{
  "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": 4.390243902439025,
  "eval_steps": 30,
  "global_step": 270,
  "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
    }
  ],
  "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
}