 0.04274769499897957, "fcm_dpo/margin": 50.27285385131836, "fcm_dpo/q_t": 0.4180905520915985, "grad_norm": 6.6651201248168945, "learning_rate": 1.1151998403347243e-07, "logits/chosen": 1.3839824199676514, "logits/rejected": 1.3616328239440918, "logps/chosen": -186.7384033203125, "logps/ref_chosen": -93.72602844238281, "logps/ref_rejected": -94.390625, "logps/rejected": -237.67584228515625, "loss": 1.1399, "margin_dpo/margin_mean": 50.272850036621094, "margin_dpo/margin_std": 82.08091735839844, "step": 476 }, { "epoch": 0.7210884353741497, "fcm_dpo/beta": 0.007278554607182741, "fcm_dpo/delta": 0.12292856723070145, "fcm_dpo/margin": 38.535186767578125, "fcm_dpo/q_t": 0.43708232045173645, "grad_norm": 7.889428615570068, "learning_rate": 1.1042108616837692e-07, "logits/chosen": 1.58128023147583, "logits/rejected": 1.484431266784668, "logps/chosen": -189.0590057373047, "logps/ref_chosen": -76.51399993896484, "logps/ref_rejected": -99.14356231689453, "logps/rejected": -250.2237548828125, "loss": 1.2394, "margin_dpo/margin_mean": 38.535194396972656, "margin_dpo/margin_std": 94.74784088134766, "step": 477 }, { "epoch": 0.7226001511715797, "fcm_dpo/beta": 0.007414321415126324, "fcm_dpo/delta": 0.09388057887554169, "fcm_dpo/margin": 41.69120788574219, "fcm_dpo/q_t": 0.43154817819595337, "grad_norm": 9.232916831970215, "learning_rate": 1.0932609262554746e-07, "logits/chosen": 1.4031165838241577, "logits/rejected": 1.3811558485031128, "logps/chosen": -168.47891235351562, "logps/ref_chosen": -77.95186614990234, "logps/ref_rejected": -69.77754211425781, "logps/rejected": -201.99581909179688, "loss": 1.2187, "margin_dpo/margin_mean": 41.69120788574219, "margin_dpo/margin_std": 95.53366088867188, "step": 478 }, { "epoch": 0.7241118669690099, "fcm_dpo/beta": 0.007548708468675613, "fcm_dpo/delta": 0.08790592849254608, "fcm_dpo/margin": 41.721351623535156, "fcm_dpo/q_t": 0.4289689362049103, "grad_norm": 9.109953880310059, "learning_rate": 1.0823503403430734e-07, "logits/chosen": 1.3494288921356201, "logits/rejected": 1.2263519763946533, "logps/chosen": -164.5086212158203, "logps/ref_chosen": -76.56551361083984, "logps/ref_rejected": -84.33758544921875, "logps/rejected": -214.00204467773438, "loss": 1.2174, "margin_dpo/margin_mean": 41.721351623535156, "margin_dpo/margin_std": 95.05142211914062, "step": 479 }, { "epoch": 0.7256235827664399, "fcm_dpo/beta": 0.007575510535389185, "fcm_dpo/delta": -0.04130330681800842, "fcm_dpo/margin": 57.996116638183594, "fcm_dpo/q_t": 0.4007851481437683, "grad_norm": 11.577788352966309, "learning_rate": 1.0714794091391072e-07, "logits/chosen": 1.257918119430542, "logits/rejected": 1.2538151741027832, "logps/chosen": -165.85855102539062, "logps/ref_chosen": -80.15884399414062, "logps/ref_rejected": -84.88697814941406, "logps/rejected": -228.58279418945312, "loss": 1.1053, "margin_dpo/margin_mean": 57.996116638183594, "margin_dpo/margin_std": 90.42449951171875, "step": 480 }, { "epoch": 0.72713529856387, "fcm_dpo/beta": 0.007553757634013891, "fcm_dpo/delta": 0.018801141530275345, "fcm_dpo/margin": 50.558876037597656, "fcm_dpo/q_t": 0.41458386182785034, "grad_norm": 8.55683422088623, "learning_rate": 1.0606484367268906e-07, "logits/chosen": 1.2538020610809326, "logits/rejected": 1.2463691234588623, "logps/chosen": -175.15777587890625, "logps/ref_chosen": -84.56254577636719, "logps/ref_rejected": -90.06451416015625, "logps/rejected": -231.2186279296875, "loss": 1.1431, "margin_dpo/margin_mean": 50.558876037597656, "margin_dpo/margin_std": 88.14126586914062, "step": 481 }, { "epoch": 0.7286470143613001, "fcm_dpo/beta": 0.007597423158586025, "fcm_dpo/delta": 0.01751277782022953, "fcm_dpo/margin": 50.41091537475586, "fcm_dpo/q_t": 0.415740966796875, "grad_norm": 6.092764377593994, "learning_rate": 1.0498577260720048e-07, "logits/chosen": 1.6070960760116577, "logits/rejected": 1.2131727933883667, "logps/chosen": -184.42752075195312, "logps/ref_chosen": -78.88141632080078, "logps/ref_rejected": -125.41990661621094, "logps/rejected": -281.3769226074219, "loss": 1.1738, "margin_dpo/margin_mean": 50.41091537475586, "margin_dpo/margin_std": 100.31005859375, "step": 482 }, { "epoch": 0.7301587301587301, "fcm_dpo/beta": 0.007529154419898987, "fcm_dpo/delta": -0.043737735599279404, "fcm_dpo/margin": 58.672149658203125, "fcm_dpo/q_t": 0.39993715286254883, "grad_norm": 10.599023818969727, "learning_rate": 1.0391075790138232e-07, "logits/chosen": 1.8194961547851562, "logits/rejected": 1.504716396331787, "logps/chosen": -157.94146728515625, "logps/ref_chosen": -72.690185546875, "logps/ref_rejected": -98.37237548828125, "logps/rejected": -242.29580688476562, "loss": 1.0924, "margin_dpo/margin_mean": 58.672149658203125, "margin_dpo/margin_std": 87.27351379394531, "step": 483 }, { "epoch": 0.7316704459561603, "fcm_dpo/beta": 0.007634430192410946, "fcm_dpo/delta": 0.08491643518209457, "fcm_dpo/margin": 41.58033752441406, "fcm_dpo/q_t": 0.42818331718444824, "grad_norm": 7.452567100524902, "learning_rate": 1.0283982962570681e-07, "logits/chosen": 1.5895819664001465, "logits/rejected": 1.5464513301849365, "logps/chosen": -158.65072631835938, "logps/ref_chosen": -73.98435974121094, "logps/ref_rejected": -89.99178314208984, "logps/rejected": -216.23846435546875, "loss": 1.1752, "margin_dpo/margin_mean": 41.58033752441406, "margin_dpo/margin_std": 76.23760986328125, "step": 484 }, { "epoch": 0.7331821617535903, "fcm_dpo/beta": 0.0076524196192622185, "fcm_dpo/delta": 0.046526044607162476, "fcm_dpo/margin": 46.33942413330078, "fcm_dpo/q_t": 0.4194195866584778, "grad_norm": 7.961450099945068, "learning_rate": 1.0177301773633992e-07, "logits/chosen": 1.7761842012405396, "logits/rejected": 1.681747555732727, "logps/chosen": -166.32699584960938, "logps/ref_chosen": -78.0927963256836, "logps/ref_rejected": -89.14010620117188, "logps/rejected": -223.7137451171875, "loss": 1.1499, "margin_dpo/margin_mean": 46.33942413330078, "margin_dpo/margin_std": 77.6761474609375, "step": 485 }, { "epoch": 0.7346938775510204, "fcm_dpo/beta": 0.007767094299197197, "fcm_dpo/delta": 0.02391296997666359, "fcm_dpo/margin": 48.51192092895508, "fcm_dpo/q_t": 0.41680556535720825, "grad_norm": 6.9542646408081055, "learning_rate": 1.007103520743035e-07, "logits/chosen": 1.6766993999481201, "logits/rejected": 1.3588069677352905, "logps/chosen": -170.6450653076172, "logps/ref_chosen": -73.74685668945312, "logps/ref_rejected": -107.752685546875, "logps/rejected": -253.16281127929688, "loss": 1.1663, "margin_dpo/margin_mean": 48.51192092895508, "margin_dpo/margin_std": 93.03460693359375, "step": 486 }, { "epoch": 0.7362055933484505, "fcm_dpo/beta": 0.007746942341327667, "fcm_dpo/delta": -0.003563001286238432, "fcm_dpo/margin": 52.0731201171875, "fcm_dpo/q_t": 0.41003572940826416, "grad_norm": 7.556940078735352, "learning_rate": 9.965186236464046e-08, "logits/chosen": 1.468240737915039, "logits/rejected": 1.282604455947876, "logps/chosen": -176.71240234375, "logps/ref_chosen": -79.57780456542969, "logps/ref_rejected": -102.2916259765625, "logps/rejected": -251.4993438720703, "loss": 1.1183, "margin_dpo/margin_mean": 52.0731201171875, "margin_dpo/margin_std": 83.100341796875, "step": 487 }, { "epoch": 0.7377173091458806, "fcm_dpo/beta": 0.007788753602653742, "fcm_dpo/delta": 0.0015588030219078064, "fcm_dpo/margin": 51.109127044677734, "fcm_dpo/q_t": 0.4141284227371216, "grad_norm": 20.479114532470703, "learning_rate": 9.859757821558337e-08, "logits/chosen": 1.504908561706543, "logits/rejected": 1.3210365772247314, "logps/chosen": -168.25746154785156, "logps/ref_chosen": -80.62767791748047, "logps/ref_rejected": -100.4541015625, "logps/rejected": -239.19302368164062, "loss": 1.1584, "margin_dpo/margin_mean": 51.10912322998047, "margin_dpo/margin_std": 96.0980453491211, "step": 488 }, { "epoch": 0.7392290249433107, "fcm_dpo/beta": 0.0077533237636089325, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 24.302478790283203, "fcm_dpo/q_t": 0.4580850899219513, "grad_norm": 13.592652320861816, "learning_rate": 9.754752911772615e-08, "logits/chosen": 1.2843654155731201, "logits/rejected": 1.170314073562622, "logps/chosen": -185.48910522460938, "logps/ref_chosen": -85.39521026611328, "logps/ref_rejected": -101.97309875488281, "logps/rejected": -226.36947631835938, "loss": 1.2946, "margin_dpo/margin_mean": 24.302478790283203, "margin_dpo/margin_std": 78.64244079589844, "step": 489 }, { "epoch": 0.7407407407407407, "fcm_dpo/beta": 0.007696731481701136, "fcm_dpo/delta": -0.017715483903884888, "fcm_dpo/margin": 54.10951614379883, "fcm_dpo/q_t": 0.4072234034538269, "grad_norm": 9.231451034545898, "learning_rate": 9.650174444319956e-08, "logits/chosen": 1.945656418800354, "logits/rejected": 1.905264973640442, "logps/chosen": -161.941650390625, "logps/ref_chosen": -77.75590515136719, "logps/ref_rejected": -88.98885345458984, "logps/rejected": -227.2841033935547, "loss": 1.1338, "margin_dpo/margin_mean": 54.10951614379883, "margin_dpo/margin_std": 93.16372680664062, "step": 490 }, { "epoch": 0.7422524565381708, "fcm_dpo/beta": 0.007663640193641186, "fcm_dpo/delta": -0.013870414346456528, "fcm_dpo/margin": 53.826045989990234, "fcm_dpo/q_t": 0.40752020478248596, "grad_norm": 8.06557559967041, "learning_rate": 9.546025344484868e-08, "logits/chosen": 1.3035305738449097, "logits/rejected": 1.1545779705047607, "logps/chosen": -157.3299560546875, "logps/ref_chosen": -74.33360290527344, "logps/ref_rejected": -91.4105224609375, "logps/rejected": -228.23292541503906, "loss": 1.1018, "margin_dpo/margin_mean": 53.826053619384766, "margin_dpo/margin_std": 78.12969970703125, "step": 491 }, { "epoch": 0.7437641723356009, "fcm_dpo/beta": 0.00771558191627264, "fcm_dpo/delta": 0.01436400692909956, "fcm_dpo/margin": 36.33604431152344, "fcm_dpo/q_t": 0.4382849335670471, "grad_norm": 12.463248252868652, "learning_rate": 9.442308525541589e-08, "logits/chosen": 1.6404715776443481, "logits/rejected": 1.3996877670288086, "logps/chosen": -193.6622314453125, "logps/ref_chosen": -85.14178466796875, "logps/ref_rejected": -103.44204711914062, "logps/rejected": -248.29855346679688, "loss": 1.2292, "margin_dpo/margin_mean": 36.33604431152344, "margin_dpo/margin_std": 85.1468734741211, "step": 492 }, { "epoch": 0.745275888133031, "fcm_dpo/beta": 0.007784358225762844, "fcm_dpo/delta": 0.026980683207511902, "fcm_dpo/margin": 48.0078010559082, "fcm_dpo/q_t": 0.41681572794914246, "grad_norm": 9.322619438171387, "learning_rate": 9.339026888672468e-08, "logits/chosen": 1.6487233638763428, "logits/rejected": 1.4259631633758545, "logps/chosen": -165.03854370117188, "logps/ref_chosen": -75.81439208984375, "logps/ref_rejected": -95.30766296386719, "logps/rejected": -232.53961181640625, "loss": 1.175, "margin_dpo/margin_mean": 48.00779724121094, "margin_dpo/margin_std": 94.79415130615234, "step": 493 }, { "epoch": 0.7467876039304611, "fcm_dpo/beta": 0.007813823409378529, "fcm_dpo/delta": 0.060058847069740295, "fcm_dpo/margin": 43.76633071899414, "fcm_dpo/q_t": 0.4248165488243103, "grad_norm": 10.274767875671387, "learning_rate": 9.236183322886945e-08, "logits/chosen": 1.3294415473937988, "logits/rejected": 1.2298643589019775, "logps/chosen": -182.89959716796875, "logps/ref_chosen": -93.83562469482422, "logps/ref_rejected": -112.21142578125, "logps/rejected": -245.04171752929688, "loss": 1.1956, "margin_dpo/margin_mean": 43.766326904296875, "margin_dpo/margin_std": 92.66915893554688, "step": 494 }, { "epoch": 0.7482993197278912, "fcm_dpo/beta": 0.007944008335471153, "fcm_dpo/delta": 0.050161540508270264, "fcm_dpo/margin": 44.22233200073242, "fcm_dpo/q_t": 0.4226170778274536, "grad_norm": 14.485008239746094, "learning_rate": 9.133780704940594e-08, "logits/chosen": 1.640815258026123, "logits/rejected": 1.4514687061309814, "logps/chosen": -155.20675659179688, "logps/ref_chosen": -68.52467346191406, "logps/ref_rejected": -89.65379333496094, "logps/rejected": -220.55819702148438, "loss": 1.1922, "margin_dpo/margin_mean": 44.22233581542969, "margin_dpo/margin_std": 91.24566650390625, "step": 495 }, { "epoch": 0.7498110355253212, "fcm_dpo/beta": 0.007922867313027382, "fcm_dpo/delta": -0.0013457629829645157, "fcm_dpo/margin": 50.63785934448242, "fcm_dpo/q_t": 0.41145336627960205, "grad_norm": 7.610332489013672, "learning_rate": 9.031821899254797e-08, "logits/chosen": 1.7609965801239014, "logits/rejected": 1.4800307750701904, "logps/chosen": -164.26663208007812, "logps/ref_chosen": -73.13618469238281, "logps/ref_rejected": -111.50930786132812, "logps/rejected": -253.27760314941406, "loss": 1.1438, "margin_dpo/margin_mean": 50.63785934448242, "margin_dpo/margin_std": 89.98391723632812, "step": 496 }, { "epoch": 0.7513227513227513, "fcm_dpo/beta": 0.007826825603842735, "fcm_dpo/delta": -0.06445329636335373, "fcm_dpo/margin": 58.869712829589844, "fcm_dpo/q_t": 0.39559072256088257, "grad_norm": 12.443891525268555, "learning_rate": 8.930309757836516e-08, "logits/chosen": 1.779658555984497, "logits/rejected": 1.6461775302886963, "logps/chosen": -184.48745727539062, "logps/ref_chosen": -88.71475219726562, "logps/ref_rejected": -105.74935913085938, "logps/rejected": -260.39178466796875, "loss": 1.0692, "margin_dpo/margin_mean": 58.869712829589844, "margin_dpo/margin_std": 79.18601989746094, "step": 497 }, { "epoch": 0.7528344671201814, "fcm_dpo/beta": 0.007826920598745346, "fcm_dpo/delta": -0.021468058228492737, "fcm_dpo/margin": 53.729766845703125, "fcm_dpo/q_t": 0.4065396189689636, "grad_norm": 6.615968227386475, "learning_rate": 8.829247120198563e-08, "logits/chosen": 1.4853155612945557, "logits/rejected": 1.4078007936477661, "logps/chosen": -163.5498046875, "logps/ref_chosen": -83.3353271484375, "logps/ref_rejected": -89.34941864013672, "logps/rejected": -223.29367065429688, "loss": 1.1105, "margin_dpo/margin_mean": 53.729766845703125, "margin_dpo/margin_std": 84.90679931640625, "step": 498 }, { "epoch": 0.7543461829176115, "fcm_dpo/beta": 0.007880712859332561, "fcm_dpo/delta": 0.05969448387622833, "fcm_dpo/margin": 43.433326721191406, "fcm_dpo/q_t": 0.425603449344635, "grad_norm": 10.157211303710938, "learning_rate": 8.728636813280163e-08, "logits/chosen": 1.4775807857513428, "logits/rejected": 1.2861871719360352, "logps/chosen": -172.53701782226562, "logps/ref_chosen": -79.373779296875, "logps/ref_rejected": -104.62533569335938, "logps/rejected": -241.22189331054688, "loss": 1.2216, "margin_dpo/margin_mean": 43.433326721191406, "margin_dpo/margin_std": 101.27740478515625, "step": 499 }, { "epoch": 0.7558578987150416, "fcm_dpo/beta": 0.007957592606544495, "fcm_dpo/delta": 0.006692651659250259, "fcm_dpo/margin": 49.34884262084961, "fcm_dpo/q_t": 0.41281697154045105, "grad_norm": 8.680094718933105, "learning_rate": 8.628481651367875e-08, "logits/chosen": 1.2721607685089111, "logits/rejected": 1.2470263242721558, "logps/chosen": -177.0205078125, "logps/ref_chosen": -85.953857421875, "logps/ref_rejected": -90.40995788574219, "logps/rejected": -230.82546997070312, "loss": 1.1309, "margin_dpo/margin_mean": 49.34884262084961, "margin_dpo/margin_std": 80.7742919921875, "step": 500 }, { "epoch": 0.7558578987150416, "eval_fcm_dpo/beta": 0.007908526808023453, "eval_logits/chosen": 1.5455379486083984, "eval_logits/rejected": 1.4084484577178955, "eval_logps/chosen": -169.92417907714844, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -230.74034118652344, "eval_loss": 0.5676109790802002, "eval_margin_dpo/margin_mean": 51.021575927734375, "eval_margin_dpo/margin_std": 86.49874114990234, "eval_runtime": 42.2831, "eval_samples_per_second": 54.466, "eval_steps_per_second": 1.703, "step": 500 }, { "epoch": 0.7573696145124716, "fcm_dpo/beta": 0.008018089458346367, "fcm_dpo/delta": 0.06569428741931915, "fcm_dpo/margin": 41.89936065673828, "fcm_dpo/q_t": 0.4245343804359436, "grad_norm": 7.522352695465088, "learning_rate": 8.528784436016878e-08, "logits/chosen": 1.4151782989501953, "logits/rejected": 1.424743413925171, "logps/chosen": -168.38230895996094, "logps/ref_chosen": -81.22268676757812, "logps/ref_rejected": -86.97892761230469, "logps/rejected": -216.03790283203125, "loss": 1.1462, "margin_dpo/margin_mean": 41.89936065673828, "margin_dpo/margin_std": 66.22032165527344, "step": 501 }, { "epoch": 0.7588813303099018, "fcm_dpo/beta": 0.008128372952342033, "fcm_dpo/delta": 0.08888862282037735, "fcm_dpo/margin": 38.60469436645508, "fcm_dpo/q_t": 0.42894744873046875, "grad_norm": 10.359253883361816, "learning_rate": 8.4295479559726e-08, "logits/chosen": 1.2615420818328857, "logits/rejected": 1.1646403074264526, "logps/chosen": -168.2417755126953, "logps/ref_chosen": -83.1567611694336, "logps/ref_rejected": -106.74440002441406, "logps/rejected": -230.4341278076172, "loss": 1.1725, "margin_dpo/margin_mean": 38.604698181152344, "margin_dpo/margin_std": 69.67572784423828, "step": 502 }, { "epoch": 0.7603930461073318, "fcm_dpo/beta": 0.008050322532653809, "fcm_dpo/delta": -0.08986638486385345, "fcm_dpo/margin": 60.310672760009766, "fcm_dpo/q_t": 0.3902174234390259, "grad_norm": 7.535645484924316, "learning_rate": 8.330774987092712e-08, "logits/chosen": 1.522834062576294, "logits/rejected": 1.5648481845855713, "logps/chosen": -146.69717407226562, "logps/ref_chosen": -68.51583862304688, "logps/ref_rejected": -75.02178955078125, "logps/rejected": -213.5137939453125, "loss": 1.0616, "margin_dpo/margin_mean": 60.31067657470703, "margin_dpo/margin_std": 82.89733123779297, "step": 503 }, { "epoch": 0.7619047619047619, "fcm_dpo/beta": 0.007913358509540558, "fcm_dpo/delta": -0.12686683237552643, "fcm_dpo/margin": 65.74114990234375, "fcm_dpo/q_t": 0.3832342326641083, "grad_norm": 7.837214469909668, "learning_rate": 8.232468292269479e-08, "logits/chosen": 1.6095514297485352, "logits/rejected": 1.5772470235824585, "logps/chosen": -164.8587646484375, "logps/ref_chosen": -85.15829467773438, "logps/ref_rejected": -96.16879272460938, "logps/rejected": -241.6103973388672, "loss": 1.0304, "margin_dpo/margin_mean": 65.74114990234375, "margin_dpo/margin_std": 80.77653503417969, "step": 504 }, { "epoch": 0.763416477702192, "fcm_dpo/beta": 0.007828344590961933, "fcm_dpo/delta": 0.011198666878044605, "fcm_dpo/margin": 34.941402435302734, "fcm_dpo/q_t": 0.4389588236808777, "grad_norm": 18.425174713134766, "learning_rate": 8.134630621352483e-08, "logits/chosen": 1.3816214799880981, "logits/rejected": 1.2274703979492188, "logps/chosen": -167.62496948242188, "logps/ref_chosen": -79.26185607910156, "logps/ref_rejected": -96.34947967529297, "logps/rejected": -219.65399169921875, "loss": 1.2728, "margin_dpo/margin_mean": 34.941402435302734, "margin_dpo/margin_std": 98.05575561523438, "step": 505 }, { "epoch": 0.764928193499622, "fcm_dpo/beta": 0.007884200662374496, "fcm_dpo/delta": 0.059605300426483154, "fcm_dpo/margin": 43.43827819824219, "fcm_dpo/q_t": 0.42400920391082764, "grad_norm": 8.992940902709961, "learning_rate": 8.037264711071698e-08, "logits/chosen": 1.498244047164917, "logits/rejected": 1.4761946201324463, "logps/chosen": -175.23486328125, "logps/ref_chosen": -88.192626953125, "logps/ref_rejected": -100.86880493164062, "logps/rejected": -231.3493194580078, "loss": 1.2296, "margin_dpo/margin_mean": 43.43827819824219, "margin_dpo/margin_std": 104.13943481445312, "step": 506 }, { "epoch": 0.7664399092970522, "fcm_dpo/beta": 0.00795232504606247, "fcm_dpo/delta": 0.02266060933470726, "fcm_dpo/margin": 47.55640411376953, "fcm_dpo/q_t": 0.41735711693763733, "grad_norm": 7.954833507537842, "learning_rate": 7.940373284960933e-08, "logits/chosen": 1.1958003044128418, "logits/rejected": 1.0543439388275146, "logps/chosen": -181.73443603515625, "logps/ref_chosen": -86.04632568359375, "logps/ref_rejected": -111.44412994384766, "logps/rejected": -254.68865966796875, "loss": 1.1584, "margin_dpo/margin_mean": 47.5564079284668, "margin_dpo/margin_std": 88.17842102050781, "step": 507 }, { "epoch": 0.7679516250944822, "fcm_dpo/beta": 0.007975887507200241, "fcm_dpo/delta": 0.0014539193361997604, "fcm_dpo/margin": 49.96595001220703, "fcm_dpo/q_t": 0.4134010374546051, "grad_norm": 9.42202091217041, "learning_rate": 7.843959053281663e-08, "logits/chosen": 1.2900254726409912, "logits/rejected": 0.9407453536987305, "logps/chosen": -164.1695556640625, "logps/ref_chosen": -79.25038146972656, "logps/ref_rejected": -118.49089813232422, "logps/rejected": -253.37603759765625, "loss": 1.1535, "margin_dpo/margin_mean": 49.96595001220703, "margin_dpo/margin_std": 93.14352416992188, "step": 508 }, { "epoch": 0.7694633408919124, "fcm_dpo/beta": 0.007849051617085934, "fcm_dpo/delta": -0.06434465199708939, "fcm_dpo/margin": 58.69524383544922, "fcm_dpo/q_t": 0.39460235834121704, "grad_norm": 8.181747436523438, "learning_rate": 7.748024712947204e-08, "logits/chosen": 1.2476528882980347, "logits/rejected": 1.179711103439331, "logps/chosen": -157.0579376220703, "logps/ref_chosen": -80.7039566040039, "logps/ref_rejected": -90.50444793701172, "logps/rejected": -225.55368041992188, "loss": 1.0765, "margin_dpo/margin_mean": 58.69524002075195, "margin_dpo/margin_std": 82.17289733886719, "step": 509 }, { "epoch": 0.7709750566893424, "fcm_dpo/beta": 0.0077163660898804665, "fcm_dpo/delta": -0.0882585346698761, "fcm_dpo/margin": 62.59996795654297, "fcm_dpo/q_t": 0.39215773344039917, "grad_norm": 7.753228187561035, "learning_rate": 7.652572947447272e-08, "logits/chosen": 1.672702670097351, "logits/rejected": 1.4109150171279907, "logps/chosen": -145.37037658691406, "logps/ref_chosen": -67.64491271972656, "logps/ref_rejected": -108.92274475097656, "logps/rejected": -249.2481689453125, "loss": 1.0962, "margin_dpo/margin_mean": 62.59996795654297, "margin_dpo/margin_std": 98.73306274414062, "step": 510 }, { "epoch": 0.7724867724867724, "fcm_dpo/beta": 0.00761351129040122, "fcm_dpo/delta": -0.11252570152282715, "fcm_dpo/margin": 66.58784484863281, "fcm_dpo/q_t": 0.3877149224281311, "grad_norm": 8.772534370422363, "learning_rate": 7.557606426772961e-08, "logits/chosen": 1.4028170108795166, "logits/rejected": 1.2960667610168457, "logps/chosen": -158.74371337890625, "logps/ref_chosen": -75.66263580322266, "logps/ref_rejected": -104.26296997070312, "logps/rejected": -253.931884765625, "loss": 1.0694, "margin_dpo/margin_mean": 66.58784484863281, "margin_dpo/margin_std": 97.47923278808594, "step": 511 }, { "epoch": 0.7739984882842026, "fcm_dpo/beta": 0.00763990543782711, "fcm_dpo/delta": 0.09289787709712982, "fcm_dpo/margin": 40.58736038208008, "fcm_dpo/q_t": 0.43213844299316406, "grad_norm": 9.161316871643066, "learning_rate": 7.463127807341966e-08, "logits/chosen": 1.1914927959442139, "logits/rejected": 1.1217591762542725, "logps/chosen": -163.9389190673828, "logps/ref_chosen": -79.31925964355469, "logps/ref_rejected": -82.22052001953125, "logps/rejected": -207.4275360107422, "loss": 1.2055, "margin_dpo/margin_mean": 40.58736038208008, "margin_dpo/margin_std": 87.80073547363281, "step": 512 }, { "epoch": 0.7755102040816326, "fcm_dpo/beta": 0.007657117675989866, "fcm_dpo/delta": 0.0008791796863079071, "fcm_dpo/margin": 52.076133728027344, "fcm_dpo/q_t": 0.4106717109680176, "grad_norm": 7.0931172370910645, "learning_rate": 7.369139731924401e-08, "logits/chosen": 1.6813411712646484, "logits/rejected": 1.5264880657196045, "logps/chosen": -152.16217041015625, "logps/ref_chosen": -72.02534484863281, "logps/ref_rejected": -86.56224060058594, "logps/rejected": -218.77520751953125, "loss": 1.1117, "margin_dpo/margin_mean": 52.07612991333008, "margin_dpo/margin_std": 78.48776245117188, "step": 513 }, { "epoch": 0.7770219198790628, "fcm_dpo/beta": 0.007597388233989477, "fcm_dpo/delta": -0.09755103290081024, "fcm_dpo/margin": 64.87118530273438, "fcm_dpo/q_t": 0.3874030113220215, "grad_norm": 8.100028038024902, "learning_rate": 7.275644829568747e-08, "logits/chosen": 1.5403776168823242, "logits/rejected": 1.4848439693450928, "logps/chosen": -169.60714721679688, "logps/ref_chosen": -84.94093322753906, "logps/ref_rejected": -102.44367980957031, "logps/rejected": -251.98109436035156, "loss": 1.0548, "margin_dpo/margin_mean": 64.87118530273438, "margin_dpo/margin_std": 87.02154541015625, "step": 514 }, { "epoch": 0.7785336356764928, "fcm_dpo/beta": 0.007553492672741413, "fcm_dpo/delta": 0.01082646381109953, "fcm_dpo/margin": 51.578346252441406, "fcm_dpo/q_t": 0.41361895203590393, "grad_norm": 13.685517311096191, "learning_rate": 7.182645715528435e-08, "logits/chosen": 1.8065004348754883, "logits/rejected": 1.5909342765808105, "logps/chosen": -162.07525634765625, "logps/ref_chosen": -72.9662094116211, "logps/ref_rejected": -102.53651428222656, "logps/rejected": -243.22390747070312, "loss": 1.1433, "margin_dpo/margin_mean": 51.57834243774414, "margin_dpo/margin_std": 90.6058349609375, "step": 515 }, { "epoch": 0.780045351473923, "fcm_dpo/beta": 0.0076246303506195545, "fcm_dpo/delta": 0.048703186213970184, "fcm_dpo/margin": 46.287757873535156, "fcm_dpo/q_t": 0.41831153631210327, "grad_norm": 15.306377410888672, "learning_rate": 7.090144991188568e-08, "logits/chosen": 1.5644421577453613, "logits/rejected": 1.3997728824615479, "logps/chosen": -148.07150268554688, "logps/ref_chosen": -76.63414001464844, "logps/ref_rejected": -91.01750183105469, "logps/rejected": -208.74264526367188, "loss": 1.1623, "margin_dpo/margin_mean": 46.28776168823242, "margin_dpo/margin_std": 84.75020599365234, "step": 516 }, { "epoch": 0.781557067271353, "fcm_dpo/beta": 0.007668118923902512, "fcm_dpo/delta": 0.044410329312086105, "fcm_dpo/margin": 31.92813491821289, "fcm_dpo/q_t": 0.44710972905158997, "grad_norm": 8.951191902160645, "learning_rate": 6.998145243993284e-08, "logits/chosen": 1.6496453285217285, "logits/rejected": 1.6587225198745728, "logps/chosen": -174.9427032470703, "logps/ref_chosen": -77.06817626953125, "logps/ref_rejected": -80.048583984375, "logps/rejected": -209.8512420654297, "loss": 1.2711, "margin_dpo/margin_mean": 31.928133010864258, "margin_dpo/margin_std": 91.1561279296875, "step": 517 }, { "epoch": 0.783068783068783, "fcm_dpo/beta": 0.00768995750695467, "fcm_dpo/delta": -0.0029140058904886246, "fcm_dpo/margin": 52.375755310058594, "fcm_dpo/q_t": 0.4096565544605255, "grad_norm": 9.834497451782227, "learning_rate": 6.906649047373245e-08, "logits/chosen": 1.4923443794250488, "logits/rejected": 1.3131781816482544, "logps/chosen": -149.75588989257812, "logps/ref_chosen": -78.69026184082031, "logps/ref_rejected": -97.58124542236328, "logps/rejected": -221.02264404296875, "loss": 1.1372, "margin_dpo/margin_mean": 52.375755310058594, "margin_dpo/margin_std": 90.88835144042969, "step": 518 }, { "epoch": 0.7845804988662132, "fcm_dpo/beta": 0.007714019622653723, "fcm_dpo/delta": 0.02117512747645378, "fcm_dpo/margin": 31.934837341308594, "fcm_dpo/q_t": 0.4449980854988098, "grad_norm": 8.429184913635254, "learning_rate": 6.815658960673781e-08, "logits/chosen": 1.5910195112228394, "logits/rejected": 1.467621088027954, "logps/chosen": -176.11871337890625, "logps/ref_chosen": -78.35087585449219, "logps/ref_rejected": -95.79212188720703, "logps/rejected": -225.4947967529297, "loss": 1.3203, "margin_dpo/margin_mean": 31.934839248657227, "margin_dpo/margin_std": 107.92922973632812, "step": 519 }, { "epoch": 0.7860922146636432, "fcm_dpo/beta": 0.007716222666203976, "fcm_dpo/delta": 0.03524667024612427, "fcm_dpo/margin": 47.34004211425781, "fcm_dpo/q_t": 0.41651666164398193, "grad_norm": 8.534330368041992, "learning_rate": 6.725177529083209e-08, "logits/chosen": 1.6108531951904297, "logits/rejected": 1.4225859642028809, "logps/chosen": -163.86770629882812, "logps/ref_chosen": -80.40513610839844, "logps/ref_rejected": -93.02791595458984, "logps/rejected": -223.83053588867188, "loss": 1.1325, "margin_dpo/margin_mean": 47.34003829956055, "margin_dpo/margin_std": 72.44415283203125, "step": 520 }, { "epoch": 0.7876039304610734, "fcm_dpo/beta": 0.007807170040905476, "fcm_dpo/delta": -0.018284276127815247, "fcm_dpo/margin": 53.414710998535156, "fcm_dpo/q_t": 0.4060171842575073, "grad_norm": 8.058144569396973, "learning_rate": 6.63520728356167e-08, "logits/chosen": 1.2749577760696411, "logits/rejected": 1.0321365594863892, "logps/chosen": -174.77325439453125, "logps/ref_chosen": -86.5218276977539, "logps/ref_rejected": -109.20257568359375, "logps/rejected": -250.86871337890625, "loss": 1.1165, "margin_dpo/margin_mean": 53.414710998535156, "margin_dpo/margin_std": 84.53794860839844, "step": 521 }, { "epoch": 0.7891156462585034, "fcm_dpo/beta": 0.007875899784266949, "fcm_dpo/delta": 0.10553185641765594, "fcm_dpo/margin": 37.79329299926758, "fcm_dpo/q_t": 0.4356454014778137, "grad_norm": 9.020870208740234, "learning_rate": 6.545750740770336e-08, "logits/chosen": 1.2026581764221191, "logits/rejected": 1.2267944812774658, "logps/chosen": -168.311767578125, "logps/ref_chosen": -78.24254608154297, "logps/ref_rejected": -85.23554992675781, "logps/rejected": -213.09805297851562, "loss": 1.2329, "margin_dpo/margin_mean": 37.793296813964844, "margin_dpo/margin_std": 92.2118911743164, "step": 522 }, { "epoch": 0.7906273620559335, "fcm_dpo/beta": 0.007950318977236748, "fcm_dpo/delta": 0.054845184087753296, "fcm_dpo/margin": 43.630802154541016, "fcm_dpo/q_t": 0.4217304587364197, "grad_norm": 10.220833778381348, "learning_rate": 6.456810403001012e-08, "logits/chosen": 1.435563087463379, "logits/rejected": 1.1382131576538086, "logps/chosen": -179.95175170898438, "logps/ref_chosen": -83.50096893310547, "logps/ref_rejected": -117.45217895507812, "logps/rejected": -257.53375244140625, "loss": 1.1755, "margin_dpo/margin_mean": 43.63079833984375, "margin_dpo/margin_std": 83.78480529785156, "step": 523 }, { "epoch": 0.7921390778533636, "fcm_dpo/beta": 0.008069904521107674, "fcm_dpo/delta": 0.020586442202329636, "fcm_dpo/margin": 47.05378723144531, "fcm_dpo/q_t": 0.41369175910949707, "grad_norm": 8.774334907531738, "learning_rate": 6.368388758106134e-08, "logits/chosen": 1.3084217309951782, "logits/rejected": 1.2617387771606445, "logps/chosen": -165.7652130126953, "logps/ref_chosen": -93.22590637207031, "logps/ref_rejected": -108.17863464355469, "logps/rejected": -227.771728515625, "loss": 1.1598, "margin_dpo/margin_mean": 47.05378723144531, "margin_dpo/margin_std": 87.1861343383789, "step": 524 }, { "epoch": 0.7936507936507936, "fcm_dpo/beta": 0.008036179468035698, "fcm_dpo/delta": -0.003930669277906418, "fcm_dpo/margin": 30.131698608398438, "fcm_dpo/q_t": 0.4459991455078125, "grad_norm": 10.040858268737793, "learning_rate": 6.280488279429185e-08, "logits/chosen": 0.972198486328125, "logits/rejected": 0.968452513217926, "logps/chosen": -189.3087158203125, "logps/ref_chosen": -94.08831787109375, "logps/ref_rejected": -100.682373046875, "logps/rejected": -226.03445434570312, "loss": 1.2665, "margin_dpo/margin_mean": 30.131698608398438, "margin_dpo/margin_std": 81.79655456542969, "step": 525 }, { "epoch": 0.7951625094482238, "fcm_dpo/beta": 0.007959752343595028, "fcm_dpo/delta": -0.047779276967048645, "fcm_dpo/margin": 39.6178092956543, "fcm_dpo/q_t": 0.42807871103286743, "grad_norm": 12.535789489746094, "learning_rate": 6.193111425735515e-08, "logits/chosen": 1.7268145084381104, "logits/rejected": 1.5319491624832153, "logps/chosen": -172.51361083984375, "logps/ref_chosen": -77.78373718261719, "logps/ref_rejected": -100.29583740234375, "logps/rejected": -234.64352416992188, "loss": 1.1984, "margin_dpo/margin_mean": 39.61780548095703, "margin_dpo/margin_std": 80.48246765136719, "step": 526 }, { "epoch": 0.7966742252456538, "fcm_dpo/beta": 0.008134608156979084, "fcm_dpo/delta": 0.12869685888290405, "fcm_dpo/margin": 33.732025146484375, "fcm_dpo/q_t": 0.4389396607875824, "grad_norm": 12.428853988647461, "learning_rate": 6.106260641143546e-08, "logits/chosen": 1.9265590906143188, "logits/rejected": 1.6606621742248535, "logps/chosen": -182.11387634277344, "logps/ref_chosen": -76.695068359375, "logps/ref_rejected": -107.68281555175781, "logps/rejected": -246.83364868164062, "loss": 1.2233, "margin_dpo/margin_mean": 33.732025146484375, "margin_dpo/margin_std": 77.11158752441406, "step": 527 }, { "epoch": 0.7981859410430839, "fcm_dpo/beta": 0.008256562054157257, "fcm_dpo/delta": 0.07902608066797256, "fcm_dpo/margin": 39.19126892089844, "fcm_dpo/q_t": 0.42815977334976196, "grad_norm": 13.860108375549316, "learning_rate": 6.019938355056422e-08, "logits/chosen": 1.4714868068695068, "logits/rejected": 1.2811636924743652, "logps/chosen": -157.07546997070312, "logps/ref_chosen": -75.0361328125, "logps/ref_rejected": -94.67579650878906, "logps/rejected": -215.90640258789062, "loss": 1.2167, "margin_dpo/margin_mean": 39.19126892089844, "margin_dpo/margin_std": 90.16360473632812, "step": 528 }, { "epoch": 0.799697656840514, "fcm_dpo/beta": 0.007970629259943962, "fcm_dpo/delta": -0.23854364454746246, "fcm_dpo/margin": 78.05899810791016, "fcm_dpo/q_t": 0.3612556457519531, "grad_norm": 5.713718414306641, "learning_rate": 5.934146982094049e-08, "logits/chosen": 1.3859180212020874, "logits/rejected": 1.2408912181854248, "logps/chosen": -147.29428100585938, "logps/ref_chosen": -72.84869384765625, "logps/ref_rejected": -93.25855255126953, "logps/rejected": -245.76315307617188, "loss": 0.979, "margin_dpo/margin_mean": 78.05899810791016, "margin_dpo/margin_std": 91.38536071777344, "step": 529 }, { "epoch": 0.8012093726379441, "fcm_dpo/beta": 0.007819782942533493, "fcm_dpo/delta": -0.05217093229293823, "fcm_dpo/margin": 57.46356964111328, "fcm_dpo/q_t": 0.4001220464706421, "grad_norm": 7.648416996002197, "learning_rate": 5.848888922025552e-08, "logits/chosen": 1.516282558441162, "logits/rejected": 1.4072870016098022, "logps/chosen": -155.42422485351562, "logps/ref_chosen": -79.4971694946289, "logps/ref_rejected": -93.59564208984375, "logps/rejected": -226.98626708984375, "loss": 1.0785, "margin_dpo/margin_mean": 57.46356201171875, "margin_dpo/margin_std": 80.2945327758789, "step": 530 }, { "epoch": 0.8027210884353742, "fcm_dpo/beta": 0.007811452262103558, "fcm_dpo/delta": -0.022111359983682632, "fcm_dpo/margin": 41.554649353027344, "fcm_dpo/q_t": 0.42616134881973267, "grad_norm": 7.592843532562256, "learning_rate": 5.7641665597021435e-08, "logits/chosen": 1.5034434795379639, "logits/rejected": 1.340606689453125, "logps/chosen": -156.27935791015625, "logps/ref_chosen": -69.45396423339844, "logps/ref_rejected": -96.30017852783203, "logps/rejected": -224.68020629882812, "loss": 1.1692, "margin_dpo/margin_mean": 41.554649353027344, "margin_dpo/margin_std": 73.87443542480469, "step": 531 }, { "epoch": 0.8042328042328042, "fcm_dpo/beta": 0.007778765633702278, "fcm_dpo/delta": 0.014322098344564438, "fcm_dpo/margin": 49.61630630493164, "fcm_dpo/q_t": 0.41433119773864746, "grad_norm": 8.264182090759277, "learning_rate": 5.679982264990424e-08, "logits/chosen": 1.2624887228012085, "logits/rejected": 1.1222734451293945, "logps/chosen": -179.19329833984375, "logps/ref_chosen": -76.52011108398438, "logps/ref_rejected": -94.79593658447266, "logps/rejected": -247.08544921875, "loss": 1.1447, "margin_dpo/margin_mean": 49.61630630493164, "margin_dpo/margin_std": 86.81497192382812, "step": 532 }, { "epoch": 0.8057445200302343, "fcm_dpo/beta": 0.007856892421841621, "fcm_dpo/delta": 0.01767205074429512, "fcm_dpo/margin": 48.72636032104492, "fcm_dpo/q_t": 0.4150225520133972, "grad_norm": 9.463761329650879, "learning_rate": 5.596338392706076e-08, "logits/chosen": 1.8482627868652344, "logits/rejected": 1.6290550231933594, "logps/chosen": -143.453369140625, "logps/ref_chosen": -72.31800842285156, "logps/ref_rejected": -89.26652526855469, "logps/rejected": -209.12826538085938, "loss": 1.1506, "margin_dpo/margin_mean": 48.72636032104492, "margin_dpo/margin_std": 87.32180786132812, "step": 533 }, { "epoch": 0.8072562358276644, "fcm_dpo/beta": 0.007759833708405495, "fcm_dpo/delta": -0.03004516288638115, "fcm_dpo/margin": 55.12841033935547, "fcm_dpo/q_t": 0.40312954783439636, "grad_norm": 8.84256649017334, "learning_rate": 5.513237282548033e-08, "logits/chosen": 1.4910259246826172, "logits/rejected": 1.4467054605484009, "logps/chosen": -148.83465576171875, "logps/ref_chosen": -77.87559509277344, "logps/ref_rejected": -92.21171569824219, "logps/rejected": -218.29917907714844, "loss": 1.1012, "margin_dpo/margin_mean": 55.12840270996094, "margin_dpo/margin_std": 81.7762451171875, "step": 534 }, { "epoch": 0.8087679516250945, "fcm_dpo/beta": 0.00789577979594469, "fcm_dpo/delta": 0.0677485466003418, "fcm_dpo/margin": 42.331016540527344, "fcm_dpo/q_t": 0.42601895332336426, "grad_norm": 7.4685587882995605, "learning_rate": 5.430681259032957e-08, "logits/chosen": 1.137467622756958, "logits/rejected": 0.9851101636886597, "logps/chosen": -168.10491943359375, "logps/ref_chosen": -78.16358184814453, "logps/ref_rejected": -97.78164672851562, "logps/rejected": -230.0540008544922, "loss": 1.1951, "margin_dpo/margin_mean": 42.331016540527344, "margin_dpo/margin_std": 88.60259246826172, "step": 535 }, { "epoch": 0.8102796674225246, "fcm_dpo/beta": 0.0078198853880167, "fcm_dpo/delta": -0.08114311099052429, "fcm_dpo/margin": 61.04181671142578, "fcm_dpo/q_t": 0.3922014832496643, "grad_norm": 6.651211738586426, "learning_rate": 5.3486726314303175e-08, "logits/chosen": 1.4360030889511108, "logits/rejected": 1.2131242752075195, "logps/chosen": -145.51263427734375, "logps/ref_chosen": -66.65623474121094, "logps/ref_rejected": -89.49085998535156, "logps/rejected": -229.38906860351562, "loss": 1.0404, "margin_dpo/margin_mean": 61.04182434082031, "margin_dpo/margin_std": 73.56864166259766, "step": 536 }, { "epoch": 0.8117913832199547, "fcm_dpo/beta": 0.007776736747473478, "fcm_dpo/delta": 0.02339194528758526, "fcm_dpo/margin": 48.52098846435547, "fcm_dpo/q_t": 0.4167477488517761, "grad_norm": 6.910667896270752, "learning_rate": 5.267213693697695e-08, "logits/chosen": 1.7525105476379395, "logits/rejected": 1.4899015426635742, "logps/chosen": -174.97885131835938, "logps/ref_chosen": -74.99390411376953, "logps/ref_rejected": -110.6627197265625, "logps/rejected": -259.16864013671875, "loss": 1.1461, "margin_dpo/margin_mean": 48.52098846435547, "margin_dpo/margin_std": 84.7010498046875, "step": 537 }, { "epoch": 0.8133030990173847, "fcm_dpo/beta": 0.007822012528777122, "fcm_dpo/delta": -0.037601351737976074, "fcm_dpo/margin": 55.653079986572266, "fcm_dpo/q_t": 0.4035933315753937, "grad_norm": 6.632546424865723, "learning_rate": 5.1863067244167144e-08, "logits/chosen": 1.4656171798706055, "logits/rejected": 1.3803694248199463, "logps/chosen": -179.07177734375, "logps/ref_chosen": -87.61151123046875, "logps/ref_rejected": -98.1150131225586, "logps/rejected": -245.22833251953125, "loss": 1.1038, "margin_dpo/margin_mean": 55.653079986572266, "margin_dpo/margin_std": 84.92758178710938, "step": 538 }, { "epoch": 0.8148148148148148, "fcm_dpo/beta": 0.007802900858223438, "fcm_dpo/delta": 0.05218294635415077, "fcm_dpo/margin": 44.81316375732422, "fcm_dpo/q_t": 0.4222569763660431, "grad_norm": 8.919053077697754, "learning_rate": 5.105953986729195e-08, "logits/chosen": 1.5954413414001465, "logits/rejected": 1.256216049194336, "logps/chosen": -169.81942749023438, "logps/ref_chosen": -78.86482238769531, "logps/ref_rejected": -100.84349822998047, "logps/rejected": -236.61126708984375, "loss": 1.1518, "margin_dpo/margin_mean": 44.813167572021484, "margin_dpo/margin_std": 77.50193786621094, "step": 539 }, { "epoch": 0.8163265306122449, "fcm_dpo/beta": 0.007814684882760048, "fcm_dpo/delta": -0.043207671493291855, "fcm_dpo/margin": 56.455596923828125, "fcm_dpo/q_t": 0.4015369117259979, "grad_norm": 7.1346659660339355, "learning_rate": 5.026157728273966e-08, "logits/chosen": 1.431809425354004, "logits/rejected": 1.1509695053100586, "logps/chosen": -171.71527099609375, "logps/ref_chosen": -83.66409301757812, "logps/ref_rejected": -114.8860092163086, "logps/rejected": -259.39276123046875, "loss": 1.1013, "margin_dpo/margin_mean": 56.45559310913086, "margin_dpo/margin_std": 87.17050170898438, "step": 540 }, { "epoch": 0.817838246409675, "fcm_dpo/beta": 0.007632635533809662, "fcm_dpo/delta": -0.09920863062143326, "fcm_dpo/margin": 64.70645141601562, "fcm_dpo/q_t": 0.38797062635421753, "grad_norm": 6.952167510986328, "learning_rate": 4.9469201811239035e-08, "logits/chosen": 1.3315198421478271, "logits/rejected": 1.3776156902313232, "logps/chosen": -161.97076416015625, "logps/ref_chosen": -83.12225341796875, "logps/ref_rejected": -74.80526733398438, "logps/rejected": -218.3602294921875, "loss": 1.0381, "margin_dpo/margin_mean": 64.70645141601562, "margin_dpo/margin_std": 79.60298156738281, "step": 541 }, { "epoch": 0.8193499622071051, "fcm_dpo/beta": 0.007459321990609169, "fcm_dpo/delta": -0.1281038075685501, "fcm_dpo/margin": 69.86752319335938, "fcm_dpo/q_t": 0.38251811265945435, "grad_norm": 6.959683895111084, "learning_rate": 4.868243561723534e-08, "logits/chosen": 1.6677457094192505, "logits/rejected": 1.5448417663574219, "logps/chosen": -135.40603637695312, "logps/ref_chosen": -66.3132553100586, "logps/ref_rejected": -83.24588012695312, "logps/rejected": -222.2061767578125, "loss": 1.033, "margin_dpo/margin_mean": 69.86752319335938, "margin_dpo/margin_std": 89.15379333496094, "step": 542 }, { "epoch": 0.8208616780045351, "fcm_dpo/beta": 0.0074135130271315575, "fcm_dpo/delta": -0.023007860407233238, "fcm_dpo/margin": 56.92228317260742, "fcm_dpo/q_t": 0.4051409065723419, "grad_norm": 8.201580047607422, "learning_rate": 4.790130070827028e-08, "logits/chosen": 1.3210530281066895, "logits/rejected": 1.05372953414917, "logps/chosen": -150.39297485351562, "logps/ref_chosen": -68.11429595947266, "logps/ref_rejected": -94.62380981445312, "logps/rejected": -233.8247833251953, "loss": 1.1143, "margin_dpo/margin_mean": 56.92228698730469, "margin_dpo/margin_std": 91.03446960449219, "step": 543 }, { "epoch": 0.8223733938019653, "fcm_dpo/beta": 0.0073410142213106155, "fcm_dpo/delta": -0.03817659243941307, "fcm_dpo/margin": 59.45600891113281, "fcm_dpo/q_t": 0.4023340344429016, "grad_norm": 7.93362283706665, "learning_rate": 4.7125818934366454e-08, "logits/chosen": 1.6760622262954712, "logits/rejected": 1.4469034671783447, "logps/chosen": -167.59967041015625, "logps/ref_chosen": -81.187255859375, "logps/ref_rejected": -105.84722900390625, "logps/rejected": -251.71563720703125, "loss": 1.0968, "margin_dpo/margin_mean": 59.45600891113281, "margin_dpo/margin_std": 89.44548034667969, "step": 544 }, { "epoch": 0.8238851095993953, "fcm_dpo/beta": 0.007404550909996033, "fcm_dpo/delta": 0.08629242330789566, "fcm_dpo/margin": 42.74488067626953, "fcm_dpo/q_t": 0.4291386604309082, "grad_norm": 7.565622329711914, "learning_rate": 4.635601198741607e-08, "logits/chosen": 1.3802976608276367, "logits/rejected": 1.1980688571929932, "logps/chosen": -170.48873901367188, "logps/ref_chosen": -78.81717681884766, "logps/ref_rejected": -98.65876770019531, "logps/rejected": -233.07521057128906, "loss": 1.1938, "margin_dpo/margin_mean": 42.74488067626953, "margin_dpo/margin_std": 86.98738098144531, "step": 545 }, { "epoch": 0.8253968253968254, "fcm_dpo/beta": 0.0075421445071697235, "fcm_dpo/delta": 0.05457151308655739, "fcm_dpo/margin": 46.011558532714844, "fcm_dpo/q_t": 0.4226982593536377, "grad_norm": 7.30482816696167, "learning_rate": 4.559190140057428e-08, "logits/chosen": 1.5349187850952148, "logits/rejected": 1.5369720458984375, "logps/chosen": -154.75668334960938, "logps/ref_chosen": -74.2529296875, "logps/ref_rejected": -80.32308959960938, "logps/rejected": -206.83840942382812, "loss": 1.1535, "margin_dpo/margin_mean": 46.011558532714844, "margin_dpo/margin_std": 78.93978881835938, "step": 546 }, { "epoch": 0.8269085411942555, "fcm_dpo/beta": 0.007430747617036104, "fcm_dpo/delta": -0.08842149376869202, "fcm_dpo/margin": 65.12631225585938, "fcm_dpo/q_t": 0.38987237215042114, "grad_norm": 9.399640083312988, "learning_rate": 4.483350854765672e-08, "logits/chosen": 1.1584597826004028, "logits/rejected": 0.9976880550384521, "logps/chosen": -144.39535522460938, "logps/ref_chosen": -69.9368896484375, "logps/ref_rejected": -90.25672912597656, "logps/rejected": -229.84152221679688, "loss": 1.058, "margin_dpo/margin_mean": 65.12631225585938, "margin_dpo/margin_std": 86.3273696899414, "step": 547 }, { "epoch": 0.8284202569916855, "fcm_dpo/beta": 0.007527903653681278, "fcm_dpo/delta": 0.07334420830011368, "fcm_dpo/margin": 43.605491638183594, "fcm_dpo/q_t": 0.42615681886672974, "grad_norm": 11.522804260253906, "learning_rate": 4.4080854642541826e-08, "logits/chosen": 1.3264061212539673, "logits/rejected": 1.1371994018554688, "logps/chosen": -176.6475067138672, "logps/ref_chosen": -81.1605224609375, "logps/ref_rejected": -99.7246322631836, "logps/rejected": -238.81710815429688, "loss": 1.1772, "margin_dpo/margin_mean": 43.60548782348633, "margin_dpo/margin_std": 82.16632080078125, "step": 548 }, { "epoch": 0.8299319727891157, "fcm_dpo/beta": 0.007619916927069426, "fcm_dpo/delta": 0.0813249796628952, "fcm_dpo/margin": 42.149147033691406, "fcm_dpo/q_t": 0.4282255172729492, "grad_norm": 14.586856842041016, "learning_rate": 4.333396073857723e-08, "logits/chosen": 1.9198331832885742, "logits/rejected": 1.7107844352722168, "logps/chosen": -170.18838500976562, "logps/ref_chosen": -80.49800872802734, "logps/ref_rejected": -113.20750427246094, "logps/rejected": -245.04702758789062, "loss": 1.2076, "margin_dpo/margin_mean": 42.149147033691406, "margin_dpo/margin_std": 92.20016479492188, "step": 549 }, { "epoch": 0.8314436885865457, "fcm_dpo/beta": 0.007620522286742926, "fcm_dpo/delta": -0.01459870021790266, "fcm_dpo/margin": 40.88159942626953, "fcm_dpo/q_t": 0.42959415912628174, "grad_norm": 11.19546127319336, "learning_rate": 4.259284772799099e-08, "logits/chosen": 1.6284606456756592, "logits/rejected": 1.5507255792617798, "logps/chosen": -167.57029724121094, "logps/ref_chosen": -75.13760375976562, "logps/ref_rejected": -79.04876708984375, "logps/rejected": -212.36305236816406, "loss": 1.2136, "margin_dpo/margin_mean": 40.881595611572266, "margin_dpo/margin_std": 90.40715789794922, "step": 550 }, { "epoch": 0.8329554043839759, "fcm_dpo/beta": 0.007725295145064592, "fcm_dpo/delta": 0.06760243326425552, "fcm_dpo/margin": 43.256893157958984, "fcm_dpo/q_t": 0.42407017946243286, "grad_norm": 13.680648803710938, "learning_rate": 4.1857536341307176e-08, "logits/chosen": 1.9498487710952759, "logits/rejected": 1.7900187969207764, "logps/chosen": -177.50527954101562, "logps/ref_chosen": -85.4496078491211, "logps/ref_rejected": -103.48530578613281, "logps/rejected": -238.7978515625, "loss": 1.1569, "margin_dpo/margin_mean": 43.256893157958984, "margin_dpo/margin_std": 74.33500671386719, "step": 551 }, { "epoch": 0.8344671201814059, "fcm_dpo/beta": 0.007839495316147804, "fcm_dpo/delta": 0.06313008815050125, "fcm_dpo/margin": 43.128273010253906, "fcm_dpo/q_t": 0.42163771390914917, "grad_norm": 8.0436372756958, "learning_rate": 4.112804714676593e-08, "logits/chosen": 1.6295115947723389, "logits/rejected": 1.4652934074401855, "logps/chosen": -169.71463012695312, "logps/ref_chosen": -82.01036071777344, "logps/ref_rejected": -101.61884307861328, "logps/rejected": -232.45138549804688, "loss": 1.1854, "margin_dpo/margin_mean": 43.128273010253906, "margin_dpo/margin_std": 85.137451171875, "step": 552 }, { "epoch": 0.8359788359788359, "fcm_dpo/beta": 0.007808818481862545, "fcm_dpo/delta": -0.0008383337408304214, "fcm_dpo/margin": 51.32048797607422, "fcm_dpo/q_t": 0.41101810336112976, "grad_norm": 7.586859703063965, "learning_rate": 4.0404400549748144e-08, "logits/chosen": 1.6329729557037354, "logits/rejected": 1.3663992881774902, "logps/chosen": -168.96517944335938, "logps/ref_chosen": -73.81416320800781, "logps/ref_rejected": -104.27050018310547, "logps/rejected": -250.7420196533203, "loss": 1.1658, "margin_dpo/margin_mean": 51.32048797607422, "margin_dpo/margin_std": 98.8353271484375, "step": 553 }, { "epoch": 0.8374905517762661, "fcm_dpo/beta": 0.007749027572572231, "fcm_dpo/delta": -0.03996382653713226, "fcm_dpo/margin": 56.50690460205078, "fcm_dpo/q_t": 0.4017537832260132, "grad_norm": 8.198256492614746, "learning_rate": 3.968661679220467e-08, "logits/chosen": 1.1512682437896729, "logits/rejected": 1.1083042621612549, "logps/chosen": -169.98992919921875, "logps/ref_chosen": -81.43980407714844, "logps/ref_rejected": -89.32518005371094, "logps/rejected": -234.3822021484375, "loss": 1.105, "margin_dpo/margin_mean": 56.50690460205078, "margin_dpo/margin_std": 87.52005767822266, "step": 554 }, { "epoch": 0.8390022675736961, "fcm_dpo/beta": 0.007797827012836933, "fcm_dpo/delta": -0.008863821625709534, "fcm_dpo/margin": 52.293487548828125, "fcm_dpo/q_t": 0.4080876410007477, "grad_norm": 7.238057613372803, "learning_rate": 3.89747159520904e-08, "logits/chosen": 1.528789758682251, "logits/rejected": 1.4679286479949951, "logps/chosen": -178.78054809570312, "logps/ref_chosen": -81.66071319580078, "logps/ref_rejected": -87.20857238769531, "logps/rejected": -236.6219024658203, "loss": 1.154, "margin_dpo/margin_mean": 52.293487548828125, "margin_dpo/margin_std": 95.19200134277344, "step": 555 }, { "epoch": 0.8405139833711263, "fcm_dpo/beta": 0.00775865837931633, "fcm_dpo/delta": 0.0233792494982481, "fcm_dpo/margin": 48.65314483642578, "fcm_dpo/q_t": 0.41721731424331665, "grad_norm": 10.24924373626709, "learning_rate": 3.826871794280192e-08, "logits/chosen": 1.7879140377044678, "logits/rejected": 1.6586774587631226, "logps/chosen": -162.71484375, "logps/ref_chosen": -66.02448272705078, "logps/ref_rejected": -82.74746704101562, "logps/rejected": -228.09097290039062, "loss": 1.1462, "margin_dpo/margin_mean": 48.653141021728516, "margin_dpo/margin_std": 85.63908386230469, "step": 556 }, { "epoch": 0.8420256991685563, "fcm_dpo/beta": 0.0076392171904444695, "fcm_dpo/delta": -0.0801071897149086, "fcm_dpo/margin": 62.17573928833008, "fcm_dpo/q_t": 0.3935287594795227, "grad_norm": 6.463259220123291, "learning_rate": 3.756864251262143e-08, "logits/chosen": 1.7464919090270996, "logits/rejected": 1.4219131469726562, "logps/chosen": -164.66787719726562, "logps/ref_chosen": -73.08985900878906, "logps/ref_rejected": -97.43034362792969, "logps/rejected": -251.18409729003906, "loss": 1.0599, "margin_dpo/margin_mean": 62.17573928833008, "margin_dpo/margin_std": 81.93167114257812, "step": 557 }, { "epoch": 0.8435374149659864, "fcm_dpo/beta": 0.007571034599095583, "fcm_dpo/delta": -0.04616940766572952, "fcm_dpo/margin": 58.581756591796875, "fcm_dpo/q_t": 0.39998599886894226, "grad_norm": 6.480743408203125, "learning_rate": 3.687450924416341e-08, "logits/chosen": 1.7585268020629883, "logits/rejected": 1.6162996292114258, "logps/chosen": -171.26419067382812, "logps/ref_chosen": -80.1357192993164, "logps/ref_rejected": -106.65797424316406, "logps/rejected": -256.3681945800781, "loss": 1.0993, "margin_dpo/margin_mean": 58.581756591796875, "margin_dpo/margin_std": 88.86549377441406, "step": 558 }, { "epoch": 0.8450491307634165, "fcm_dpo/beta": 0.007496877107769251, "fcm_dpo/delta": -0.05996110662817955, "fcm_dpo/margin": 49.02497863769531, "fcm_dpo/q_t": 0.41891011595726013, "grad_norm": 7.072001934051514, "learning_rate": 3.6186337553827743e-08, "logits/chosen": 1.4914745092391968, "logits/rejected": 1.255955696105957, "logps/chosen": -169.67742919921875, "logps/ref_chosen": -79.42267608642578, "logps/ref_rejected": -98.59402465820312, "logps/rejected": -237.87374877929688, "loss": 1.1862, "margin_dpo/margin_mean": 49.02497863769531, "margin_dpo/margin_std": 99.9202880859375, "step": 559 }, { "epoch": 0.8465608465608465, "fcm_dpo/beta": 0.007487626746296883, "fcm_dpo/delta": -0.02866279147565365, "fcm_dpo/margin": 57.073089599609375, "fcm_dpo/q_t": 0.40354275703430176, "grad_norm": 12.783280372619629, "learning_rate": 3.550414669125573e-08, "logits/chosen": 1.37535560131073, "logits/rejected": 1.2821707725524902, "logps/chosen": -170.34442138671875, "logps/ref_chosen": -77.49559020996094, "logps/ref_rejected": -92.61347961425781, "logps/rejected": -242.535400390625, "loss": 1.0977, "margin_dpo/margin_mean": 57.073089599609375, "margin_dpo/margin_std": 84.51661682128906, "step": 560 }, { "epoch": 0.8480725623582767, "fcm_dpo/beta": 0.007413958199322224, "fcm_dpo/delta": -0.018094100058078766, "fcm_dpo/margin": 56.26551818847656, "fcm_dpo/q_t": 0.40612316131591797, "grad_norm": 5.807117462158203, "learning_rate": 3.482795573879241e-08, "logits/chosen": 1.424013376235962, "logits/rejected": 1.3693034648895264, "logps/chosen": -168.86117553710938, "logps/ref_chosen": -79.20771789550781, "logps/ref_rejected": -93.46514892578125, "logps/rejected": -239.38412475585938, "loss": 1.0969, "margin_dpo/margin_mean": 56.26551818847656, "margin_dpo/margin_std": 80.80874633789062, "step": 561 }, { "epoch": 0.8495842781557067, "fcm_dpo/beta": 0.0073739904910326, "fcm_dpo/delta": -0.012268077582120895, "fcm_dpo/margin": 55.75514221191406, "fcm_dpo/q_t": 0.40779781341552734, "grad_norm": 7.490271091461182, "learning_rate": 3.415778361095226e-08, "logits/chosen": 1.7233574390411377, "logits/rejected": 1.5865188837051392, "logps/chosen": -188.96682739257812, "logps/ref_chosen": -94.88652801513672, "logps/ref_rejected": -109.33815002441406, "logps/rejected": -259.173583984375, "loss": 1.1011, "margin_dpo/margin_mean": 55.7551383972168, "margin_dpo/margin_std": 80.49266052246094, "step": 562 }, { "epoch": 0.8510959939531368, "fcm_dpo/beta": 0.0073908064514398575, "fcm_dpo/delta": -0.026350384578108788, "fcm_dpo/margin": 57.53425216674805, "fcm_dpo/q_t": 0.4067358374595642, "grad_norm": 11.612875938415527, "learning_rate": 3.349364905389032e-08, "logits/chosen": 1.7460877895355225, "logits/rejected": 1.5561182498931885, "logps/chosen": -144.20834350585938, "logps/ref_chosen": -65.90719604492188, "logps/ref_rejected": -84.07121276855469, "logps/rejected": -219.9066162109375, "loss": 1.1289, "margin_dpo/margin_mean": 57.53425216674805, "margin_dpo/margin_std": 98.30337524414062, "step": 563 }, { "epoch": 0.8526077097505669, "fcm_dpo/beta": 0.007390887942165136, "fcm_dpo/delta": -0.003784339874982834, "fcm_dpo/margin": 54.58319854736328, "fcm_dpo/q_t": 0.40921688079833984, "grad_norm": 7.469838619232178, "learning_rate": 3.283557064487785e-08, "logits/chosen": 1.5153069496154785, "logits/rejected": 1.4310331344604492, "logps/chosen": -151.17416381835938, "logps/ref_chosen": -72.32071685791016, "logps/ref_rejected": -88.05014038085938, "logps/rejected": -221.48675537109375, "loss": 1.1491, "margin_dpo/margin_mean": 54.58319854736328, "margin_dpo/margin_std": 98.81858825683594, "step": 564 }, { "epoch": 0.854119425547997, "fcm_dpo/beta": 0.007420201785862446, "fcm_dpo/delta": 0.0206848606467247, "fcm_dpo/margin": 51.16043472290039, "fcm_dpo/q_t": 0.41406577825546265, "grad_norm": 8.247408866882324, "learning_rate": 3.218356679178252e-08, "logits/chosen": 1.2496565580368042, "logits/rejected": 1.1596516370773315, "logps/chosen": -176.40371704101562, "logps/ref_chosen": -80.18453979492188, "logps/ref_rejected": -99.55126953125, "logps/rejected": -246.93087768554688, "loss": 1.1314, "margin_dpo/margin_mean": 51.16043472290039, "margin_dpo/margin_std": 82.370361328125, "step": 565 }, { "epoch": 0.8556311413454271, "fcm_dpo/beta": 0.007455028593540192, "fcm_dpo/delta": 0.0031773000955581665, "fcm_dpo/margin": 53.09089279174805, "fcm_dpo/q_t": 0.4105803370475769, "grad_norm": 8.502391815185547, "learning_rate": 3.1537655732553764e-08, "logits/chosen": 1.427902102470398, "logits/rejected": 1.3613345623016357, "logps/chosen": -168.61880493164062, "logps/ref_chosen": -88.0877914428711, "logps/ref_rejected": -87.7589111328125, "logps/rejected": -221.3808135986328, "loss": 1.1421, "margin_dpo/margin_mean": 53.09088897705078, "margin_dpo/margin_std": 90.66483306884766, "step": 566 }, { "epoch": 0.8571428571428571, "fcm_dpo/beta": 0.007287529297173023, "fcm_dpo/delta": -0.06993226706981659, "fcm_dpo/margin": 63.92971420288086, "fcm_dpo/q_t": 0.39457574486732483, "grad_norm": 11.652644157409668, "learning_rate": 3.089785553471233e-08, "logits/chosen": 1.7592835426330566, "logits/rejected": 1.4737862348556519, "logps/chosen": -156.90435791015625, "logps/ref_chosen": -69.93267822265625, "logps/ref_rejected": -95.71786499023438, "logps/rejected": -246.6192626953125, "loss": 1.0628, "margin_dpo/margin_mean": 63.92971420288086, "margin_dpo/margin_std": 84.06944274902344, "step": 567 }, { "epoch": 0.8586545729402872, "fcm_dpo/beta": 0.007296562194824219, "fcm_dpo/delta": -0.05656176805496216, "fcm_dpo/margin": 62.0971565246582, "fcm_dpo/q_t": 0.3984963893890381, "grad_norm": 7.199716091156006, "learning_rate": 3.026418409484513e-08, "logits/chosen": 1.3262768983840942, "logits/rejected": 1.1121933460235596, "logps/chosen": -143.99913024902344, "logps/ref_chosen": -70.33343505859375, "logps/ref_rejected": -108.86271667480469, "logps/rejected": -244.62554931640625, "loss": 1.0754, "margin_dpo/margin_mean": 62.09714889526367, "margin_dpo/margin_std": 80.93890380859375, "step": 568 }, { "epoch": 0.8601662887377173, "fcm_dpo/beta": 0.007188756484538317, "fcm_dpo/delta": -0.01805015653371811, "fcm_dpo/margin": 41.79746627807617, "fcm_dpo/q_t": 0.4310176372528076, "grad_norm": 8.492779731750488, "learning_rate": 2.963665913810451e-08, "logits/chosen": 1.2412662506103516, "logits/rejected": 1.2150421142578125, "logps/chosen": -167.37234497070312, "logps/ref_chosen": -80.85043334960938, "logps/ref_rejected": -92.77810668945312, "logps/rejected": -221.0974884033203, "loss": 1.2138, "margin_dpo/margin_mean": 41.797462463378906, "margin_dpo/margin_std": 90.91089630126953, "step": 569 }, { "epoch": 0.8616780045351474, "fcm_dpo/beta": 0.007139476947486401, "fcm_dpo/delta": -0.08498235046863556, "fcm_dpo/margin": 67.33297729492188, "fcm_dpo/q_t": 0.3915640115737915, "grad_norm": 7.1222615242004395, "learning_rate": 2.9015298217712453e-08, "logits/chosen": 1.2934246063232422, "logits/rejected": 1.1038811206817627, "logps/chosen": -145.34852600097656, "logps/ref_chosen": -69.94769287109375, "logps/ref_rejected": -97.37059020996094, "logps/rejected": -240.10440063476562, "loss": 1.0557, "margin_dpo/margin_mean": 67.33297729492188, "margin_dpo/margin_std": 86.74455261230469, "step": 570 }, { "epoch": 0.8631897203325775, "fcm_dpo/beta": 0.0072095394134521484, "fcm_dpo/delta": 0.12205064296722412, "fcm_dpo/margin": 38.994258880615234, "fcm_dpo/q_t": 0.435294508934021, "grad_norm": 24.326377868652344, "learning_rate": 2.840011871446962e-08, "logits/chosen": 1.3863130807876587, "logits/rejected": 1.3189308643341064, "logps/chosen": -157.417724609375, "logps/ref_chosen": -72.28555297851562, "logps/ref_rejected": -84.57748413085938, "logps/rejected": -208.70391845703125, "loss": 1.2304, "margin_dpo/margin_mean": 38.994258880615234, "margin_dpo/margin_std": 92.22761535644531, "step": 571 }, { "epoch": 0.8647014361300076, "fcm_dpo/beta": 0.007278360426425934, "fcm_dpo/delta": 0.041402481496334076, "fcm_dpo/margin": 49.476531982421875, "fcm_dpo/q_t": 0.41775819659233093, "grad_norm": 6.691201686859131, "learning_rate": 2.7791137836269158e-08, "logits/chosen": 1.612381935119629, "logits/rejected": 1.7202537059783936, "logps/chosen": -181.08970642089844, "logps/ref_chosen": -91.4906997680664, "logps/ref_rejected": -80.44602966308594, "logps/rejected": -219.52157592773438, "loss": 1.1267, "margin_dpo/margin_mean": 49.476531982421875, "margin_dpo/margin_std": 74.04133605957031, "step": 572 }, { "epoch": 0.8662131519274376, "fcm_dpo/beta": 0.007352625019848347, "fcm_dpo/delta": 0.05982336774468422, "fcm_dpo/margin": 46.549224853515625, "fcm_dpo/q_t": 0.4252585470676422, "grad_norm": 7.987117290496826, "learning_rate": 2.718837261761528e-08, "logits/chosen": 1.4434504508972168, "logits/rejected": 1.321394443511963, "logps/chosen": -180.029052734375, "logps/ref_chosen": -87.54232788085938, "logps/ref_rejected": -104.32984924316406, "logps/rejected": -243.36578369140625, "loss": 1.2021, "margin_dpo/margin_mean": 46.549224853515625, "margin_dpo/margin_std": 101.46734619140625, "step": 573 }, { "epoch": 0.8677248677248677, "fcm_dpo/beta": 0.007343747653067112, "fcm_dpo/delta": -0.07292570173740387, "fcm_dpo/margin": 63.92116165161133, "fcm_dpo/q_t": 0.3938440978527069, "grad_norm": 6.1711297035217285, "learning_rate": 2.659183991914696e-08, "logits/chosen": 1.590470790863037, "logits/rejected": 1.4732091426849365, "logps/chosen": -159.10311889648438, "logps/ref_chosen": -75.36632537841797, "logps/ref_rejected": -103.27328491210938, "logps/rejected": -250.93124389648438, "loss": 1.0603, "margin_dpo/margin_mean": 63.92115783691406, "margin_dpo/margin_std": 83.52648162841797, "step": 574 }, { "epoch": 0.8692365835222978, "fcm_dpo/beta": 0.007190365344285965, "fcm_dpo/delta": -0.06424938142299652, "fcm_dpo/margin": 43.404239654541016, "fcm_dpo/q_t": 0.43036943674087524, "grad_norm": 7.720411777496338, "learning_rate": 2.600155642716606e-08, "logits/chosen": 1.527137279510498, "logits/rejected": 1.3130396604537964, "logps/chosen": -173.9998779296875, "logps/ref_chosen": -81.678466796875, "logps/ref_rejected": -112.84233093261719, "logps/rejected": -248.56800842285156, "loss": 1.2034, "margin_dpo/margin_mean": 43.404239654541016, "margin_dpo/margin_std": 88.2564926147461, "step": 575 }, { "epoch": 0.8707482993197279, "fcm_dpo/beta": 0.007088257931172848, "fcm_dpo/delta": -0.07014288008213043, "fcm_dpo/margin": 65.78793334960938, "fcm_dpo/q_t": 0.3944045901298523, "grad_norm": 9.954662322998047, "learning_rate": 2.5417538653170754e-08, "logits/chosen": 1.54152250289917, "logits/rejected": 1.2846912145614624, "logps/chosen": -145.27874755859375, "logps/ref_chosen": -68.78944396972656, "logps/ref_rejected": -102.79037475585938, "logps/rejected": -245.06761169433594, "loss": 1.0648, "margin_dpo/margin_mean": 65.78793334960938, "margin_dpo/margin_std": 87.97828674316406, "step": 576 }, { "epoch": 0.872260015117158, "fcm_dpo/beta": 0.0072364904917776585, "fcm_dpo/delta": 0.13765141367912292, "fcm_dpo/margin": 36.76244354248047, "fcm_dpo/q_t": 0.43941205739974976, "grad_norm": 6.795727729797363, "learning_rate": 2.4839802933393607e-08, "logits/chosen": 1.2348496913909912, "logits/rejected": 1.201481819152832, "logps/chosen": -163.54574584960938, "logps/ref_chosen": -79.84675598144531, "logps/ref_rejected": -84.08309936523438, "logps/rejected": -204.54454040527344, "loss": 1.2285, "margin_dpo/margin_mean": 36.76244354248047, "margin_dpo/margin_std": 85.11286163330078, "step": 577 }, { "epoch": 0.873771730914588, "fcm_dpo/beta": 0.00742256548255682, "fcm_dpo/delta": 0.13279461860656738, "fcm_dpo/margin": 36.504695892333984, "fcm_dpo/q_t": 0.4408913850784302, "grad_norm": 8.053617477416992, "learning_rate": 2.4268365428344733e-08, "logits/chosen": 1.703123688697815, "logits/rejected": 1.610666036605835, "logps/chosen": -160.46142578125, "logps/ref_chosen": -74.91357421875, "logps/ref_rejected": -83.64881896972656, "logps/rejected": -205.7013702392578, "loss": 1.215, "margin_dpo/margin_mean": 36.504695892333984, "margin_dpo/margin_std": 79.07931518554688, "step": 578 }, { "epoch": 0.8752834467120182, "fcm_dpo/beta": 0.007423688657581806, "fcm_dpo/delta": -0.02768631838262081, "fcm_dpo/margin": 57.395477294921875, "fcm_dpo/q_t": 0.40147894620895386, "grad_norm": 6.850172519683838, "learning_rate": 2.3703242122359357e-08, "logits/chosen": 1.2945516109466553, "logits/rejected": 1.2207485437393188, "logps/chosen": -168.03199768066406, "logps/ref_chosen": -75.51022338867188, "logps/ref_rejected": -84.83192443847656, "logps/rejected": -234.74917602539062, "loss": 1.087, "margin_dpo/margin_mean": 57.395477294921875, "margin_dpo/margin_std": 79.23515319824219, "step": 579 }, { "epoch": 0.8767951625094482, "fcm_dpo/beta": 0.007472585886716843, "fcm_dpo/delta": 0.005576103925704956, "fcm_dpo/margin": 52.784088134765625, "fcm_dpo/q_t": 0.4105369448661804, "grad_norm": 6.64032506942749, "learning_rate": 2.3144448823151392e-08, "logits/chosen": 1.4655518531799316, "logits/rejected": 1.2879977226257324, "logps/chosen": -152.16595458984375, "logps/ref_chosen": -76.61564636230469, "logps/ref_rejected": -97.09959411621094, "logps/rejected": -225.43399047851562, "loss": 1.1362, "margin_dpo/margin_mean": 52.784088134765625, "margin_dpo/margin_std": 89.77684020996094, "step": 580 }, { "epoch": 0.8783068783068783, "fcm_dpo/beta": 0.007484517525881529, "fcm_dpo/delta": 0.05809904262423515, "fcm_dpo/margin": 45.93260955810547, "fcm_dpo/q_t": 0.4221350848674774, "grad_norm": 7.525079727172852, "learning_rate": 2.259200116137039e-08, "logits/chosen": 1.4739665985107422, "logits/rejected": 1.3425915241241455, "logps/chosen": -169.97055053710938, "logps/ref_chosen": -74.8531265258789, "logps/ref_rejected": -101.5344009399414, "logps/rejected": -242.58445739746094, "loss": 1.1751, "margin_dpo/margin_mean": 45.93260955810547, "margin_dpo/margin_std": 87.29202270507812, "step": 581 }, { "epoch": 0.8798185941043084, "fcm_dpo/beta": 0.007569895125925541, "fcm_dpo/delta": 0.01635417342185974, "fcm_dpo/margin": 50.751991271972656, "fcm_dpo/q_t": 0.4134330749511719, "grad_norm": 8.547701835632324, "learning_rate": 2.204591459016525e-08, "logits/chosen": 1.5489530563354492, "logits/rejected": 1.6596481800079346, "logps/chosen": -170.80535888671875, "logps/ref_chosen": -81.07638549804688, "logps/ref_rejected": -72.83570861816406, "logps/rejected": -213.3166961669922, "loss": 1.1526, "margin_dpo/margin_mean": 50.751991271972656, "margin_dpo/margin_std": 92.16818237304688, "step": 582 }, { "epoch": 0.8813303099017384, "fcm_dpo/beta": 0.007639412768185139, "fcm_dpo/delta": 0.06253516674041748, "fcm_dpo/margin": 44.445091247558594, "fcm_dpo/q_t": 0.4248436987400055, "grad_norm": 9.364850044250488, "learning_rate": 2.1506204384751064e-08, "logits/chosen": 1.6426869630813599, "logits/rejected": 1.3243718147277832, "logps/chosen": -159.094482421875, "logps/ref_chosen": -66.78465270996094, "logps/ref_rejected": -106.45825958251953, "logps/rejected": -243.21319580078125, "loss": 1.2079, "margin_dpo/margin_mean": 44.44509506225586, "margin_dpo/margin_std": 98.5951919555664, "step": 583 }, { "epoch": 0.8828420256991686, "fcm_dpo/beta": 0.007633395027369261, "fcm_dpo/delta": 0.02281120792031288, "fcm_dpo/margin": 49.43035888671875, "fcm_dpo/q_t": 0.41661393642425537, "grad_norm": 8.11761474609375, "learning_rate": 2.09728856419826e-08, "logits/chosen": 2.000200033187866, "logits/rejected": 1.7143186330795288, "logps/chosen": -134.8970184326172, "logps/ref_chosen": -60.802913665771484, "logps/ref_rejected": -99.45012664794922, "logps/rejected": -222.97457885742188, "loss": 1.17, "margin_dpo/margin_mean": 49.43035888671875, "margin_dpo/margin_std": 94.57049560546875, "step": 584 }, { "epoch": 0.8843537414965986, "fcm_dpo/beta": 0.007718199864029884, "fcm_dpo/delta": 0.03235515207052231, "fcm_dpo/margin": 34.23884201049805, "fcm_dpo/q_t": 0.440193235874176, "grad_norm": 6.512060642242432, "learning_rate": 2.044597327993153e-08, "logits/chosen": 1.4816594123840332, "logits/rejected": 1.3898038864135742, "logps/chosen": -161.7969207763672, "logps/ref_chosen": -75.92616271972656, "logps/ref_rejected": -94.47601318359375, "logps/rejected": -214.58563232421875, "loss": 1.2388, "margin_dpo/margin_mean": 34.23883819580078, "margin_dpo/margin_std": 82.98875427246094, "step": 585 }, { "epoch": 0.8858654572940288, "fcm_dpo/beta": 0.007636943832039833, "fcm_dpo/delta": -0.08058130741119385, "fcm_dpo/margin": 62.398765563964844, "fcm_dpo/q_t": 0.3913051187992096, "grad_norm": 9.164061546325684, "learning_rate": 1.9925482037469187e-08, "logits/chosen": 1.5348619222640991, "logits/rejected": 1.399658441543579, "logps/chosen": -147.88909912109375, "logps/ref_chosen": -68.62062072753906, "logps/ref_rejected": -81.98324584960938, "logps/rejected": -223.65048217773438, "loss": 1.0567, "margin_dpo/margin_mean": 62.398765563964844, "margin_dpo/margin_std": 82.34362030029297, "step": 586 }, { "epoch": 0.8873771730914588, "fcm_dpo/beta": 0.007494157180190086, "fcm_dpo/delta": -0.10239773988723755, "fcm_dpo/margin": 66.33705139160156, "fcm_dpo/q_t": 0.38791346549987793, "grad_norm": 12.113510131835938, "learning_rate": 1.9411426473854687e-08, "logits/chosen": 1.3303208351135254, "logits/rejected": 1.3321213722229004, "logps/chosen": -153.1464385986328, "logps/ref_chosen": -77.67031860351562, "logps/ref_rejected": -79.35327911376953, "logps/rejected": -221.1664581298828, "loss": 1.0714, "margin_dpo/margin_mean": 66.33705139160156, "margin_dpo/margin_std": 96.68234252929688, "step": 587 }, { "epoch": 0.8888888888888888, "fcm_dpo/beta": 0.007391571067273617, "fcm_dpo/delta": -0.03514527529478073, "fcm_dpo/margin": 58.586280822753906, "fcm_dpo/q_t": 0.40127915143966675, "grad_norm": 10.57682991027832, "learning_rate": 1.890382096832699e-08, "logits/chosen": 1.4161376953125, "logits/rejected": 1.3039031028747559, "logps/chosen": -164.74606323242188, "logps/ref_chosen": -77.94320678710938, "logps/ref_rejected": -98.41210174560547, "logps/rejected": -243.8012237548828, "loss": 1.0924, "margin_dpo/margin_mean": 58.586280822753906, "margin_dpo/margin_std": 84.51554870605469, "step": 588 }, { "epoch": 0.890400604686319, "fcm_dpo/beta": 0.00733354315161705, "fcm_dpo/delta": -0.0650940015912056, "fcm_dpo/margin": 62.99828338623047, "fcm_dpo/q_t": 0.3934873938560486, "grad_norm": 10.318829536437988, "learning_rate": 1.840267971970344e-08, "logits/chosen": 1.322197437286377, "logits/rejected": 1.2743580341339111, "logps/chosen": -153.74691772460938, "logps/ref_chosen": -75.18646240234375, "logps/ref_rejected": -93.35910034179688, "logps/rejected": -234.9178466796875, "loss": 1.0548, "margin_dpo/margin_mean": 62.99828338623047, "margin_dpo/margin_std": 78.32982635498047, "step": 589 }, { "epoch": 0.891912320483749, "fcm_dpo/beta": 0.007300875149667263, "fcm_dpo/delta": -0.027903899550437927, "fcm_dpo/margin": 58.443626403808594, "fcm_dpo/q_t": 0.4029507637023926, "grad_norm": 6.3697028160095215, "learning_rate": 1.7908016745981856e-08, "logits/chosen": 1.296320915222168, "logits/rejected": 1.2430548667907715, "logps/chosen": -173.72471618652344, "logps/ref_chosen": -86.9908447265625, "logps/ref_rejected": -100.61723327636719, "logps/rejected": -245.7947235107422, "loss": 1.097, "margin_dpo/margin_mean": 58.443626403808594, "margin_dpo/margin_std": 86.44003295898438, "step": 590 }, { "epoch": 0.8934240362811792, "fcm_dpo/beta": 0.007098148111253977, "fcm_dpo/delta": -0.12376554310321808, "fcm_dpo/margin": 57.82063674926758, "fcm_dpo/q_t": 0.4067874252796173, "grad_norm": 9.336226463317871, "learning_rate": 1.7419845883949098e-08, "logits/chosen": 1.7065235376358032, "logits/rejected": 1.542229413986206, "logps/chosen": -146.87442016601562, "logps/ref_chosen": -74.85809326171875, "logps/ref_rejected": -102.75840759277344, "logps/rejected": -232.59536743164062, "loss": 1.1219, "margin_dpo/margin_mean": 57.82063293457031, "margin_dpo/margin_std": 89.06134796142578, "step": 591 }, { "epoch": 0.8949357520786092, "fcm_dpo/beta": 0.007119194604456425, "fcm_dpo/delta": 0.0307568721473217, "fcm_dpo/margin": 52.02532958984375, "fcm_dpo/q_t": 0.4162050783634186, "grad_norm": 6.295296669006348, "learning_rate": 1.6938180788793556e-08, "logits/chosen": 1.7110962867736816, "logits/rejected": 1.4122859239578247, "logps/chosen": -148.53353881835938, "logps/ref_chosen": -67.90579223632812, "logps/ref_rejected": -100.35234069824219, "logps/rejected": -233.00543212890625, "loss": 1.1321, "margin_dpo/margin_mean": 52.02532958984375, "margin_dpo/margin_std": 80.80274963378906, "step": 592 }, { "epoch": 0.8964474678760394, "fcm_dpo/beta": 0.007164083421230316, "fcm_dpo/delta": 0.02655250020325184, "fcm_dpo/margin": 52.26756286621094, "fcm_dpo/q_t": 0.4153831899166107, "grad_norm": 10.239151954650879, "learning_rate": 1.6463034933723336e-08, "logits/chosen": 1.348867416381836, "logits/rejected": 1.1232930421829224, "logps/chosen": -131.43345642089844, "logps/ref_chosen": -59.29489517211914, "logps/ref_rejected": -85.31307983398438, "logps/rejected": -209.71920776367188, "loss": 1.1429, "margin_dpo/margin_mean": 52.26756286621094, "margin_dpo/margin_std": 89.36260986328125, "step": 593 }, { "epoch": 0.8979591836734694, "fcm_dpo/beta": 0.007193025201559067, "fcm_dpo/delta": 0.029274439439177513, "fcm_dpo/margin": 51.68307876586914, "fcm_dpo/q_t": 0.41542428731918335, "grad_norm": 10.818907737731934, "learning_rate": 1.5994421609589385e-08, "logits/chosen": 1.2572638988494873, "logits/rejected": 1.2137444019317627, "logps/chosen": -168.90765380859375, "logps/ref_chosen": -83.14643859863281, "logps/ref_rejected": -88.201904296875, "logps/rejected": -225.64620971679688, "loss": 1.1315, "margin_dpo/margin_mean": 51.68307876586914, "margin_dpo/margin_std": 82.45545959472656, "step": 594 }, { "epoch": 0.8994708994708994, "fcm_dpo/beta": 0.007129493169486523, "fcm_dpo/delta": -0.09650059044361115, "fcm_dpo/margin": 68.9874267578125, "fcm_dpo/q_t": 0.38809216022491455, "grad_norm": 10.010942459106445, "learning_rate": 1.553235392451377e-08, "logits/chosen": 1.7847068309783936, "logits/rejected": 1.529867172241211, "logps/chosen": -150.77587890625, "logps/ref_chosen": -70.40016174316406, "logps/ref_rejected": -103.95550537109375, "logps/rejected": -253.31863403320312, "loss": 1.0697, "margin_dpo/margin_mean": 68.9874267578125, "margin_dpo/margin_std": 98.64369201660156, "step": 595 }, { "epoch": 0.9009826152683296, "fcm_dpo/beta": 0.007084083743393421, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 19.771663665771484, "fcm_dpo/q_t": 0.469125896692276, "grad_norm": 6.528999328613281, "learning_rate": 1.507684480352292e-08, "logits/chosen": 1.183467149734497, "logits/rejected": 1.2314069271087646, "logps/chosen": -183.73483276367188, "logps/ref_chosen": -86.083740234375, "logps/ref_rejected": -78.41991424560547, "logps/rejected": -195.84266662597656, "loss": 1.3362, "margin_dpo/margin_mean": 19.771663665771484, "margin_dpo/margin_std": 84.07432556152344, "step": 596 }, { "epoch": 0.9024943310657596, "fcm_dpo/beta": 0.007151216268539429, "fcm_dpo/delta": 0.05845186114311218, "fcm_dpo/margin": 48.03459167480469, "fcm_dpo/q_t": 0.42271360754966736, "grad_norm": 7.565585613250732, "learning_rate": 1.4627906988186111e-08, "logits/chosen": 1.5501301288604736, "logits/rejected": 1.5076608657836914, "logps/chosen": -140.9503173828125, "logps/ref_chosen": -67.8086166381836, "logps/ref_rejected": -71.09245300292969, "logps/rejected": -192.26873779296875, "loss": 1.1646, "margin_dpo/margin_mean": 48.03459167480469, "margin_dpo/margin_std": 87.82294464111328, "step": 597 }, { "epoch": 0.9040060468631897, "fcm_dpo/beta": 0.007216842845082283, "fcm_dpo/delta": 0.06853266060352325, "fcm_dpo/margin": 31.869421005249023, "fcm_dpo/q_t": 0.44815605878829956, "grad_norm": 6.673587322235107, "learning_rate": 1.4185553036259095e-08, "logits/chosen": 1.628659725189209, "logits/rejected": 1.4073009490966797, "logps/chosen": -178.23072814941406, "logps/ref_chosen": -74.31095886230469, "logps/ref_rejected": -98.08122253417969, "logps/rejected": -233.87042236328125, "loss": 1.2428, "margin_dpo/margin_mean": 31.869421005249023, "margin_dpo/margin_std": 77.04680633544922, "step": 598 }, { "epoch": 0.9055177626606198, "fcm_dpo/beta": 0.0073530226945877075, "fcm_dpo/delta": 0.07715515047311783, "fcm_dpo/margin": 44.24391555786133, "fcm_dpo/q_t": 0.4264746904373169, "grad_norm": 9.405556678771973, "learning_rate": 1.3749795321332885e-08, "logits/chosen": 1.7087280750274658, "logits/rejected": 1.6005630493164062, "logps/chosen": -167.94850158691406, "logps/ref_chosen": -74.21861267089844, "logps/ref_rejected": -90.1492919921875, "logps/rejected": -228.12310791015625, "loss": 1.1816, "margin_dpo/margin_mean": 44.24391555786133, "margin_dpo/margin_std": 86.17495727539062, "step": 599 }, { "epoch": 0.9070294784580499, "fcm_dpo/beta": 0.007373571861535311, "fcm_dpo/delta": -0.007764637470245361, "fcm_dpo/margin": 41.892059326171875, "fcm_dpo/q_t": 0.4296872019767761, "grad_norm": 7.145091533660889, "learning_rate": 1.3320646032487393e-08, "logits/chosen": 1.3918439149856567, "logits/rejected": 1.212207317352295, "logps/chosen": -168.78567504882812, "logps/ref_chosen": -79.34190368652344, "logps/ref_rejected": -97.0519790649414, "logps/rejected": -228.3878173828125, "loss": 1.183, "margin_dpo/margin_mean": 41.89206314086914, "margin_dpo/margin_std": 78.34820556640625, "step": 600 }, { "epoch": 0.9070294784580499, "eval_fcm_dpo/beta": 0.007367846090346575, "eval_logits/chosen": 1.3822048902511597, "eval_logits/rejected": 1.2497957944869995, "eval_logps/chosen": -172.7694549560547, "eval_logps/ref_chosen": -86.90177917480469, "eval_logps/ref_rejected": -96.69639587402344, "eval_logps/rejected": -234.10513305664062, "eval_loss": 0.5698250532150269, "eval_margin_dpo/margin_mean": 51.54107666015625, "eval_margin_dpo/margin_std": 86.13619995117188, "eval_runtime": 42.2834, "eval_samples_per_second": 54.466, "eval_steps_per_second": 1.703, "step": 600 }, { "epoch": 0.90854119425548, "fcm_dpo/beta": 0.007286690175533295, "fcm_dpo/delta": -0.06402283906936646, "fcm_dpo/margin": 63.2564697265625, "fcm_dpo/q_t": 0.3962099254131317, "grad_norm": 8.408620834350586, "learning_rate": 1.2898117173950868e-08, "logits/chosen": 1.2801477909088135, "logits/rejected": 1.0866169929504395, "logps/chosen": -146.13392639160156, "logps/ref_chosen": -72.06497192382812, "logps/ref_rejected": -97.60928344726562, "logps/rejected": -234.93472290039062, "loss": 1.0767, "margin_dpo/margin_mean": 63.256465911865234, "margin_dpo/margin_std": 89.92544555664062, "step": 601 }, { "epoch": 0.91005291005291, "fcm_dpo/beta": 0.00729843694716692, "fcm_dpo/delta": -0.006445102393627167, "fcm_dpo/margin": 55.607749938964844, "fcm_dpo/q_t": 0.40874192118644714, "grad_norm": 9.119906425476074, "learning_rate": 1.2482220564763667e-08, "logits/chosen": 1.881563425064087, "logits/rejected": 1.7863805294036865, "logps/chosen": -146.75225830078125, "logps/ref_chosen": -77.80416870117188, "logps/ref_rejected": -89.05026245117188, "logps/rejected": -213.60609436035156, "loss": 1.1081, "margin_dpo/margin_mean": 55.607749938964844, "margin_dpo/margin_std": 82.802001953125, "step": 602 }, { "epoch": 0.9115646258503401, "fcm_dpo/beta": 0.007275847718119621, "fcm_dpo/delta": 0.03916401416063309, "fcm_dpo/margin": 49.760406494140625, "fcm_dpo/q_t": 0.41777119040489197, "grad_norm": 6.798327445983887, "learning_rate": 1.2072967838448051e-08, "logits/chosen": 1.402630090713501, "logits/rejected": 1.265625, "logps/chosen": -154.68133544921875, "logps/ref_chosen": -68.30155944824219, "logps/ref_rejected": -90.542724609375, "logps/rejected": -226.68292236328125, "loss": 1.1616, "margin_dpo/margin_mean": 49.760406494140625, "margin_dpo/margin_std": 90.02642822265625, "step": 603 }, { "epoch": 0.9130763416477702, "fcm_dpo/beta": 0.0073940567672252655, "fcm_dpo/delta": 0.04941772669553757, "fcm_dpo/margin": 47.617759704589844, "fcm_dpo/q_t": 0.4212024509906769, "grad_norm": 5.99643611907959, "learning_rate": 1.1670370442682459e-08, "logits/chosen": 1.3572053909301758, "logits/rejected": 1.3480005264282227, "logps/chosen": -162.58450317382812, "logps/ref_chosen": -90.55952453613281, "logps/ref_rejected": -84.6327133178711, "logps/rejected": -204.27545166015625, "loss": 1.1818, "margin_dpo/margin_mean": 47.617759704589844, "margin_dpo/margin_std": 95.67184448242188, "step": 604 }, { "epoch": 0.9145880574452003, "fcm_dpo/beta": 0.007479371502995491, "fcm_dpo/delta": 0.06939111649990082, "fcm_dpo/margin": 44.499267578125, "fcm_dpo/q_t": 0.4240415096282959, "grad_norm": 10.035806655883789, "learning_rate": 1.1274439638981532e-08, "logits/chosen": 1.2821848392486572, "logits/rejected": 1.1488748788833618, "logps/chosen": -173.67697143554688, "logps/ref_chosen": -80.26661682128906, "logps/ref_rejected": -100.26485443115234, "logps/rejected": -238.17446899414062, "loss": 1.1637, "margin_dpo/margin_mean": 44.499267578125, "margin_dpo/margin_std": 79.42837524414062, "step": 605 }, { "epoch": 0.9160997732426304, "fcm_dpo/beta": 0.007518238388001919, "fcm_dpo/delta": -0.0017635505646467209, "fcm_dpo/margin": 53.40687561035156, "fcm_dpo/q_t": 0.4092721939086914, "grad_norm": 9.322007179260254, "learning_rate": 1.0885186502381016e-08, "logits/chosen": 1.525209665298462, "logits/rejected": 1.3565664291381836, "logps/chosen": -146.97242736816406, "logps/ref_chosen": -70.73554229736328, "logps/ref_rejected": -95.9410400390625, "logps/rejected": -225.58480834960938, "loss": 1.106, "margin_dpo/margin_mean": 53.406883239746094, "margin_dpo/margin_std": 78.14736938476562, "step": 606 }, { "epoch": 0.9176114890400605, "fcm_dpo/beta": 0.00732019217684865, "fcm_dpo/delta": -0.11818597465753555, "fcm_dpo/margin": 53.20615768432617, "fcm_dpo/q_t": 0.4098881781101227, "grad_norm": 9.862732887268066, "learning_rate": 1.0502621921127774e-08, "logits/chosen": 1.524349570274353, "logits/rejected": 1.4376423358917236, "logps/chosen": -171.41452026367188, "logps/ref_chosen": -81.26203918457031, "logps/ref_rejected": -92.71575927734375, "logps/rejected": -236.07440185546875, "loss": 1.1186, "margin_dpo/margin_mean": 53.20615768432617, "margin_dpo/margin_std": 75.97901916503906, "step": 607 }, { "epoch": 0.9191232048374905, "fcm_dpo/beta": 0.007443271577358246, "fcm_dpo/delta": 0.10113994777202606, "fcm_dpo/margin": 40.53437042236328, "fcm_dpo/q_t": 0.43184107542037964, "grad_norm": 9.41500473022461, "learning_rate": 1.0126756596375685e-08, "logits/chosen": 1.4544808864593506, "logits/rejected": 1.1771209239959717, "logps/chosen": -175.67120361328125, "logps/ref_chosen": -82.6530990600586, "logps/ref_rejected": -110.64334106445312, "logps/rejected": -244.19581604003906, "loss": 1.1933, "margin_dpo/margin_mean": 40.53437042236328, "margin_dpo/margin_std": 81.21488952636719, "step": 608 }, { "epoch": 0.9206349206349206, "fcm_dpo/beta": 0.007462367881089449, "fcm_dpo/delta": -0.0018095960840582848, "fcm_dpo/margin": 53.83350372314453, "fcm_dpo/q_t": 0.40612369775772095, "grad_norm": 8.645689010620117, "learning_rate": 9.757601041885694e-09, "logits/chosen": 1.7371234893798828, "logits/rejected": 1.6018296480178833, "logps/chosen": -146.37277221679688, "logps/ref_chosen": -68.20232391357422, "logps/ref_rejected": -81.90515899658203, "logps/rejected": -213.90911865234375, "loss": 1.0875, "margin_dpo/margin_mean": 53.83350372314453, "margin_dpo/margin_std": 68.90634155273438, "step": 609 }, { "epoch": 0.9221466364323507, "fcm_dpo/beta": 0.007475072983652353, "fcm_dpo/delta": 0.019946957007050514, "fcm_dpo/margin": 50.93968963623047, "fcm_dpo/q_t": 0.4158932864665985, "grad_norm": 14.192397117614746, "learning_rate": 9.395165583732379e-09, "logits/chosen": 1.5707309246063232, "logits/rejected": 1.5060572624206543, "logps/chosen": -187.23873901367188, "logps/ref_chosen": -99.01324462890625, "logps/ref_rejected": -102.26054382324219, "logps/rejected": -241.42572021484375, "loss": 1.1659, "margin_dpo/margin_mean": 50.93968963623047, "margin_dpo/margin_std": 97.81775665283203, "step": 610 }, { "epoch": 0.9236583522297808, "fcm_dpo/beta": 0.007563972845673561, "fcm_dpo/delta": 0.08229602128267288, "fcm_dpo/margin": 42.35655212402344, "fcm_dpo/q_t": 0.42811521887779236, "grad_norm": 7.965384006500244, "learning_rate": 9.03946036001449e-09, "logits/chosen": 1.5573077201843262, "logits/rejected": 1.4318535327911377, "logps/chosen": -144.16986083984375, "logps/ref_chosen": -66.36254119873047, "logps/ref_rejected": -88.74557495117188, "logps/rejected": -208.90943908691406, "loss": 1.1632, "margin_dpo/margin_mean": 42.35655212402344, "margin_dpo/margin_std": 72.27166748046875, "step": 611 }, { "epoch": 0.9251700680272109, "fcm_dpo/beta": 0.007552753668278456, "fcm_dpo/delta": -0.07210865616798401, "fcm_dpo/margin": 62.068572998046875, "fcm_dpo/q_t": 0.3943250775337219, "grad_norm": 6.521274089813232, "learning_rate": 8.690495320571839e-09, "logits/chosen": 1.2143969535827637, "logits/rejected": 1.0262000560760498, "logps/chosen": -170.40911865234375, "logps/ref_chosen": -78.6339111328125, "logps/ref_rejected": -108.34969329833984, "logps/rejected": -262.1934814453125, "loss": 1.0867, "margin_dpo/margin_mean": 62.06857681274414, "margin_dpo/margin_std": 93.37065887451172, "step": 612 }, { "epoch": 0.926681783824641, "fcm_dpo/beta": 0.0074174664914608, "fcm_dpo/delta": -0.1009131669998169, "fcm_dpo/margin": 66.87256622314453, "fcm_dpo/q_t": 0.387722373008728, "grad_norm": 7.430573463439941, "learning_rate": 8.348280226706722e-09, "logits/chosen": 1.249620795249939, "logits/rejected": 1.2631680965423584, "logps/chosen": -143.5052032470703, "logps/ref_chosen": -73.3539047241211, "logps/ref_rejected": -76.91837310791016, "logps/rejected": -213.94223022460938, "loss": 1.0475, "margin_dpo/margin_mean": 66.87255859375, "margin_dpo/margin_std": 87.33309173583984, "step": 613 }, { "epoch": 0.9281934996220711, "fcm_dpo/beta": 0.007404108997434378, "fcm_dpo/delta": 0.002595767378807068, "fcm_dpo/margin": 53.611568450927734, "fcm_dpo/q_t": 0.40840423107147217, "grad_norm": 8.825176239013672, "learning_rate": 8.012824650910937e-09, "logits/chosen": 1.506474494934082, "logits/rejected": 1.480654001235962, "logps/chosen": -165.64871215820312, "logps/ref_chosen": -77.80007934570312, "logps/ref_rejected": -89.05572509765625, "logps/rejected": -230.51593017578125, "loss": 1.1024, "margin_dpo/margin_mean": 53.61156463623047, "margin_dpo/margin_std": 74.10795593261719, "step": 614 }, { "epoch": 0.9297052154195011, "fcm_dpo/beta": 0.007292012684047222, "fcm_dpo/delta": -0.0258035846054554, "fcm_dpo/margin": 58.11692428588867, "fcm_dpo/q_t": 0.4042222499847412, "grad_norm": 7.2233381271362305, "learning_rate": 7.684137976598088e-09, "logits/chosen": 1.276944637298584, "logits/rejected": 1.1412029266357422, "logps/chosen": -183.43215942382812, "logps/ref_chosen": -90.06971740722656, "logps/ref_rejected": -118.7764892578125, "logps/rejected": -270.255859375, "loss": 1.1177, "margin_dpo/margin_mean": 58.116920471191406, "margin_dpo/margin_std": 92.6316909790039, "step": 615 }, { "epoch": 0.9312169312169312, "fcm_dpo/beta": 0.0073302434757351875, "fcm_dpo/delta": 0.01808079145848751, "fcm_dpo/margin": 52.18651580810547, "fcm_dpo/q_t": 0.4146103262901306, "grad_norm": 10.137645721435547, "learning_rate": 7.36222939784098e-09, "logits/chosen": 1.5233194828033447, "logits/rejected": 1.325713872909546, "logps/chosen": -160.01788330078125, "logps/ref_chosen": -74.62954711914062, "logps/ref_rejected": -93.655029296875, "logps/rejected": -231.22987365722656, "loss": 1.126, "margin_dpo/margin_mean": 52.18651580810547, "margin_dpo/margin_std": 82.57902526855469, "step": 616 }, { "epoch": 0.9327286470143613, "fcm_dpo/beta": 0.007304855156689882, "fcm_dpo/delta": -0.033911511301994324, "fcm_dpo/margin": 44.39569854736328, "fcm_dpo/q_t": 0.4244207739830017, "grad_norm": 9.833024024963379, "learning_rate": 7.047107919114586e-09, "logits/chosen": 1.3680877685546875, "logits/rejected": 1.2418947219848633, "logps/chosen": -173.10882568359375, "logps/ref_chosen": -75.98182678222656, "logps/ref_rejected": -97.1640625, "logps/rejected": -238.68673706054688, "loss": 1.1656, "margin_dpo/margin_mean": 44.39570236206055, "margin_dpo/margin_std": 76.2586669921875, "step": 617 }, { "epoch": 0.9342403628117913, "fcm_dpo/beta": 0.0073931096121668816, "fcm_dpo/delta": 0.08461872488260269, "fcm_dpo/margin": 43.02881622314453, "fcm_dpo/q_t": 0.4286499619483948, "grad_norm": 16.107576370239258, "learning_rate": 6.738782355044048e-09, "logits/chosen": 1.7296488285064697, "logits/rejected": 1.44431471824646, "logps/chosen": -155.86195373535156, "logps/ref_chosen": -74.47208404541016, "logps/ref_rejected": -107.09980773925781, "logps/rejected": -231.51849365234375, "loss": 1.1778, "margin_dpo/margin_mean": 43.02881622314453, "margin_dpo/margin_std": 81.21162414550781, "step": 618 }, { "epoch": 0.9357520786092215, "fcm_dpo/beta": 0.007361262571066618, "fcm_dpo/delta": -0.04189985245466232, "fcm_dpo/margin": 59.74002456665039, "fcm_dpo/q_t": 0.40103021264076233, "grad_norm": 9.501195907592773, "learning_rate": 6.437261330158206e-09, "logits/chosen": 1.698243260383606, "logits/rejected": 1.468477487564087, "logps/chosen": -154.20974731445312, "logps/ref_chosen": -70.84220886230469, "logps/ref_rejected": -98.07801818847656, "logps/rejected": -241.18557739257812, "loss": 1.0963, "margin_dpo/margin_mean": 59.740020751953125, "margin_dpo/margin_std": 90.09310913085938, "step": 619 }, { "epoch": 0.9372637944066515, "fcm_dpo/beta": 0.007336704060435295, "fcm_dpo/delta": -0.042020995169878006, "fcm_dpo/margin": 44.221435546875, "fcm_dpo/q_t": 0.4262590706348419, "grad_norm": 7.753275394439697, "learning_rate": 6.142553278648238e-09, "logits/chosen": 1.2661397457122803, "logits/rejected": 1.2839014530181885, "logps/chosen": -156.5693817138672, "logps/ref_chosen": -76.93606567382812, "logps/ref_rejected": -81.28453063964844, "logps/rejected": -205.1392822265625, "loss": 1.1765, "margin_dpo/margin_mean": 44.221435546875, "margin_dpo/margin_std": 80.71849060058594, "step": 620 }, { "epoch": 0.9387755102040817, "fcm_dpo/beta": 0.007401227951049805, "fcm_dpo/delta": 0.08091681450605392, "fcm_dpo/margin": 43.45226287841797, "fcm_dpo/q_t": 0.42756223678588867, "grad_norm": 7.846070766448975, "learning_rate": 5.854666444131934e-09, "logits/chosen": 1.8054823875427246, "logits/rejected": 1.4816551208496094, "logps/chosen": -157.64730834960938, "logps/ref_chosen": -69.87464904785156, "logps/ref_rejected": -105.61328887939453, "logps/rejected": -236.83822631835938, "loss": 1.1834, "margin_dpo/margin_mean": 43.452266693115234, "margin_dpo/margin_std": 85.07792663574219, "step": 621 }, { "epoch": 0.9402872260015117, "fcm_dpo/beta": 0.007493887562304735, "fcm_dpo/delta": 0.0487191379070282, "fcm_dpo/margin": 47.08264923095703, "fcm_dpo/q_t": 0.4192439317703247, "grad_norm": 10.554768562316895, "learning_rate": 5.573608879422875e-09, "logits/chosen": 1.321624994277954, "logits/rejected": 1.184661626815796, "logps/chosen": -166.62425231933594, "logps/ref_chosen": -78.9598388671875, "logps/ref_rejected": -97.90648651123047, "logps/rejected": -232.65354919433594, "loss": 1.1468, "margin_dpo/margin_mean": 47.08264923095703, "margin_dpo/margin_std": 78.495361328125, "step": 622 }, { "epoch": 0.9417989417989417, "fcm_dpo/beta": 0.007525959052145481, "fcm_dpo/delta": 0.02474692091345787, "fcm_dpo/margin": 49.984649658203125, "fcm_dpo/q_t": 0.41398096084594727, "grad_norm": 6.809837818145752, "learning_rate": 5.299388446305342e-09, "logits/chosen": 1.6094965934753418, "logits/rejected": 1.4611616134643555, "logps/chosen": -181.2794189453125, "logps/ref_chosen": -83.22647094726562, "logps/ref_rejected": -105.1362533569336, "logps/rejected": -253.17385864257812, "loss": 1.126, "margin_dpo/margin_mean": 49.984649658203125, "margin_dpo/margin_std": 77.9920654296875, "step": 623 }, { "epoch": 0.9433106575963719, "fcm_dpo/beta": 0.007394228130578995, "fcm_dpo/delta": -0.10044856369495392, "fcm_dpo/margin": 66.91777038574219, "fcm_dpo/q_t": 0.3872142434120178, "grad_norm": 6.696838855743408, "learning_rate": 5.03201281531429e-09, "logits/chosen": 1.518819808959961, "logits/rejected": 1.2560850381851196, "logps/chosen": -140.40402221679688, "logps/ref_chosen": -66.10560607910156, "logps/ref_rejected": -91.66778564453125, "logps/rejected": -232.88397216796875, "loss": 1.0408, "margin_dpo/margin_mean": 66.91777038574219, "margin_dpo/margin_std": 82.78280639648438, "step": 624 }, { "epoch": 0.9448223733938019, "fcm_dpo/beta": 0.007477354723960161, "fcm_dpo/delta": 0.06206696480512619, "fcm_dpo/margin": 32.61128616333008, "fcm_dpo/q_t": 0.4454808235168457, "grad_norm": 11.197325706481934, "learning_rate": 4.7714894655209174e-09, "logits/chosen": 1.6123887300491333, "logits/rejected": 1.3841365575790405, "logps/chosen": -161.52566528320312, "logps/ref_chosen": -73.20295715332031, "logps/ref_rejected": -105.31025695800781, "logps/rejected": -226.2442626953125, "loss": 1.2569, "margin_dpo/margin_mean": 32.61128616333008, "margin_dpo/margin_std": 85.51419067382812, "step": 625 }, { "epoch": 0.9463340891912321, "fcm_dpo/beta": 0.007401737850159407, "fcm_dpo/delta": -0.0417199544608593, "fcm_dpo/margin": 59.367549896240234, "fcm_dpo/q_t": 0.4038395881652832, "grad_norm": 8.29497241973877, "learning_rate": 4.517825684323323e-09, "logits/chosen": 1.704930067062378, "logits/rejected": 1.3351404666900635, "logps/chosen": -140.19259643554688, "logps/ref_chosen": -62.181278228759766, "logps/ref_rejected": -108.17747497558594, "logps/rejected": -245.55633544921875, "loss": 1.1167, "margin_dpo/margin_mean": 59.367549896240234, "margin_dpo/margin_std": 97.91468048095703, "step": 626 }, { "epoch": 0.9478458049886621, "fcm_dpo/beta": 0.007266830187290907, "fcm_dpo/delta": -0.12542851269245148, "fcm_dpo/margin": 71.37725830078125, "fcm_dpo/q_t": 0.3819334805011749, "grad_norm": 7.506706714630127, "learning_rate": 4.271028567242818e-09, "logits/chosen": 1.4920332431793213, "logits/rejected": 1.1570463180541992, "logps/chosen": -160.24082946777344, "logps/ref_chosen": -77.72123718261719, "logps/ref_rejected": -114.40547180175781, "logps/rejected": -268.30230712890625, "loss": 1.0246, "margin_dpo/margin_mean": 71.37725830078125, "margin_dpo/margin_std": 86.36345672607422, "step": 627 }, { "epoch": 0.9493575207860923, "fcm_dpo/beta": 0.007203748449683189, "fcm_dpo/delta": -0.09114761650562286, "fcm_dpo/margin": 67.43045806884766, "fcm_dpo/q_t": 0.38849127292633057, "grad_norm": 7.820545673370361, "learning_rate": 4.0311050177251895e-09, "logits/chosen": 1.194248080253601, "logits/rejected": 1.1566791534423828, "logps/chosen": -151.1189422607422, "logps/ref_chosen": -70.71195983886719, "logps/ref_rejected": -93.85909271240234, "logps/rejected": -241.696533203125, "loss": 1.0798, "margin_dpo/margin_mean": 67.43045806884766, "margin_dpo/margin_std": 92.61256408691406, "step": 628 }, { "epoch": 0.9508692365835223, "fcm_dpo/beta": 0.007152793928980827, "fcm_dpo/delta": 0.03857763484120369, "fcm_dpo/margin": 50.71005630493164, "fcm_dpo/q_t": 0.4172513782978058, "grad_norm": 8.010543823242188, "learning_rate": 3.798061746947995e-09, "logits/chosen": 1.3519011735916138, "logits/rejected": 1.3062744140625, "logps/chosen": -169.06600952148438, "logps/ref_chosen": -88.66283416748047, "logps/ref_rejected": -94.67845153808594, "logps/rejected": -225.79168701171875, "loss": 1.1216, "margin_dpo/margin_mean": 50.71005630493164, "margin_dpo/margin_std": 73.14350891113281, "step": 629 }, { "epoch": 0.9523809523809523, "fcm_dpo/beta": 0.007133721373975277, "fcm_dpo/delta": -0.020191561430692673, "fcm_dpo/margin": 58.78317642211914, "fcm_dpo/q_t": 0.40502220392227173, "grad_norm": 6.478476524353027, "learning_rate": 3.5719052736323806e-09, "logits/chosen": 1.4322681427001953, "logits/rejected": 1.270760416984558, "logps/chosen": -153.3521728515625, "logps/ref_chosen": -72.94979858398438, "logps/ref_rejected": -92.7632827758789, "logps/rejected": -231.9488525390625, "loss": 1.0856, "margin_dpo/margin_mean": 58.78317642211914, "margin_dpo/margin_std": 79.745849609375, "step": 630 }, { "epoch": 0.9538926681783825, "fcm_dpo/beta": 0.0070120044983923435, "fcm_dpo/delta": -0.06671465933322906, "fcm_dpo/margin": 65.94947052001953, "fcm_dpo/q_t": 0.39557164907455444, "grad_norm": 8.37364673614502, "learning_rate": 3.352641923861144e-09, "logits/chosen": 1.8200433254241943, "logits/rejected": 1.4821150302886963, "logps/chosen": -156.98353576660156, "logps/ref_chosen": -78.58656311035156, "logps/ref_rejected": -115.38685607910156, "logps/rejected": -259.7332763671875, "loss": 1.0733, "margin_dpo/margin_mean": 65.949462890625, "margin_dpo/margin_std": 90.15409851074219, "step": 631 }, { "epoch": 0.9554043839758125, "fcm_dpo/beta": 0.007028430234640837, "fcm_dpo/delta": -0.010720066726207733, "fcm_dpo/margin": 58.3730354309082, "fcm_dpo/q_t": 0.40583336353302, "grad_norm": 8.03717041015625, "learning_rate": 3.140277830901428e-09, "logits/chosen": 1.498687744140625, "logits/rejected": 1.4070721864700317, "logps/chosen": -154.97467041015625, "logps/ref_chosen": -75.24861907958984, "logps/ref_rejected": -82.98665618896484, "logps/rejected": -221.08575439453125, "loss": 1.099, "margin_dpo/margin_mean": 58.37303924560547, "margin_dpo/margin_std": 83.79042053222656, "step": 632 }, { "epoch": 0.9569160997732427, "fcm_dpo/beta": 0.007006094790995121, "fcm_dpo/delta": -0.03479746729135513, "fcm_dpo/margin": 61.82661437988281, "fcm_dpo/q_t": 0.40178823471069336, "grad_norm": 12.49810791015625, "learning_rate": 2.9348189350335007e-09, "logits/chosen": 1.363844394683838, "logits/rejected": 1.2224321365356445, "logps/chosen": -131.18251037597656, "logps/ref_chosen": -68.8402099609375, "logps/ref_rejected": -84.64610290527344, "logps/rejected": -208.81500244140625, "loss": 1.0818, "margin_dpo/margin_mean": 61.82661056518555, "margin_dpo/margin_std": 84.4749755859375, "step": 633 }, { "epoch": 0.9584278155706727, "fcm_dpo/beta": 0.006968853063881397, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 17.51637840270996, "fcm_dpo/q_t": 0.4729679226875305, "grad_norm": 10.013653755187988, "learning_rate": 2.736270983384276e-09, "logits/chosen": 1.642852783203125, "logits/rejected": 1.6877248287200928, "logps/chosen": -173.70953369140625, "logps/ref_chosen": -77.0589599609375, "logps/ref_rejected": -74.37579345703125, "logps/rejected": -188.54275512695312, "loss": 1.3598, "margin_dpo/margin_mean": 17.516376495361328, "margin_dpo/margin_std": 88.91429138183594, "step": 634 }, { "epoch": 0.9599395313681028, "fcm_dpo/beta": 0.0071297320537269115, "fcm_dpo/delta": 0.1415490359067917, "fcm_dpo/margin": 36.72964859008789, "fcm_dpo/q_t": 0.4411892592906952, "grad_norm": 7.8840765953063965, "learning_rate": 2.5446395297668287e-09, "logits/chosen": 1.1347073316574097, "logits/rejected": 0.9766046404838562, "logps/chosen": -190.28985595703125, "logps/ref_chosen": -85.60243225097656, "logps/ref_rejected": -104.29497528076172, "logps/rejected": -245.71206665039062, "loss": 1.2396, "margin_dpo/margin_mean": 36.72964859008789, "margin_dpo/margin_std": 90.49435424804688, "step": 635 }, { "epoch": 0.9614512471655329, "fcm_dpo/beta": 0.007117372006177902, "fcm_dpo/delta": -0.05001338571310043, "fcm_dpo/margin": 62.914146423339844, "fcm_dpo/q_t": 0.39747586846351624, "grad_norm": 7.932164669036865, "learning_rate": 2.359929934524829e-09, "logits/chosen": 1.6017391681671143, "logits/rejected": 1.3030370473861694, "logps/chosen": -145.52099609375, "logps/ref_chosen": -68.72154235839844, "logps/ref_rejected": -97.44863891601562, "logps/rejected": -237.1622314453125, "loss": 1.0621, "margin_dpo/margin_mean": 62.914146423339844, "margin_dpo/margin_std": 79.58000183105469, "step": 636 }, { "epoch": 0.9629629629629629, "fcm_dpo/beta": 0.007132566533982754, "fcm_dpo/delta": 0.01812063902616501, "fcm_dpo/margin": 53.61853790283203, "fcm_dpo/q_t": 0.41387444734573364, "grad_norm": 6.419241428375244, "learning_rate": 2.1821473643827137e-09, "logits/chosen": 1.409173846244812, "logits/rejected": 1.2437349557876587, "logps/chosen": -193.94113159179688, "logps/ref_chosen": -92.38919067382812, "logps/ref_rejected": -103.70460510253906, "logps/rejected": -258.8750915527344, "loss": 1.1161, "margin_dpo/margin_mean": 53.61853790283203, "margin_dpo/margin_std": 80.25310516357422, "step": 637 }, { "epoch": 0.9644746787603931, "fcm_dpo/beta": 0.007207995280623436, "fcm_dpo/delta": 0.06300410628318787, "fcm_dpo/margin": 47.013023376464844, "fcm_dpo/q_t": 0.42337724566459656, "grad_norm": 10.596990585327148, "learning_rate": 2.0112967923011646e-09, "logits/chosen": 1.3792505264282227, "logits/rejected": 1.233081340789795, "logps/chosen": -178.09756469726562, "logps/ref_chosen": -83.36921691894531, "logps/ref_rejected": -103.04508209228516, "logps/rejected": -244.78646850585938, "loss": 1.1589, "margin_dpo/margin_mean": 47.01301956176758, "margin_dpo/margin_std": 82.29072570800781, "step": 638 }, { "epoch": 0.9659863945578231, "fcm_dpo/beta": 0.007193556986749172, "fcm_dpo/delta": -0.041456226259469986, "fcm_dpo/margin": 61.10279083251953, "fcm_dpo/q_t": 0.40119677782058716, "grad_norm": 6.433756351470947, "learning_rate": 1.847382997337943e-09, "logits/chosen": 1.2101722955703735, "logits/rejected": 0.983842670917511, "logps/chosen": -148.6090850830078, "logps/ref_chosen": -70.45247650146484, "logps/ref_rejected": -93.77748107910156, "logps/rejected": -233.036865234375, "loss": 1.0853, "margin_dpo/margin_mean": 61.102783203125, "margin_dpo/margin_std": 86.17239379882812, "step": 639 }, { "epoch": 0.9674981103552532, "fcm_dpo/beta": 0.007202021777629852, "fcm_dpo/delta": 0.06750090420246124, "fcm_dpo/margin": 46.48023986816406, "fcm_dpo/q_t": 0.4226430058479309, "grad_norm": 7.745254039764404, "learning_rate": 1.690410564514244e-09, "logits/chosen": 1.6796314716339111, "logits/rejected": 1.4730074405670166, "logps/chosen": -152.3632354736328, "logps/ref_chosen": -68.51570129394531, "logps/ref_rejected": -92.35081481933594, "logps/rejected": -222.6785888671875, "loss": 1.1784, "margin_dpo/margin_mean": 46.48023986816406, "margin_dpo/margin_std": 89.2449951171875, "step": 640 }, { "epoch": 0.9690098261526833, "fcm_dpo/beta": 0.007284895051270723, "fcm_dpo/delta": 0.058470651507377625, "fcm_dpo/margin": 47.1434326171875, "fcm_dpo/q_t": 0.4222804605960846, "grad_norm": 8.309812545776367, "learning_rate": 1.5403838846864692e-09, "logits/chosen": 1.6926078796386719, "logits/rejected": 1.6602896451950073, "logps/chosen": -186.4867401123047, "logps/ref_chosen": -92.35102844238281, "logps/ref_rejected": -102.4269790649414, "logps/rejected": -243.7061309814453, "loss": 1.1561, "margin_dpo/margin_mean": 47.14342498779297, "margin_dpo/margin_std": 81.37835693359375, "step": 641 }, { "epoch": 0.9705215419501134, "fcm_dpo/beta": 0.007320537231862545, "fcm_dpo/delta": -0.011010588146746159, "fcm_dpo/margin": 38.06249237060547, "fcm_dpo/q_t": 0.4363037943840027, "grad_norm": 8.437037467956543, "learning_rate": 1.3973071544233218e-09, "logits/chosen": 1.2161636352539062, "logits/rejected": 1.2543673515319824, "logps/chosen": -184.1721954345703, "logps/ref_chosen": -88.39617919921875, "logps/ref_rejected": -88.73035430908203, "logps/rejected": -222.56884765625, "loss": 1.2144, "margin_dpo/margin_mean": 38.06249237060547, "margin_dpo/margin_std": 80.98915100097656, "step": 642 }, { "epoch": 0.9720332577475435, "fcm_dpo/beta": 0.007371140643954277, "fcm_dpo/delta": 0.07876811921596527, "fcm_dpo/margin": 43.91210174560547, "fcm_dpo/q_t": 0.42799311876296997, "grad_norm": 9.573797225952148, "learning_rate": 1.261184375888541e-09, "logits/chosen": 1.276777982711792, "logits/rejected": 0.9570169448852539, "logps/chosen": -174.10198974609375, "logps/ref_chosen": -84.83087158203125, "logps/ref_rejected": -105.31499481201172, "logps/rejected": -238.4982147216797, "loss": 1.2043, "margin_dpo/margin_mean": 43.91210174560547, "margin_dpo/margin_std": 94.48892974853516, "step": 643 }, { "epoch": 0.9735449735449735, "fcm_dpo/beta": 0.0075573730282485485, "fcm_dpo/delta": 0.09459188580513, "fcm_dpo/margin": 40.763572692871094, "fcm_dpo/q_t": 0.431037575006485, "grad_norm": 8.969107627868652, "learning_rate": 1.1320193567288527e-09, "logits/chosen": 1.6555099487304688, "logits/rejected": 1.5394855737686157, "logps/chosen": -154.88946533203125, "logps/ref_chosen": -65.11122131347656, "logps/ref_rejected": -80.4027328491211, "logps/rejected": -210.94454956054688, "loss": 1.222, "margin_dpo/margin_mean": 40.763572692871094, "margin_dpo/margin_std": 94.83735656738281, "step": 644 }, { "epoch": 0.9750566893424036, "fcm_dpo/beta": 0.007558923214673996, "fcm_dpo/delta": -0.014460157603025436, "fcm_dpo/margin": 54.74996566772461, "fcm_dpo/q_t": 0.4060879647731781, "grad_norm": 8.757953643798828, "learning_rate": 1.0098157099674987e-09, "logits/chosen": 1.608795404434204, "logits/rejected": 1.5974090099334717, "logps/chosen": -162.4353790283203, "logps/ref_chosen": -76.93634033203125, "logps/ref_rejected": -89.14311981201172, "logps/rejected": -229.39212036132812, "loss": 1.1055, "margin_dpo/margin_mean": 54.749969482421875, "margin_dpo/margin_std": 82.67251586914062, "step": 645 }, { "epoch": 0.9765684051398337, "fcm_dpo/beta": 0.007531238719820976, "fcm_dpo/delta": -0.0364893302321434, "fcm_dpo/margin": 57.74522018432617, "fcm_dpo/q_t": 0.4003961682319641, "grad_norm": 8.327498435974121, "learning_rate": 8.945768539031783e-10, "logits/chosen": 1.7237350940704346, "logits/rejected": 1.6151888370513916, "logps/chosen": -174.32347106933594, "logps/ref_chosen": -77.69122314453125, "logps/ref_rejected": -98.14374542236328, "logps/rejected": -252.52120971679688, "loss": 1.0839, "margin_dpo/margin_mean": 57.74522399902344, "margin_dpo/margin_std": 80.33204650878906, "step": 646 }, { "epoch": 0.9780801209372638, "fcm_dpo/beta": 0.007370360661298037, "fcm_dpo/delta": -0.13900147378444672, "fcm_dpo/margin": 72.1575698852539, "fcm_dpo/q_t": 0.3773247301578522, "grad_norm": 8.999638557434082, "learning_rate": 7.863060120144316e-10, "logits/chosen": 1.6755530834197998, "logits/rejected": 1.4901223182678223, "logps/chosen": -178.73919677734375, "logps/ref_chosen": -83.79997253417969, "logps/ref_rejected": -116.81965637207031, "logps/rejected": -283.91644287109375, "loss": 0.9998, "margin_dpo/margin_mean": 72.1575698852539, "margin_dpo/margin_std": 77.4002914428711, "step": 647 }, { "epoch": 0.9795918367346939, "fcm_dpo/beta": 0.007339120376855135, "fcm_dpo/delta": 0.05002519488334656, "fcm_dpo/margin": 47.92961120605469, "fcm_dpo/q_t": 0.4191440939903259, "grad_norm": 6.385339736938477, "learning_rate": 6.850062128694045e-10, "logits/chosen": 1.3917063474655151, "logits/rejected": 1.2107430696487427, "logps/chosen": -182.20465087890625, "logps/ref_chosen": -85.9629898071289, "logps/ref_rejected": -101.36552429199219, "logps/rejected": -245.53680419921875, "loss": 1.1435, "margin_dpo/margin_mean": 47.92961120605469, "margin_dpo/margin_std": 76.83576965332031, "step": 648 }, { "epoch": 0.981103552532124, "fcm_dpo/beta": 0.007368366699665785, "fcm_dpo/delta": 0.004403694532811642, "fcm_dpo/margin": 53.710758209228516, "fcm_dpo/q_t": 0.4111127257347107, "grad_norm": 9.27660083770752, "learning_rate": 5.906802900412788e-10, "logits/chosen": 1.7626551389694214, "logits/rejected": 1.6102287769317627, "logps/chosen": -153.03140258789062, "logps/ref_chosen": -68.64892578125, "logps/ref_rejected": -89.84898376464844, "logps/rejected": -227.94223022460938, "loss": 1.1353, "margin_dpo/margin_mean": 53.710758209228516, "margin_dpo/margin_std": 91.26134490966797, "step": 649 }, { "epoch": 0.982615268329554, "fcm_dpo/beta": 0.007361288648098707, "fcm_dpo/delta": -0.014930552802979946, "fcm_dpo/margin": 56.28282165527344, "fcm_dpo/q_t": 0.40710097551345825, "grad_norm": 7.917013645172119, "learning_rate": 5.033308820289184e-10, "logits/chosen": 1.457055687904358, "logits/rejected": 1.2520674467086792, "logps/chosen": -147.81912231445312, "logps/ref_chosen": -72.97265625, "logps/ref_rejected": -93.04617309570312, "logps/rejected": -224.17544555664062, "loss": 1.1136, "margin_dpo/margin_mean": 56.282814025878906, "margin_dpo/margin_std": 87.57786560058594, "step": 650 }, { "epoch": 0.9841269841269841, "fcm_dpo/beta": 0.007448974065482616, "fcm_dpo/delta": 0.06296257674694061, "fcm_dpo/margin": 45.47364044189453, "fcm_dpo/q_t": 0.4245682954788208, "grad_norm": 14.473260879516602, "learning_rate": 4.2296043218295606e-10, "logits/chosen": 1.6711803674697876, "logits/rejected": 1.4281296730041504, "logps/chosen": -158.29653930664062, "logps/ref_chosen": -71.05281066894531, "logps/ref_rejected": -94.23469543457031, "logps/rejected": -226.95205688476562, "loss": 1.1638, "margin_dpo/margin_mean": 45.47364044189453, "margin_dpo/margin_std": 79.72322082519531, "step": 651 }, { "epoch": 0.9856386999244142, "fcm_dpo/beta": 0.0074330344796180725, "fcm_dpo/delta": -0.005656237713992596, "fcm_dpo/margin": 54.540061950683594, "fcm_dpo/q_t": 0.40991270542144775, "grad_norm": 9.309069633483887, "learning_rate": 3.4957118863768176e-10, "logits/chosen": 1.4886457920074463, "logits/rejected": 1.407999873161316, "logps/chosen": -172.19561767578125, "logps/ref_chosen": -80.06941223144531, "logps/ref_rejected": -99.22327423095703, "logps/rejected": -245.88955688476562, "loss": 1.1204, "margin_dpo/margin_mean": 54.540061950683594, "margin_dpo/margin_std": 87.85244750976562, "step": 652 }, { "epoch": 0.9871504157218443, "fcm_dpo/beta": 0.007382528856396675, "fcm_dpo/delta": -0.08395875990390778, "fcm_dpo/margin": 64.99937438964844, "fcm_dpo/q_t": 0.3910355567932129, "grad_norm": 10.658775329589844, "learning_rate": 2.831652042480093e-10, "logits/chosen": 1.7752074003219604, "logits/rejected": 1.6271909475326538, "logps/chosen": -158.2850341796875, "logps/ref_chosen": -80.35701751708984, "logps/ref_rejected": -92.1295394897461, "logps/rejected": -235.05691528320312, "loss": 1.0567, "margin_dpo/margin_mean": 64.99937438964844, "margin_dpo/margin_std": 84.79908752441406, "step": 653 }, { "epoch": 0.9886621315192744, "fcm_dpo/beta": 0.00728489737957716, "fcm_dpo/delta": -0.02065638080239296, "fcm_dpo/margin": 43.905029296875, "fcm_dpo/q_t": 0.4288891553878784, "grad_norm": 9.484676361083984, "learning_rate": 2.2374433653205016e-10, "logits/chosen": 1.6096107959747314, "logits/rejected": 1.3091274499893188, "logps/chosen": -171.35845947265625, "logps/ref_chosen": -78.06475830078125, "logps/ref_rejected": -106.05763244628906, "logps/rejected": -243.25636291503906, "loss": 1.1841, "margin_dpo/margin_mean": 43.905029296875, "margin_dpo/margin_std": 84.46098327636719, "step": 654 }, { "epoch": 0.9901738473167044, "fcm_dpo/beta": 0.007229020819067955, "fcm_dpo/delta": -0.002755559980869293, "fcm_dpo/margin": 55.5506591796875, "fcm_dpo/q_t": 0.4089524447917938, "grad_norm": 8.344823837280273, "learning_rate": 1.7131024761923852e-10, "logits/chosen": 1.263197660446167, "logits/rejected": 0.9553810358047485, "logps/chosen": -145.0635528564453, "logps/ref_chosen": -67.03407287597656, "logps/ref_rejected": -97.57197570800781, "logps/rejected": -231.152099609375, "loss": 1.0989, "margin_dpo/margin_mean": 55.5506591796875, "margin_dpo/margin_std": 75.53077697753906, "step": 655 }, { "epoch": 0.9916855631141346, "fcm_dpo/beta": 0.007296122610569, "fcm_dpo/delta": 0.006786551792174578, "fcm_dpo/margin": 53.92597961425781, "fcm_dpo/q_t": 0.4102938175201416, "grad_norm": 5.894443035125732, "learning_rate": 1.2586440420372934e-10, "logits/chosen": 1.2040374279022217, "logits/rejected": 1.0824060440063477, "logps/chosen": -184.591796875, "logps/ref_chosen": -89.31463623046875, "logps/ref_rejected": -105.14315795898438, "logps/rejected": -254.3463134765625, "loss": 1.1228, "margin_dpo/margin_mean": 53.92597961425781, "margin_dpo/margin_std": 85.65472412109375, "step": 656 }, { "epoch": 0.9931972789115646, "fcm_dpo/beta": 0.0071443431079387665, "fcm_dpo/delta": -0.13093850016593933, "fcm_dpo/margin": 73.339111328125, "fcm_dpo/q_t": 0.38050127029418945, "grad_norm": 7.297959327697754, "learning_rate": 8.740807750345913e-11, "logits/chosen": 1.6534512042999268, "logits/rejected": 1.423713207244873, "logps/chosen": -142.54873657226562, "logps/ref_chosen": -64.89747619628906, "logps/ref_rejected": -94.21998596191406, "logps/rejected": -245.21035766601562, "loss": 1.0233, "margin_dpo/margin_mean": 73.339111328125, "margin_dpo/margin_std": 88.55533599853516, "step": 657 }, { "epoch": 0.9947089947089947, "fcm_dpo/beta": 0.007135612890124321, "fcm_dpo/delta": -0.005358390510082245, "fcm_dpo/margin": 56.69953155517578, "fcm_dpo/q_t": 0.4094886779785156, "grad_norm": 12.389694213867188, "learning_rate": 5.594234322453539e-11, "logits/chosen": 1.448069453239441, "logits/rejected": 1.3515777587890625, "logps/chosen": -162.52061462402344, "logps/ref_chosen": -81.16606140136719, "logps/ref_rejected": -97.72825622558594, "logps/rejected": -235.7823486328125, "loss": 1.1347, "margin_dpo/margin_mean": 56.69953155517578, "margin_dpo/margin_std": 96.20867919921875, "step": 658 }, { "epoch": 0.9962207105064248, "fcm_dpo/beta": 0.007101975381374359, "fcm_dpo/delta": 0.005114157218486071, "fcm_dpo/margin": 29.31315803527832, "fcm_dpo/q_t": 0.4529067277908325, "grad_norm": 7.469188690185547, "learning_rate": 3.146808153123293e-11, "logits/chosen": 1.7898869514465332, "logits/rejected": 1.5685615539550781, "logps/chosen": -173.17181396484375, "logps/ref_chosen": -74.42193603515625, "logps/ref_rejected": -87.81561279296875, "logps/rejected": -215.87863159179688, "loss": 1.2881, "margin_dpo/margin_mean": 29.313159942626953, "margin_dpo/margin_std": 86.78317260742188, "step": 659 }, { "epoch": 0.9977324263038548, "fcm_dpo/beta": 0.007041741628199816, "fcm_dpo/delta": -0.042580414563417435, "fcm_dpo/margin": 62.558837890625, "fcm_dpo/q_t": 0.39910098910331726, "grad_norm": 8.689871788024902, "learning_rate": 1.3985977021235829e-11, "logits/chosen": 1.7107985019683838, "logits/rejected": 1.5310373306274414, "logps/chosen": -157.3565216064453, "logps/ref_chosen": -71.68511962890625, "logps/ref_rejected": -98.01472473144531, "logps/rejected": -246.24496459960938, "loss": 1.0657, "margin_dpo/margin_mean": 62.558837890625, "margin_dpo/margin_std": 78.82179260253906, "step": 660 }, { "epoch": 0.999244142101285, "fcm_dpo/beta": 0.007080161012709141, "fcm_dpo/delta": 0.05424977466464043, "fcm_dpo/margin": 36.04890060424805, "fcm_dpo/q_t": 0.4424479603767395, "grad_norm": 7.011692523956299, "learning_rate": 3.4965187065971735e-12, "logits/chosen": 1.4969191551208496, "logits/rejected": 1.2757140398025513, "logps/chosen": -187.23583984375, "logps/ref_chosen": -78.35111999511719, "logps/ref_rejected": -99.47113037109375, "logps/rejected": -244.40475463867188, "loss": 1.251, "margin_dpo/margin_mean": 36.04889678955078, "margin_dpo/margin_std": 93.42096710205078, "step": 661 }, { "epoch": 0.999244142101285, "step": 661, "total_flos": 0.0, "train_loss": 1.1795702667712444, "train_runtime": 2124.2784, "train_samples_per_second": 19.93, "train_steps_per_second": 0.311 } ], "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 }