{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 200, "global_step": 681, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0014684287812041115, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02287006378173828, "fcm_dpo/q_t": 0.5005706548690796, "grad_norm": 83.51022338867188, "learning_rate": 0.0, "logits/chosen": -0.4974287748336792, "logits/rejected": -0.43299180269241333, "logps/chosen": -50.1435661315918, "logps/ref_chosen": -50.14883804321289, "logps/ref_rejected": -74.1280517578125, "logps/rejected": -74.09991455078125, "loss": 1.389, "margin_dpo/margin_mean": -0.02287048101425171, "margin_dpo/margin_std": 0.41920793056488037, "step": 1 }, { "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.007774516940116882, "fcm_dpo/q_t": 0.500194251537323, "grad_norm": 89.51734924316406, "learning_rate": 2.898550724637681e-08, "logits/chosen": -0.4900396764278412, "logits/rejected": -0.4533483386039734, "logps/chosen": -56.05889892578125, "logps/ref_chosen": -56.05734634399414, "logps/ref_rejected": -78.69325256347656, "logps/rejected": -78.68702697753906, "loss": 1.3874, "margin_dpo/margin_mean": -0.007774434983730316, "margin_dpo/margin_std": 0.37205374240875244, "step": 5 }, { "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.017793914303183556, "fcm_dpo/q_t": 0.49955543875694275, "grad_norm": 70.37712097167969, "learning_rate": 6.521739130434782e-08, "logits/chosen": -0.49457064270973206, "logits/rejected": -0.4558273255825043, "logps/chosen": -59.53853225708008, "logps/ref_chosen": -59.54457473754883, "logps/ref_rejected": -91.17041778564453, "logps/rejected": -91.18217468261719, "loss": 1.3849, "margin_dpo/margin_mean": 0.017794013023376465, "margin_dpo/margin_std": 0.40502458810806274, "step": 10 }, { "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.013358664698898792, "fcm_dpo/q_t": 0.4996658265590668, "grad_norm": 64.22943878173828, "learning_rate": 1.0144927536231885e-07, "logits/chosen": -0.48777151107788086, "logits/rejected": -0.4582846164703369, "logps/chosen": -58.85475540161133, "logps/ref_chosen": -58.83195877075195, "logps/ref_rejected": -92.93949890136719, "logps/rejected": -92.97565460205078, "loss": 1.3854, "margin_dpo/margin_mean": 0.01335877738893032, "margin_dpo/margin_std": 0.3935544788837433, "step": 15 }, { "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.09099045395851135, "fcm_dpo/q_t": 0.49772581458091736, "grad_norm": 73.8072509765625, "learning_rate": 1.3768115942028986e-07, "logits/chosen": -0.5034265518188477, "logits/rejected": -0.45439839363098145, "logps/chosen": -59.618431091308594, "logps/ref_chosen": -59.6396598815918, "logps/ref_rejected": -82.76026916503906, "logps/rejected": -82.83003234863281, "loss": 1.3775, "margin_dpo/margin_mean": 0.09099055081605911, "margin_dpo/margin_std": 0.35805386304855347, "step": 20 }, { "epoch": 0.03671071953010279, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.21624115109443665, "fcm_dpo/q_t": 0.49459710717201233, "grad_norm": 73.70997619628906, "learning_rate": 1.7391304347826085e-07, "logits/chosen": -0.5030972361564636, "logits/rejected": -0.47600775957107544, "logps/chosen": -53.1678352355957, "logps/ref_chosen": -53.205284118652344, "logps/ref_rejected": -88.99608612060547, "logps/rejected": -89.17488098144531, "loss": 1.3652, "margin_dpo/margin_mean": 0.21624140441417694, "margin_dpo/margin_std": 0.4134984016418457, "step": 25 }, { "epoch": 0.04405286343612335, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4783309996128082, "fcm_dpo/q_t": 0.4880555272102356, "grad_norm": 87.57738494873047, "learning_rate": 2.1014492753623187e-07, "logits/chosen": -0.5126634240150452, "logits/rejected": -0.4724017083644867, "logps/chosen": -53.46284103393555, "logps/ref_chosen": -53.5526008605957, "logps/ref_rejected": -97.87371826171875, "logps/rejected": -98.26229858398438, "loss": 1.3399, "margin_dpo/margin_mean": 0.47833117842674255, "margin_dpo/margin_std": 0.5532158017158508, "step": 30 }, { "epoch": 0.0513950073421439, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7770208120346069, "fcm_dpo/q_t": 0.48065242171287537, "grad_norm": 82.80741119384766, "learning_rate": 2.463768115942029e-07, "logits/chosen": -0.5122802257537842, "logits/rejected": -0.4792337417602539, "logps/chosen": -56.210533142089844, "logps/ref_chosen": -56.3298454284668, "logps/ref_rejected": -91.76858520507812, "logps/rejected": -92.4262924194336, "loss": 1.3121, "margin_dpo/margin_mean": 0.7770206332206726, "margin_dpo/margin_std": 0.8431359529495239, "step": 35 }, { "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.3613355159759521, "fcm_dpo/q_t": 0.46629637479782104, "grad_norm": 60.205745697021484, "learning_rate": 2.8260869565217386e-07, "logits/chosen": -0.5219647884368896, "logits/rejected": -0.4850676953792572, "logps/chosen": -54.278953552246094, "logps/ref_chosen": -54.38492965698242, "logps/ref_rejected": -82.94353485107422, "logps/rejected": -84.19891357421875, "loss": 1.2605, "margin_dpo/margin_mean": 1.3613355159759521, "margin_dpo/margin_std": 1.4631270170211792, "step": 40 }, { "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.10669933259487152, "fcm_dpo/delta": 0.12330988794565201, "fcm_dpo/margin": 2.4058146476745605, "fcm_dpo/q_t": 0.4380177855491638, "grad_norm": 68.1404800415039, "learning_rate": 3.188405797101449e-07, "logits/chosen": -0.5121681094169617, "logits/rejected": -0.4841387867927551, "logps/chosen": -54.63410186767578, "logps/ref_chosen": -54.862335205078125, "logps/ref_rejected": -98.0264663696289, "logps/rejected": -100.20404052734375, "loss": 1.1601, "margin_dpo/margin_mean": 2.4058141708374023, "margin_dpo/margin_std": 2.224526882171631, "step": 45 }, { "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.11535553634166718, "fcm_dpo/delta": 0.0067912861704826355, "fcm_dpo/margin": 3.4076931476593018, "fcm_dpo/q_t": 0.4073707163333893, "grad_norm": 68.53203582763672, "learning_rate": 3.5507246376811595e-07, "logits/chosen": -0.562477707862854, "logits/rejected": -0.5116110444068909, "logps/chosen": -58.12030029296875, "logps/ref_chosen": -58.304595947265625, "logps/ref_rejected": -91.69480895996094, "logps/rejected": -94.91819763183594, "loss": 1.0694, "margin_dpo/margin_mean": 3.4076931476593018, "margin_dpo/margin_std": 3.365309953689575, "step": 50 }, { "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.10359964519739151, "fcm_dpo/delta": -0.19395628571510315, "fcm_dpo/margin": 5.653986930847168, "fcm_dpo/q_t": 0.3696584403514862, "grad_norm": 52.22880935668945, "learning_rate": 3.9130434782608694e-07, "logits/chosen": -0.5961982011795044, "logits/rejected": -0.5513918399810791, "logps/chosen": -56.34511184692383, "logps/ref_chosen": -56.06591796875, "logps/ref_rejected": -85.69367980957031, "logps/rejected": -91.62686157226562, "loss": 0.968, "margin_dpo/margin_mean": 5.653987407684326, "margin_dpo/margin_std": 5.778587818145752, "step": 55 }, { "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.08351375162601471, "fcm_dpo/delta": -0.2116357386112213, "fcm_dpo/margin": 7.191043853759766, "fcm_dpo/q_t": 0.3656514883041382, "grad_norm": 53.557796478271484, "learning_rate": 4.2753623188405794e-07, "logits/chosen": -0.614372193813324, "logits/rejected": -0.5657563209533691, "logps/chosen": -62.0140266418457, "logps/ref_chosen": -60.6871337890625, "logps/ref_rejected": -89.72715759277344, "logps/rejected": -98.24510192871094, "loss": 0.9663, "margin_dpo/margin_mean": 7.191043853759766, "margin_dpo/margin_std": 7.640544891357422, "step": 60 }, { "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.06861739605665207, "fcm_dpo/delta": -0.19986796379089355, "fcm_dpo/margin": 8.604007720947266, "fcm_dpo/q_t": 0.37094324827194214, "grad_norm": 41.2183723449707, "learning_rate": 4.63768115942029e-07, "logits/chosen": -0.6187667846679688, "logits/rejected": -0.5872984528541565, "logps/chosen": -63.8182487487793, "logps/ref_chosen": -61.75325393676758, "logps/ref_rejected": -93.30108642578125, "logps/rejected": -103.9700927734375, "loss": 0.9904, "margin_dpo/margin_mean": 8.604008674621582, "margin_dpo/margin_std": 10.206991195678711, "step": 65 }, { "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.05661847069859505, "fcm_dpo/delta": -0.1891469657421112, "fcm_dpo/margin": 10.259188652038574, "fcm_dpo/q_t": 0.3718249201774597, "grad_norm": 39.173946380615234, "learning_rate": 5e-07, "logits/chosen": -0.6201697587966919, "logits/rejected": -0.5813671350479126, "logps/chosen": -63.28239059448242, "logps/ref_chosen": -59.548004150390625, "logps/ref_rejected": -84.01609802246094, "logps/rejected": -98.00967407226562, "loss": 0.98, "margin_dpo/margin_mean": 10.259187698364258, "margin_dpo/margin_std": 11.377448081970215, "step": 70 }, { "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.04313134402036667, "fcm_dpo/delta": -0.354682594537735, "fcm_dpo/margin": 17.1324405670166, "fcm_dpo/q_t": 0.34925174713134766, "grad_norm": 31.45458221435547, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.6573521494865417, "logits/rejected": -0.6307379603385925, "logps/chosen": -67.56647491455078, "logps/ref_chosen": -59.86931228637695, "logps/ref_rejected": -98.05613708496094, "logps/rejected": -122.8857421875, "loss": 0.9415, "margin_dpo/margin_mean": 17.132436752319336, "margin_dpo/margin_std": 21.05853843688965, "step": 75 }, { "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.03168341889977455, "fcm_dpo/delta": -0.22349825501441956, "fcm_dpo/margin": 19.243648529052734, "fcm_dpo/q_t": 0.3687431514263153, "grad_norm": 28.407453536987305, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.6761894822120667, "logits/rejected": -0.6410226821899414, "logps/chosen": -66.81375122070312, "logps/ref_chosen": -56.18925857543945, "logps/ref_rejected": -86.42393493652344, "logps/rejected": -116.2920913696289, "loss": 0.9939, "margin_dpo/margin_mean": 19.243648529052734, "margin_dpo/margin_std": 23.982927322387695, "step": 80 }, { "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.025587420910596848, "fcm_dpo/delta": -0.18879233300685883, "fcm_dpo/margin": 22.462379455566406, "fcm_dpo/q_t": 0.3756103515625, "grad_norm": 26.530624389648438, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.6672359704971313, "logits/rejected": -0.6387097835540771, "logps/chosen": -73.97752380371094, "logps/ref_chosen": -60.018287658691406, "logps/ref_rejected": -98.01185607910156, "logps/rejected": -134.43350219726562, "loss": 1.0175, "margin_dpo/margin_mean": 22.462383270263672, "margin_dpo/margin_std": 29.141366958618164, "step": 85 }, { "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.02217969484627247, "fcm_dpo/delta": -0.15027864277362823, "fcm_dpo/margin": 24.47287368774414, "fcm_dpo/q_t": 0.38759851455688477, "grad_norm": 26.1925106048584, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.6896823644638062, "logits/rejected": -0.6781142354011536, "logps/chosen": -78.13359069824219, "logps/ref_chosen": -59.8709831237793, "logps/ref_rejected": -96.78519439697266, "logps/rejected": -139.5206756591797, "loss": 1.0705, "margin_dpo/margin_mean": 24.47287368774414, "margin_dpo/margin_std": 37.618186950683594, "step": 90 }, { "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.019353795796632767, "fcm_dpo/delta": -0.10450585186481476, "fcm_dpo/margin": 25.787700653076172, "fcm_dpo/q_t": 0.3899272084236145, "grad_norm": 21.816125869750977, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.690959095954895, "logits/rejected": -0.6617642641067505, "logps/chosen": -74.7010269165039, "logps/ref_chosen": -55.94385528564453, "logps/ref_rejected": -83.6790542602539, "logps/rejected": -128.2239227294922, "loss": 1.0485, "margin_dpo/margin_mean": 25.787700653076172, "margin_dpo/margin_std": 34.392967224121094, "step": 95 }, { "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.017355434596538544, "fcm_dpo/delta": -0.09411351382732391, "fcm_dpo/margin": 28.09613037109375, "fcm_dpo/q_t": 0.3915005326271057, "grad_norm": 23.495582580566406, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.6664688587188721, "logits/rejected": -0.6419214010238647, "logps/chosen": -78.07569885253906, "logps/ref_chosen": -57.05888748168945, "logps/ref_rejected": -86.11727142333984, "logps/rejected": -135.23020935058594, "loss": 1.057, "margin_dpo/margin_mean": 28.09613037109375, "margin_dpo/margin_std": 37.85677719116211, "step": 100 }, { "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.01605215296149254, "fcm_dpo/delta": -0.14830544590950012, "fcm_dpo/margin": 32.540122985839844, "fcm_dpo/q_t": 0.38781845569610596, "grad_norm": 22.918745040893555, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.6817123889923096, "logits/rejected": -0.6505634188652039, "logps/chosen": -84.82267761230469, "logps/ref_chosen": -59.20774459838867, "logps/ref_rejected": -86.49754333496094, "logps/rejected": -144.65260314941406, "loss": 1.052, "margin_dpo/margin_mean": 32.540122985839844, "margin_dpo/margin_std": 44.27156066894531, "step": 105 }, { "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.013644491322338581, "fcm_dpo/delta": -0.1275637000799179, "fcm_dpo/margin": 38.23952102661133, "fcm_dpo/q_t": 0.38538408279418945, "grad_norm": 18.83212661743164, "learning_rate": 4.947482930773511e-07, "logits/chosen": -0.6280543804168701, "logits/rejected": -0.5978196859359741, "logps/chosen": -88.46613311767578, "logps/ref_chosen": -60.437957763671875, "logps/ref_rejected": -90.83917999267578, "logps/rejected": -157.1068878173828, "loss": 1.0403, "margin_dpo/margin_mean": 38.23952102661133, "margin_dpo/margin_std": 50.03200149536133, "step": 110 }, { "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.012269817292690277, "fcm_dpo/delta": -0.06500422954559326, "fcm_dpo/margin": 37.55778121948242, "fcm_dpo/q_t": 0.39965710043907166, "grad_norm": 27.971012115478516, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.6194974184036255, "logits/rejected": -0.5893136262893677, "logps/chosen": -96.05851745605469, "logps/ref_chosen": -61.7908821105957, "logps/ref_rejected": -85.36943054199219, "logps/rejected": -157.19485473632812, "loss": 1.0871, "margin_dpo/margin_mean": 37.55778121948242, "margin_dpo/margin_std": 56.132057189941406, "step": 115 }, { "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.01178265642374754, "fcm_dpo/delta": -0.05727796629071236, "fcm_dpo/margin": 38.61918640136719, "fcm_dpo/q_t": 0.39924171566963196, "grad_norm": 21.68465805053711, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6018606424331665, "logits/rejected": -0.580163836479187, "logps/chosen": -104.28592681884766, "logps/ref_chosen": -65.3261489868164, "logps/ref_rejected": -86.75518798828125, "logps/rejected": -164.33416748046875, "loss": 1.0756, "margin_dpo/margin_mean": 38.61919021606445, "margin_dpo/margin_std": 53.98603057861328, "step": 120 }, { "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.009986856020987034, "fcm_dpo/delta": -0.2121940553188324, "fcm_dpo/margin": 59.94220733642578, "fcm_dpo/q_t": 0.3725253939628601, "grad_norm": 19.076047897338867, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.5440986752510071, "logits/rejected": -0.5444971323013306, "logps/chosen": -102.51204681396484, "logps/ref_chosen": -58.323204040527344, "logps/ref_rejected": -101.2106704711914, "logps/rejected": -205.34170532226562, "loss": 1.0102, "margin_dpo/margin_mean": 59.94221115112305, "margin_dpo/margin_std": 79.08087158203125, "step": 125 }, { "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.00899023748934269, "fcm_dpo/delta": -0.0574115514755249, "fcm_dpo/margin": 50.56972122192383, "fcm_dpo/q_t": 0.4008238911628723, "grad_norm": 21.011878967285156, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.5073991417884827, "logits/rejected": -0.49552878737449646, "logps/chosen": -103.4196548461914, "logps/ref_chosen": -56.38518524169922, "logps/ref_rejected": -86.15767669677734, "logps/rejected": -183.76187133789062, "loss": 1.0949, "margin_dpo/margin_mean": 50.569725036621094, "margin_dpo/margin_std": 77.31275939941406, "step": 130 }, { "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.00818575732409954, "fcm_dpo/delta": -0.10119257867336273, "fcm_dpo/margin": 60.596702575683594, "fcm_dpo/q_t": 0.39124470949172974, "grad_norm": 21.952133178710938, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.487496942281723, "logits/rejected": -0.4898650050163269, "logps/chosen": -103.59019470214844, "logps/ref_chosen": -54.59065628051758, "logps/ref_rejected": -95.26080322265625, "logps/rejected": -204.8570556640625, "loss": 1.0502, "margin_dpo/margin_mean": 60.596702575683594, "margin_dpo/margin_std": 81.37339782714844, "step": 135 }, { "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.007442955859005451, "fcm_dpo/delta": -0.09183903783559799, "fcm_dpo/margin": 65.49043273925781, "fcm_dpo/q_t": 0.39486831426620483, "grad_norm": 20.813894271850586, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.4298696517944336, "logits/rejected": -0.4265991747379303, "logps/chosen": -114.61155700683594, "logps/ref_chosen": -56.04347610473633, "logps/ref_rejected": -93.27880859375, "logps/rejected": -217.3373565673828, "loss": 1.064, "margin_dpo/margin_mean": 65.49043273925781, "margin_dpo/margin_std": 92.36592102050781, "step": 140 }, { "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.007036598864942789, "fcm_dpo/delta": -0.012454366311430931, "fcm_dpo/margin": 58.48834991455078, "fcm_dpo/q_t": 0.40865078568458557, "grad_norm": 23.93450927734375, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.40944308042526245, "logits/rejected": -0.39997661113739014, "logps/chosen": -129.96218872070312, "logps/ref_chosen": -61.4414176940918, "logps/ref_rejected": -86.32813262939453, "logps/rejected": -213.33724975585938, "loss": 1.115, "margin_dpo/margin_mean": 58.48834991455078, "margin_dpo/margin_std": 92.11347961425781, "step": 145 }, { "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.006964278873056173, "fcm_dpo/delta": -0.03296690061688423, "fcm_dpo/margin": 59.02332305908203, "fcm_dpo/q_t": 0.4086775779724121, "grad_norm": 18.16049575805664, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.44094228744506836, "logits/rejected": -0.42822694778442383, "logps/chosen": -123.40377044677734, "logps/ref_chosen": -57.70451736450195, "logps/ref_rejected": -87.76991271972656, "logps/rejected": -212.4925079345703, "loss": 1.1066, "margin_dpo/margin_mean": 59.02332305908203, "margin_dpo/margin_std": 87.62327575683594, "step": 150 }, { "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.006572508253157139, "fcm_dpo/delta": -0.058337144553661346, "fcm_dpo/margin": 69.28470611572266, "fcm_dpo/q_t": 0.3986854553222656, "grad_norm": 19.04676628112793, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.4826192855834961, "logits/rejected": -0.4734383523464203, "logps/chosen": -126.15814208984375, "logps/ref_chosen": -62.08925247192383, "logps/ref_rejected": -94.79930114746094, "logps/rejected": -228.1529083251953, "loss": 1.076, "margin_dpo/margin_mean": 69.28469848632812, "margin_dpo/margin_std": 97.56148529052734, "step": 155 }, { "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.006233358755707741, "fcm_dpo/delta": -0.06433447450399399, "fcm_dpo/margin": 70.75523376464844, "fcm_dpo/q_t": 0.4046873450279236, "grad_norm": 23.905847549438477, "learning_rate": 4.737908228387656e-07, "logits/chosen": -0.37852075695991516, "logits/rejected": -0.37446096539497375, "logps/chosen": -155.4814453125, "logps/ref_chosen": -67.15288543701172, "logps/ref_rejected": -96.92537689208984, "logps/rejected": -256.0091552734375, "loss": 1.1205, "margin_dpo/margin_mean": 70.75523376464844, "margin_dpo/margin_std": 118.20479583740234, "step": 160 }, { "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.005916006397455931, "fcm_dpo/delta": -0.016961436718702316, "fcm_dpo/margin": 70.23262023925781, "fcm_dpo/q_t": 0.4076644480228424, "grad_norm": 21.921010971069336, "learning_rate": 4.708572792802069e-07, "logits/chosen": -0.38327422738075256, "logits/rejected": -0.369274765253067, "logps/chosen": -138.60523986816406, "logps/ref_chosen": -57.40401077270508, "logps/ref_rejected": -80.31498718261719, "logps/rejected": -231.7488555908203, "loss": 1.1054, "margin_dpo/margin_mean": 70.23262023925781, "margin_dpo/margin_std": 105.99822998046875, "step": 165 }, { "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.005466562695801258, "fcm_dpo/delta": -0.10773168504238129, "fcm_dpo/margin": 87.78046417236328, "fcm_dpo/q_t": 0.3945610821247101, "grad_norm": 17.49802589416504, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -0.3227120041847229, "logits/rejected": -0.3182465136051178, "logps/chosen": -131.82830810546875, "logps/ref_chosen": -52.029144287109375, "logps/ref_rejected": -85.73944091796875, "logps/rejected": -253.31906127929688, "loss": 1.0613, "margin_dpo/margin_mean": 87.78046417236328, "margin_dpo/margin_std": 120.57357025146484, "step": 170 }, { "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.0053649842739105225, "fcm_dpo/delta": -0.02835622988641262, "fcm_dpo/margin": 79.64360809326172, "fcm_dpo/q_t": 0.40541133284568787, "grad_norm": 29.300512313842773, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.3354695737361908, "logits/rejected": -0.3250008523464203, "logps/chosen": -156.86700439453125, "logps/ref_chosen": -62.996971130371094, "logps/ref_rejected": -92.98394012451172, "logps/rejected": -266.4975891113281, "loss": 1.0959, "margin_dpo/margin_mean": 79.64360046386719, "margin_dpo/margin_std": 116.92179107666016, "step": 175 }, { "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.004915366414934397, "fcm_dpo/delta": -0.08844007551670074, "fcm_dpo/margin": 98.45795440673828, "fcm_dpo/q_t": 0.39298662543296814, "grad_norm": 18.73993492126465, "learning_rate": 4.611919330113591e-07, "logits/chosen": -0.2953912019729614, "logits/rejected": -0.29680582880973816, "logps/chosen": -155.4091796875, "logps/ref_chosen": -57.0670280456543, "logps/ref_rejected": -97.1115493774414, "logps/rejected": -293.9116516113281, "loss": 1.0541, "margin_dpo/margin_mean": 98.45795440673828, "margin_dpo/margin_std": 131.67587280273438, "step": 180 }, { "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.00469121104106307, "fcm_dpo/delta": -0.007265460677444935, "fcm_dpo/margin": 72.3204574584961, "fcm_dpo/q_t": 0.423015296459198, "grad_norm": 19.846059799194336, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -0.3211399018764496, "logits/rejected": -0.31003421545028687, "logps/chosen": -150.46461486816406, "logps/ref_chosen": -54.840736389160156, "logps/ref_rejected": -75.51002502441406, "logps/rejected": -243.45437622070312, "loss": 1.1531, "margin_dpo/margin_mean": 72.3204574584961, "margin_dpo/margin_std": 118.7593994140625, "step": 185 }, { "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.004630076698958874, "fcm_dpo/delta": -0.02935381792485714, "fcm_dpo/margin": 92.37378692626953, "fcm_dpo/q_t": 0.4038674235343933, "grad_norm": 18.63943862915039, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -0.3224960267543793, "logits/rejected": -0.31152957677841187, "logps/chosen": -155.59190368652344, "logps/ref_chosen": -57.72148895263672, "logps/ref_rejected": -86.85997009277344, "logps/rejected": -277.10418701171875, "loss": 1.0829, "margin_dpo/margin_mean": 92.37378692626953, "margin_dpo/margin_std": 125.8380126953125, "step": 190 }, { "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.004372724797576666, "fcm_dpo/delta": -0.041835661977529526, "fcm_dpo/margin": 94.63658142089844, "fcm_dpo/q_t": 0.4060741066932678, "grad_norm": 19.838130950927734, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -0.2853700518608093, "logits/rejected": -0.2791460156440735, "logps/chosen": -171.98052978515625, "logps/ref_chosen": -58.26164627075195, "logps/ref_rejected": -89.46485900878906, "logps/rejected": -297.82037353515625, "loss": 1.0971, "margin_dpo/margin_mean": 94.63658142089844, "margin_dpo/margin_std": 133.46847534179688, "step": 195 }, { "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.0042917924001812935, "fcm_dpo/delta": -0.03731077164411545, "fcm_dpo/margin": 101.45069122314453, "fcm_dpo/q_t": 0.40244197845458984, "grad_norm": 21.282487869262695, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -0.29983407258987427, "logits/rejected": -0.2885088324546814, "logps/chosen": -161.27999877929688, "logps/ref_chosen": -55.71953201293945, "logps/ref_rejected": -83.15235137939453, "logps/rejected": -290.16351318359375, "loss": 1.0801, "margin_dpo/margin_mean": 101.45067596435547, "margin_dpo/margin_std": 138.32688903808594, "step": 200 }, { "epoch": 0.2936857562408223, "eval_fcm_dpo/beta": 0.004899847786873579, "eval_fcm_dpo/delta": 0.03208634629845619, "eval_fcm_dpo/margin": 58.008628845214844, "eval_fcm_dpo/q_t": 0.4392388164997101, "eval_logits/chosen": -0.3472006916999817, "eval_logits/rejected": -0.3329962491989136, "eval_logps/chosen": -225.8389892578125, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -291.5945129394531, "eval_loss": 0.6239348649978638, "eval_margin_dpo/margin_mean": 58.008628845214844, "eval_margin_dpo/margin_std": 147.6545867919922, "eval_runtime": 39.6889, "eval_samples_per_second": 58.933, "eval_steps_per_second": 1.864, "step": 200 }, { "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.005112385377287865, "fcm_dpo/delta": -0.07373825460672379, "fcm_dpo/margin": 91.88821411132812, "fcm_dpo/q_t": 0.39707884192466736, "grad_norm": 24.034622192382812, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -0.3471163213253021, "logits/rejected": -0.3326929211616516, "logps/chosen": -181.0006103515625, "logps/ref_chosen": -65.13258361816406, "logps/ref_rejected": -92.10203552246094, "logps/rejected": -299.8582763671875, "loss": 1.0848, "margin_dpo/margin_mean": 91.88821411132812, "margin_dpo/margin_std": 137.34268188476562, "step": 205 }, { "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.00456308713182807, "fcm_dpo/delta": -0.12305764853954315, "fcm_dpo/margin": 112.89747619628906, "fcm_dpo/q_t": 0.3846554160118103, "grad_norm": 25.71703338623047, "learning_rate": 4.381713366536311e-07, "logits/chosen": -0.30507659912109375, "logits/rejected": -0.2925286889076233, "logps/chosen": -175.38168334960938, "logps/ref_chosen": -54.52837371826172, "logps/ref_rejected": -87.06227111816406, "logps/rejected": -320.81304931640625, "loss": 1.0332, "margin_dpo/margin_mean": 112.89747619628906, "margin_dpo/margin_std": 143.22720336914062, "step": 210 }, { "epoch": 0.315712187958884, "fcm_dpo/beta": 0.004155799280852079, "fcm_dpo/delta": -0.09338673204183578, "fcm_dpo/margin": 117.60514068603516, "fcm_dpo/q_t": 0.39112934470176697, "grad_norm": 27.700122833251953, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -0.2812441289424896, "logits/rejected": -0.265227735042572, "logps/chosen": -197.89137268066406, "logps/ref_chosen": -59.905250549316406, "logps/ref_rejected": -90.25511932373047, "logps/rejected": -345.84637451171875, "loss": 1.0502, "margin_dpo/margin_mean": 117.60514068603516, "margin_dpo/margin_std": 151.13796997070312, "step": 215 }, { "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.0038633563090115786, "fcm_dpo/delta": -0.0188637413084507, "fcm_dpo/margin": 107.99397277832031, "fcm_dpo/q_t": 0.406075656414032, "grad_norm": 27.14980697631836, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -0.2910076379776001, "logits/rejected": -0.27931129932403564, "logps/chosen": -209.11544799804688, "logps/ref_chosen": -57.68498611450195, "logps/ref_rejected": -87.72792053222656, "logps/rejected": -347.15234375, "loss": 1.0982, "margin_dpo/margin_mean": 107.99397277832031, "margin_dpo/margin_std": 155.56427001953125, "step": 220 }, { "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.003635880770161748, "fcm_dpo/delta": -0.11107522249221802, "fcm_dpo/margin": 138.54534912109375, "fcm_dpo/q_t": 0.3866313397884369, "grad_norm": 21.774181365966797, "learning_rate": 4.249525076191759e-07, "logits/chosen": -0.26980072259902954, "logits/rejected": -0.2639961540699005, "logps/chosen": -216.45687866210938, "logps/ref_chosen": -54.47245407104492, "logps/ref_rejected": -93.26266479492188, "logps/rejected": -393.79248046875, "loss": 1.0374, "margin_dpo/margin_mean": 138.5453643798828, "margin_dpo/margin_std": 173.9648895263672, "step": 225 }, { "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.003416923340409994, "fcm_dpo/delta": -0.05279744789004326, "fcm_dpo/margin": 131.7572021484375, "fcm_dpo/q_t": 0.3975502550601959, "grad_norm": 29.499502182006836, "learning_rate": 4.203117865141635e-07, "logits/chosen": -0.3394869565963745, "logits/rejected": -0.31970351934432983, "logps/chosen": -219.03005981445312, "logps/ref_chosen": -58.7701301574707, "logps/ref_rejected": -87.85963439941406, "logps/rejected": -379.87677001953125, "loss": 1.0656, "margin_dpo/margin_mean": 131.7572021484375, "margin_dpo/margin_std": 169.88063049316406, "step": 230 }, { "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.003319287672638893, "fcm_dpo/delta": -0.0038651395589113235, "fcm_dpo/margin": 121.47023010253906, "fcm_dpo/q_t": 0.40905800461769104, "grad_norm": 27.062564849853516, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -0.3577764332294464, "logits/rejected": -0.3489416241645813, "logps/chosen": -223.79031372070312, "logps/ref_chosen": -59.0481071472168, "logps/ref_rejected": -91.19654846191406, "logps/rejected": -377.40899658203125, "loss": 1.0973, "margin_dpo/margin_mean": 121.47023010253906, "margin_dpo/margin_std": 166.64047241210938, "step": 235 }, { "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.0032308667432516813, "fcm_dpo/delta": -0.02104497328400612, "fcm_dpo/margin": 129.8544921875, "fcm_dpo/q_t": 0.4038158357143402, "grad_norm": 31.66817855834961, "learning_rate": 4.106969024216348e-07, "logits/chosen": -0.3540635108947754, "logits/rejected": -0.35047775506973267, "logps/chosen": -231.42440795898438, "logps/ref_chosen": -55.238983154296875, "logps/ref_rejected": -91.08428955078125, "logps/rejected": -397.12420654296875, "loss": 1.0799, "margin_dpo/margin_mean": 129.85450744628906, "margin_dpo/margin_std": 166.92564392089844, "step": 240 }, { "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.003068047808483243, "fcm_dpo/delta": -0.07209217548370361, "fcm_dpo/margin": 152.28811645507812, "fcm_dpo/q_t": 0.3937745690345764, "grad_norm": 28.177356719970703, "learning_rate": 4.057290731287531e-07, "logits/chosen": -0.3967796266078949, "logits/rejected": -0.37566548585891724, "logps/chosen": -252.69760131835938, "logps/ref_chosen": -65.08844757080078, "logps/ref_rejected": -86.05777740478516, "logps/rejected": -425.95501708984375, "loss": 1.0455, "margin_dpo/margin_mean": 152.28811645507812, "margin_dpo/margin_std": 179.84130859375, "step": 245 }, { "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.0029506587889045477, "fcm_dpo/delta": -0.029526233673095703, "fcm_dpo/margin": 144.90142822265625, "fcm_dpo/q_t": 0.4018786549568176, "grad_norm": 30.520957946777344, "learning_rate": 4.006586590948141e-07, "logits/chosen": -0.3914863169193268, "logits/rejected": -0.3723214566707611, "logps/chosen": -273.7362365722656, "logps/ref_chosen": -59.08491897583008, "logps/ref_rejected": -87.36727142333984, "logps/rejected": -446.9200134277344, "loss": 1.0748, "margin_dpo/margin_mean": 144.9014434814453, "margin_dpo/margin_std": 182.38096618652344, "step": 250 }, { "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.0029482415411621332, "fcm_dpo/delta": 0.010621277615427971, "fcm_dpo/margin": 132.09812927246094, "fcm_dpo/q_t": 0.4101809859275818, "grad_norm": 21.41288185119629, "learning_rate": 3.954890003969163e-07, "logits/chosen": -0.38516297936439514, "logits/rejected": -0.36370107531547546, "logps/chosen": -264.5665588378906, "logps/ref_chosen": -61.85979461669922, "logps/ref_rejected": -88.32804107666016, "logps/rejected": -423.1329650878906, "loss": 1.1051, "margin_dpo/margin_mean": 132.09812927246094, "margin_dpo/margin_std": 182.64788818359375, "step": 255 }, { "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.0029091238975524902, "fcm_dpo/delta": -0.0311284102499485, "fcm_dpo/margin": 147.6251678466797, "fcm_dpo/q_t": 0.4018276631832123, "grad_norm": 31.76065444946289, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -0.3353942334651947, "logits/rejected": -0.33657923340797424, "logps/chosen": -256.07330322265625, "logps/ref_chosen": -52.843467712402344, "logps/ref_rejected": -90.4744873046875, "logps/rejected": -441.3294982910156, "loss": 1.0767, "margin_dpo/margin_mean": 147.6251678466797, "margin_dpo/margin_std": 192.7305908203125, "step": 260 }, { "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.002804348012432456, "fcm_dpo/delta": -0.036855168640613556, "fcm_dpo/margin": 155.0548095703125, "fcm_dpo/q_t": 0.4009890556335449, "grad_norm": 26.897104263305664, "learning_rate": 3.848656339557562e-07, "logits/chosen": -0.3249055743217468, "logits/rejected": -0.3102591931819916, "logps/chosen": -295.2470703125, "logps/ref_chosen": -59.35320281982422, "logps/ref_rejected": -90.73350524902344, "logps/rejected": -481.68218994140625, "loss": 1.0752, "margin_dpo/margin_mean": 155.0548095703125, "margin_dpo/margin_std": 203.52064514160156, "step": 265 }, { "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.002779749920591712, "fcm_dpo/delta": 0.0037582286167889833, "fcm_dpo/margin": 142.50425720214844, "fcm_dpo/q_t": 0.40962162613868713, "grad_norm": 36.368804931640625, "learning_rate": 3.794189242333106e-07, "logits/chosen": -0.3710918426513672, "logits/rejected": -0.3549976944923401, "logps/chosen": -305.8779602050781, "logps/ref_chosen": -66.30875396728516, "logps/ref_rejected": -95.4130630493164, "logps/rejected": -477.486572265625, "loss": 1.1067, "margin_dpo/margin_mean": 142.50425720214844, "margin_dpo/margin_std": 207.22396850585938, "step": 270 }, { "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.002664657775312662, "fcm_dpo/delta": -0.05992863327264786, "fcm_dpo/margin": 171.36749267578125, "fcm_dpo/q_t": 0.3942517042160034, "grad_norm": 29.650026321411133, "learning_rate": 3.738869612786737e-07, "logits/chosen": -0.3030190169811249, "logits/rejected": -0.2977046072483063, "logps/chosen": -267.1299743652344, "logps/ref_chosen": -54.69990921020508, "logps/ref_rejected": -92.23838806152344, "logps/rejected": -476.03594970703125, "loss": 1.0428, "margin_dpo/margin_mean": 171.36749267578125, "margin_dpo/margin_std": 191.60519409179688, "step": 275 }, { "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.002544180955737829, "fcm_dpo/delta": -0.059092938899993896, "fcm_dpo/margin": 179.50125122070312, "fcm_dpo/q_t": 0.395501971244812, "grad_norm": 27.062313079833984, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -0.2739403247833252, "logits/rejected": -0.26840710639953613, "logps/chosen": -304.5663146972656, "logps/ref_chosen": -54.64586639404297, "logps/ref_rejected": -88.19416809082031, "logps/rejected": -517.6158447265625, "loss": 1.053, "margin_dpo/margin_mean": 179.5012664794922, "margin_dpo/margin_std": 217.075439453125, "step": 280 }, { "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.002449630293995142, "fcm_dpo/delta": -0.01916874758899212, "fcm_dpo/margin": 170.6898956298828, "fcm_dpo/q_t": 0.40265077352523804, "grad_norm": 27.657169342041016, "learning_rate": 3.625819059005228e-07, "logits/chosen": -0.3483652174472809, "logits/rejected": -0.33659619092941284, "logps/chosen": -324.3221130371094, "logps/ref_chosen": -63.02496337890625, "logps/ref_rejected": -93.16323852539062, "logps/rejected": -525.1502685546875, "loss": 1.0686, "margin_dpo/margin_mean": 170.68988037109375, "margin_dpo/margin_std": 199.1321563720703, "step": 285 }, { "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.0023679626174271107, "fcm_dpo/delta": -0.05114912986755371, "fcm_dpo/margin": 189.60928344726562, "fcm_dpo/q_t": 0.39646488428115845, "grad_norm": 24.63570785522461, "learning_rate": 3.568162605525952e-07, "logits/chosen": -0.30854520201683044, "logits/rejected": -0.30314746499061584, "logps/chosen": -322.7059631347656, "logps/ref_chosen": -58.37105178833008, "logps/ref_rejected": -91.59428405761719, "logps/rejected": -545.5384521484375, "loss": 1.0504, "margin_dpo/margin_mean": 189.60928344726562, "margin_dpo/margin_std": 216.8871307373047, "step": 290 }, { "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.00226641446352005, "fcm_dpo/delta": -0.026808306574821472, "fcm_dpo/margin": 187.75306701660156, "fcm_dpo/q_t": 0.4028984606266022, "grad_norm": 37.57793045043945, "learning_rate": 3.509802512179737e-07, "logits/chosen": -0.3241910934448242, "logits/rejected": -0.31886667013168335, "logps/chosen": -314.74871826171875, "logps/ref_chosen": -55.113426208496094, "logps/ref_rejected": -85.29283905029297, "logps/rejected": -532.68115234375, "loss": 1.0814, "margin_dpo/margin_mean": 187.75306701660156, "margin_dpo/margin_std": 250.9693603515625, "step": 295 }, { "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.0022166285198181868, "fcm_dpo/delta": -0.03155837580561638, "fcm_dpo/margin": 183.79110717773438, "fcm_dpo/q_t": 0.40807557106018066, "grad_norm": 29.412609100341797, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -0.37595048546791077, "logits/rejected": -0.3784215748310089, "logps/chosen": -391.8227844238281, "logps/ref_chosen": -59.46582794189453, "logps/ref_rejected": -97.03690338134766, "logps/rejected": -613.1849975585938, "loss": 1.107, "margin_dpo/margin_mean": 183.79110717773438, "margin_dpo/margin_std": 271.849853515625, "step": 300 }, { "epoch": 0.447870778267254, "fcm_dpo/beta": 0.0021604837384074926, "fcm_dpo/delta": 0.016617389395833015, "fcm_dpo/margin": 177.69085693359375, "fcm_dpo/q_t": 0.41120848059654236, "grad_norm": 26.565410614013672, "learning_rate": 3.391125620245535e-07, "logits/chosen": -0.4044717252254486, "logits/rejected": -0.3968522846698761, "logps/chosen": -303.1658935546875, "logps/ref_chosen": -62.78144454956055, "logps/ref_rejected": -91.95039367675781, "logps/rejected": -510.02569580078125, "loss": 1.0938, "margin_dpo/margin_mean": 177.69085693359375, "margin_dpo/margin_std": 218.7647247314453, "step": 305 }, { "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.0022380652371793985, "fcm_dpo/delta": 0.01699766516685486, "fcm_dpo/margin": 171.28904724121094, "fcm_dpo/q_t": 0.4098632335662842, "grad_norm": 25.57453155517578, "learning_rate": 3.3308869986991487e-07, "logits/chosen": -0.3871304988861084, "logits/rejected": -0.3712696135044098, "logps/chosen": -297.95452880859375, "logps/ref_chosen": -61.359039306640625, "logps/ref_rejected": -82.75496673583984, "logps/rejected": -490.6394958496094, "loss": 1.0861, "margin_dpo/margin_mean": 171.2890625, "margin_dpo/margin_std": 196.2567138671875, "step": 310 }, { "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.0022226418368518353, "fcm_dpo/delta": -0.002875490579754114, "fcm_dpo/margin": 181.02659606933594, "fcm_dpo/q_t": 0.4087609350681305, "grad_norm": 23.077590942382812, "learning_rate": 3.270101039870797e-07, "logits/chosen": -0.3677603602409363, "logits/rejected": -0.36049968004226685, "logps/chosen": -327.0533447265625, "logps/ref_chosen": -51.77602005004883, "logps/ref_rejected": -84.58292388916016, "logps/rejected": -540.8868408203125, "loss": 1.0964, "margin_dpo/margin_mean": 181.02658081054688, "margin_dpo/margin_std": 249.53970336914062, "step": 315 }, { "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.002118379110470414, "fcm_dpo/delta": -0.09106224775314331, "fcm_dpo/margin": 229.5352325439453, "fcm_dpo/q_t": 0.38847747445106506, "grad_norm": 26.46944236755371, "learning_rate": 3.208807785813777e-07, "logits/chosen": -0.40023693442344666, "logits/rejected": -0.3959748148918152, "logps/chosen": -333.487548828125, "logps/ref_chosen": -56.777862548828125, "logps/ref_rejected": -99.26368713378906, "logps/rejected": -605.5086059570312, "loss": 1.0268, "margin_dpo/margin_mean": 229.5352325439453, "margin_dpo/margin_std": 254.9644317626953, "step": 320 }, { "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.002015223726630211, "fcm_dpo/delta": 0.0017573073273524642, "fcm_dpo/margin": 197.41522216796875, "fcm_dpo/q_t": 0.4082733690738678, "grad_norm": 27.207300186157227, "learning_rate": 3.147047612756302e-07, "logits/chosen": -0.4213026165962219, "logits/rejected": -0.4106159210205078, "logps/chosen": -381.36676025390625, "logps/ref_chosen": -58.28468704223633, "logps/ref_rejected": -83.80326843261719, "logps/rejected": -604.3004760742188, "loss": 1.0923, "margin_dpo/margin_mean": 197.41522216796875, "margin_dpo/margin_std": 251.11337280273438, "step": 325 }, { "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.002033454831689596, "fcm_dpo/delta": -0.02132582850754261, "fcm_dpo/margin": 206.72314453125, "fcm_dpo/q_t": 0.40213847160339355, "grad_norm": 28.69426727294922, "learning_rate": 3.084861204504122e-07, "logits/chosen": -0.42850977182388306, "logits/rejected": -0.42162972688674927, "logps/chosen": -349.00677490234375, "logps/ref_chosen": -62.75822067260742, "logps/ref_rejected": -94.04203033447266, "logps/rejected": -587.0137939453125, "loss": 1.0667, "margin_dpo/margin_mean": 206.72314453125, "margin_dpo/margin_std": 237.6807403564453, "step": 330 }, { "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.0019993516616523266, "fcm_dpo/delta": 0.017332084476947784, "fcm_dpo/margin": 191.62435913085938, "fcm_dpo/q_t": 0.41195163130760193, "grad_norm": 32.573978424072266, "learning_rate": 3.022289525640531e-07, "logits/chosen": -0.46553975343704224, "logits/rejected": -0.454815149307251, "logps/chosen": -367.4043273925781, "logps/ref_chosen": -58.59650421142578, "logps/ref_rejected": -88.69586944580078, "logps/rejected": -589.1280517578125, "loss": 1.1035, "margin_dpo/margin_mean": 191.62435913085938, "margin_dpo/margin_std": 254.1148223876953, "step": 335 }, { "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.001957540400326252, "fcm_dpo/delta": -0.052527785301208496, "fcm_dpo/margin": 229.53335571289062, "fcm_dpo/q_t": 0.3968558609485626, "grad_norm": 28.016321182250977, "learning_rate": 2.959373794541426e-07, "logits/chosen": -0.45491886138916016, "logits/rejected": -0.44332224130630493, "logps/chosen": -372.17889404296875, "logps/ref_chosen": -58.18162155151367, "logps/ref_rejected": -94.44358825683594, "logps/rejected": -637.9742431640625, "loss": 1.0561, "margin_dpo/margin_mean": 229.53335571289062, "margin_dpo/margin_std": 272.4295959472656, "step": 340 }, { "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.0019028617534786463, "fcm_dpo/delta": -0.032430313527584076, "fcm_dpo/margin": 226.35147094726562, "fcm_dpo/q_t": 0.4004201292991638, "grad_norm": 25.25508689880371, "learning_rate": 2.896155456223163e-07, "logits/chosen": -0.44945335388183594, "logits/rejected": -0.4429987072944641, "logps/chosen": -372.24151611328125, "logps/ref_chosen": -57.9904899597168, "logps/ref_rejected": -99.11092376708984, "logps/rejected": -639.7134399414062, "loss": 1.0594, "margin_dpo/margin_mean": 226.3514862060547, "margin_dpo/margin_std": 255.4004364013672, "step": 345 }, { "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.0018661676440387964, "fcm_dpo/delta": 0.005433606915175915, "fcm_dpo/margin": 211.39328002929688, "fcm_dpo/q_t": 0.40922412276268005, "grad_norm": 30.192829132080078, "learning_rate": 2.8326761550411346e-07, "logits/chosen": -0.442319393157959, "logits/rejected": -0.4401112198829651, "logps/chosen": -384.77960205078125, "logps/ref_chosen": -58.29923629760742, "logps/ref_rejected": -89.25711822509766, "logps/rejected": -627.1307373046875, "loss": 1.0959, "margin_dpo/margin_mean": 211.3932647705078, "margin_dpo/margin_std": 280.6878967285156, "step": 350 }, { "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.001845445716753602, "fcm_dpo/delta": -0.019822590053081512, "fcm_dpo/margin": 216.8295135498047, "fcm_dpo/q_t": 0.4076092839241028, "grad_norm": 37.005271911621094, "learning_rate": 2.7689777072570284e-07, "logits/chosen": -0.4728454649448395, "logits/rejected": -0.46268409490585327, "logps/chosen": -382.65460205078125, "logps/ref_chosen": -60.788482666015625, "logps/ref_rejected": -85.94129943847656, "logps/rejected": -624.6368408203125, "loss": 1.0868, "margin_dpo/margin_mean": 216.8295135498047, "margin_dpo/margin_std": 270.96405029296875, "step": 355 }, { "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.0018649225821718574, "fcm_dpo/delta": 0.030804917216300964, "fcm_dpo/margin": 183.98890686035156, "fcm_dpo/q_t": 0.4203580319881439, "grad_norm": 36.66780471801758, "learning_rate": 2.7051020734928443e-07, "logits/chosen": -0.4746428430080414, "logits/rejected": -0.4617390036582947, "logps/chosen": -384.0989074707031, "logps/ref_chosen": -57.6871337890625, "logps/ref_rejected": -80.62527465820312, "logps/rejected": -591.0260009765625, "loss": 1.1282, "margin_dpo/margin_mean": 183.98892211914062, "margin_dpo/margin_std": 254.2589569091797, "step": 360 }, { "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.0018712872406467795, "fcm_dpo/delta": -0.026755044236779213, "fcm_dpo/margin": 227.2862091064453, "fcm_dpo/q_t": 0.40148013830184937, "grad_norm": 26.54566764831543, "learning_rate": 2.641091331089811e-07, "logits/chosen": -0.473797470331192, "logits/rejected": -0.47837966680526733, "logps/chosen": -337.5777893066406, "logps/ref_chosen": -51.490867614746094, "logps/ref_rejected": -91.02871704101562, "logps/rejected": -604.4019165039062, "loss": 1.0628, "margin_dpo/margin_mean": 227.28622436523438, "margin_dpo/margin_std": 260.75958251953125, "step": 365 }, { "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.0018645223462954164, "fcm_dpo/delta": 0.038537509739398956, "fcm_dpo/margin": 194.5872039794922, "fcm_dpo/q_t": 0.41541823744773865, "grad_norm": 38.50527572631836, "learning_rate": 2.5769876463904263e-07, "logits/chosen": -0.47979289293289185, "logits/rejected": -0.4759935438632965, "logps/chosen": -354.23822021484375, "logps/ref_chosen": -58.113502502441406, "logps/ref_rejected": -89.43451690673828, "logps/rejected": -580.1464233398438, "loss": 1.1143, "margin_dpo/margin_mean": 194.5872039794922, "margin_dpo/margin_std": 256.3455505371094, "step": 370 }, { "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.0018788978923112154, "fcm_dpo/delta": -0.028660928830504417, "fcm_dpo/margin": 227.169677734375, "fcm_dpo/q_t": 0.403217077255249, "grad_norm": 42.09442901611328, "learning_rate": 2.512833246961859e-07, "logits/chosen": -0.5087894201278687, "logits/rejected": -0.494729608297348, "logps/chosen": -409.12677001953125, "logps/ref_chosen": -65.23600769042969, "logps/ref_rejected": -89.24995422363281, "logps/rejected": -660.3104248046875, "loss": 1.0839, "margin_dpo/margin_mean": 227.169677734375, "margin_dpo/margin_std": 307.71136474609375, "step": 375 }, { "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.001791507238522172, "fcm_dpo/delta": -0.07822184264659882, "fcm_dpo/margin": 264.9973449707031, "fcm_dpo/q_t": 0.39022326469421387, "grad_norm": 27.727758407592773, "learning_rate": 2.4486703937790243e-07, "logits/chosen": -0.4783501625061035, "logits/rejected": -0.49099501967430115, "logps/chosen": -376.74932861328125, "logps/ref_chosen": -53.33893966674805, "logps/ref_rejected": -102.15375518798828, "logps/rejected": -690.5615234375, "loss": 1.034, "margin_dpo/margin_mean": 264.9973449707031, "margin_dpo/margin_std": 298.83721923828125, "step": 380 }, { "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.0017382428050041199, "fcm_dpo/delta": 0.007922522723674774, "fcm_dpo/margin": 225.6304931640625, "fcm_dpo/q_t": 0.40880244970321655, "grad_norm": 39.414588928222656, "learning_rate": 2.3845413533856514e-07, "logits/chosen": -0.4913402497768402, "logits/rejected": -0.4843069911003113, "logps/chosen": -414.2498474121094, "logps/ref_chosen": -58.36262130737305, "logps/ref_rejected": -89.44685363769531, "logps/rejected": -670.964599609375, "loss": 1.0946, "margin_dpo/margin_mean": 225.6304931640625, "margin_dpo/margin_std": 290.14666748046875, "step": 385 }, { "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.0017035829368978739, "fcm_dpo/delta": -0.029813114553689957, "fcm_dpo/margin": 251.324462890625, "fcm_dpo/q_t": 0.4024105966091156, "grad_norm": 26.369205474853516, "learning_rate": 2.320488370051681e-07, "logits/chosen": -0.486684650182724, "logits/rejected": -0.4810120165348053, "logps/chosen": -435.18115234375, "logps/ref_chosen": -56.380653381347656, "logps/ref_rejected": -90.47447204589844, "logps/rejected": -720.5994873046875, "loss": 1.0741, "margin_dpo/margin_mean": 251.324462890625, "margin_dpo/margin_std": 322.6702575683594, "step": 390 }, { "epoch": 0.580029368575624, "fcm_dpo/beta": 0.0017207658383995295, "fcm_dpo/delta": 0.012584102340042591, "fcm_dpo/margin": 225.0916748046875, "fcm_dpo/q_t": 0.41174229979515076, "grad_norm": 31.417821884155273, "learning_rate": 2.2565536379453404e-07, "logits/chosen": -0.47851499915122986, "logits/rejected": -0.4759146273136139, "logps/chosen": -402.8694763183594, "logps/ref_chosen": -55.95304489135742, "logps/ref_rejected": -87.13162994384766, "logps/rejected": -659.1397705078125, "loss": 1.1036, "margin_dpo/margin_mean": 225.0916748046875, "margin_dpo/margin_std": 308.83575439453125, "step": 395 }, { "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.0016688309842720628, "fcm_dpo/delta": -0.05934460088610649, "fcm_dpo/margin": 273.747314453125, "fcm_dpo/q_t": 0.39453184604644775, "grad_norm": 22.06127166748047, "learning_rate": 2.192779273338215e-07, "logits/chosen": -0.48663657903671265, "logits/rejected": -0.4787791669368744, "logps/chosen": -424.73455810546875, "logps/ref_chosen": -64.59160614013672, "logps/ref_rejected": -96.700927734375, "logps/rejected": -730.5912475585938, "loss": 1.0433, "margin_dpo/margin_mean": 273.747314453125, "margin_dpo/margin_std": 306.27801513671875, "step": 400 }, { "epoch": 0.5873715124816447, "eval_fcm_dpo/beta": 0.0018981158500537276, "eval_fcm_dpo/delta": 0.03762635216116905, "eval_fcm_dpo/margin": 156.90065002441406, "eval_fcm_dpo/q_t": 0.43355923891067505, "eval_logits/chosen": -0.5229986310005188, "eval_logits/rejected": -0.5119041800498962, "eval_logps/chosen": -532.2332763671875, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -696.880859375, "eval_loss": 0.6033138036727905, "eval_margin_dpo/margin_mean": 156.90061950683594, "eval_margin_dpo/margin_std": 328.45404052734375, "eval_runtime": 39.7397, "eval_samples_per_second": 58.858, "eval_steps_per_second": 1.862, "step": 400 }, { "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.001960620516911149, "fcm_dpo/delta": -0.1365078240633011, "fcm_dpo/margin": 260.3297119140625, "fcm_dpo/q_t": 0.3874695897102356, "grad_norm": 38.00107955932617, "learning_rate": 2.129207286861638e-07, "logits/chosen": -0.505179226398468, "logits/rejected": -0.5002223253250122, "logps/chosen": -406.5534973144531, "logps/ref_chosen": -53.61777877807617, "logps/ref_rejected": -81.28938293457031, "logps/rejected": -694.5548095703125, "loss": 1.0528, "margin_dpo/margin_mean": 260.3297119140625, "margin_dpo/margin_std": 351.9696044921875, "step": 405 }, { "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.0018232803558930755, "fcm_dpo/delta": -0.01396404393017292, "fcm_dpo/margin": 216.00399780273438, "fcm_dpo/q_t": 0.4089049696922302, "grad_norm": 22.69110679626465, "learning_rate": 2.065879555832674e-07, "logits/chosen": -0.5198922157287598, "logits/rejected": -0.508955717086792, "logps/chosen": -396.2334899902344, "logps/ref_chosen": -58.9287223815918, "logps/ref_rejected": -85.55818176269531, "logps/rejected": -638.8670043945312, "loss": 1.096, "margin_dpo/margin_mean": 216.00399780273438, "margin_dpo/margin_std": 284.52081298828125, "step": 410 }, { "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.0017951425397768617, "fcm_dpo/delta": -0.023637911304831505, "fcm_dpo/margin": 235.01473999023438, "fcm_dpo/q_t": 0.4024273753166199, "grad_norm": 26.751333236694336, "learning_rate": 2.002837796667909e-07, "logits/chosen": -0.4968322813510895, "logits/rejected": -0.5035668015480042, "logps/chosen": -417.4717712402344, "logps/ref_chosen": -58.45662307739258, "logps/ref_rejected": -93.67063903808594, "logps/rejected": -687.7005615234375, "loss": 1.0778, "margin_dpo/margin_mean": 235.01473999023438, "margin_dpo/margin_std": 297.6578063964844, "step": 415 }, { "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.001714545302093029, "fcm_dpo/delta": -0.04269330948591232, "fcm_dpo/margin": 256.39959716796875, "fcm_dpo/q_t": 0.39849138259887695, "grad_norm": 58.22108459472656, "learning_rate": 1.9401235374032425e-07, "logits/chosen": -0.47421860694885254, "logits/rejected": -0.46059077978134155, "logps/chosen": -428.48443603515625, "logps/ref_chosen": -64.2349853515625, "logps/ref_rejected": -84.86299133300781, "logps/rejected": -705.5120849609375, "loss": 1.062, "margin_dpo/margin_mean": 256.39959716796875, "margin_dpo/margin_std": 306.3155822753906, "step": 420 }, { "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.0017159763956442475, "fcm_dpo/delta": 0.010966436937451363, "fcm_dpo/margin": 226.81314086914062, "fcm_dpo/q_t": 0.4094250798225403, "grad_norm": 45.61664962768555, "learning_rate": 1.8777780903377732e-07, "logits/chosen": -0.5051223635673523, "logits/rejected": -0.4999346137046814, "logps/chosen": -442.1536560058594, "logps/ref_chosen": -56.054161071777344, "logps/ref_rejected": -85.17119598388672, "logps/rejected": -698.0838012695312, "loss": 1.1005, "margin_dpo/margin_mean": 226.81314086914062, "margin_dpo/margin_std": 304.298583984375, "step": 425 }, { "epoch": 0.631424375917768, "fcm_dpo/beta": 0.0017323114443570375, "fcm_dpo/delta": 0.009447330608963966, "fcm_dpo/margin": 213.61093139648438, "fcm_dpo/q_t": 0.414214551448822, "grad_norm": 40.23195266723633, "learning_rate": 1.8158425248197928e-07, "logits/chosen": -0.5632457733154297, "logits/rejected": -0.5552536249160767, "logps/chosen": -472.23919677734375, "logps/ref_chosen": -69.24568176269531, "logps/ref_rejected": -91.8664321899414, "logps/rejected": -708.4708862304688, "loss": 1.1182, "margin_dpo/margin_mean": 213.61093139648438, "margin_dpo/margin_std": 301.84185791015625, "step": 430 }, { "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.0017033193726092577, "fcm_dpo/delta": -0.027609745040535927, "fcm_dpo/margin": 249.95333862304688, "fcm_dpo/q_t": 0.4024946093559265, "grad_norm": 34.37434768676758, "learning_rate": 1.7543576401928218e-07, "logits/chosen": -0.618207573890686, "logits/rejected": -0.6184796690940857, "logps/chosen": -466.94012451171875, "logps/ref_chosen": -60.03449630737305, "logps/ref_rejected": -90.6872329711914, "logps/rejected": -747.5462036132812, "loss": 1.0796, "margin_dpo/margin_mean": 249.95333862304688, "margin_dpo/margin_std": 326.78277587890625, "step": 435 }, { "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.0016936672618612647, "fcm_dpo/delta": -0.009997454471886158, "fcm_dpo/margin": 227.880126953125, "fcm_dpo/q_t": 0.41129302978515625, "grad_norm": 52.6486930847168, "learning_rate": 1.6933639389195134e-07, "logits/chosen": -0.647782027721405, "logits/rejected": -0.6332479119300842, "logps/chosen": -471.6587829589844, "logps/ref_chosen": -65.50349426269531, "logps/ref_rejected": -85.66627502441406, "logps/rejected": -719.70166015625, "loss": 1.1022, "margin_dpo/margin_mean": 227.88015747070312, "margin_dpo/margin_std": 307.6328125, "step": 440 }, { "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.001662442460656166, "fcm_dpo/delta": -0.03005291521549225, "fcm_dpo/margin": 245.63253784179688, "fcm_dpo/q_t": 0.4081791043281555, "grad_norm": 32.25020980834961, "learning_rate": 1.6329015999011182e-07, "logits/chosen": -0.6261739730834961, "logits/rejected": -0.6163330674171448, "logps/chosen": -485.8421936035156, "logps/ref_chosen": -60.72443389892578, "logps/ref_rejected": -89.9255142211914, "logps/rejected": -760.6757202148438, "loss": 1.1036, "margin_dpo/margin_mean": 245.63253784179688, "margin_dpo/margin_std": 356.2879943847656, "step": 445 }, { "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.0015721891541033983, "fcm_dpo/delta": -0.07007580995559692, "fcm_dpo/margin": 297.0084533691406, "fcm_dpo/q_t": 0.3933987319469452, "grad_norm": 37.675498962402344, "learning_rate": 1.573010452010098e-07, "logits/chosen": -0.674242377281189, "logits/rejected": -0.6845596432685852, "logps/chosen": -466.7557067871094, "logps/ref_chosen": -59.96248245239258, "logps/ref_rejected": -98.71509552001953, "logps/rejected": -802.5167846679688, "loss": 1.0453, "margin_dpo/margin_mean": 297.0084533691406, "margin_dpo/margin_std": 355.6662292480469, "step": 450 }, { "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.0015043766470625997, "fcm_dpo/delta": -0.04557330161333084, "fcm_dpo/margin": 295.05352783203125, "fcm_dpo/q_t": 0.3984234929084778, "grad_norm": 46.07245635986328, "learning_rate": 1.5137299478533064e-07, "logits/chosen": -0.7228642702102661, "logits/rejected": -0.7254277467727661, "logps/chosen": -525.2667846679688, "logps/ref_chosen": -54.48131561279297, "logps/ref_rejected": -90.6321029663086, "logps/rejected": -856.4710693359375, "loss": 1.0728, "margin_dpo/margin_mean": 295.05352783203125, "margin_dpo/margin_std": 389.12469482421875, "step": 455 }, { "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.0013938735937699676, "fcm_dpo/delta": -0.0647268071770668, "fcm_dpo/margin": 331.0070495605469, "fcm_dpo/q_t": 0.3938430845737457, "grad_norm": 38.70931625366211, "learning_rate": 1.4550991377830423e-07, "logits/chosen": -0.6967940926551819, "logits/rejected": -0.7114230394363403, "logps/chosen": -526.4164428710938, "logps/ref_chosen": -52.97611618041992, "logps/ref_rejected": -95.65971374511719, "logps/rejected": -900.1070556640625, "loss": 1.0402, "margin_dpo/margin_mean": 331.0070495605469, "margin_dpo/margin_std": 373.8094177246094, "step": 460 }, { "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.0013582499232143164, "fcm_dpo/delta": -0.040646448731422424, "fcm_dpo/margin": 323.32489013671875, "fcm_dpo/q_t": 0.3995057940483093, "grad_norm": 29.232101440429688, "learning_rate": 1.3971566441730714e-07, "logits/chosen": -0.7399168014526367, "logits/rejected": -0.7570127248764038, "logps/chosen": -553.7732543945312, "logps/ref_chosen": -58.2827033996582, "logps/ref_rejected": -94.15567779541016, "logps/rejected": -912.9710693359375, "loss": 1.0657, "margin_dpo/margin_mean": 323.32489013671875, "margin_dpo/margin_std": 396.70892333984375, "step": 465 }, { "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.0012759026139974594, "fcm_dpo/delta": -0.050003357231616974, "fcm_dpo/margin": 350.7779235839844, "fcm_dpo/q_t": 0.3979727625846863, "grad_norm": 27.062070846557617, "learning_rate": 1.339940635976592e-07, "logits/chosen": -0.7901474833488464, "logits/rejected": -0.7927638292312622, "logps/chosen": -599.5657348632812, "logps/ref_chosen": -62.69774627685547, "logps/ref_rejected": -96.33873748779297, "logps/rejected": -983.9846801757812, "loss": 1.0702, "margin_dpo/margin_mean": 350.7779846191406, "margin_dpo/margin_std": 457.7704162597656, "step": 470 }, { "epoch": 0.697503671071953, "fcm_dpo/beta": 0.0012454693205654621, "fcm_dpo/delta": -0.008181336335837841, "fcm_dpo/margin": 326.77447509765625, "fcm_dpo/q_t": 0.41024675965309143, "grad_norm": 34.630069732666016, "learning_rate": 1.2834888035828596e-07, "logits/chosen": -0.9200963973999023, "logits/rejected": -0.9176486730575562, "logps/chosen": -705.766845703125, "logps/ref_chosen": -61.12194061279297, "logps/ref_rejected": -92.42192077636719, "logps/rejected": -1063.84130859375, "loss": 1.1387, "margin_dpo/margin_mean": 326.7744140625, "margin_dpo/margin_std": 567.3615112304688, "step": 475 }, { "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.0012077607680112123, "fcm_dpo/delta": -0.04966871812939644, "fcm_dpo/margin": 349.90106201171875, "fcm_dpo/q_t": 0.4062713086605072, "grad_norm": 62.070411682128906, "learning_rate": 1.227838333989088e-07, "logits/chosen": -0.9689761996269226, "logits/rejected": -0.9785006642341614, "logps/chosen": -730.4820556640625, "logps/ref_chosen": -53.550628662109375, "logps/ref_rejected": -82.65167999267578, "logps/rejected": -1109.484130859375, "loss": 1.111, "margin_dpo/margin_mean": 349.90106201171875, "margin_dpo/margin_std": 551.5509033203125, "step": 480 }, { "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.001141466898843646, "fcm_dpo/delta": -0.044795047491788864, "fcm_dpo/margin": 370.09796142578125, "fcm_dpo/q_t": 0.404446542263031, "grad_norm": 33.84178924560547, "learning_rate": 1.1730258863039347e-07, "logits/chosen": -0.9285829663276672, "logits/rejected": -0.9435798525810242, "logps/chosen": -689.6338500976562, "logps/ref_chosen": -60.76704788208008, "logps/ref_rejected": -92.70649719238281, "logps/rejected": -1091.671142578125, "loss": 1.0989, "margin_dpo/margin_mean": 370.09796142578125, "margin_dpo/margin_std": 542.5574951171875, "step": 485 }, { "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.0011192201636731625, "fcm_dpo/delta": 0.0058375876396894455, "fcm_dpo/margin": 352.23004150390625, "fcm_dpo/q_t": 0.4108501374721527, "grad_norm": 50.5614128112793, "learning_rate": 1.1190875675987355e-07, "logits/chosen": -0.9131494760513306, "logits/rejected": -0.9286991953849792, "logps/chosen": -628.578369140625, "logps/ref_chosen": -53.78142547607422, "logps/ref_rejected": -91.67438507080078, "logps/rejected": -1018.7013549804688, "loss": 1.1102, "margin_dpo/margin_mean": 352.2300109863281, "margin_dpo/margin_std": 520.4669189453125, "step": 490 }, { "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.0011620967416092753, "fcm_dpo/delta": 0.019140174612402916, "fcm_dpo/margin": 309.65850830078125, "fcm_dpo/q_t": 0.418142706155777, "grad_norm": 35.54931640625, "learning_rate": 1.0660589091223854e-07, "logits/chosen": -0.9482976794242859, "logits/rejected": -0.9485372304916382, "logps/chosen": -638.1903076171875, "logps/ref_chosen": -58.9004020690918, "logps/ref_rejected": -82.15424346923828, "logps/rejected": -971.1026611328125, "loss": 1.1356, "margin_dpo/margin_mean": 309.6585388183594, "margin_dpo/margin_std": 488.75909423828125, "step": 495 }, { "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.0011855836492031813, "fcm_dpo/delta": 0.024493711069226265, "fcm_dpo/margin": 291.8643798828125, "fcm_dpo/q_t": 0.42091307044029236, "grad_norm": 30.435544967651367, "learning_rate": 1.0139748428955333e-07, "logits/chosen": -0.9160359501838684, "logits/rejected": -0.91932213306427, "logps/chosen": -699.1871948242188, "logps/ref_chosen": -62.13483810424805, "logps/ref_rejected": -87.91773223876953, "logps/rejected": -1016.83447265625, "loss": 1.1519, "margin_dpo/margin_mean": 291.8643798828125, "margin_dpo/margin_std": 487.0228576660156, "step": 500 }, { "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.0011691536055877805, "fcm_dpo/delta": 0.006966643966734409, "fcm_dpo/margin": 336.22186279296875, "fcm_dpo/q_t": 0.41082048416137695, "grad_norm": 31.930696487426758, "learning_rate": 9.628696786995188e-08, "logits/chosen": -0.8171493411064148, "logits/rejected": -0.8128056526184082, "logps/chosen": -598.7544555664062, "logps/ref_chosen": -62.631813049316406, "logps/ref_rejected": -87.59168243408203, "logps/rejected": -959.9362182617188, "loss": 1.1017, "margin_dpo/margin_mean": 336.2218322753906, "margin_dpo/margin_std": 463.255615234375, "step": 505 }, { "epoch": 0.748898678414097, "fcm_dpo/beta": 0.0012091896496713161, "fcm_dpo/delta": 0.028515011072158813, "fcm_dpo/margin": 307.8777770996094, "fcm_dpo/q_t": 0.4152832627296448, "grad_norm": 37.425514221191406, "learning_rate": 9.127770814751932e-08, "logits/chosen": -0.7757030129432678, "logits/rejected": -0.7864035367965698, "logps/chosen": -574.5384521484375, "logps/ref_chosen": -60.552574157714844, "logps/ref_rejected": -91.0874252319336, "logps/rejected": -912.951171875, "loss": 1.1211, "margin_dpo/margin_mean": 307.8777770996094, "margin_dpo/margin_std": 457.5035705566406, "step": 510 }, { "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.0012351378099992871, "fcm_dpo/delta": 0.04284387826919556, "fcm_dpo/margin": 290.5696105957031, "fcm_dpo/q_t": 0.41743749380111694, "grad_norm": 42.793312072753906, "learning_rate": 8.637300491465272e-08, "logits/chosen": -0.7973334789276123, "logits/rejected": -0.8029496073722839, "logps/chosen": -538.7898559570312, "logps/ref_chosen": -60.9382438659668, "logps/ref_rejected": -83.36767578125, "logps/rejected": -851.7888793945312, "loss": 1.1273, "margin_dpo/margin_mean": 290.5696105957031, "margin_dpo/margin_std": 429.46759033203125, "step": 515 }, { "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.0012580296024680138, "fcm_dpo/delta": 0.001387386815622449, "fcm_dpo/margin": 316.6277770996094, "fcm_dpo/q_t": 0.4091448187828064, "grad_norm": 41.04801559448242, "learning_rate": 8.15760890883607e-08, "logits/chosen": -0.8406698107719421, "logits/rejected": -0.8473473787307739, "logps/chosen": -598.2073974609375, "logps/ref_chosen": -65.47642517089844, "logps/ref_rejected": -98.10872650146484, "logps/rejected": -947.4674072265625, "loss": 1.1012, "margin_dpo/margin_mean": 316.6277770996094, "margin_dpo/margin_std": 440.81640625, "step": 520 }, { "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.0012629171833395958, "fcm_dpo/delta": -0.03817005455493927, "fcm_dpo/margin": 345.6103820800781, "fcm_dpo/q_t": 0.4016591012477875, "grad_norm": 33.16600799560547, "learning_rate": 7.689012058193384e-08, "logits/chosen": -0.9183955192565918, "logits/rejected": -0.9257911443710327, "logps/chosen": -636.746337890625, "logps/ref_chosen": -59.072021484375, "logps/ref_rejected": -82.8821792602539, "logps/rejected": -1006.1668701171875, "loss": 1.0892, "margin_dpo/margin_mean": 345.6103515625, "margin_dpo/margin_std": 494.670654296875, "step": 525 }, { "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.001211377326399088, "fcm_dpo/delta": -0.020752830430865288, "fcm_dpo/margin": 330.8208923339844, "fcm_dpo/q_t": 0.4099728465080261, "grad_norm": 49.311241149902344, "learning_rate": 7.231818622338822e-08, "logits/chosen": -0.9991754293441772, "logits/rejected": -0.9904956817626953, "logps/chosen": -721.9043579101562, "logps/ref_chosen": -61.11234664916992, "logps/ref_rejected": -87.01112365722656, "logps/rejected": -1078.6241455078125, "loss": 1.1306, "margin_dpo/margin_mean": 330.8208923339844, "margin_dpo/margin_std": 549.4489135742188, "step": 530 }, { "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.001194678363390267, "fcm_dpo/delta": -0.01992225833237171, "fcm_dpo/margin": 334.68017578125, "fcm_dpo/q_t": 0.4102866053581238, "grad_norm": 28.87725067138672, "learning_rate": 6.786329772205246e-08, "logits/chosen": -0.9631765484809875, "logits/rejected": -0.9504965543746948, "logps/chosen": -703.7914428710938, "logps/ref_chosen": -60.96736526489258, "logps/ref_rejected": -81.727294921875, "logps/rejected": -1059.231689453125, "loss": 1.1238, "margin_dpo/margin_mean": 334.68017578125, "margin_dpo/margin_std": 534.3411865234375, "step": 535 }, { "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.001148082548752427, "fcm_dpo/delta": -0.046564746648073196, "fcm_dpo/margin": 386.79718017578125, "fcm_dpo/q_t": 0.4005100131034851, "grad_norm": 40.99845886230469, "learning_rate": 6.352838968463919e-08, "logits/chosen": -0.8846408128738403, "logits/rejected": -0.8896158337593079, "logps/chosen": -599.884765625, "logps/ref_chosen": -58.64385986328125, "logps/ref_rejected": -90.40965270996094, "logps/rejected": -1018.4476318359375, "loss": 1.0741, "margin_dpo/margin_mean": 386.7971496582031, "margin_dpo/margin_std": 522.71826171875, "step": 540 }, { "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.0011082093697041273, "fcm_dpo/delta": -0.027572477236390114, "fcm_dpo/margin": 361.2355651855469, "fcm_dpo/q_t": 0.4120791554450989, "grad_norm": 31.77776527404785, "learning_rate": 5.9316317682106294e-08, "logits/chosen": -0.9308038949966431, "logits/rejected": -0.9519888758659363, "logps/chosen": -702.5291137695312, "logps/ref_chosen": -64.73474884033203, "logps/ref_rejected": -100.44208526611328, "logps/rejected": -1099.471923828125, "loss": 1.1263, "margin_dpo/margin_mean": 361.2355651855469, "margin_dpo/margin_std": 592.0833740234375, "step": 545 }, { "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.0011139644775539637, "fcm_dpo/delta": 0.027384355664253235, "fcm_dpo/margin": 318.26593017578125, "fcm_dpo/q_t": 0.42001286149024963, "grad_norm": 41.989044189453125, "learning_rate": 5.5229856368582376e-08, "logits/chosen": -0.9026668667793274, "logits/rejected": -0.9054571390151978, "logps/chosen": -672.8770751953125, "logps/ref_chosen": -59.13951873779297, "logps/ref_rejected": -87.15635681152344, "logps/rejected": -1019.1598510742188, "loss": 1.1374, "margin_dpo/margin_mean": 318.26593017578125, "margin_dpo/margin_std": 509.5707092285156, "step": 550 }, { "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.0010551275918260217, "fcm_dpo/delta": -0.08985067903995514, "fcm_dpo/margin": 458.0889587402344, "fcm_dpo/q_t": 0.39341944456100464, "grad_norm": 37.768455505371094, "learning_rate": 5.127169765359515e-08, "logits/chosen": -0.9196065664291382, "logits/rejected": -0.9442132115364075, "logps/chosen": -678.6885986328125, "logps/ref_chosen": -62.1995849609375, "logps/ref_rejected": -102.51883697509766, "logps/rejected": -1177.096923828125, "loss": 1.0596, "margin_dpo/margin_mean": 458.0889587402344, "margin_dpo/margin_std": 621.8514404296875, "step": 555 }, { "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.001013588160276413, "fcm_dpo/delta": -0.03401713818311691, "fcm_dpo/margin": 426.43377685546875, "fcm_dpo/q_t": 0.4025075435638428, "grad_norm": 32.85615158081055, "learning_rate": 4.7444448928806615e-08, "logits/chosen": -0.9791582226753235, "logits/rejected": -0.987125039100647, "logps/chosen": -729.31787109375, "logps/ref_chosen": -61.541908264160156, "logps/ref_rejected": -97.37491607666016, "logps/rejected": -1191.584716796875, "loss": 1.0859, "margin_dpo/margin_mean": 426.43377685546875, "margin_dpo/margin_std": 586.039794921875, "step": 560 }, { "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.0010072619188576937, "fcm_dpo/delta": 0.004320817068219185, "fcm_dpo/margin": 365.18194580078125, "fcm_dpo/q_t": 0.41634687781333923, "grad_norm": 38.367191314697266, "learning_rate": 4.375063135042445e-08, "logits/chosen": -1.0174771547317505, "logits/rejected": -1.0257703065872192, "logps/chosen": -805.9225463867188, "logps/ref_chosen": -62.85475540161133, "logps/ref_rejected": -93.8392105102539, "logps/rejected": -1202.0889892578125, "loss": 1.1532, "margin_dpo/margin_mean": 365.18194580078125, "margin_dpo/margin_std": 646.3736572265625, "step": 565 }, { "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.000991632230579853, "fcm_dpo/delta": -0.003437476698309183, "fcm_dpo/margin": 406.2334899902344, "fcm_dpo/q_t": 0.40820297598838806, "grad_norm": 32.0954475402832, "learning_rate": 4.019267817841834e-08, "logits/chosen": -1.057476282119751, "logits/rejected": -1.0744601488113403, "logps/chosen": -756.536865234375, "logps/ref_chosen": -57.98622512817383, "logps/ref_rejected": -87.91555786132812, "logps/rejected": -1192.6998291015625, "loss": 1.108, "margin_dpo/margin_mean": 406.2334899902344, "margin_dpo/margin_std": 592.2410278320312, "step": 570 }, { "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.000984189915470779, "fcm_dpo/delta": -0.0074859424494206905, "fcm_dpo/margin": 413.28948974609375, "fcm_dpo/q_t": 0.4089413583278656, "grad_norm": 39.946693420410156, "learning_rate": 3.677293317363864e-08, "logits/chosen": -1.012502908706665, "logits/rejected": -1.0302174091339111, "logps/chosen": -763.88623046875, "logps/ref_chosen": -55.194114685058594, "logps/ref_rejected": -89.68229675292969, "logps/rejected": -1211.663818359375, "loss": 1.1112, "margin_dpo/margin_mean": 413.28948974609375, "margin_dpo/margin_std": 633.7578735351562, "step": 575 }, { "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.0010015666484832764, "fcm_dpo/delta": 0.005687057971954346, "fcm_dpo/margin": 373.99749755859375, "fcm_dpo/q_t": 0.41545066237449646, "grad_norm": 36.99066925048828, "learning_rate": 3.349364905389032e-08, "logits/chosen": -0.9414408802986145, "logits/rejected": -0.9566031694412231, "logps/chosen": -681.4942626953125, "logps/ref_chosen": -54.605796813964844, "logps/ref_rejected": -88.9614486694336, "logps/rejected": -1089.847412109375, "loss": 1.1256, "margin_dpo/margin_mean": 373.99749755859375, "margin_dpo/margin_std": 568.8309326171875, "step": 580 }, { "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.0009860583813861012, "fcm_dpo/delta": -0.024760950356721878, "fcm_dpo/margin": 374.35772705078125, "fcm_dpo/q_t": 0.4164944589138031, "grad_norm": 30.979644775390625, "learning_rate": 3.035698600998121e-08, "logits/chosen": -1.005885362625122, "logits/rejected": -1.0213685035705566, "logps/chosen": -722.11083984375, "logps/ref_chosen": -59.03770065307617, "logps/ref_rejected": -94.4029541015625, "logps/rejected": -1131.833984375, "loss": 1.1365, "margin_dpo/margin_mean": 374.35772705078125, "margin_dpo/margin_std": 598.1573486328125, "step": 585 }, { "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.0009923167526721954, "fcm_dpo/delta": 0.020330144092440605, "fcm_dpo/margin": 382.85650634765625, "fcm_dpo/q_t": 0.41268691420555115, "grad_norm": 26.06329917907715, "learning_rate": 2.736501028272095e-08, "logits/chosen": -1.001636266708374, "logits/rejected": -1.0243651866912842, "logps/chosen": -699.3173828125, "logps/ref_chosen": -53.5163688659668, "logps/ref_rejected": -99.90290832519531, "logps/rejected": -1128.5604248046875, "loss": 1.1114, "margin_dpo/margin_mean": 382.85650634765625, "margin_dpo/margin_std": 544.3261108398438, "step": 590 }, { "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.0009959937306120992, "fcm_dpo/delta": 0.008491059765219688, "fcm_dpo/margin": 356.68133544921875, "fcm_dpo/q_t": 0.4196888506412506, "grad_norm": 51.35660934448242, "learning_rate": 2.451969280180849e-08, "logits/chosen": -0.9872266054153442, "logits/rejected": -0.9868671298027039, "logps/chosen": -701.3692626953125, "logps/ref_chosen": -51.44538497924805, "logps/ref_rejected": -77.43083190917969, "logps/rejected": -1084.035888671875, "loss": 1.1503, "margin_dpo/margin_mean": 356.68133544921875, "margin_dpo/margin_std": 606.1689453125, "step": 595 }, { "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.0010021307971328497, "fcm_dpo/delta": 0.01705138385295868, "fcm_dpo/margin": 358.8981018066406, "fcm_dpo/q_t": 0.418224960565567, "grad_norm": 36.674869537353516, "learning_rate": 2.1822907887504932e-08, "logits/chosen": -0.9832857251167297, "logits/rejected": -0.9899972677230835, "logps/chosen": -721.6881103515625, "logps/ref_chosen": -57.161705017089844, "logps/ref_rejected": -87.73274230957031, "logps/rejected": -1111.1573486328125, "loss": 1.1393, "margin_dpo/margin_mean": 358.89813232421875, "margin_dpo/margin_std": 566.36669921875, "step": 600 }, { "epoch": 0.8810572687224669, "eval_fcm_dpo/beta": 0.0011259522289037704, "eval_fcm_dpo/delta": 0.023851953446865082, "eval_fcm_dpo/margin": 270.7278747558594, "eval_fcm_dpo/q_t": 0.4331645667552948, "eval_logits/chosen": -1.0197498798370361, "eval_logits/rejected": -1.0176793336868286, "eval_logps/chosen": -823.232421875, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -1101.70703125, "eval_loss": 0.6074439883232117, "eval_margin_dpo/margin_mean": 270.7278747558594, "eval_margin_dpo/margin_std": 601.7490234375, "eval_runtime": 39.7407, "eval_samples_per_second": 58.856, "eval_steps_per_second": 1.862, "step": 600 }, { "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.0011384881800040603, "fcm_dpo/delta": -0.09710332006216049, "fcm_dpo/margin": 432.0416564941406, "fcm_dpo/q_t": 0.39160969853401184, "grad_norm": 36.93719482421875, "learning_rate": 1.9276432015946446e-08, "logits/chosen": -0.9972618222236633, "logits/rejected": -1.01743483543396, "logps/chosen": -690.4810791015625, "logps/ref_chosen": -58.169830322265625, "logps/ref_rejected": -95.36891174316406, "logps/rejected": -1159.7216796875, "loss": 1.0561, "margin_dpo/margin_mean": 432.0416564941406, "margin_dpo/margin_std": 594.4193725585938, "step": 605 }, { "epoch": 0.895741556534508, "fcm_dpo/beta": 0.0010870896512642503, "fcm_dpo/delta": -0.01724116876721382, "fcm_dpo/margin": 382.8441467285156, "fcm_dpo/q_t": 0.4077000617980957, "grad_norm": 48.549068450927734, "learning_rate": 1.6881942648911074e-08, "logits/chosen": -0.9739276766777039, "logits/rejected": -0.978052020072937, "logps/chosen": -703.338623046875, "logps/ref_chosen": -58.97087860107422, "logps/ref_rejected": -89.0286865234375, "logps/rejected": -1116.2406005859375, "loss": 1.102, "margin_dpo/margin_mean": 382.8440856933594, "margin_dpo/margin_std": 576.8626708984375, "step": 610 }, { "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.001023916294798255, "fcm_dpo/delta": -0.07400115579366684, "fcm_dpo/margin": 458.7464294433594, "fcm_dpo/q_t": 0.39545583724975586, "grad_norm": 26.998958587646484, "learning_rate": 1.4641017128809801e-08, "logits/chosen": -0.9687691926956177, "logits/rejected": -1.0111548900604248, "logps/chosen": -723.0494384765625, "logps/ref_chosen": -58.081878662109375, "logps/ref_rejected": -95.92155456542969, "logps/rejected": -1219.635498046875, "loss": 1.0677, "margin_dpo/margin_mean": 458.7464904785156, "margin_dpo/margin_std": 639.427734375, "step": 615 }, { "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.0010129540460184216, "fcm_dpo/delta": 0.012466437183320522, "fcm_dpo/margin": 362.10418701171875, "fcm_dpo/q_t": 0.4198682904243469, "grad_norm": 31.997928619384766, "learning_rate": 1.2555131639630567e-08, "logits/chosen": -0.9928318858146667, "logits/rejected": -0.9981027841567993, "logps/chosen": -801.2032470703125, "logps/ref_chosen": -62.203094482421875, "logps/ref_rejected": -80.53683471679688, "logps/rejected": -1181.641357421875, "loss": 1.1594, "margin_dpo/margin_mean": 362.10418701171875, "margin_dpo/margin_std": 658.1754150390625, "step": 620 }, { "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.0009900578297674656, "fcm_dpo/delta": -0.04029404744505882, "fcm_dpo/margin": 442.16802978515625, "fcm_dpo/q_t": 0.4008522033691406, "grad_norm": 67.80354309082031, "learning_rate": 1.0625660234518913e-08, "logits/chosen": -0.9926285743713379, "logits/rejected": -1.0111424922943115, "logps/chosen": -743.4299926757812, "logps/ref_chosen": -61.727455139160156, "logps/ref_rejected": -88.4387435913086, "logps/rejected": -1212.309326171875, "loss": 1.0786, "margin_dpo/margin_mean": 442.1680603027344, "margin_dpo/margin_std": 596.7000732421875, "step": 625 }, { "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.0009705528500489891, "fcm_dpo/delta": -0.0021920218132436275, "fcm_dpo/margin": 385.10406494140625, "fcm_dpo/q_t": 0.41799044609069824, "grad_norm": 57.19618225097656, "learning_rate": 8.85387393063622e-09, "logits/chosen": -0.981094479560852, "logits/rejected": -0.999189019203186, "logps/chosen": -764.3931884765625, "logps/ref_chosen": -61.30865478515625, "logps/ref_rejected": -96.54997253417969, "logps/rejected": -1184.7386474609375, "loss": 1.1426, "margin_dpo/margin_mean": 385.10406494140625, "margin_dpo/margin_std": 646.9403076171875, "step": 630 }, { "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.000991217908449471, "fcm_dpo/delta": 0.04439081624150276, "fcm_dpo/margin": 321.2152099609375, "fcm_dpo/q_t": 0.42787250876426697, "grad_norm": 41.87347412109375, "learning_rate": 7.240939871891699e-09, "logits/chosen": -1.0177139043807983, "logits/rejected": -1.0183391571044922, "logps/chosen": -819.9927978515625, "logps/ref_chosen": -63.7315673828125, "logps/ref_rejected": -89.66435241699219, "logps/rejected": -1167.1407470703125, "loss": 1.1768, "margin_dpo/margin_mean": 321.2152404785156, "margin_dpo/margin_std": 599.1561889648438, "step": 635 }, { "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.0009926257189363241, "fcm_dpo/delta": -0.05684474855661392, "fcm_dpo/margin": 380.9329528808594, "fcm_dpo/q_t": 0.41502299904823303, "grad_norm": 40.166175842285156, "learning_rate": 5.7879205600998296e-09, "logits/chosen": -0.9638550877571106, "logits/rejected": -0.9688556790351868, "logps/chosen": -797.8259887695312, "logps/ref_chosen": -59.17915725708008, "logps/ref_rejected": -88.51210021972656, "logps/rejected": -1208.0919189453125, "loss": 1.1439, "margin_dpo/margin_mean": 380.9329528808594, "margin_dpo/margin_std": 651.566162109375, "step": 640 }, { "epoch": 0.947136563876652, "fcm_dpo/beta": 0.0009744317503646016, "fcm_dpo/delta": 0.0341024175286293, "fcm_dpo/margin": 321.48162841796875, "fcm_dpo/q_t": 0.43180274963378906, "grad_norm": 79.06712341308594, "learning_rate": 4.495773155069299e-09, "logits/chosen": -0.9938424229621887, "logits/rejected": -1.0045819282531738, "logps/chosen": -791.4851684570312, "logps/ref_chosen": -59.50596237182617, "logps/ref_rejected": -93.92404174804688, "logps/rejected": -1147.3848876953125, "loss": 1.1994, "margin_dpo/margin_mean": 321.48162841796875, "margin_dpo/margin_std": 670.0167846679688, "step": 645 }, { "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.0009998297318816185, "fcm_dpo/delta": 0.0030700929928570986, "fcm_dpo/margin": 377.16595458984375, "fcm_dpo/q_t": 0.41648992896080017, "grad_norm": 39.774078369140625, "learning_rate": 3.3653488440851253e-09, "logits/chosen": -0.994001567363739, "logits/rejected": -1.0127675533294678, "logps/chosen": -719.2337036132812, "logps/ref_chosen": -57.774566650390625, "logps/ref_rejected": -89.61600494384766, "logps/rejected": -1128.2410888671875, "loss": 1.1354, "margin_dpo/margin_mean": 377.16595458984375, "margin_dpo/margin_std": 618.4581298828125, "step": 650 }, { "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.0009819023543968797, "fcm_dpo/delta": 0.001074020517989993, "fcm_dpo/margin": 406.14300537109375, "fcm_dpo/q_t": 0.41048040986061096, "grad_norm": 35.741294860839844, "learning_rate": 2.397392281198729e-09, "logits/chosen": -1.0151567459106445, "logits/rejected": -1.0480637550354004, "logps/chosen": -682.8621826171875, "logps/ref_chosen": -55.68403244018555, "logps/ref_rejected": -102.4081802368164, "logps/rejected": -1135.7293701171875, "loss": 1.1068, "margin_dpo/margin_mean": 406.1430358886719, "margin_dpo/margin_std": 598.8343505859375, "step": 655 }, { "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.0009497703285887837, "fcm_dpo/delta": -0.056020140647888184, "fcm_dpo/margin": 475.952880859375, "fcm_dpo/q_t": 0.39766284823417664, "grad_norm": 36.36381149291992, "learning_rate": 1.592541096695571e-09, "logits/chosen": -0.988078773021698, "logits/rejected": -1.017830491065979, "logps/chosen": -716.3291625976562, "logps/ref_chosen": -59.19981002807617, "logps/ref_rejected": -94.19200134277344, "logps/rejected": -1227.274169921875, "loss": 1.0617, "margin_dpo/margin_mean": 475.95281982421875, "margin_dpo/margin_std": 605.3031005859375, "step": 660 }, { "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.0009469041833654046, "fcm_dpo/delta": 0.0041985055431723595, "fcm_dpo/margin": 417.9197692871094, "fcm_dpo/q_t": 0.41134023666381836, "grad_norm": 40.71209716796875, "learning_rate": 9.513254770636137e-10, "logits/chosen": -1.0182468891143799, "logits/rejected": -1.0391613245010376, "logps/chosen": -698.357421875, "logps/ref_chosen": -61.2533073425293, "logps/ref_rejected": -95.86351013183594, "logps/rejected": -1150.887451171875, "loss": 1.1093, "margin_dpo/margin_mean": 417.9197692871094, "margin_dpo/margin_std": 626.8549194335938, "step": 665 }, { "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.0009450761717744172, "fcm_dpo/delta": -0.012223686091601849, "fcm_dpo/margin": 435.3887634277344, "fcm_dpo/q_t": 0.40688472986221313, "grad_norm": 29.223600387573242, "learning_rate": 4.741678157389739e-10, "logits/chosen": -0.9618457555770874, "logits/rejected": -0.9741120338439941, "logps/chosen": -724.4750366210938, "logps/ref_chosen": -62.95263671875, "logps/ref_rejected": -92.4662094116211, "logps/rejected": -1189.3773193359375, "loss": 1.1, "margin_dpo/margin_mean": 435.3887634277344, "margin_dpo/margin_std": 627.5145263671875, "step": 670 }, { "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.000940955535043031, "fcm_dpo/delta": -0.006233499385416508, "fcm_dpo/margin": 431.116455078125, "fcm_dpo/q_t": 0.4094417095184326, "grad_norm": 23.441814422607422, "learning_rate": 1.6138243485910863e-10, "logits/chosen": -1.0078109502792358, "logits/rejected": -1.02387535572052, "logps/chosen": -741.0542602539062, "logps/ref_chosen": -48.5856819152832, "logps/ref_rejected": -81.27871704101562, "logps/rejected": -1204.8636474609375, "loss": 1.1069, "margin_dpo/margin_mean": 431.116455078125, "margin_dpo/margin_std": 644.87255859375, "step": 675 }, { "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.0009158365428447723, "fcm_dpo/delta": -0.01913156732916832, "fcm_dpo/margin": 456.44354248046875, "fcm_dpo/q_t": 0.40588778257369995, "grad_norm": 31.682743072509766, "learning_rate": 1.31753782067201e-11, "logits/chosen": -0.9820996522903442, "logits/rejected": -0.9966181516647339, "logps/chosen": -744.6751708984375, "logps/ref_chosen": -60.25421905517578, "logps/ref_rejected": -87.23457336425781, "logps/rejected": -1228.09912109375, "loss": 1.0916, "margin_dpo/margin_mean": 456.44354248046875, "margin_dpo/margin_std": 645.1212768554688, "step": 680 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 1.1011297496229893, "train_runtime": 1861.3017, "train_samples_per_second": 23.423, "train_steps_per_second": 0.366 } ], "logging_steps": 5, "max_steps": 681, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": false, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }