| { | |
| "best_metric": 0.42961153388023376, | |
| "best_model_checkpoint": "./mistral/19-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.4-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-732_batch_4_2024-04-19_ppid_9/checkpoint-270", | |
| "epoch": 10.24390243902439, | |
| "eval_steps": 30, | |
| "global_step": 630, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.16, | |
| "grad_norm": 38.021175384521484, | |
| "learning_rate": 1.590909090909091e-06, | |
| "logits/chosen": -1.7808929681777954, | |
| "logits/rejected": -1.8136215209960938, | |
| "logps/chosen": -52.2663688659668, | |
| "logps/rejected": -88.76155853271484, | |
| "loss": 0.6927, | |
| "rewards/accuracies": 0.20000000298023224, | |
| "rewards/chosen": 0.0008833981119096279, | |
| "rewards/margins": 0.0009520913590677083, | |
| "rewards/rejected": -6.869318895041943e-05, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "grad_norm": 19.505126953125, | |
| "learning_rate": 3.863636363636364e-06, | |
| "logits/chosen": -1.865220308303833, | |
| "logits/rejected": -1.880090355873108, | |
| "logps/chosen": -34.3599967956543, | |
| "logps/rejected": -53.656646728515625, | |
| "loss": 0.6846, | |
| "rewards/accuracies": 0.22499999403953552, | |
| "rewards/chosen": 0.019346294924616814, | |
| "rewards/margins": 0.017682768404483795, | |
| "rewards/rejected": 0.0016635276842862368, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "grad_norm": 16.67194938659668, | |
| "learning_rate": 4.964788732394366e-06, | |
| "logits/chosen": -1.7692855596542358, | |
| "logits/rejected": -1.7932851314544678, | |
| "logps/chosen": -38.679237365722656, | |
| "logps/rejected": -79.23713684082031, | |
| "loss": 0.665, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.07158979028463364, | |
| "rewards/margins": 0.061105482280254364, | |
| "rewards/rejected": 0.010484304279088974, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_logits/chosen": -1.7801028490066528, | |
| "eval_logits/rejected": -1.7981160879135132, | |
| "eval_logps/chosen": -54.71597671508789, | |
| "eval_logps/rejected": -95.75775909423828, | |
| "eval_loss": 0.6322354674339294, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 0.14569346606731415, | |
| "eval_rewards/margins": 0.13385750353336334, | |
| "eval_rewards/rejected": 0.011835969984531403, | |
| "eval_runtime": 8.5124, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.645, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "grad_norm": 22.69721031188965, | |
| "learning_rate": 4.894366197183099e-06, | |
| "logits/chosen": -1.8458446264266968, | |
| "logits/rejected": -1.8594070672988892, | |
| "logps/chosen": -39.180992126464844, | |
| "logps/rejected": -57.04741287231445, | |
| "loss": 0.6164, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 0.19042351841926575, | |
| "rewards/margins": 0.18086455762386322, | |
| "rewards/rejected": 0.009558964520692825, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "grad_norm": 0.0, | |
| "learning_rate": 4.823943661971832e-06, | |
| "logits/chosen": -1.8411815166473389, | |
| "logits/rejected": -1.8659855127334595, | |
| "logps/chosen": -31.194744110107422, | |
| "logps/rejected": -58.17326736450195, | |
| "loss": 0.6138, | |
| "rewards/accuracies": 0.30000001192092896, | |
| "rewards/chosen": 0.19823171198368073, | |
| "rewards/margins": 0.19669394195079803, | |
| "rewards/rejected": 0.0015377718955278397, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "grad_norm": 10.77029800415039, | |
| "learning_rate": 4.753521126760564e-06, | |
| "logits/chosen": -1.816936731338501, | |
| "logits/rejected": -1.8421777486801147, | |
| "logps/chosen": -50.517478942871094, | |
| "logps/rejected": -80.6562728881836, | |
| "loss": 0.5445, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 0.4567543864250183, | |
| "rewards/margins": 0.42180705070495605, | |
| "rewards/rejected": 0.03494739532470703, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_logits/chosen": -1.784444808959961, | |
| "eval_logits/rejected": -1.8030898571014404, | |
| "eval_logps/chosen": -53.55432891845703, | |
| "eval_logps/rejected": -95.73511505126953, | |
| "eval_loss": 0.527186930179596, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 0.4941868484020233, | |
| "eval_rewards/margins": 0.47555941343307495, | |
| "eval_rewards/rejected": 0.018627464771270752, | |
| "eval_runtime": 8.5149, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "grad_norm": 8.583382606506348, | |
| "learning_rate": 4.683098591549296e-06, | |
| "logits/chosen": -1.7579460144042969, | |
| "logits/rejected": -1.776581048965454, | |
| "logps/chosen": -54.420066833496094, | |
| "logps/rejected": -79.0608901977539, | |
| "loss": 0.451, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.8353117108345032, | |
| "rewards/margins": 0.8680375814437866, | |
| "rewards/rejected": -0.032725896686315536, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "grad_norm": 2.3610572814941406, | |
| "learning_rate": 4.612676056338028e-06, | |
| "logits/chosen": -1.8833833932876587, | |
| "logits/rejected": -1.9092212915420532, | |
| "logps/chosen": -25.5800724029541, | |
| "logps/rejected": -64.96099853515625, | |
| "loss": 0.5302, | |
| "rewards/accuracies": 0.30000001192092896, | |
| "rewards/chosen": 0.5431563258171082, | |
| "rewards/margins": 0.6763826012611389, | |
| "rewards/rejected": -0.13322624564170837, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "grad_norm": 3.739367723464966, | |
| "learning_rate": 4.542253521126761e-06, | |
| "logits/chosen": -1.8194096088409424, | |
| "logits/rejected": -1.8343127965927124, | |
| "logps/chosen": -41.80801773071289, | |
| "logps/rejected": -63.005592346191406, | |
| "loss": 0.4616, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.830731213092804, | |
| "rewards/margins": 0.9403120875358582, | |
| "rewards/rejected": -0.109580859541893, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "eval_logits/chosen": -1.7902206182479858, | |
| "eval_logits/rejected": -1.8099161386489868, | |
| "eval_logps/chosen": -52.01105880737305, | |
| "eval_logps/rejected": -95.83198547363281, | |
| "eval_loss": 0.46216800808906555, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 0.957168459892273, | |
| "eval_rewards/margins": 0.9676021337509155, | |
| "eval_rewards/rejected": -0.010433738119900227, | |
| "eval_runtime": 8.5144, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "grad_norm": 0.5382364988327026, | |
| "learning_rate": 4.471830985915494e-06, | |
| "logits/chosen": -1.845868468284607, | |
| "logits/rejected": -1.8715919256210327, | |
| "logps/chosen": -40.285335540771484, | |
| "logps/rejected": -73.16102600097656, | |
| "loss": 0.4611, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 1.0631928443908691, | |
| "rewards/margins": 1.3067411184310913, | |
| "rewards/rejected": -0.24354824423789978, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "grad_norm": 12.538246154785156, | |
| "learning_rate": 4.401408450704226e-06, | |
| "logits/chosen": -1.8342373371124268, | |
| "logits/rejected": -1.862433671951294, | |
| "logps/chosen": -29.438589096069336, | |
| "logps/rejected": -62.31974411010742, | |
| "loss": 0.5266, | |
| "rewards/accuracies": 0.30000001192092896, | |
| "rewards/chosen": 0.7012811899185181, | |
| "rewards/margins": 0.8140069246292114, | |
| "rewards/rejected": -0.11272567510604858, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "grad_norm": 2.3604652881622314, | |
| "learning_rate": 4.3309859154929575e-06, | |
| "logits/chosen": -1.8481504917144775, | |
| "logits/rejected": -1.8528194427490234, | |
| "logps/chosen": -35.794334411621094, | |
| "logps/rejected": -56.22735595703125, | |
| "loss": 0.5327, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 1.078013300895691, | |
| "rewards/margins": 0.9431617856025696, | |
| "rewards/rejected": 0.1348516196012497, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "eval_logits/chosen": -1.7962983846664429, | |
| "eval_logits/rejected": -1.8171442747116089, | |
| "eval_logps/chosen": -50.74287796020508, | |
| "eval_logps/rejected": -96.0649185180664, | |
| "eval_loss": 0.4424423575401306, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 1.337624192237854, | |
| "eval_rewards/margins": 1.4179356098175049, | |
| "eval_rewards/rejected": -0.08031141757965088, | |
| "eval_runtime": 8.5188, | |
| "eval_samples_per_second": 3.287, | |
| "eval_steps_per_second": 1.643, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "grad_norm": 5.3829216957092285, | |
| "learning_rate": 4.26056338028169e-06, | |
| "logits/chosen": -1.8153988122940063, | |
| "logits/rejected": -1.853118658065796, | |
| "logps/chosen": -44.251441955566406, | |
| "logps/rejected": -85.38258361816406, | |
| "loss": 0.4011, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 1.5001299381256104, | |
| "rewards/margins": 1.9703232049942017, | |
| "rewards/rejected": -0.4701933264732361, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "grad_norm": 0.6282079815864563, | |
| "learning_rate": 4.190140845070423e-06, | |
| "logits/chosen": -1.8527545928955078, | |
| "logits/rejected": -1.8757705688476562, | |
| "logps/chosen": -35.6223258972168, | |
| "logps/rejected": -80.99830627441406, | |
| "loss": 0.4423, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 1.484961748123169, | |
| "rewards/margins": 1.8964126110076904, | |
| "rewards/rejected": -0.41145071387290955, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "grad_norm": 0.0, | |
| "learning_rate": 4.119718309859156e-06, | |
| "logits/chosen": -1.9129869937896729, | |
| "logits/rejected": -1.927869439125061, | |
| "logps/chosen": -24.400936126708984, | |
| "logps/rejected": -50.861839294433594, | |
| "loss": 0.5459, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 0.8408910036087036, | |
| "rewards/margins": 1.1376488208770752, | |
| "rewards/rejected": -0.2967577874660492, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "eval_logits/chosen": -1.8025389909744263, | |
| "eval_logits/rejected": -1.824258804321289, | |
| "eval_logps/chosen": -49.72334671020508, | |
| "eval_logps/rejected": -96.41252899169922, | |
| "eval_loss": 0.4335039258003235, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 1.6434847116470337, | |
| "eval_rewards/margins": 1.828076720237732, | |
| "eval_rewards/rejected": -0.1845920830965042, | |
| "eval_runtime": 8.5149, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "grad_norm": 0.7459288239479065, | |
| "learning_rate": 4.0492957746478875e-06, | |
| "logits/chosen": -1.786231279373169, | |
| "logits/rejected": -1.8223682641983032, | |
| "logps/chosen": -40.68749237060547, | |
| "logps/rejected": -78.1619873046875, | |
| "loss": 0.4348, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 1.9040056467056274, | |
| "rewards/margins": 2.482433795928955, | |
| "rewards/rejected": -0.5784280896186829, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "grad_norm": 1.4669339656829834, | |
| "learning_rate": 3.97887323943662e-06, | |
| "logits/chosen": -1.8449195623397827, | |
| "logits/rejected": -1.871626615524292, | |
| "logps/chosen": -46.335182189941406, | |
| "logps/rejected": -86.10281372070312, | |
| "loss": 0.4232, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 1.8989893198013306, | |
| "rewards/margins": 2.384274482727051, | |
| "rewards/rejected": -0.4852851927280426, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "grad_norm": 0.0301869735121727, | |
| "learning_rate": 3.908450704225352e-06, | |
| "logits/chosen": -1.8375946283340454, | |
| "logits/rejected": -1.8527040481567383, | |
| "logps/chosen": -39.01136016845703, | |
| "logps/rejected": -65.78593444824219, | |
| "loss": 0.4055, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 1.9841420650482178, | |
| "rewards/margins": 2.3045599460601807, | |
| "rewards/rejected": -0.32041770219802856, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "eval_logits/chosen": -1.8074016571044922, | |
| "eval_logits/rejected": -1.830051064491272, | |
| "eval_logps/chosen": -48.993629455566406, | |
| "eval_logps/rejected": -96.92731475830078, | |
| "eval_loss": 0.4326101839542389, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 1.8623981475830078, | |
| "eval_rewards/margins": 2.201425075531006, | |
| "eval_rewards/rejected": -0.3390272557735443, | |
| "eval_runtime": 8.5144, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "grad_norm": 2.224493980407715, | |
| "learning_rate": 3.838028169014085e-06, | |
| "logits/chosen": -1.8459789752960205, | |
| "logits/rejected": -1.8675992488861084, | |
| "logps/chosen": -32.56737518310547, | |
| "logps/rejected": -61.4633674621582, | |
| "loss": 0.4576, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 1.5155870914459229, | |
| "rewards/margins": 1.9375633001327515, | |
| "rewards/rejected": -0.42197614908218384, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "grad_norm": 16.606130599975586, | |
| "learning_rate": 3.767605633802817e-06, | |
| "logits/chosen": -1.8678531646728516, | |
| "logits/rejected": -1.8987194299697876, | |
| "logps/chosen": -36.32281494140625, | |
| "logps/rejected": -82.02101135253906, | |
| "loss": 0.4697, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 1.7492707967758179, | |
| "rewards/margins": 2.534116268157959, | |
| "rewards/rejected": -0.7848455309867859, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "grad_norm": 2.050732135772705, | |
| "learning_rate": 3.69718309859155e-06, | |
| "logits/chosen": -1.8438726663589478, | |
| "logits/rejected": -1.8704954385757446, | |
| "logps/chosen": -31.156131744384766, | |
| "logps/rejected": -71.87327575683594, | |
| "loss": 0.4694, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 1.7663137912750244, | |
| "rewards/margins": 2.468672513961792, | |
| "rewards/rejected": -0.702358603477478, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "eval_logits/chosen": -1.8135889768600464, | |
| "eval_logits/rejected": -1.836788535118103, | |
| "eval_logps/chosen": -48.54452133178711, | |
| "eval_logps/rejected": -97.27555084228516, | |
| "eval_loss": 0.43106651306152344, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 1.99713134765625, | |
| "eval_rewards/margins": 2.4406321048736572, | |
| "eval_rewards/rejected": -0.4435007870197296, | |
| "eval_runtime": 8.5161, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "grad_norm": 1.3447190523147583, | |
| "learning_rate": 3.626760563380282e-06, | |
| "logits/chosen": -1.9027411937713623, | |
| "logits/rejected": -1.9270541667938232, | |
| "logps/chosen": -31.44101333618164, | |
| "logps/rejected": -62.2841911315918, | |
| "loss": 0.4686, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 1.8702812194824219, | |
| "rewards/margins": 2.6481127738952637, | |
| "rewards/rejected": -0.7778315544128418, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "grad_norm": 0.8471798300743103, | |
| "learning_rate": 3.5563380281690144e-06, | |
| "logits/chosen": -1.7770761251449585, | |
| "logits/rejected": -1.8111178874969482, | |
| "logps/chosen": -48.4028434753418, | |
| "logps/rejected": -93.4535903930664, | |
| "loss": 0.3062, | |
| "rewards/accuracies": 0.574999988079071, | |
| "rewards/chosen": 2.7809433937072754, | |
| "rewards/margins": 3.951282501220703, | |
| "rewards/rejected": -1.1703392267227173, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "grad_norm": 0.017692934721708298, | |
| "learning_rate": 3.4859154929577467e-06, | |
| "logits/chosen": -1.8823143243789673, | |
| "logits/rejected": -1.908074975013733, | |
| "logps/chosen": -24.35366439819336, | |
| "logps/rejected": -55.66550827026367, | |
| "loss": 0.5431, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 1.2430522441864014, | |
| "rewards/margins": 1.8208128213882446, | |
| "rewards/rejected": -0.577760636806488, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "eval_logits/chosen": -1.8164188861846924, | |
| "eval_logits/rejected": -1.840105652809143, | |
| "eval_logps/chosen": -48.241355895996094, | |
| "eval_logps/rejected": -97.6273422241211, | |
| "eval_loss": 0.4247165322303772, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.0880820751190186, | |
| "eval_rewards/margins": 2.6371231079101562, | |
| "eval_rewards/rejected": -0.5490409731864929, | |
| "eval_runtime": 8.5143, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "grad_norm": 0.06990335881710052, | |
| "learning_rate": 3.415492957746479e-06, | |
| "logits/chosen": -1.855215311050415, | |
| "logits/rejected": -1.8647253513336182, | |
| "logps/chosen": -41.163124084472656, | |
| "logps/rejected": -63.0202751159668, | |
| "loss": 0.3954, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 2.336212635040283, | |
| "rewards/margins": 2.88411283493042, | |
| "rewards/rejected": -0.5479003190994263, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "grad_norm": 0.5815927386283875, | |
| "learning_rate": 3.3450704225352113e-06, | |
| "logits/chosen": -1.9256658554077148, | |
| "logits/rejected": -1.930415153503418, | |
| "logps/chosen": -24.930946350097656, | |
| "logps/rejected": -40.15483474731445, | |
| "loss": 0.5089, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 1.2408175468444824, | |
| "rewards/margins": 1.7317492961883545, | |
| "rewards/rejected": -0.49093180894851685, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "grad_norm": 0.38188156485557556, | |
| "learning_rate": 3.274647887323944e-06, | |
| "logits/chosen": -1.8994340896606445, | |
| "logits/rejected": -1.9313939809799194, | |
| "logps/chosen": -27.106714248657227, | |
| "logps/rejected": -67.90898132324219, | |
| "loss": 0.4547, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 1.7479994297027588, | |
| "rewards/margins": 2.5788474082946777, | |
| "rewards/rejected": -0.830848217010498, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "eval_logits/chosen": -1.8190696239471436, | |
| "eval_logits/rejected": -1.8433464765548706, | |
| "eval_logps/chosen": -47.96824264526367, | |
| "eval_logps/rejected": -98.09535217285156, | |
| "eval_loss": 0.42961153388023376, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.170013666152954, | |
| "eval_rewards/margins": 2.8594539165496826, | |
| "eval_rewards/rejected": -0.6894403696060181, | |
| "eval_runtime": 8.5148, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "grad_norm": 0.3809558153152466, | |
| "learning_rate": 3.2042253521126764e-06, | |
| "logits/chosen": -1.861537218093872, | |
| "logits/rejected": -1.880858063697815, | |
| "logps/chosen": -33.568603515625, | |
| "logps/rejected": -57.01904296875, | |
| "loss": 0.4508, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 2.471442699432373, | |
| "rewards/margins": 3.4677319526672363, | |
| "rewards/rejected": -0.996289074420929, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "grad_norm": 0.03738081455230713, | |
| "learning_rate": 3.133802816901409e-06, | |
| "logits/chosen": -1.8528019189834595, | |
| "logits/rejected": -1.8946154117584229, | |
| "logps/chosen": -35.256317138671875, | |
| "logps/rejected": -99.2815933227539, | |
| "loss": 0.442, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 1.5687124729156494, | |
| "rewards/margins": 3.0280191898345947, | |
| "rewards/rejected": -1.4593069553375244, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "grad_norm": 0.007441267836838961, | |
| "learning_rate": 3.063380281690141e-06, | |
| "logits/chosen": -1.7671687602996826, | |
| "logits/rejected": -1.7900078296661377, | |
| "logps/chosen": -58.0703239440918, | |
| "logps/rejected": -107.0584487915039, | |
| "loss": 0.3606, | |
| "rewards/accuracies": 0.550000011920929, | |
| "rewards/chosen": 3.007930278778076, | |
| "rewards/margins": 4.793817043304443, | |
| "rewards/rejected": -1.7858864068984985, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "eval_logits/chosen": -1.8213108777999878, | |
| "eval_logits/rejected": -1.8460028171539307, | |
| "eval_logps/chosen": -47.7896842956543, | |
| "eval_logps/rejected": -98.4369125366211, | |
| "eval_loss": 0.4289851188659668, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.223581314086914, | |
| "eval_rewards/margins": 3.015495777130127, | |
| "eval_rewards/rejected": -0.7919142842292786, | |
| "eval_runtime": 8.5154, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "grad_norm": 0.037768810987472534, | |
| "learning_rate": 2.9929577464788733e-06, | |
| "logits/chosen": -1.8851121664047241, | |
| "logits/rejected": -1.9122593402862549, | |
| "logps/chosen": -21.110910415649414, | |
| "logps/rejected": -48.944602966308594, | |
| "loss": 0.4731, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 1.653512954711914, | |
| "rewards/margins": 2.38820219039917, | |
| "rewards/rejected": -0.7346888780593872, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "grad_norm": 0.0, | |
| "learning_rate": 2.922535211267606e-06, | |
| "logits/chosen": -1.8363854885101318, | |
| "logits/rejected": -1.869361162185669, | |
| "logps/chosen": -36.65054702758789, | |
| "logps/rejected": -84.7163314819336, | |
| "loss": 0.3992, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 2.3141231536865234, | |
| "rewards/margins": 3.9065093994140625, | |
| "rewards/rejected": -1.5923866033554077, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "grad_norm": 0.08495225757360458, | |
| "learning_rate": 2.8521126760563383e-06, | |
| "logits/chosen": -1.854119062423706, | |
| "logits/rejected": -1.8729274272918701, | |
| "logps/chosen": -40.588775634765625, | |
| "logps/rejected": -68.68611145019531, | |
| "loss": 0.4021, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 2.146328926086426, | |
| "rewards/margins": 3.4630661010742188, | |
| "rewards/rejected": -1.3167372941970825, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "eval_logits/chosen": -1.8219343423843384, | |
| "eval_logits/rejected": -1.8470803499221802, | |
| "eval_logps/chosen": -47.683902740478516, | |
| "eval_logps/rejected": -98.87833404541016, | |
| "eval_loss": 0.4302048683166504, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.255316734313965, | |
| "eval_rewards/margins": 3.179651975631714, | |
| "eval_rewards/rejected": -0.9243355989456177, | |
| "eval_runtime": 8.5149, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "grad_norm": 0.00047089156578294933, | |
| "learning_rate": 2.781690140845071e-06, | |
| "logits/chosen": -1.9298601150512695, | |
| "logits/rejected": -1.9418340921401978, | |
| "logps/chosen": -28.818872451782227, | |
| "logps/rejected": -59.13349151611328, | |
| "loss": 0.5413, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 1.8383833169937134, | |
| "rewards/margins": 2.883132219314575, | |
| "rewards/rejected": -1.0447486639022827, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "grad_norm": 2.729531764984131, | |
| "learning_rate": 2.7112676056338033e-06, | |
| "logits/chosen": -1.8659921884536743, | |
| "logits/rejected": -1.905010461807251, | |
| "logps/chosen": -27.63974380493164, | |
| "logps/rejected": -73.26609802246094, | |
| "loss": 0.4368, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 2.2738237380981445, | |
| "rewards/margins": 3.4326729774475098, | |
| "rewards/rejected": -1.1588493585586548, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "grad_norm": 0.01013993564993143, | |
| "learning_rate": 2.640845070422535e-06, | |
| "logits/chosen": -1.8495924472808838, | |
| "logits/rejected": -1.8761341571807861, | |
| "logps/chosen": -33.9546012878418, | |
| "logps/rejected": -70.13398742675781, | |
| "loss": 0.419, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 2.0099234580993652, | |
| "rewards/margins": 3.498170852661133, | |
| "rewards/rejected": -1.488247275352478, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "eval_logits/chosen": -1.8214120864868164, | |
| "eval_logits/rejected": -1.8469568490982056, | |
| "eval_logps/chosen": -47.67514419555664, | |
| "eval_logps/rejected": -99.1513900756836, | |
| "eval_loss": 0.43358945846557617, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.257944107055664, | |
| "eval_rewards/margins": 3.264200210571289, | |
| "eval_rewards/rejected": -1.0062559843063354, | |
| "eval_runtime": 8.5148, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "grad_norm": 0.10058386623859406, | |
| "learning_rate": 2.570422535211268e-06, | |
| "logits/chosen": -1.8099803924560547, | |
| "logits/rejected": -1.8450464010238647, | |
| "logps/chosen": -40.64691925048828, | |
| "logps/rejected": -88.50006103515625, | |
| "loss": 0.4072, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 2.367983341217041, | |
| "rewards/margins": 3.652095079421997, | |
| "rewards/rejected": -1.2841118574142456, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "grad_norm": 0.0, | |
| "learning_rate": 2.5e-06, | |
| "logits/chosen": -1.8781477212905884, | |
| "logits/rejected": -1.9100782871246338, | |
| "logps/chosen": -33.296836853027344, | |
| "logps/rejected": -83.65415954589844, | |
| "loss": 0.4006, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 2.6570353507995605, | |
| "rewards/margins": 4.220401763916016, | |
| "rewards/rejected": -1.5633666515350342, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "grad_norm": 0.6534411907196045, | |
| "learning_rate": 2.4295774647887325e-06, | |
| "logits/chosen": -1.8289324045181274, | |
| "logits/rejected": -1.8389921188354492, | |
| "logps/chosen": -46.14112091064453, | |
| "logps/rejected": -61.8495979309082, | |
| "loss": 0.3984, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 2.3760623931884766, | |
| "rewards/margins": 3.9815852642059326, | |
| "rewards/rejected": -1.605522871017456, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "eval_logits/chosen": -1.824257493019104, | |
| "eval_logits/rejected": -1.8499394655227661, | |
| "eval_logps/chosen": -47.629581451416016, | |
| "eval_logps/rejected": -99.36784362792969, | |
| "eval_loss": 0.42913129925727844, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2716121673583984, | |
| "eval_rewards/margins": 3.3427982330322266, | |
| "eval_rewards/rejected": -1.0711863040924072, | |
| "eval_runtime": 8.5204, | |
| "eval_samples_per_second": 3.286, | |
| "eval_steps_per_second": 1.643, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "grad_norm": 0.10979699343442917, | |
| "learning_rate": 2.359154929577465e-06, | |
| "logits/chosen": -1.8595476150512695, | |
| "logits/rejected": -1.8786380290985107, | |
| "logps/chosen": -36.14017105102539, | |
| "logps/rejected": -69.46199798583984, | |
| "loss": 0.4672, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 2.0604913234710693, | |
| "rewards/margins": 3.718773365020752, | |
| "rewards/rejected": -1.658281683921814, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "grad_norm": 0.00019151422020513564, | |
| "learning_rate": 2.2887323943661975e-06, | |
| "logits/chosen": -1.8760582208633423, | |
| "logits/rejected": -1.903466820716858, | |
| "logps/chosen": -28.268178939819336, | |
| "logps/rejected": -71.25038146972656, | |
| "loss": 0.468, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 2.309755802154541, | |
| "rewards/margins": 3.500605344772339, | |
| "rewards/rejected": -1.190850019454956, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "grad_norm": 0.015070440247654915, | |
| "learning_rate": 2.21830985915493e-06, | |
| "logits/chosen": -1.877672553062439, | |
| "logits/rejected": -1.8992106914520264, | |
| "logps/chosen": -34.51100158691406, | |
| "logps/rejected": -69.979248046875, | |
| "loss": 0.435, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 2.112297534942627, | |
| "rewards/margins": 3.443708896636963, | |
| "rewards/rejected": -1.3314111232757568, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "eval_logits/chosen": -1.8236370086669922, | |
| "eval_logits/rejected": -1.8495099544525146, | |
| "eval_logps/chosen": -47.62683868408203, | |
| "eval_logps/rejected": -99.54405212402344, | |
| "eval_loss": 0.4284569323062897, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.272435426712036, | |
| "eval_rewards/margins": 3.396484613418579, | |
| "eval_rewards/rejected": -1.1240490674972534, | |
| "eval_runtime": 8.5146, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "grad_norm": 0.027868641540408134, | |
| "learning_rate": 2.147887323943662e-06, | |
| "logits/chosen": -1.8732540607452393, | |
| "logits/rejected": -1.9188286066055298, | |
| "logps/chosen": -28.66534996032715, | |
| "logps/rejected": -87.85456085205078, | |
| "loss": 0.4359, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 1.747658133506775, | |
| "rewards/margins": 3.2419276237487793, | |
| "rewards/rejected": -1.494269847869873, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "grad_norm": 1.6011111736297607, | |
| "learning_rate": 2.0774647887323944e-06, | |
| "logits/chosen": -1.8680269718170166, | |
| "logits/rejected": -1.883652925491333, | |
| "logps/chosen": -22.022457122802734, | |
| "logps/rejected": -51.710731506347656, | |
| "loss": 0.4878, | |
| "rewards/accuracies": 0.30000001192092896, | |
| "rewards/chosen": 1.6784204244613647, | |
| "rewards/margins": 2.697260856628418, | |
| "rewards/rejected": -1.0188405513763428, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "grad_norm": 0.005799587815999985, | |
| "learning_rate": 2.007042253521127e-06, | |
| "logits/chosen": -1.9159187078475952, | |
| "logits/rejected": -1.942246675491333, | |
| "logps/chosen": -26.426944732666016, | |
| "logps/rejected": -63.260643005371094, | |
| "loss": 0.5148, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 1.5094796419143677, | |
| "rewards/margins": 2.8935487270355225, | |
| "rewards/rejected": -1.3840690851211548, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "eval_logits/chosen": -1.8220092058181763, | |
| "eval_logits/rejected": -1.8481791019439697, | |
| "eval_logps/chosen": -47.637298583984375, | |
| "eval_logps/rejected": -99.83494567871094, | |
| "eval_loss": 0.43089213967323303, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2692975997924805, | |
| "eval_rewards/margins": 3.480616331100464, | |
| "eval_rewards/rejected": -1.211318850517273, | |
| "eval_runtime": 8.5151, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "grad_norm": 1.8945941925048828, | |
| "learning_rate": 1.936619718309859e-06, | |
| "logits/chosen": -1.890688180923462, | |
| "logits/rejected": -1.9237228631973267, | |
| "logps/chosen": -24.1690616607666, | |
| "logps/rejected": -72.56752014160156, | |
| "loss": 0.4519, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 1.6932684183120728, | |
| "rewards/margins": 3.1507351398468018, | |
| "rewards/rejected": -1.4574666023254395, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "grad_norm": 0.004997989162802696, | |
| "learning_rate": 1.8661971830985917e-06, | |
| "logits/chosen": -1.8222097158432007, | |
| "logits/rejected": -1.8574409484863281, | |
| "logps/chosen": -47.20992660522461, | |
| "logps/rejected": -87.4762954711914, | |
| "loss": 0.364, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 2.4281115531921387, | |
| "rewards/margins": 4.541991710662842, | |
| "rewards/rejected": -2.1138806343078613, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "grad_norm": 0.4863984286785126, | |
| "learning_rate": 1.7957746478873243e-06, | |
| "logits/chosen": -1.8273519277572632, | |
| "logits/rejected": -1.8548485040664673, | |
| "logps/chosen": -42.0764045715332, | |
| "logps/rejected": -92.58042907714844, | |
| "loss": 0.412, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 2.7775075435638428, | |
| "rewards/margins": 4.632546424865723, | |
| "rewards/rejected": -1.8550386428833008, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "eval_logits/chosen": -1.8217111825942993, | |
| "eval_logits/rejected": -1.848099946975708, | |
| "eval_logps/chosen": -47.652671813964844, | |
| "eval_logps/rejected": -100.00603485107422, | |
| "eval_loss": 0.43079689145088196, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2646849155426025, | |
| "eval_rewards/margins": 3.527329206466675, | |
| "eval_rewards/rejected": -1.2626441717147827, | |
| "eval_runtime": 8.5152, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "grad_norm": 0.15775924921035767, | |
| "learning_rate": 1.7253521126760563e-06, | |
| "logits/chosen": -1.8239513635635376, | |
| "logits/rejected": -1.8541948795318604, | |
| "logps/chosen": -45.210750579833984, | |
| "logps/rejected": -90.68567657470703, | |
| "loss": 0.3297, | |
| "rewards/accuracies": 0.5249999761581421, | |
| "rewards/chosen": 3.3941636085510254, | |
| "rewards/margins": 5.666165351867676, | |
| "rewards/rejected": -2.2720017433166504, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "grad_norm": 0.0004808221128769219, | |
| "learning_rate": 1.6549295774647889e-06, | |
| "logits/chosen": -1.8781496286392212, | |
| "logits/rejected": -1.907164216041565, | |
| "logps/chosen": -29.897720336914062, | |
| "logps/rejected": -69.37852478027344, | |
| "loss": 0.417, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 2.1767258644104004, | |
| "rewards/margins": 3.8588624000549316, | |
| "rewards/rejected": -1.6821367740631104, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "grad_norm": 0.016680508852005005, | |
| "learning_rate": 1.5845070422535214e-06, | |
| "logits/chosen": -1.8918203115463257, | |
| "logits/rejected": -1.8992207050323486, | |
| "logps/chosen": -33.8150634765625, | |
| "logps/rejected": -48.536460876464844, | |
| "loss": 0.4911, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 1.9511134624481201, | |
| "rewards/margins": 3.087284564971924, | |
| "rewards/rejected": -1.1361711025238037, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "eval_logits/chosen": -1.8200041055679321, | |
| "eval_logits/rejected": -1.8465949296951294, | |
| "eval_logps/chosen": -47.68352508544922, | |
| "eval_logps/rejected": -100.16289520263672, | |
| "eval_loss": 0.43305009603500366, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2554304599761963, | |
| "eval_rewards/margins": 3.5651323795318604, | |
| "eval_rewards/rejected": -1.3097018003463745, | |
| "eval_runtime": 8.5147, | |
| "eval_samples_per_second": 3.288, | |
| "eval_steps_per_second": 1.644, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "grad_norm": 0.3721207082271576, | |
| "learning_rate": 1.5140845070422537e-06, | |
| "logits/chosen": -1.8607041835784912, | |
| "logits/rejected": -1.888972520828247, | |
| "logps/chosen": -34.54069900512695, | |
| "logps/rejected": -72.07875061035156, | |
| "loss": 0.4, | |
| "rewards/accuracies": 0.42500001192092896, | |
| "rewards/chosen": 2.3984122276306152, | |
| "rewards/margins": 4.267429351806641, | |
| "rewards/rejected": -1.869017243385315, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "grad_norm": 0.04920463636517525, | |
| "learning_rate": 1.443661971830986e-06, | |
| "logits/chosen": -1.8621752262115479, | |
| "logits/rejected": -1.8955097198486328, | |
| "logps/chosen": -33.04378128051758, | |
| "logps/rejected": -88.19145202636719, | |
| "loss": 0.4161, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 2.082742214202881, | |
| "rewards/margins": 4.248581886291504, | |
| "rewards/rejected": -2.165839672088623, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "grad_norm": 0.0, | |
| "learning_rate": 1.3732394366197185e-06, | |
| "logits/chosen": -1.8557628393173218, | |
| "logits/rejected": -1.877593994140625, | |
| "logps/chosen": -37.17829895019531, | |
| "logps/rejected": -76.12166595458984, | |
| "loss": 0.4433, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 1.7849305868148804, | |
| "rewards/margins": 3.114356517791748, | |
| "rewards/rejected": -1.3294259309768677, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "eval_logits/chosen": -1.8201100826263428, | |
| "eval_logits/rejected": -1.8468064069747925, | |
| "eval_logps/chosen": -47.717445373535156, | |
| "eval_logps/rejected": -100.2647933959961, | |
| "eval_loss": 0.4317461848258972, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.245253562927246, | |
| "eval_rewards/margins": 3.5855274200439453, | |
| "eval_rewards/rejected": -1.3402734994888306, | |
| "eval_runtime": 8.5142, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "grad_norm": 0.008082494139671326, | |
| "learning_rate": 1.3028169014084508e-06, | |
| "logits/chosen": -1.867776870727539, | |
| "logits/rejected": -1.9175941944122314, | |
| "logps/chosen": -32.92365646362305, | |
| "logps/rejected": -90.70045471191406, | |
| "loss": 0.4336, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 2.273036479949951, | |
| "rewards/margins": 4.531402111053467, | |
| "rewards/rejected": -2.258366107940674, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "grad_norm": 0.045851148664951324, | |
| "learning_rate": 1.2323943661971833e-06, | |
| "logits/chosen": -1.8460557460784912, | |
| "logits/rejected": -1.8667150735855103, | |
| "logps/chosen": -42.35684585571289, | |
| "logps/rejected": -79.35388946533203, | |
| "loss": 0.3643, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 3.1248345375061035, | |
| "rewards/margins": 5.3199896812438965, | |
| "rewards/rejected": -2.195154905319214, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "grad_norm": 0.010707350447773933, | |
| "learning_rate": 1.1619718309859156e-06, | |
| "logits/chosen": -1.8079864978790283, | |
| "logits/rejected": -1.8591219186782837, | |
| "logps/chosen": -33.473018646240234, | |
| "logps/rejected": -108.78472900390625, | |
| "loss": 0.3813, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 3.130399465560913, | |
| "rewards/margins": 5.355044364929199, | |
| "rewards/rejected": -2.224644660949707, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "eval_logits/chosen": -1.8204511404037476, | |
| "eval_logits/rejected": -1.8472681045532227, | |
| "eval_logps/chosen": -47.7364501953125, | |
| "eval_logps/rejected": -100.41536712646484, | |
| "eval_loss": 0.4337718188762665, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2395517826080322, | |
| "eval_rewards/margins": 3.624999523162842, | |
| "eval_rewards/rejected": -1.3854477405548096, | |
| "eval_runtime": 8.514, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 9.43, | |
| "grad_norm": 0.002776271430775523, | |
| "learning_rate": 1.0915492957746479e-06, | |
| "logits/chosen": -1.8287931680679321, | |
| "logits/rejected": -1.8448396921157837, | |
| "logps/chosen": -42.72909927368164, | |
| "logps/rejected": -65.72794342041016, | |
| "loss": 0.373, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 2.085402011871338, | |
| "rewards/margins": 3.9838340282440186, | |
| "rewards/rejected": -1.8984321355819702, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 9.59, | |
| "grad_norm": 0.1402006894350052, | |
| "learning_rate": 1.0211267605633804e-06, | |
| "logits/chosen": -1.830857515335083, | |
| "logits/rejected": -1.8767585754394531, | |
| "logps/chosen": -36.405574798583984, | |
| "logps/rejected": -95.58689880371094, | |
| "loss": 0.4341, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 2.1797754764556885, | |
| "rewards/margins": 4.409985065460205, | |
| "rewards/rejected": -2.2302088737487793, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "grad_norm": 0.0, | |
| "learning_rate": 9.507042253521127e-07, | |
| "logits/chosen": -1.9259939193725586, | |
| "logits/rejected": -1.9520877599716187, | |
| "logps/chosen": -26.32663345336914, | |
| "logps/rejected": -62.07328414916992, | |
| "loss": 0.5026, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 1.7667009830474854, | |
| "rewards/margins": 2.9083454608917236, | |
| "rewards/rejected": -1.1416443586349487, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "eval_logits/chosen": -1.8202803134918213, | |
| "eval_logits/rejected": -1.8471726179122925, | |
| "eval_logps/chosen": -47.739723205566406, | |
| "eval_logps/rejected": -100.47122955322266, | |
| "eval_loss": 0.4333479106426239, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2385685443878174, | |
| "eval_rewards/margins": 3.640772581100464, | |
| "eval_rewards/rejected": -1.4022036790847778, | |
| "eval_runtime": 8.5176, | |
| "eval_samples_per_second": 3.287, | |
| "eval_steps_per_second": 1.644, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "grad_norm": 0.0, | |
| "learning_rate": 8.802816901408451e-07, | |
| "logits/chosen": -1.917443037033081, | |
| "logits/rejected": -1.9254251718521118, | |
| "logps/chosen": -21.112218856811523, | |
| "logps/rejected": -43.751461029052734, | |
| "loss": 0.5207, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 1.3523606061935425, | |
| "rewards/margins": 2.597020387649536, | |
| "rewards/rejected": -1.2446597814559937, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "grad_norm": 0.01932019367814064, | |
| "learning_rate": 8.098591549295776e-07, | |
| "logits/chosen": -1.851624846458435, | |
| "logits/rejected": -1.8719497919082642, | |
| "logps/chosen": -41.65423583984375, | |
| "logps/rejected": -61.135986328125, | |
| "loss": 0.3882, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 2.643747091293335, | |
| "rewards/margins": 4.554553031921387, | |
| "rewards/rejected": -1.9108059406280518, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 10.24, | |
| "grad_norm": 0.022428497672080994, | |
| "learning_rate": 7.394366197183099e-07, | |
| "logits/chosen": -1.7906919717788696, | |
| "logits/rejected": -1.8341833353042603, | |
| "logps/chosen": -50.148780822753906, | |
| "logps/rejected": -102.08969116210938, | |
| "loss": 0.3121, | |
| "rewards/accuracies": 0.550000011920929, | |
| "rewards/chosen": 2.8718371391296387, | |
| "rewards/margins": 5.628880023956299, | |
| "rewards/rejected": -2.757042646408081, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 10.24, | |
| "eval_logits/chosen": -1.819262146949768, | |
| "eval_logits/rejected": -1.8461952209472656, | |
| "eval_logps/chosen": -47.75532150268555, | |
| "eval_logps/rejected": -100.51660919189453, | |
| "eval_loss": 0.43236425518989563, | |
| "eval_rewards/accuracies": 0.4285714328289032, | |
| "eval_rewards/chosen": 2.2338900566101074, | |
| "eval_rewards/margins": 3.6497082710266113, | |
| "eval_rewards/rejected": -1.415818452835083, | |
| "eval_runtime": 8.5143, | |
| "eval_samples_per_second": 3.289, | |
| "eval_steps_per_second": 1.644, | |
| "step": 630 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 732, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 12, | |
| "save_steps": 90, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |