{
  "best_metric": 0.3756321966648102,
  "best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-180",
  "epoch": 1.1020408163265305,
  "eval_steps": 30,
  "global_step": 270,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.04,
      "grad_norm": 83.44550323486328,
      "learning_rate": 6.666666666666667e-07,
      "logits/chosen": -1.7825796604156494,
      "logits/rejected": -1.7886196374893188,
      "logps/chosen": -169.71481323242188,
      "logps/rejected": -167.6764373779297,
      "loss": 0.6937,
      "rewards/accuracies": 0.20000000298023224,
      "rewards/chosen": -0.0002815247280523181,
      "rewards/margins": -0.0010146332206204534,
      "rewards/rejected": 0.0007331084925681353,
      "step": 10
    },
    {
      "epoch": 0.08,
      "grad_norm": 87.56982421875,
      "learning_rate": 1.777777777777778e-06,
      "logits/chosen": -1.623265027999878,
      "logits/rejected": -1.6968040466308594,
      "logps/chosen": -128.7312469482422,
      "logps/rejected": -184.20187377929688,
      "loss": 0.6912,
      "rewards/accuracies": 0.5,
      "rewards/chosen": 0.0030528828501701355,
      "rewards/margins": 0.004004458896815777,
      "rewards/rejected": -0.0009515761630609632,
      "step": 20
    },
    {
      "epoch": 0.12,
      "grad_norm": 113.4741439819336,
      "learning_rate": 2.888888888888889e-06,
      "logits/chosen": -1.7152118682861328,
      "logits/rejected": -1.758681058883667,
      "logps/chosen": -146.91494750976562,
      "logps/rejected": -213.5966339111328,
      "loss": 0.6781,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.03251449763774872,
      "rewards/margins": 0.030697176232933998,
      "rewards/rejected": 0.0018173219868913293,
      "step": 30
    },
    {
      "epoch": 0.12,
      "eval_logits/chosen": -1.795080304145813,
      "eval_logits/rejected": -1.8307874202728271,
      "eval_logps/chosen": -194.3749542236328,
      "eval_logps/rejected": -243.1332244873047,
      "eval_loss": 0.6762310266494751,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 0.05039369314908981,
      "eval_rewards/margins": 0.03467999026179314,
      "eval_rewards/rejected": 0.01571369729936123,
      "eval_runtime": 26.2365,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 30
    },
    {
      "epoch": 0.16,
      "grad_norm": 50.695308685302734,
      "learning_rate": 4.000000000000001e-06,
      "logits/chosen": -1.9357620477676392,
      "logits/rejected": -1.9375683069229126,
      "logps/chosen": -183.8218231201172,
      "logps/rejected": -115.80632019042969,
      "loss": 0.6768,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.056698836386203766,
      "rewards/margins": 0.03413978964090347,
      "rewards/rejected": 0.02255905233323574,
      "step": 40
    },
    {
      "epoch": 0.2,
      "grad_norm": 71.6834716796875,
      "learning_rate": 4.996491228070176e-06,
      "logits/chosen": -1.7773185968399048,
      "logits/rejected": -1.8331743478775024,
      "logps/chosen": -167.78533935546875,
      "logps/rejected": -227.0582733154297,
      "loss": 0.6205,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.1449195146560669,
      "rewards/margins": 0.15578031539916992,
      "rewards/rejected": -0.010860783979296684,
      "step": 50
    },
    {
      "epoch": 0.24,
      "grad_norm": 93.83502197265625,
      "learning_rate": 4.96140350877193e-06,
      "logits/chosen": -1.7739536762237549,
      "logits/rejected": -1.8414790630340576,
      "logps/chosen": -153.57481384277344,
      "logps/rejected": -194.3485870361328,
      "loss": 0.5918,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.2464807778596878,
      "rewards/margins": 0.22958576679229736,
      "rewards/rejected": 0.016895027831196785,
      "step": 60
    },
    {
      "epoch": 0.24,
      "eval_logits/chosen": -1.7974774837493896,
      "eval_logits/rejected": -1.8332849740982056,
      "eval_logps/chosen": -193.71743774414062,
      "eval_logps/rejected": -243.05783081054688,
      "eval_loss": 0.5998325943946838,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.24764502048492432,
      "eval_rewards/margins": 0.20930658280849457,
      "eval_rewards/rejected": 0.03833846375346184,
      "eval_runtime": 26.2349,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 60
    },
    {
      "epoch": 0.29,
      "grad_norm": 109.71602630615234,
      "learning_rate": 4.926315789473685e-06,
      "logits/chosen": -1.8950660228729248,
      "logits/rejected": -1.9148075580596924,
      "logps/chosen": -169.89028930664062,
      "logps/rejected": -177.87301635742188,
      "loss": 0.5992,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.15271134674549103,
      "rewards/margins": 0.21031668782234192,
      "rewards/rejected": -0.0576053187251091,
      "step": 70
    },
    {
      "epoch": 0.33,
      "grad_norm": 90.3881607055664,
      "learning_rate": 4.8912280701754386e-06,
      "logits/chosen": -1.798600196838379,
      "logits/rejected": -1.855521559715271,
      "logps/chosen": -160.3217315673828,
      "logps/rejected": -247.6388397216797,
      "loss": 0.5551,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.29141485691070557,
      "rewards/margins": 0.3369868993759155,
      "rewards/rejected": -0.04557197913527489,
      "step": 80
    },
    {
      "epoch": 0.37,
      "grad_norm": 147.27755737304688,
      "learning_rate": 4.856140350877193e-06,
      "logits/chosen": -1.8193047046661377,
      "logits/rejected": -1.8773219585418701,
      "logps/chosen": -155.37123107910156,
      "logps/rejected": -245.6790008544922,
      "loss": 0.4932,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.5351110696792603,
      "rewards/margins": 0.5927553176879883,
      "rewards/rejected": -0.05764435604214668,
      "step": 90
    },
    {
      "epoch": 0.37,
      "eval_logits/chosen": -1.8004387617111206,
      "eval_logits/rejected": -1.836446762084961,
      "eval_logps/chosen": -192.66908264160156,
      "eval_logps/rejected": -242.95899963378906,
      "eval_loss": 0.5072147846221924,
      "eval_rewards/accuracies": 0.8214285969734192,
      "eval_rewards/chosen": 0.5621528029441833,
      "eval_rewards/margins": 0.4941706657409668,
      "eval_rewards/rejected": 0.06798211485147476,
      "eval_runtime": 26.2422,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 90
    },
    {
      "epoch": 0.41,
      "grad_norm": 62.75928497314453,
      "learning_rate": 4.821052631578948e-06,
      "logits/chosen": -1.7835988998413086,
      "logits/rejected": -1.8885362148284912,
      "logps/chosen": -109.94624328613281,
      "logps/rejected": -232.5241241455078,
      "loss": 0.4252,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.5394761562347412,
      "rewards/margins": 0.787738025188446,
      "rewards/rejected": -0.248261958360672,
      "step": 100
    },
    {
      "epoch": 0.45,
      "grad_norm": 84.42039489746094,
      "learning_rate": 4.785964912280702e-06,
      "logits/chosen": -1.7081111669540405,
      "logits/rejected": -1.7244077920913696,
      "logps/chosen": -92.51998138427734,
      "logps/rejected": -135.5617218017578,
      "loss": 0.5077,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.6028972864151001,
      "rewards/margins": 0.511993408203125,
      "rewards/rejected": 0.09090389311313629,
      "step": 110
    },
    {
      "epoch": 0.49,
      "grad_norm": 26.61499786376953,
      "learning_rate": 4.754385964912281e-06,
      "logits/chosen": -1.9250166416168213,
      "logits/rejected": -1.9426968097686768,
      "logps/chosen": -152.0164337158203,
      "logps/rejected": -154.2359161376953,
      "loss": 0.4391,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.7938548922538757,
      "rewards/margins": 0.8368040919303894,
      "rewards/rejected": -0.042949117720127106,
      "step": 120
    },
    {
      "epoch": 0.49,
      "eval_logits/chosen": -1.8051280975341797,
      "eval_logits/rejected": -1.8413046598434448,
      "eval_logps/chosen": -191.2982177734375,
      "eval_logps/rejected": -242.81198120117188,
      "eval_loss": 0.4335751235485077,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.9734136462211609,
      "eval_rewards/margins": 0.8613229990005493,
      "eval_rewards/rejected": 0.11209066957235336,
      "eval_runtime": 26.2345,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 120
    },
    {
      "epoch": 0.53,
      "grad_norm": 90.6368637084961,
      "learning_rate": 4.7192982456140355e-06,
      "logits/chosen": -1.865144968032837,
      "logits/rejected": -1.9209120273590088,
      "logps/chosen": -149.78990173339844,
      "logps/rejected": -293.8749084472656,
      "loss": 0.4133,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.4554510712623596,
      "rewards/margins": 0.9285024404525757,
      "rewards/rejected": -0.47305139899253845,
      "step": 130
    },
    {
      "epoch": 0.57,
      "grad_norm": 6.8597941398620605,
      "learning_rate": 4.68421052631579e-06,
      "logits/chosen": -1.8089487552642822,
      "logits/rejected": -1.8955650329589844,
      "logps/chosen": -100.26250457763672,
      "logps/rejected": -209.1724090576172,
      "loss": 0.3414,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.9389357566833496,
      "rewards/margins": 1.521246314048767,
      "rewards/rejected": -0.5823107361793518,
      "step": 140
    },
    {
      "epoch": 0.61,
      "grad_norm": 79.9189453125,
      "learning_rate": 4.652631578947368e-06,
      "logits/chosen": -1.760371446609497,
      "logits/rejected": -1.8096182346343994,
      "logps/chosen": -193.62579345703125,
      "logps/rejected": -166.35800170898438,
      "loss": 0.3208,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.2600924968719482,
      "rewards/margins": 1.5845091342926025,
      "rewards/rejected": -0.32441645860671997,
      "step": 150
    },
    {
      "epoch": 0.61,
      "eval_logits/chosen": -1.8129692077636719,
      "eval_logits/rejected": -1.849239706993103,
      "eval_logps/chosen": -189.88929748535156,
      "eval_logps/rejected": -242.91104125976562,
      "eval_loss": 0.39326220750808716,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.3960883617401123,
      "eval_rewards/margins": 1.3137177228927612,
      "eval_rewards/rejected": 0.08237060159444809,
      "eval_runtime": 26.3372,
      "eval_samples_per_second": 1.063,
      "eval_steps_per_second": 1.063,
      "step": 150
    },
    {
      "epoch": 0.65,
      "grad_norm": 3.151426315307617,
      "learning_rate": 4.617543859649123e-06,
      "logits/chosen": -1.7185112237930298,
      "logits/rejected": -1.749427080154419,
      "logps/chosen": -178.8656463623047,
      "logps/rejected": -201.32858276367188,
      "loss": 0.4418,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 1.3884060382843018,
      "rewards/margins": 1.4542597532272339,
      "rewards/rejected": -0.06585375964641571,
      "step": 160
    },
    {
      "epoch": 0.69,
      "grad_norm": 37.79099655151367,
      "learning_rate": 4.582456140350878e-06,
      "logits/chosen": -1.730787992477417,
      "logits/rejected": -1.8099721670150757,
      "logps/chosen": -158.50277709960938,
      "logps/rejected": -185.48129272460938,
      "loss": 0.3912,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 1.4253065586090088,
      "rewards/margins": 1.2418479919433594,
      "rewards/rejected": 0.18345877528190613,
      "step": 170
    },
    {
      "epoch": 0.73,
      "grad_norm": 18.49593734741211,
      "learning_rate": 4.547368421052632e-06,
      "logits/chosen": -1.842390775680542,
      "logits/rejected": -1.8834030628204346,
      "logps/chosen": -125.84483337402344,
      "logps/rejected": -182.78854370117188,
      "loss": 0.3215,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.5408955812454224,
      "rewards/margins": 1.4677695035934448,
      "rewards/rejected": 0.07312598079442978,
      "step": 180
    },
    {
      "epoch": 0.73,
      "eval_logits/chosen": -1.8194406032562256,
      "eval_logits/rejected": -1.8561830520629883,
      "eval_logps/chosen": -188.3820037841797,
      "eval_logps/rejected": -243.13536071777344,
      "eval_loss": 0.3756321966648102,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.8482797145843506,
      "eval_rewards/margins": 1.8331985473632812,
      "eval_rewards/rejected": 0.015081183984875679,
      "eval_runtime": 26.2381,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 180
    },
    {
      "epoch": 0.78,
      "grad_norm": 106.56973266601562,
      "learning_rate": 4.512280701754386e-06,
      "logits/chosen": -1.8318649530410767,
      "logits/rejected": -1.8742272853851318,
      "logps/chosen": -197.66619873046875,
      "logps/rejected": -247.83560180664062,
      "loss": 0.3303,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.4889655113220215,
      "rewards/margins": 2.536695957183838,
      "rewards/rejected": -0.047730110585689545,
      "step": 190
    },
    {
      "epoch": 0.82,
      "grad_norm": 1.9826362133026123,
      "learning_rate": 4.47719298245614e-06,
      "logits/chosen": -1.7579128742218018,
      "logits/rejected": -1.7890180349349976,
      "logps/chosen": -129.37191772460938,
      "logps/rejected": -166.67636108398438,
      "loss": 0.3292,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.5646984577178955,
      "rewards/margins": 2.6955490112304688,
      "rewards/rejected": -0.13085035979747772,
      "step": 200
    },
    {
      "epoch": 0.86,
      "grad_norm": 6.777231693267822,
      "learning_rate": 4.442105263157896e-06,
      "logits/chosen": -1.7335002422332764,
      "logits/rejected": -1.8272641897201538,
      "logps/chosen": -114.22731018066406,
      "logps/rejected": -215.8419647216797,
      "loss": 0.0817,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.278646469116211,
      "rewards/margins": 4.009958267211914,
      "rewards/rejected": -0.7313117980957031,
      "step": 210
    },
    {
      "epoch": 0.86,
      "eval_logits/chosen": -1.826608419418335,
      "eval_logits/rejected": -1.8640671968460083,
      "eval_logps/chosen": -186.82992553710938,
      "eval_logps/rejected": -243.80209350585938,
      "eval_loss": 0.3834719657897949,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 2.3139071464538574,
      "eval_rewards/margins": 2.4988515377044678,
      "eval_rewards/rejected": -0.1849442571401596,
      "eval_runtime": 26.2309,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 210
    },
    {
      "epoch": 0.9,
      "grad_norm": 1.0230931043624878,
      "learning_rate": 4.40701754385965e-06,
      "logits/chosen": -1.7561031579971313,
      "logits/rejected": -1.8601219654083252,
      "logps/chosen": -134.11428833007812,
      "logps/rejected": -273.5272216796875,
      "loss": 0.1796,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.9340410232543945,
      "rewards/margins": 3.8275578022003174,
      "rewards/rejected": -1.8935168981552124,
      "step": 220
    },
    {
      "epoch": 0.94,
      "grad_norm": 54.65635299682617,
      "learning_rate": 4.371929824561404e-06,
      "logits/chosen": -1.8512550592422485,
      "logits/rejected": -1.900559663772583,
      "logps/chosen": -210.9802703857422,
      "logps/rejected": -240.60269165039062,
      "loss": 0.5004,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 1.5826597213745117,
      "rewards/margins": 2.925461530685425,
      "rewards/rejected": -1.342801809310913,
      "step": 230
    },
    {
      "epoch": 0.98,
      "grad_norm": 18.797134399414062,
      "learning_rate": 4.336842105263158e-06,
      "logits/chosen": -1.7781873941421509,
      "logits/rejected": -1.8780953884124756,
      "logps/chosen": -136.18722534179688,
      "logps/rejected": -257.26495361328125,
      "loss": 0.137,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.9192733764648438,
      "rewards/margins": 3.8388113975524902,
      "rewards/rejected": -1.9195384979248047,
      "step": 240
    },
    {
      "epoch": 0.98,
      "eval_logits/chosen": -1.83432936668396,
      "eval_logits/rejected": -1.8722076416015625,
      "eval_logps/chosen": -185.8831329345703,
      "eval_logps/rejected": -244.8594207763672,
      "eval_loss": 0.41323208808898926,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.5979421138763428,
      "eval_rewards/margins": 3.1000888347625732,
      "eval_rewards/rejected": -0.5021467208862305,
      "eval_runtime": 26.2323,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 240
    },
    {
      "epoch": 1.02,
      "grad_norm": 132.51947021484375,
      "learning_rate": 4.301754385964912e-06,
      "logits/chosen": -1.8787914514541626,
      "logits/rejected": -1.9520515203475952,
      "logps/chosen": -186.46717834472656,
      "logps/rejected": -250.008056640625,
      "loss": 0.3531,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.192800283432007,
      "rewards/margins": 3.485269069671631,
      "rewards/rejected": -1.2924686670303345,
      "step": 250
    },
    {
      "epoch": 1.06,
      "grad_norm": 6.529983043670654,
      "learning_rate": 4.266666666666668e-06,
      "logits/chosen": -1.7029941082000732,
      "logits/rejected": -1.7871618270874023,
      "logps/chosen": -128.5960235595703,
      "logps/rejected": -209.27603149414062,
      "loss": 0.0883,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.7405991554260254,
      "rewards/margins": 5.148273468017578,
      "rewards/rejected": -2.4076738357543945,
      "step": 260
    },
    {
      "epoch": 1.1,
      "grad_norm": 0.27644646167755127,
      "learning_rate": 4.2315789473684215e-06,
      "logits/chosen": -1.7499496936798096,
      "logits/rejected": -1.8194109201431274,
      "logps/chosen": -119.34013366699219,
      "logps/rejected": -230.69229125976562,
      "loss": 0.0997,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 4.146358013153076,
      "rewards/margins": 5.72900390625,
      "rewards/rejected": -1.5826464891433716,
      "step": 270
    },
    {
      "epoch": 1.1,
      "eval_logits/chosen": -1.842972755432129,
      "eval_logits/rejected": -1.8816190958023071,
      "eval_logps/chosen": -185.4148406982422,
      "eval_logps/rejected": -246.53672790527344,
      "eval_loss": 0.46570485830307007,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.7384250164031982,
      "eval_rewards/margins": 3.743762969970703,
      "eval_rewards/rejected": -1.0053375959396362,
      "eval_runtime": 26.2348,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 270
    }
  ],
  "logging_steps": 10,
  "max_steps": 1470,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 90,
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}