checkpoint-1440/trainer_state.json

{
  "best_metric": 0.09226309508085251,
  "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-1440",
  "epoch": 5.877551020408164,
  "eval_steps": 30,
  "global_step": 1440,
  "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
    },
    {
      "epoch": 1.14,
      "grad_norm": 31.16436195373535,
      "learning_rate": 4.196491228070176e-06,
      "logits/chosen": -1.8344509601593018,
      "logits/rejected": -1.919553518295288,
      "logps/chosen": -131.51156616210938,
      "logps/rejected": -246.3345184326172,
      "loss": 0.138,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.982485771179199,
      "rewards/margins": 6.261662483215332,
      "rewards/rejected": -3.2791759967803955,
      "step": 280
    },
    {
      "epoch": 1.18,
      "grad_norm": 0.08183232694864273,
      "learning_rate": 4.16140350877193e-06,
      "logits/chosen": -1.847378134727478,
      "logits/rejected": -1.8774774074554443,
      "logps/chosen": -127.52336120605469,
      "logps/rejected": -157.3633270263672,
      "loss": 0.3883,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.7564169764518738,
      "rewards/margins": 2.541962146759033,
      "rewards/rejected": -1.7855453491210938,
      "step": 290
    },
    {
      "epoch": 1.22,
      "grad_norm": 10.61989974975586,
      "learning_rate": 4.126315789473685e-06,
      "logits/chosen": -1.7770633697509766,
      "logits/rejected": -1.8337452411651611,
      "logps/chosen": -112.7281265258789,
      "logps/rejected": -195.87020874023438,
      "loss": 0.0432,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.176725387573242,
      "rewards/margins": 6.761993408203125,
      "rewards/rejected": -3.5852675437927246,
      "step": 300
    },
    {
      "epoch": 1.22,
      "eval_logits/chosen": -1.849483609199524,
      "eval_logits/rejected": -1.888432502746582,
      "eval_logps/chosen": -185.52931213378906,
      "eval_logps/rejected": -248.10926818847656,
      "eval_loss": 0.501145601272583,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.704090118408203,
      "eval_rewards/margins": 4.181191921234131,
      "eval_rewards/rejected": -1.477101445198059,
      "eval_runtime": 26.2437,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 300
    },
    {
      "epoch": 1.27,
      "grad_norm": 327.02899169921875,
      "learning_rate": 4.091228070175439e-06,
      "logits/chosen": -1.8802036046981812,
      "logits/rejected": -1.9435373544692993,
      "logps/chosen": -164.2046661376953,
      "logps/rejected": -269.22235107421875,
      "loss": 0.4088,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 3.3527073860168457,
      "rewards/margins": 5.9436140060424805,
      "rewards/rejected": -2.5909066200256348,
      "step": 310
    },
    {
      "epoch": 1.31,
      "grad_norm": 6.565308094024658,
      "learning_rate": 4.056140350877193e-06,
      "logits/chosen": -1.9250032901763916,
      "logits/rejected": -1.9860813617706299,
      "logps/chosen": -197.4870147705078,
      "logps/rejected": -207.25009155273438,
      "loss": 0.4969,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.32615354657173157,
      "rewards/margins": 2.2872400283813477,
      "rewards/rejected": -1.9610865116119385,
      "step": 320
    },
    {
      "epoch": 1.35,
      "grad_norm": 33.34074020385742,
      "learning_rate": 4.021052631578948e-06,
      "logits/chosen": -1.8470687866210938,
      "logits/rejected": -1.8675216436386108,
      "logps/chosen": -117.5759506225586,
      "logps/rejected": -151.83091735839844,
      "loss": 0.1819,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.878262758255005,
      "rewards/margins": 5.000906467437744,
      "rewards/rejected": -2.12264347076416,
      "step": 330
    },
    {
      "epoch": 1.35,
      "eval_logits/chosen": -1.8486573696136475,
      "eval_logits/rejected": -1.8877869844436646,
      "eval_logps/chosen": -185.54177856445312,
      "eval_logps/rejected": -249.2687530517578,
      "eval_loss": 0.478495329618454,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.7003517150878906,
      "eval_rewards/margins": 4.525294780731201,
      "eval_rewards/rejected": -1.8249431848526,
      "eval_runtime": 26.2362,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 330
    },
    {
      "epoch": 1.39,
      "grad_norm": 0.6777727603912354,
      "learning_rate": 3.985964912280702e-06,
      "logits/chosen": -1.8584426641464233,
      "logits/rejected": -1.884598970413208,
      "logps/chosen": -78.69309997558594,
      "logps/rejected": -126.11468505859375,
      "loss": 0.3154,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.5596954822540283,
      "rewards/margins": 4.089261054992676,
      "rewards/rejected": -1.5295660495758057,
      "step": 340
    },
    {
      "epoch": 1.43,
      "grad_norm": 45.252723693847656,
      "learning_rate": 3.950877192982457e-06,
      "logits/chosen": -1.8422210216522217,
      "logits/rejected": -1.8741470575332642,
      "logps/chosen": -198.17526245117188,
      "logps/rejected": -179.26759338378906,
      "loss": 0.2128,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.886838912963867,
      "rewards/margins": 4.228543758392334,
      "rewards/rejected": -1.3417048454284668,
      "step": 350
    },
    {
      "epoch": 1.47,
      "grad_norm": 0.0007244591251946986,
      "learning_rate": 3.9157894736842104e-06,
      "logits/chosen": -1.8951257467269897,
      "logits/rejected": -1.9684120416641235,
      "logps/chosen": -123.7894515991211,
      "logps/rejected": -278.1185607910156,
      "loss": 0.0169,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.9708969593048096,
      "rewards/margins": 9.946398735046387,
      "rewards/rejected": -5.975502014160156,
      "step": 360
    },
    {
      "epoch": 1.47,
      "eval_logits/chosen": -1.8510137796401978,
      "eval_logits/rejected": -1.8906936645507812,
      "eval_logps/chosen": -185.66189575195312,
      "eval_logps/rejected": -250.37806701660156,
      "eval_loss": 0.4872412383556366,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.664308547973633,
      "eval_rewards/margins": 4.822035789489746,
      "eval_rewards/rejected": -2.157727003097534,
      "eval_runtime": 26.2354,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 360
    },
    {
      "epoch": 1.51,
      "grad_norm": 0.001088384771719575,
      "learning_rate": 3.880701754385965e-06,
      "logits/chosen": -1.859140157699585,
      "logits/rejected": -1.903721809387207,
      "logps/chosen": -171.89047241210938,
      "logps/rejected": -257.6295471191406,
      "loss": 0.237,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.257797956466675,
      "rewards/margins": 7.423349857330322,
      "rewards/rejected": -5.165551662445068,
      "step": 370
    },
    {
      "epoch": 1.55,
      "grad_norm": 13.049817085266113,
      "learning_rate": 3.84561403508772e-06,
      "logits/chosen": -1.850102186203003,
      "logits/rejected": -1.922501564025879,
      "logps/chosen": -145.50692749023438,
      "logps/rejected": -221.7799530029297,
      "loss": 0.2906,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.9433931112289429,
      "rewards/margins": 6.384174346923828,
      "rewards/rejected": -4.440781593322754,
      "step": 380
    },
    {
      "epoch": 1.59,
      "grad_norm": 0.006105802953243256,
      "learning_rate": 3.810526315789474e-06,
      "logits/chosen": -1.9277305603027344,
      "logits/rejected": -1.9736566543579102,
      "logps/chosen": -140.36146545410156,
      "logps/rejected": -188.66836547851562,
      "loss": 0.235,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.6781504154205322,
      "rewards/margins": 5.825950622558594,
      "rewards/rejected": -3.1477997303009033,
      "step": 390
    },
    {
      "epoch": 1.59,
      "eval_logits/chosen": -1.8531960248947144,
      "eval_logits/rejected": -1.8930160999298096,
      "eval_logps/chosen": -185.68800354003906,
      "eval_logps/rejected": -251.1302032470703,
      "eval_loss": 0.48860952258110046,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.656480550765991,
      "eval_rewards/margins": 5.039853572845459,
      "eval_rewards/rejected": -2.3833730220794678,
      "eval_runtime": 26.2308,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 390
    },
    {
      "epoch": 1.63,
      "grad_norm": 0.00030716226319782436,
      "learning_rate": 3.7754385964912284e-06,
      "logits/chosen": -1.8817369937896729,
      "logits/rejected": -1.9124549627304077,
      "logps/chosen": -171.40383911132812,
      "logps/rejected": -198.42446899414062,
      "loss": 0.4322,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 1.219795823097229,
      "rewards/margins": 6.402298927307129,
      "rewards/rejected": -5.182502269744873,
      "step": 400
    },
    {
      "epoch": 1.67,
      "grad_norm": 0.0007249619229696691,
      "learning_rate": 3.7403508771929827e-06,
      "logits/chosen": -1.8018410205841064,
      "logits/rejected": -1.869511365890503,
      "logps/chosen": -168.9734649658203,
      "logps/rejected": -223.9024658203125,
      "loss": 0.9332,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 2.475252151489258,
      "rewards/margins": 5.069212913513184,
      "rewards/rejected": -2.593961238861084,
      "step": 410
    },
    {
      "epoch": 1.71,
      "grad_norm": 46.79059982299805,
      "learning_rate": 3.7052631578947374e-06,
      "logits/chosen": -1.774857759475708,
      "logits/rejected": -1.8125171661376953,
      "logps/chosen": -140.12551879882812,
      "logps/rejected": -208.7825927734375,
      "loss": 0.7551,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.244128704071045,
      "rewards/margins": 6.2417426109313965,
      "rewards/rejected": -3.997614622116089,
      "step": 420
    },
    {
      "epoch": 1.71,
      "eval_logits/chosen": -1.8527201414108276,
      "eval_logits/rejected": -1.892095923423767,
      "eval_logps/chosen": -185.46646118164062,
      "eval_logps/rejected": -251.00816345214844,
      "eval_loss": 0.438032329082489,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.7229368686676025,
      "eval_rewards/margins": 5.069703102111816,
      "eval_rewards/rejected": -2.346766233444214,
      "eval_runtime": 26.2319,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 420
    },
    {
      "epoch": 1.76,
      "grad_norm": 0.1679692417383194,
      "learning_rate": 3.6701754385964917e-06,
      "logits/chosen": -1.9766556024551392,
      "logits/rejected": -1.967247724533081,
      "logps/chosen": -181.25978088378906,
      "logps/rejected": -174.0313262939453,
      "loss": 0.5746,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.037475347518921,
      "rewards/margins": 4.661454200744629,
      "rewards/rejected": -2.623978853225708,
      "step": 430
    },
    {
      "epoch": 1.8,
      "grad_norm": 0.028290700167417526,
      "learning_rate": 3.635087719298246e-06,
      "logits/chosen": -1.7324810028076172,
      "logits/rejected": -1.8381952047348022,
      "logps/chosen": -118.8620376586914,
      "logps/rejected": -249.5945587158203,
      "loss": 0.1919,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.84242582321167,
      "rewards/margins": 8.061222076416016,
      "rewards/rejected": -5.218796730041504,
      "step": 440
    },
    {
      "epoch": 1.84,
      "grad_norm": 3.435687303543091,
      "learning_rate": 3.6000000000000003e-06,
      "logits/chosen": -1.788468599319458,
      "logits/rejected": -1.8786054849624634,
      "logps/chosen": -127.3366928100586,
      "logps/rejected": -223.4063262939453,
      "loss": 0.134,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.9258162379264832,
      "rewards/margins": 6.352460861206055,
      "rewards/rejected": -5.426644802093506,
      "step": 450
    },
    {
      "epoch": 1.84,
      "eval_logits/chosen": -1.8530861139297485,
      "eval_logits/rejected": -1.8925215005874634,
      "eval_logps/chosen": -185.6543426513672,
      "eval_logps/rejected": -251.7077178955078,
      "eval_loss": 0.4383006691932678,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.6665782928466797,
      "eval_rewards/margins": 5.223209857940674,
      "eval_rewards/rejected": -2.556631565093994,
      "eval_runtime": 26.2389,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 450
    },
    {
      "epoch": 1.88,
      "grad_norm": 0.2778886556625366,
      "learning_rate": 3.5649122807017545e-06,
      "logits/chosen": -1.8211870193481445,
      "logits/rejected": -1.8562345504760742,
      "logps/chosen": -90.34883117675781,
      "logps/rejected": -171.23648071289062,
      "loss": 0.1284,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.756679058074951,
      "rewards/margins": 7.474773406982422,
      "rewards/rejected": -3.7180938720703125,
      "step": 460
    },
    {
      "epoch": 1.92,
      "grad_norm": 0.00020280799071770161,
      "learning_rate": 3.5298245614035093e-06,
      "logits/chosen": -1.7977367639541626,
      "logits/rejected": -1.8836923837661743,
      "logps/chosen": -180.67724609375,
      "logps/rejected": -215.39608764648438,
      "loss": 0.5069,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.3001179695129395,
      "rewards/margins": 7.598956108093262,
      "rewards/rejected": -5.298838138580322,
      "step": 470
    },
    {
      "epoch": 1.96,
      "grad_norm": 0.0038736851420253515,
      "learning_rate": 3.4947368421052635e-06,
      "logits/chosen": -1.8841644525527954,
      "logits/rejected": -1.986575722694397,
      "logps/chosen": -189.42210388183594,
      "logps/rejected": -368.7510070800781,
      "loss": 0.0662,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.894408702850342,
      "rewards/margins": 9.953689575195312,
      "rewards/rejected": -7.0592803955078125,
      "step": 480
    },
    {
      "epoch": 1.96,
      "eval_logits/chosen": -1.8568991422653198,
      "eval_logits/rejected": -1.896437406539917,
      "eval_logps/chosen": -186.01429748535156,
      "eval_logps/rejected": -252.9164276123047,
      "eval_loss": 0.44481489062309265,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.55859112739563,
      "eval_rewards/margins": 5.477838039398193,
      "eval_rewards/rejected": -2.9192464351654053,
      "eval_runtime": 26.2387,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 480
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.00020024628611281514,
      "learning_rate": 3.459649122807018e-06,
      "logits/chosen": -1.7868010997772217,
      "logits/rejected": -1.8775966167449951,
      "logps/chosen": -168.01686096191406,
      "logps/rejected": -268.81329345703125,
      "loss": 0.2127,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 4.091753959655762,
      "rewards/margins": 10.252196311950684,
      "rewards/rejected": -6.160443305969238,
      "step": 490
    },
    {
      "epoch": 2.04,
      "grad_norm": 2.933464288711548,
      "learning_rate": 3.424561403508772e-06,
      "logits/chosen": -1.8647997379302979,
      "logits/rejected": -1.8972787857055664,
      "logps/chosen": -206.39712524414062,
      "logps/rejected": -255.3307647705078,
      "loss": 0.1551,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 3.2640254497528076,
      "rewards/margins": 8.8242826461792,
      "rewards/rejected": -5.560257434844971,
      "step": 500
    },
    {
      "epoch": 2.08,
      "grad_norm": 0.088715098798275,
      "learning_rate": 3.3894736842105264e-06,
      "logits/chosen": -1.8153291940689087,
      "logits/rejected": -1.907467246055603,
      "logps/chosen": -114.2393569946289,
      "logps/rejected": -245.9196319580078,
      "loss": 0.1093,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.2050167322158813,
      "rewards/margins": 7.797829627990723,
      "rewards/rejected": -6.592813014984131,
      "step": 510
    },
    {
      "epoch": 2.08,
      "eval_logits/chosen": -1.8561475276947021,
      "eval_logits/rejected": -1.8955259323120117,
      "eval_logps/chosen": -186.13938903808594,
      "eval_logps/rejected": -253.4276580810547,
      "eval_loss": 0.4262312054634094,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.521066904067993,
      "eval_rewards/margins": 5.59368371963501,
      "eval_rewards/rejected": -3.0726168155670166,
      "eval_runtime": 26.2357,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 510
    },
    {
      "epoch": 2.12,
      "grad_norm": 0.0037095127627253532,
      "learning_rate": 3.354385964912281e-06,
      "logits/chosen": -1.904370665550232,
      "logits/rejected": -1.9216587543487549,
      "logps/chosen": -173.86373901367188,
      "logps/rejected": -173.6374053955078,
      "loss": 0.1557,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.4744720458984375,
      "rewards/margins": 5.560853004455566,
      "rewards/rejected": -3.0863804817199707,
      "step": 520
    },
    {
      "epoch": 2.16,
      "grad_norm": 1.6992926532566344e-07,
      "learning_rate": 3.3192982456140354e-06,
      "logits/chosen": -1.7665284872055054,
      "logits/rejected": -1.859400987625122,
      "logps/chosen": -85.06950378417969,
      "logps/rejected": -232.3406524658203,
      "loss": 0.0056,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.268359661102295,
      "rewards/margins": 9.698001861572266,
      "rewards/rejected": -6.4296417236328125,
      "step": 530
    },
    {
      "epoch": 2.2,
      "grad_norm": 0.00015174441796261817,
      "learning_rate": 3.2842105263157897e-06,
      "logits/chosen": -1.7254387140274048,
      "logits/rejected": -1.8300654888153076,
      "logps/chosen": -126.21636962890625,
      "logps/rejected": -236.91250610351562,
      "loss": 0.1557,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.9851126670837402,
      "rewards/margins": 9.359209060668945,
      "rewards/rejected": -6.374096870422363,
      "step": 540
    },
    {
      "epoch": 2.2,
      "eval_logits/chosen": -1.85663902759552,
      "eval_logits/rejected": -1.8965102434158325,
      "eval_logps/chosen": -186.6448974609375,
      "eval_logps/rejected": -254.58482360839844,
      "eval_loss": 0.4263899326324463,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.369408369064331,
      "eval_rewards/margins": 5.789170265197754,
      "eval_rewards/rejected": -3.4197616577148438,
      "eval_runtime": 26.2541,
      "eval_samples_per_second": 1.066,
      "eval_steps_per_second": 1.066,
      "step": 540
    },
    {
      "epoch": 2.24,
      "grad_norm": 7.969775676727295,
      "learning_rate": 3.249122807017544e-06,
      "logits/chosen": -1.8786284923553467,
      "logits/rejected": -1.9631255865097046,
      "logps/chosen": -186.79092407226562,
      "logps/rejected": -295.82598876953125,
      "loss": 0.0625,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.5664806365966797,
      "rewards/margins": 6.786992073059082,
      "rewards/rejected": -6.220510959625244,
      "step": 550
    },
    {
      "epoch": 2.29,
      "grad_norm": 41.887001037597656,
      "learning_rate": 3.2140350877192982e-06,
      "logits/chosen": -1.8324702978134155,
      "logits/rejected": -1.8570420742034912,
      "logps/chosen": -198.67893981933594,
      "logps/rejected": -203.76016235351562,
      "loss": 0.389,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.470458984375,
      "rewards/margins": 8.410517692565918,
      "rewards/rejected": -5.940059185028076,
      "step": 560
    },
    {
      "epoch": 2.33,
      "grad_norm": 0.022604621946811676,
      "learning_rate": 3.178947368421053e-06,
      "logits/chosen": -1.7625017166137695,
      "logits/rejected": -1.823395013809204,
      "logps/chosen": -145.4344024658203,
      "logps/rejected": -202.81698608398438,
      "loss": 0.0962,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 3.2409331798553467,
      "rewards/margins": 7.634140968322754,
      "rewards/rejected": -4.39320707321167,
      "step": 570
    },
    {
      "epoch": 2.33,
      "eval_logits/chosen": -1.8582419157028198,
      "eval_logits/rejected": -1.897825837135315,
      "eval_logps/chosen": -186.9964141845703,
      "eval_logps/rejected": -255.54441833496094,
      "eval_loss": 0.41822776198387146,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.2639553546905518,
      "eval_rewards/margins": 5.971593379974365,
      "eval_rewards/rejected": -3.70763897895813,
      "eval_runtime": 26.2352,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 570
    },
    {
      "epoch": 2.37,
      "grad_norm": 0.0053796968422830105,
      "learning_rate": 3.1438596491228072e-06,
      "logits/chosen": -1.8564494848251343,
      "logits/rejected": -1.9009406566619873,
      "logps/chosen": -153.97903442382812,
      "logps/rejected": -182.94284057617188,
      "loss": 0.5733,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 2.0778138637542725,
      "rewards/margins": 5.466890335083008,
      "rewards/rejected": -3.389075756072998,
      "step": 580
    },
    {
      "epoch": 2.41,
      "grad_norm": 5.523962020874023,
      "learning_rate": 3.1087719298245615e-06,
      "logits/chosen": -1.8511816263198853,
      "logits/rejected": -1.853109359741211,
      "logps/chosen": -182.23892211914062,
      "logps/rejected": -120.5928955078125,
      "loss": 0.4958,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 3.659731388092041,
      "rewards/margins": 4.932890892028809,
      "rewards/rejected": -1.2731596231460571,
      "step": 590
    },
    {
      "epoch": 2.45,
      "grad_norm": 0.0013734139502048492,
      "learning_rate": 3.0736842105263158e-06,
      "logits/chosen": -1.7618497610092163,
      "logits/rejected": -1.8786863088607788,
      "logps/chosen": -114.43098449707031,
      "logps/rejected": -229.5573272705078,
      "loss": 0.0437,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.8845677375793457,
      "rewards/margins": 9.519918441772461,
      "rewards/rejected": -6.635350227355957,
      "step": 600
    },
    {
      "epoch": 2.45,
      "eval_logits/chosen": -1.8533600568771362,
      "eval_logits/rejected": -1.8932533264160156,
      "eval_logps/chosen": -187.0037384033203,
      "eval_logps/rejected": -255.77125549316406,
      "eval_loss": 0.3824370801448822,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 2.2617576122283936,
      "eval_rewards/margins": 6.037452220916748,
      "eval_rewards/rejected": -3.7756943702697754,
      "eval_runtime": 26.2326,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 600
    },
    {
      "epoch": 2.49,
      "grad_norm": 3.847404956817627,
      "learning_rate": 3.03859649122807e-06,
      "logits/chosen": -1.9108350276947021,
      "logits/rejected": -1.9136813879013062,
      "logps/chosen": -147.76666259765625,
      "logps/rejected": -151.54275512695312,
      "loss": 0.1939,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.7343037128448486,
      "rewards/margins": 7.082615852355957,
      "rewards/rejected": -4.348311424255371,
      "step": 610
    },
    {
      "epoch": 2.53,
      "grad_norm": 6.025547172328061e-09,
      "learning_rate": 3.003508771929825e-06,
      "logits/chosen": -1.9030263423919678,
      "logits/rejected": -1.964647889137268,
      "logps/chosen": -157.2318115234375,
      "logps/rejected": -335.1673583984375,
      "loss": 0.1472,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.17058677971363068,
      "rewards/margins": 10.464014053344727,
      "rewards/rejected": -10.293428421020508,
      "step": 620
    },
    {
      "epoch": 2.57,
      "grad_norm": 0.7363545298576355,
      "learning_rate": 2.9684210526315795e-06,
      "logits/chosen": -1.8167072534561157,
      "logits/rejected": -1.9096784591674805,
      "logps/chosen": -128.2840118408203,
      "logps/rejected": -278.52862548828125,
      "loss": 0.0278,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.6698997020721436,
      "rewards/margins": 10.468381881713867,
      "rewards/rejected": -7.798482418060303,
      "step": 630
    },
    {
      "epoch": 2.57,
      "eval_logits/chosen": -1.8536275625228882,
      "eval_logits/rejected": -1.8932437896728516,
      "eval_logps/chosen": -186.7086181640625,
      "eval_logps/rejected": -255.70458984375,
      "eval_loss": 0.3570733964443207,
      "eval_rewards/accuracies": 0.8571428656578064,
      "eval_rewards/chosen": 2.3502955436706543,
      "eval_rewards/margins": 6.105983257293701,
      "eval_rewards/rejected": -3.755687713623047,
      "eval_runtime": 26.2519,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 630
    },
    {
      "epoch": 2.61,
      "grad_norm": 19.793140411376953,
      "learning_rate": 2.9333333333333338e-06,
      "logits/chosen": -1.9250261783599854,
      "logits/rejected": -1.969919204711914,
      "logps/chosen": -169.236572265625,
      "logps/rejected": -216.4037322998047,
      "loss": 0.3144,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 1.5923322439193726,
      "rewards/margins": 8.081656455993652,
      "rewards/rejected": -6.48932409286499,
      "step": 640
    },
    {
      "epoch": 2.65,
      "grad_norm": 0.0,
      "learning_rate": 2.898245614035088e-06,
      "logits/chosen": -1.840488076210022,
      "logits/rejected": -1.9099540710449219,
      "logps/chosen": -188.36514282226562,
      "logps/rejected": -266.2427978515625,
      "loss": 0.0455,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.294522762298584,
      "rewards/margins": 9.861984252929688,
      "rewards/rejected": -6.5674614906311035,
      "step": 650
    },
    {
      "epoch": 2.69,
      "grad_norm": 350.0583801269531,
      "learning_rate": 2.8631578947368423e-06,
      "logits/chosen": -1.875605583190918,
      "logits/rejected": -1.9353086948394775,
      "logps/chosen": -128.51104736328125,
      "logps/rejected": -191.3477325439453,
      "loss": 0.2399,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 3.1921534538269043,
      "rewards/margins": 7.449667453765869,
      "rewards/rejected": -4.257513999938965,
      "step": 660
    },
    {
      "epoch": 2.69,
      "eval_logits/chosen": -1.851164698600769,
      "eval_logits/rejected": -1.8909271955490112,
      "eval_logps/chosen": -186.86781311035156,
      "eval_logps/rejected": -256.2709655761719,
      "eval_loss": 0.3312711715698242,
      "eval_rewards/accuracies": 0.8571428656578064,
      "eval_rewards/chosen": 2.3025403022766113,
      "eval_rewards/margins": 6.228146553039551,
      "eval_rewards/rejected": -3.9256062507629395,
      "eval_runtime": 26.2337,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 660
    },
    {
      "epoch": 2.73,
      "grad_norm": 0.0,
      "learning_rate": 2.828070175438597e-06,
      "logits/chosen": -1.7718509435653687,
      "logits/rejected": -1.8662782907485962,
      "logps/chosen": -100.0200424194336,
      "logps/rejected": -176.11337280273438,
      "loss": 0.0636,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.2194771766662598,
      "rewards/margins": 7.908571720123291,
      "rewards/rejected": -5.689094543457031,
      "step": 670
    },
    {
      "epoch": 2.78,
      "grad_norm": 9.849269554251805e-05,
      "learning_rate": 2.7929824561403513e-06,
      "logits/chosen": -1.8379380702972412,
      "logits/rejected": -1.9032844305038452,
      "logps/chosen": -147.157958984375,
      "logps/rejected": -311.342529296875,
      "loss": 0.0081,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.7090831995010376,
      "rewards/margins": 9.63591480255127,
      "rewards/rejected": -7.926831245422363,
      "step": 680
    },
    {
      "epoch": 2.82,
      "grad_norm": 5.133792877197266,
      "learning_rate": 2.7578947368421056e-06,
      "logits/chosen": -1.8350670337677002,
      "logits/rejected": -1.9039970636367798,
      "logps/chosen": -160.75314331054688,
      "logps/rejected": -247.68008422851562,
      "loss": 0.039,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 4.146931171417236,
      "rewards/margins": 11.529654502868652,
      "rewards/rejected": -7.382723808288574,
      "step": 690
    },
    {
      "epoch": 2.82,
      "eval_logits/chosen": -1.8510338068008423,
      "eval_logits/rejected": -1.8906004428863525,
      "eval_logps/chosen": -187.16348266601562,
      "eval_logps/rejected": -257.069091796875,
      "eval_loss": 0.3131271302700043,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.213836193084717,
      "eval_rewards/margins": 6.378880977630615,
      "eval_rewards/rejected": -4.165044784545898,
      "eval_runtime": 26.2333,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 690
    },
    {
      "epoch": 2.86,
      "grad_norm": 0.001812128466553986,
      "learning_rate": 2.72280701754386e-06,
      "logits/chosen": -1.888392448425293,
      "logits/rejected": -1.9376449584960938,
      "logps/chosen": -123.4749755859375,
      "logps/rejected": -221.9653778076172,
      "loss": 0.1581,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 3.1250205039978027,
      "rewards/margins": 9.35824966430664,
      "rewards/rejected": -6.23322868347168,
      "step": 700
    },
    {
      "epoch": 2.9,
      "grad_norm": 0.0007696045213378966,
      "learning_rate": 2.687719298245614e-06,
      "logits/chosen": -1.7869523763656616,
      "logits/rejected": -1.9070037603378296,
      "logps/chosen": -116.8504409790039,
      "logps/rejected": -300.51806640625,
      "loss": 0.1832,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.3490161895751953,
      "rewards/margins": 10.493500709533691,
      "rewards/rejected": -8.144485473632812,
      "step": 710
    },
    {
      "epoch": 2.94,
      "grad_norm": 83.29313659667969,
      "learning_rate": 2.652631578947369e-06,
      "logits/chosen": -1.8838169574737549,
      "logits/rejected": -1.9118610620498657,
      "logps/chosen": -104.75813293457031,
      "logps/rejected": -167.18580627441406,
      "loss": 0.3389,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.7628464698791504,
      "rewards/margins": 6.317173957824707,
      "rewards/rejected": -3.5543277263641357,
      "step": 720
    },
    {
      "epoch": 2.94,
      "eval_logits/chosen": -1.8480366468429565,
      "eval_logits/rejected": -1.8872734308242798,
      "eval_logps/chosen": -187.00791931152344,
      "eval_logps/rejected": -257.2390441894531,
      "eval_loss": 0.27630648016929626,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.2605104446411133,
      "eval_rewards/margins": 6.476548194885254,
      "eval_rewards/rejected": -4.216037750244141,
      "eval_runtime": 26.23,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 720
    },
    {
      "epoch": 2.98,
      "grad_norm": 0.030682619661092758,
      "learning_rate": 2.617543859649123e-06,
      "logits/chosen": -1.8667634725570679,
      "logits/rejected": -1.8890842199325562,
      "logps/chosen": -197.29690551757812,
      "logps/rejected": -219.12332153320312,
      "loss": 0.617,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": -0.587865948677063,
      "rewards/margins": 5.496161937713623,
      "rewards/rejected": -6.0840277671813965,
      "step": 730
    },
    {
      "epoch": 3.02,
      "grad_norm": 0.005717798601835966,
      "learning_rate": 2.5824561403508774e-06,
      "logits/chosen": -1.8484325408935547,
      "logits/rejected": -1.8692123889923096,
      "logps/chosen": -102.5673599243164,
      "logps/rejected": -141.8745880126953,
      "loss": 0.044,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.209364175796509,
      "rewards/margins": 8.334315299987793,
      "rewards/rejected": -6.124951362609863,
      "step": 740
    },
    {
      "epoch": 3.06,
      "grad_norm": 0.0006890746881254017,
      "learning_rate": 2.5473684210526317e-06,
      "logits/chosen": -1.865378737449646,
      "logits/rejected": -1.9292535781860352,
      "logps/chosen": -103.72840881347656,
      "logps/rejected": -203.67276000976562,
      "loss": 0.0154,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.9861044883728027,
      "rewards/margins": 8.630514144897461,
      "rewards/rejected": -4.644408702850342,
      "step": 750
    },
    {
      "epoch": 3.06,
      "eval_logits/chosen": -1.846961259841919,
      "eval_logits/rejected": -1.886168122291565,
      "eval_logps/chosen": -187.0342254638672,
      "eval_logps/rejected": -257.52471923828125,
      "eval_loss": 0.2703961730003357,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.252613067626953,
      "eval_rewards/margins": 6.554351806640625,
      "eval_rewards/rejected": -4.301738739013672,
      "eval_runtime": 26.236,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 750
    },
    {
      "epoch": 3.1,
      "grad_norm": 0.0,
      "learning_rate": 2.512280701754386e-06,
      "logits/chosen": -1.921618103981018,
      "logits/rejected": -1.9239141941070557,
      "logps/chosen": -248.826416015625,
      "logps/rejected": -182.67323303222656,
      "loss": 0.5385,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": -0.7813287377357483,
      "rewards/margins": 4.614243507385254,
      "rewards/rejected": -5.395572185516357,
      "step": 760
    },
    {
      "epoch": 3.14,
      "grad_norm": 20.82038688659668,
      "learning_rate": 2.4771929824561407e-06,
      "logits/chosen": -1.8723421096801758,
      "logits/rejected": -1.9096221923828125,
      "logps/chosen": -144.3488311767578,
      "logps/rejected": -200.52149963378906,
      "loss": 0.0473,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.6265076994895935,
      "rewards/margins": 5.623400688171387,
      "rewards/rejected": -4.996891975402832,
      "step": 770
    },
    {
      "epoch": 3.18,
      "grad_norm": 0.0,
      "learning_rate": 2.442105263157895e-06,
      "logits/chosen": -1.967357873916626,
      "logits/rejected": -1.995941162109375,
      "logps/chosen": -169.40643310546875,
      "logps/rejected": -225.83267211914062,
      "loss": 0.021,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.36498761177063,
      "rewards/margins": 8.846434593200684,
      "rewards/rejected": -5.481447219848633,
      "step": 780
    },
    {
      "epoch": 3.18,
      "eval_logits/chosen": -1.84483802318573,
      "eval_logits/rejected": -1.8837740421295166,
      "eval_logps/chosen": -187.02699279785156,
      "eval_logps/rejected": -257.6650085449219,
      "eval_loss": 0.24223719537258148,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.254775285720825,
      "eval_rewards/margins": 6.598587989807129,
      "eval_rewards/rejected": -4.343813419342041,
      "eval_runtime": 26.2524,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 780
    },
    {
      "epoch": 3.22,
      "grad_norm": 0.3396748900413513,
      "learning_rate": 2.4070175438596493e-06,
      "logits/chosen": -1.795336127281189,
      "logits/rejected": -1.9104626178741455,
      "logps/chosen": -127.96372985839844,
      "logps/rejected": -321.76751708984375,
      "loss": 0.0057,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.3614065647125244,
      "rewards/margins": 11.285585403442383,
      "rewards/rejected": -9.924179077148438,
      "step": 790
    },
    {
      "epoch": 3.27,
      "grad_norm": 89.25263214111328,
      "learning_rate": 2.3719298245614036e-06,
      "logits/chosen": -1.7844104766845703,
      "logits/rejected": -1.835338830947876,
      "logps/chosen": -138.7672119140625,
      "logps/rejected": -212.5260772705078,
      "loss": 0.099,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 3.200477123260498,
      "rewards/margins": 9.856096267700195,
      "rewards/rejected": -6.655619144439697,
      "step": 800
    },
    {
      "epoch": 3.31,
      "grad_norm": 0.0014283270575106144,
      "learning_rate": 2.3368421052631583e-06,
      "logits/chosen": -1.8158670663833618,
      "logits/rejected": -1.896251916885376,
      "logps/chosen": -144.15419006347656,
      "logps/rejected": -283.3548583984375,
      "loss": 0.0614,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.173565149307251,
      "rewards/margins": 12.303560256958008,
      "rewards/rejected": -9.129995346069336,
      "step": 810
    },
    {
      "epoch": 3.31,
      "eval_logits/chosen": -1.8417214155197144,
      "eval_logits/rejected": -1.8804857730865479,
      "eval_logps/chosen": -187.0991668701172,
      "eval_logps/rejected": -258.0171813964844,
      "eval_loss": 0.2144124060869217,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.2331290245056152,
      "eval_rewards/margins": 6.682602882385254,
      "eval_rewards/rejected": -4.449473857879639,
      "eval_runtime": 26.2397,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 810
    },
    {
      "epoch": 3.35,
      "grad_norm": 8.28421070764307e-06,
      "learning_rate": 2.3017543859649126e-06,
      "logits/chosen": -1.7904294729232788,
      "logits/rejected": -1.7954041957855225,
      "logps/chosen": -136.8679962158203,
      "logps/rejected": -173.52316284179688,
      "loss": 0.0305,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.9618313312530518,
      "rewards/margins": 9.469245910644531,
      "rewards/rejected": -5.507414817810059,
      "step": 820
    },
    {
      "epoch": 3.39,
      "grad_norm": 0.00032607687171548605,
      "learning_rate": 2.266666666666667e-06,
      "logits/chosen": -1.7678247690200806,
      "logits/rejected": -1.8045036792755127,
      "logps/chosen": -97.9904556274414,
      "logps/rejected": -158.86538696289062,
      "loss": 0.0114,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.4605960845947266,
      "rewards/margins": 9.087323188781738,
      "rewards/rejected": -5.626727104187012,
      "step": 830
    },
    {
      "epoch": 3.43,
      "grad_norm": 9.237614631652832,
      "learning_rate": 2.231578947368421e-06,
      "logits/chosen": -1.8179492950439453,
      "logits/rejected": -1.8643207550048828,
      "logps/chosen": -158.73011779785156,
      "logps/rejected": -175.10726928710938,
      "loss": 0.0529,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.9895429611206055,
      "rewards/margins": 7.000767707824707,
      "rewards/rejected": -5.011223793029785,
      "step": 840
    },
    {
      "epoch": 3.43,
      "eval_logits/chosen": -1.842317819595337,
      "eval_logits/rejected": -1.8808995485305786,
      "eval_logps/chosen": -187.3555145263672,
      "eval_logps/rejected": -258.76568603515625,
      "eval_loss": 0.2120712250471115,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.1562187671661377,
      "eval_rewards/margins": 6.830246925354004,
      "eval_rewards/rejected": -4.674027919769287,
      "eval_runtime": 26.238,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 840
    },
    {
      "epoch": 3.47,
      "grad_norm": 5.9564328694250435e-05,
      "learning_rate": 2.1964912280701754e-06,
      "logits/chosen": -1.782867431640625,
      "logits/rejected": -1.8579784631729126,
      "logps/chosen": -105.4168472290039,
      "logps/rejected": -259.2267761230469,
      "loss": 0.258,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 4.022895336151123,
      "rewards/margins": 12.175846099853516,
      "rewards/rejected": -8.152950286865234,
      "step": 850
    },
    {
      "epoch": 3.51,
      "grad_norm": 1.3973094610264525e-05,
      "learning_rate": 2.16140350877193e-06,
      "logits/chosen": -1.7417545318603516,
      "logits/rejected": -1.8234918117523193,
      "logps/chosen": -115.1645736694336,
      "logps/rejected": -248.0410614013672,
      "loss": 0.039,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.356029987335205,
      "rewards/margins": 12.46558952331543,
      "rewards/rejected": -9.109560012817383,
      "step": 860
    },
    {
      "epoch": 3.55,
      "grad_norm": 0.005180311389267445,
      "learning_rate": 2.1263157894736844e-06,
      "logits/chosen": -1.8123457431793213,
      "logits/rejected": -1.9022102355957031,
      "logps/chosen": -128.5855255126953,
      "logps/rejected": -246.685791015625,
      "loss": 0.001,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.4856154918670654,
      "rewards/margins": 12.966238021850586,
      "rewards/rejected": -10.480623245239258,
      "step": 870
    },
    {
      "epoch": 3.55,
      "eval_logits/chosen": -1.8409830331802368,
      "eval_logits/rejected": -1.87991464138031,
      "eval_logps/chosen": -187.5316925048828,
      "eval_logps/rejected": -259.3367614746094,
      "eval_loss": 0.2091565877199173,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 2.1033718585968018,
      "eval_rewards/margins": 6.94872522354126,
      "eval_rewards/rejected": -4.845353126525879,
      "eval_runtime": 26.2322,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 870
    },
    {
      "epoch": 3.59,
      "grad_norm": 0.3039734661579132,
      "learning_rate": 2.0912280701754387e-06,
      "logits/chosen": -1.7311649322509766,
      "logits/rejected": -1.8156330585479736,
      "logps/chosen": -146.54473876953125,
      "logps/rejected": -300.264892578125,
      "loss": 0.0179,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.358633041381836,
      "rewards/margins": 11.333723068237305,
      "rewards/rejected": -7.975089073181152,
      "step": 880
    },
    {
      "epoch": 3.63,
      "grad_norm": 0.8097418546676636,
      "learning_rate": 2.0561403508771934e-06,
      "logits/chosen": -1.8806817531585693,
      "logits/rejected": -1.922952651977539,
      "logps/chosen": -126.9049301147461,
      "logps/rejected": -184.9642333984375,
      "loss": 0.0541,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.0604941844940186,
      "rewards/margins": 6.54372501373291,
      "rewards/rejected": -5.4832305908203125,
      "step": 890
    },
    {
      "epoch": 3.67,
      "grad_norm": 0.025914475321769714,
      "learning_rate": 2.0210526315789477e-06,
      "logits/chosen": -1.8227866888046265,
      "logits/rejected": -1.9303566217422485,
      "logps/chosen": -167.85279846191406,
      "logps/rejected": -311.27655029296875,
      "loss": 0.0284,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.1594152450561523,
      "rewards/margins": 9.647980690002441,
      "rewards/rejected": -7.488564968109131,
      "step": 900
    },
    {
      "epoch": 3.67,
      "eval_logits/chosen": -1.8366106748580933,
      "eval_logits/rejected": -1.8760128021240234,
      "eval_logps/chosen": -187.9383544921875,
      "eval_logps/rejected": -260.31500244140625,
      "eval_loss": 0.20062340795993805,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 1.9813728332519531,
      "eval_rewards/margins": 7.1201934814453125,
      "eval_rewards/rejected": -5.138820171356201,
      "eval_runtime": 26.2437,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 900
    },
    {
      "epoch": 3.71,
      "grad_norm": 0.02055520936846733,
      "learning_rate": 1.985964912280702e-06,
      "logits/chosen": -1.8666080236434937,
      "logits/rejected": -1.9068281650543213,
      "logps/chosen": -171.9093475341797,
      "logps/rejected": -251.93643188476562,
      "loss": 0.2622,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.18651972711086273,
      "rewards/margins": 8.190787315368652,
      "rewards/rejected": -8.004267692565918,
      "step": 910
    },
    {
      "epoch": 3.76,
      "grad_norm": 197.38717651367188,
      "learning_rate": 1.9508771929824562e-06,
      "logits/chosen": -1.8426361083984375,
      "logits/rejected": -1.9008222818374634,
      "logps/chosen": -150.75900268554688,
      "logps/rejected": -251.9907989501953,
      "loss": 0.0599,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.7380223870277405,
      "rewards/margins": 9.459989547729492,
      "rewards/rejected": -8.721967697143555,
      "step": 920
    },
    {
      "epoch": 3.8,
      "grad_norm": 0.0,
      "learning_rate": 1.9157894736842105e-06,
      "logits/chosen": -1.8063271045684814,
      "logits/rejected": -1.8660167455673218,
      "logps/chosen": -198.50723266601562,
      "logps/rejected": -222.7967529296875,
      "loss": 0.0744,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.713536262512207,
      "rewards/margins": 9.678624153137207,
      "rewards/rejected": -7.965087890625,
      "step": 930
    },
    {
      "epoch": 3.8,
      "eval_logits/chosen": -1.833928108215332,
      "eval_logits/rejected": -1.873318076133728,
      "eval_logps/chosen": -188.06385803222656,
      "eval_logps/rejected": -260.6358337402344,
      "eval_loss": 0.1813022941350937,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 1.943717122077942,
      "eval_rewards/margins": 7.178780555725098,
      "eval_rewards/rejected": -5.2350640296936035,
      "eval_runtime": 26.2336,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 930
    },
    {
      "epoch": 3.84,
      "grad_norm": 4.8926853679631677e-08,
      "learning_rate": 1.8807017543859652e-06,
      "logits/chosen": -1.8202307224273682,
      "logits/rejected": -1.9166450500488281,
      "logps/chosen": -112.94303131103516,
      "logps/rejected": -244.4494171142578,
      "loss": 0.3721,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 2.255749225616455,
      "rewards/margins": 10.898755073547363,
      "rewards/rejected": -8.643006324768066,
      "step": 940
    },
    {
      "epoch": 3.88,
      "grad_norm": 5.456500053405762,
      "learning_rate": 1.8456140350877193e-06,
      "logits/chosen": -1.9108766317367554,
      "logits/rejected": -1.9368479251861572,
      "logps/chosen": -132.4052734375,
      "logps/rejected": -179.675537109375,
      "loss": 0.0257,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.554168701171875,
      "rewards/margins": 6.65416955947876,
      "rewards/rejected": -5.100000858306885,
      "step": 950
    },
    {
      "epoch": 3.92,
      "grad_norm": 9.124670028686523,
      "learning_rate": 1.810526315789474e-06,
      "logits/chosen": -1.7729440927505493,
      "logits/rejected": -1.8252880573272705,
      "logps/chosen": -139.13653564453125,
      "logps/rejected": -193.011962890625,
      "loss": 0.091,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.6216557025909424,
      "rewards/margins": 8.979948043823242,
      "rewards/rejected": -5.3582916259765625,
      "step": 960
    },
    {
      "epoch": 3.92,
      "eval_logits/chosen": -1.8312675952911377,
      "eval_logits/rejected": -1.8707492351531982,
      "eval_logps/chosen": -188.43190002441406,
      "eval_logps/rejected": -261.29730224609375,
      "eval_loss": 0.17215047776699066,
      "eval_rewards/accuracies": 0.8928571343421936,
      "eval_rewards/chosen": 1.8333097696304321,
      "eval_rewards/margins": 7.266810894012451,
      "eval_rewards/rejected": -5.433501243591309,
      "eval_runtime": 26.2333,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 960
    },
    {
      "epoch": 3.96,
      "grad_norm": 1.8921099353974569e-06,
      "learning_rate": 1.7754385964912283e-06,
      "logits/chosen": -1.8464349508285522,
      "logits/rejected": -1.8927011489868164,
      "logps/chosen": -172.9676971435547,
      "logps/rejected": -224.15780639648438,
      "loss": 0.0795,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.649458169937134,
      "rewards/margins": 9.076958656311035,
      "rewards/rejected": -6.4275007247924805,
      "step": 970
    },
    {
      "epoch": 4.0,
      "grad_norm": 318.0891418457031,
      "learning_rate": 1.7403508771929826e-06,
      "logits/chosen": -1.8083856105804443,
      "logits/rejected": -1.8901259899139404,
      "logps/chosen": -209.17422485351562,
      "logps/rejected": -284.4140625,
      "loss": 0.0583,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.1195216178894043,
      "rewards/margins": 12.128747940063477,
      "rewards/rejected": -10.00922679901123,
      "step": 980
    },
    {
      "epoch": 4.04,
      "grad_norm": 13.0223388671875,
      "learning_rate": 1.705263157894737e-06,
      "logits/chosen": -1.847835898399353,
      "logits/rejected": -1.868665337562561,
      "logps/chosen": -149.8837127685547,
      "logps/rejected": -161.1732940673828,
      "loss": 0.3504,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.0685189962387085,
      "rewards/margins": 6.018086910247803,
      "rewards/rejected": -4.9495673179626465,
      "step": 990
    },
    {
      "epoch": 4.04,
      "eval_logits/chosen": -1.8279262781143188,
      "eval_logits/rejected": -1.8672196865081787,
      "eval_logps/chosen": -188.3168487548828,
      "eval_logps/rejected": -261.04876708984375,
      "eval_loss": 0.14871637523174286,
      "eval_rewards/accuracies": 0.9285714030265808,
      "eval_rewards/chosen": 1.8678233623504639,
      "eval_rewards/margins": 7.22676944732666,
      "eval_rewards/rejected": -5.358946800231934,
      "eval_runtime": 26.2331,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 990
    },
    {
      "epoch": 4.08,
      "grad_norm": 0.6881242394447327,
      "learning_rate": 1.6701754385964914e-06,
      "logits/chosen": -1.9393212795257568,
      "logits/rejected": -1.9376901388168335,
      "logps/chosen": -155.8722381591797,
      "logps/rejected": -191.2816925048828,
      "loss": 0.0102,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.3404548168182373,
      "rewards/margins": 7.390922546386719,
      "rewards/rejected": -5.0504679679870605,
      "step": 1000
    },
    {
      "epoch": 4.12,
      "grad_norm": 0.0015760493697598577,
      "learning_rate": 1.6350877192982456e-06,
      "logits/chosen": -1.7732092142105103,
      "logits/rejected": -1.8348764181137085,
      "logps/chosen": -204.78118896484375,
      "logps/rejected": -285.19171142578125,
      "loss": 0.0728,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.1964905261993408,
      "rewards/margins": 11.535087585449219,
      "rewards/rejected": -10.338597297668457,
      "step": 1010
    },
    {
      "epoch": 4.16,
      "grad_norm": 5.320696345734177e-07,
      "learning_rate": 1.6000000000000001e-06,
      "logits/chosen": -1.7163060903549194,
      "logits/rejected": -1.8586156368255615,
      "logps/chosen": -123.50263977050781,
      "logps/rejected": -320.79693603515625,
      "loss": 0.0071,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.7505568265914917,
      "rewards/margins": 10.981633186340332,
      "rewards/rejected": -10.231074333190918,
      "step": 1020
    },
    {
      "epoch": 4.16,
      "eval_logits/chosen": -1.824275016784668,
      "eval_logits/rejected": -1.8636810779571533,
      "eval_logps/chosen": -188.54676818847656,
      "eval_logps/rejected": -261.5804748535156,
      "eval_loss": 0.14029009640216827,
      "eval_rewards/accuracies": 0.9285714030265808,
      "eval_rewards/chosen": 1.7988510131835938,
      "eval_rewards/margins": 7.317309856414795,
      "eval_rewards/rejected": -5.518458843231201,
      "eval_runtime": 26.2364,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1020
    },
    {
      "epoch": 4.2,
      "grad_norm": 0.002488874364644289,
      "learning_rate": 1.5649122807017544e-06,
      "logits/chosen": -1.7597904205322266,
      "logits/rejected": -1.8473920822143555,
      "logps/chosen": -165.49359130859375,
      "logps/rejected": -273.9131774902344,
      "loss": 0.0071,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.4215660095214844,
      "rewards/margins": 10.477560043334961,
      "rewards/rejected": -9.055994033813477,
      "step": 1030
    },
    {
      "epoch": 4.24,
      "grad_norm": 0.0017074132338166237,
      "learning_rate": 1.529824561403509e-06,
      "logits/chosen": -1.7763557434082031,
      "logits/rejected": -1.8936008214950562,
      "logps/chosen": -133.30006408691406,
      "logps/rejected": -336.57427978515625,
      "loss": 0.018,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 4.363164901733398,
      "rewards/margins": 14.957908630371094,
      "rewards/rejected": -10.594743728637695,
      "step": 1040
    },
    {
      "epoch": 4.29,
      "grad_norm": 66.57980346679688,
      "learning_rate": 1.4947368421052632e-06,
      "logits/chosen": -1.835628867149353,
      "logits/rejected": -1.8976163864135742,
      "logps/chosen": -177.20309448242188,
      "logps/rejected": -287.8599548339844,
      "loss": 0.0131,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.568208694458008,
      "rewards/margins": 11.586195945739746,
      "rewards/rejected": -9.017987251281738,
      "step": 1050
    },
    {
      "epoch": 4.29,
      "eval_logits/chosen": -1.8221967220306396,
      "eval_logits/rejected": -1.8615671396255493,
      "eval_logps/chosen": -188.5262451171875,
      "eval_logps/rejected": -261.6840515136719,
      "eval_loss": 0.13124893605709076,
      "eval_rewards/accuracies": 0.9285714030265808,
      "eval_rewards/chosen": 1.805002212524414,
      "eval_rewards/margins": 7.354526042938232,
      "eval_rewards/rejected": -5.549524784088135,
      "eval_runtime": 26.2328,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1050
    },
    {
      "epoch": 4.33,
      "grad_norm": 0.027719292789697647,
      "learning_rate": 1.4596491228070175e-06,
      "logits/chosen": -1.7560068368911743,
      "logits/rejected": -1.8672195672988892,
      "logps/chosen": -189.530029296875,
      "logps/rejected": -330.6167907714844,
      "loss": 0.003,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.0810093879699707,
      "rewards/margins": 9.271745681762695,
      "rewards/rejected": -8.19073486328125,
      "step": 1060
    },
    {
      "epoch": 4.37,
      "grad_norm": 0.0,
      "learning_rate": 1.4245614035087722e-06,
      "logits/chosen": -1.8792110681533813,
      "logits/rejected": -1.9164355993270874,
      "logps/chosen": -212.42788696289062,
      "logps/rejected": -238.2472686767578,
      "loss": 0.1334,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.0076087713241577,
      "rewards/margins": 9.041505813598633,
      "rewards/rejected": -8.033895492553711,
      "step": 1070
    },
    {
      "epoch": 4.41,
      "grad_norm": 3.609489340306027e-06,
      "learning_rate": 1.3894736842105263e-06,
      "logits/chosen": -1.8213157653808594,
      "logits/rejected": -1.870356798171997,
      "logps/chosen": -80.74689483642578,
      "logps/rejected": -158.96109008789062,
      "loss": 0.0868,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.509101629257202,
      "rewards/margins": 7.621730804443359,
      "rewards/rejected": -5.11262845993042,
      "step": 1080
    },
    {
      "epoch": 4.41,
      "eval_logits/chosen": -1.81948721408844,
      "eval_logits/rejected": -1.858702540397644,
      "eval_logps/chosen": -188.66748046875,
      "eval_logps/rejected": -261.94708251953125,
      "eval_loss": 0.1210150271654129,
      "eval_rewards/accuracies": 0.9285714030265808,
      "eval_rewards/chosen": 1.76263427734375,
      "eval_rewards/margins": 7.39107608795166,
      "eval_rewards/rejected": -5.62844181060791,
      "eval_runtime": 26.2393,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1080
    },
    {
      "epoch": 4.45,
      "grad_norm": 0.7020970582962036,
      "learning_rate": 1.354385964912281e-06,
      "logits/chosen": -1.8757814168930054,
      "logits/rejected": -1.9657201766967773,
      "logps/chosen": -117.080810546875,
      "logps/rejected": -241.55252075195312,
      "loss": 0.0046,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.6267999410629272,
      "rewards/margins": 10.32630729675293,
      "rewards/rejected": -8.699506759643555,
      "step": 1090
    },
    {
      "epoch": 4.49,
      "grad_norm": 0.00016405929636675864,
      "learning_rate": 1.3192982456140353e-06,
      "logits/chosen": -1.84481680393219,
      "logits/rejected": -1.8962256908416748,
      "logps/chosen": -176.97457885742188,
      "logps/rejected": -248.0132598876953,
      "loss": 0.0261,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.7715377807617188,
      "rewards/margins": 11.10531234741211,
      "rewards/rejected": -9.333776473999023,
      "step": 1100
    },
    {
      "epoch": 4.53,
      "grad_norm": 0.0024285567924380302,
      "learning_rate": 1.2842105263157895e-06,
      "logits/chosen": -1.9240976572036743,
      "logits/rejected": -1.9567447900772095,
      "logps/chosen": -148.41506958007812,
      "logps/rejected": -262.0478820800781,
      "loss": 0.0041,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.377047538757324,
      "rewards/margins": 12.98823070526123,
      "rewards/rejected": -10.611183166503906,
      "step": 1110
    },
    {
      "epoch": 4.53,
      "eval_logits/chosen": -1.8172996044158936,
      "eval_logits/rejected": -1.8566371202468872,
      "eval_logps/chosen": -188.92127990722656,
      "eval_logps/rejected": -262.44561767578125,
      "eval_loss": 0.1205989271402359,
      "eval_rewards/accuracies": 0.9285714030265808,
      "eval_rewards/chosen": 1.686492681503296,
      "eval_rewards/margins": 7.464495658874512,
      "eval_rewards/rejected": -5.778003215789795,
      "eval_runtime": 26.2482,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1110
    },
    {
      "epoch": 4.57,
      "grad_norm": 26.255502700805664,
      "learning_rate": 1.249122807017544e-06,
      "logits/chosen": -1.7752002477645874,
      "logits/rejected": -1.8067470788955688,
      "logps/chosen": -199.05264282226562,
      "logps/rejected": -198.15122985839844,
      "loss": 0.0289,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.024845600128174,
      "rewards/margins": 10.319567680358887,
      "rewards/rejected": -8.294721603393555,
      "step": 1120
    },
    {
      "epoch": 4.61,
      "grad_norm": 170.23622131347656,
      "learning_rate": 1.2140350877192983e-06,
      "logits/chosen": -1.7735440731048584,
      "logits/rejected": -1.8578115701675415,
      "logps/chosen": -140.52511596679688,
      "logps/rejected": -257.21124267578125,
      "loss": 0.0478,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.585864305496216,
      "rewards/margins": 12.626684188842773,
      "rewards/rejected": -9.04081916809082,
      "step": 1130
    },
    {
      "epoch": 4.65,
      "grad_norm": 0.09534689784049988,
      "learning_rate": 1.1789473684210526e-06,
      "logits/chosen": -1.8365780115127563,
      "logits/rejected": -1.846933364868164,
      "logps/chosen": -90.59309387207031,
      "logps/rejected": -117.09541320800781,
      "loss": 0.0107,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 4.559220314025879,
      "rewards/margins": 10.236574172973633,
      "rewards/rejected": -5.677353382110596,
      "step": 1140
    },
    {
      "epoch": 4.65,
      "eval_logits/chosen": -1.8170740604400635,
      "eval_logits/rejected": -1.8563201427459717,
      "eval_logps/chosen": -189.08616638183594,
      "eval_logps/rejected": -262.8173828125,
      "eval_loss": 0.1178145483136177,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.6370302438735962,
      "eval_rewards/margins": 7.526562213897705,
      "eval_rewards/rejected": -5.88953161239624,
      "eval_runtime": 26.2301,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1140
    },
    {
      "epoch": 4.69,
      "grad_norm": 0.0,
      "learning_rate": 1.143859649122807e-06,
      "logits/chosen": -1.7817316055297852,
      "logits/rejected": -1.8392887115478516,
      "logps/chosen": -162.9131317138672,
      "logps/rejected": -241.8101043701172,
      "loss": 0.0639,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.5432535409927368,
      "rewards/margins": 9.659387588500977,
      "rewards/rejected": -8.116134643554688,
      "step": 1150
    },
    {
      "epoch": 4.73,
      "grad_norm": 1.92129647731781,
      "learning_rate": 1.1087719298245616e-06,
      "logits/chosen": -1.856563925743103,
      "logits/rejected": -1.8815357685089111,
      "logps/chosen": -140.67437744140625,
      "logps/rejected": -180.8864288330078,
      "loss": 0.2,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.501416563987732,
      "rewards/margins": 8.230588912963867,
      "rewards/rejected": -6.7291717529296875,
      "step": 1160
    },
    {
      "epoch": 4.78,
      "grad_norm": 4.147223472595215,
      "learning_rate": 1.0736842105263159e-06,
      "logits/chosen": -1.7213938236236572,
      "logits/rejected": -1.7904603481292725,
      "logps/chosen": -125.3969497680664,
      "logps/rejected": -262.72662353515625,
      "loss": 0.0084,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.621525287628174,
      "rewards/margins": 12.910409927368164,
      "rewards/rejected": -9.288885116577148,
      "step": 1170
    },
    {
      "epoch": 4.78,
      "eval_logits/chosen": -1.815858244895935,
      "eval_logits/rejected": -1.8551596403121948,
      "eval_logps/chosen": -189.174072265625,
      "eval_logps/rejected": -262.9738464355469,
      "eval_loss": 0.11231832951307297,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.610654592514038,
      "eval_rewards/margins": 7.547122478485107,
      "eval_rewards/rejected": -5.936468601226807,
      "eval_runtime": 26.2436,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1170
    },
    {
      "epoch": 4.82,
      "grad_norm": 8.42303293779878e-09,
      "learning_rate": 1.0385964912280702e-06,
      "logits/chosen": -1.810766577720642,
      "logits/rejected": -1.8864467144012451,
      "logps/chosen": -138.02499389648438,
      "logps/rejected": -274.3287048339844,
      "loss": 0.0587,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.259162187576294,
      "rewards/margins": 12.501955032348633,
      "rewards/rejected": -10.242792129516602,
      "step": 1180
    },
    {
      "epoch": 4.86,
      "grad_norm": 1.6972613821053528e-06,
      "learning_rate": 1.0035087719298247e-06,
      "logits/chosen": -1.8087619543075562,
      "logits/rejected": -1.864861249923706,
      "logps/chosen": -106.66157531738281,
      "logps/rejected": -172.39846801757812,
      "loss": 0.0167,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.4979627132415771,
      "rewards/margins": 9.330595016479492,
      "rewards/rejected": -7.832631587982178,
      "step": 1190
    },
    {
      "epoch": 4.9,
      "grad_norm": 0.07996608316898346,
      "learning_rate": 9.68421052631579e-07,
      "logits/chosen": -1.6860878467559814,
      "logits/rejected": -1.7491111755371094,
      "logps/chosen": -86.7617416381836,
      "logps/rejected": -172.6136016845703,
      "loss": 0.0049,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.904423952102661,
      "rewards/margins": 10.647790908813477,
      "rewards/rejected": -6.743367671966553,
      "step": 1200
    },
    {
      "epoch": 4.9,
      "eval_logits/chosen": -1.8151248693466187,
      "eval_logits/rejected": -1.8545377254486084,
      "eval_logps/chosen": -189.30612182617188,
      "eval_logps/rejected": -263.3506774902344,
      "eval_loss": 0.10828591883182526,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.5710465908050537,
      "eval_rewards/margins": 7.62056303024292,
      "eval_rewards/rejected": -6.049516201019287,
      "eval_runtime": 26.2353,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1200
    },
    {
      "epoch": 4.94,
      "grad_norm": 0.0006144978106021881,
      "learning_rate": 9.333333333333334e-07,
      "logits/chosen": -1.7948280572891235,
      "logits/rejected": -1.8062083721160889,
      "logps/chosen": -201.0863037109375,
      "logps/rejected": -189.9558563232422,
      "loss": 0.0226,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.141186237335205,
      "rewards/margins": 9.003381729125977,
      "rewards/rejected": -6.862195014953613,
      "step": 1210
    },
    {
      "epoch": 4.98,
      "grad_norm": 0.004125670529901981,
      "learning_rate": 8.982456140350877e-07,
      "logits/chosen": -1.7143564224243164,
      "logits/rejected": -1.755270004272461,
      "logps/chosen": -129.1698760986328,
      "logps/rejected": -167.7386016845703,
      "loss": 0.0439,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.8997966647148132,
      "rewards/margins": 8.175169944763184,
      "rewards/rejected": -7.275373935699463,
      "step": 1220
    },
    {
      "epoch": 5.02,
      "grad_norm": 0.000474505708552897,
      "learning_rate": 8.631578947368421e-07,
      "logits/chosen": -1.8216488361358643,
      "logits/rejected": -1.880822777748108,
      "logps/chosen": -124.86578369140625,
      "logps/rejected": -239.4408721923828,
      "loss": 0.0746,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.0542094707489014,
      "rewards/margins": 9.776211738586426,
      "rewards/rejected": -8.722002029418945,
      "step": 1230
    },
    {
      "epoch": 5.02,
      "eval_logits/chosen": -1.8140125274658203,
      "eval_logits/rejected": -1.8534783124923706,
      "eval_logps/chosen": -189.43362426757812,
      "eval_logps/rejected": -263.6144104003906,
      "eval_loss": 0.10336750000715256,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.5327917337417603,
      "eval_rewards/margins": 7.6614274978637695,
      "eval_rewards/rejected": -6.128636837005615,
      "eval_runtime": 26.2322,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1230
    },
    {
      "epoch": 5.06,
      "grad_norm": 0.029490573331713676,
      "learning_rate": 8.280701754385966e-07,
      "logits/chosen": -1.778794288635254,
      "logits/rejected": -1.876511573791504,
      "logps/chosen": -113.80903625488281,
      "logps/rejected": -301.50201416015625,
      "loss": 0.0002,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.4956843852996826,
      "rewards/margins": 14.169008255004883,
      "rewards/rejected": -11.673322677612305,
      "step": 1240
    },
    {
      "epoch": 5.1,
      "grad_norm": 0.8206026554107666,
      "learning_rate": 7.92982456140351e-07,
      "logits/chosen": -1.7988386154174805,
      "logits/rejected": -1.8411651849746704,
      "logps/chosen": -80.52401733398438,
      "logps/rejected": -122.70835876464844,
      "loss": 0.0049,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.3624215126037598,
      "rewards/margins": 10.361730575561523,
      "rewards/rejected": -6.999309539794922,
      "step": 1250
    },
    {
      "epoch": 5.14,
      "grad_norm": 0.004326160531491041,
      "learning_rate": 7.578947368421054e-07,
      "logits/chosen": -1.735650658607483,
      "logits/rejected": -1.7782208919525146,
      "logps/chosen": -143.6359100341797,
      "logps/rejected": -203.25271606445312,
      "loss": 0.0091,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.7012696266174316,
      "rewards/margins": 9.915846824645996,
      "rewards/rejected": -7.214578151702881,
      "step": 1260
    },
    {
      "epoch": 5.14,
      "eval_logits/chosen": -1.8135929107666016,
      "eval_logits/rejected": -1.8531488180160522,
      "eval_logps/chosen": -189.6214599609375,
      "eval_logps/rejected": -264.03973388671875,
      "eval_loss": 0.10307549685239792,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.4764469861984253,
      "eval_rewards/margins": 7.73268461227417,
      "eval_rewards/rejected": -6.256237506866455,
      "eval_runtime": 26.2397,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1260
    },
    {
      "epoch": 5.18,
      "grad_norm": 0.05577283725142479,
      "learning_rate": 7.228070175438597e-07,
      "logits/chosen": -1.9368791580200195,
      "logits/rejected": -1.9599593877792358,
      "logps/chosen": -194.0889434814453,
      "logps/rejected": -283.97613525390625,
      "loss": 0.0003,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -0.7386444211006165,
      "rewards/margins": 11.283439636230469,
      "rewards/rejected": -12.02208423614502,
      "step": 1270
    },
    {
      "epoch": 5.22,
      "grad_norm": 6.575052975676954e-05,
      "learning_rate": 6.87719298245614e-07,
      "logits/chosen": -1.7652225494384766,
      "logits/rejected": -1.8098253011703491,
      "logps/chosen": -157.82101440429688,
      "logps/rejected": -254.159912109375,
      "loss": 0.0475,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.464066982269287,
      "rewards/margins": 12.162757873535156,
      "rewards/rejected": -8.698690414428711,
      "step": 1280
    },
    {
      "epoch": 5.27,
      "grad_norm": 0.0,
      "learning_rate": 6.526315789473684e-07,
      "logits/chosen": -1.7878305912017822,
      "logits/rejected": -1.8565266132354736,
      "logps/chosen": -204.24099731445312,
      "logps/rejected": -325.3123779296875,
      "loss": 0.0526,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.4723305702209473,
      "rewards/margins": 12.587270736694336,
      "rewards/rejected": -10.114938735961914,
      "step": 1290
    },
    {
      "epoch": 5.27,
      "eval_logits/chosen": -1.8132985830307007,
      "eval_logits/rejected": -1.8527573347091675,
      "eval_logps/chosen": -189.7008819580078,
      "eval_logps/rejected": -264.1980895996094,
      "eval_loss": 0.09966809302568436,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.4526145458221436,
      "eval_rewards/margins": 7.756356239318848,
      "eval_rewards/rejected": -6.303742408752441,
      "eval_runtime": 26.2377,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1290
    },
    {
      "epoch": 5.31,
      "grad_norm": 6.653774864417983e-09,
      "learning_rate": 6.175438596491229e-07,
      "logits/chosen": -1.7762054204940796,
      "logits/rejected": -1.883049726486206,
      "logps/chosen": -146.65390014648438,
      "logps/rejected": -295.46466064453125,
      "loss": 0.0104,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.7521862983703613,
      "rewards/margins": 11.713335037231445,
      "rewards/rejected": -10.961150169372559,
      "step": 1300
    },
    {
      "epoch": 5.35,
      "grad_norm": 2.788616418838501,
      "learning_rate": 5.824561403508772e-07,
      "logits/chosen": -1.8407360315322876,
      "logits/rejected": -1.9248844385147095,
      "logps/chosen": -156.55274963378906,
      "logps/rejected": -289.77484130859375,
      "loss": 0.2168,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 1.0223561525344849,
      "rewards/margins": 9.885419845581055,
      "rewards/rejected": -8.86306381225586,
      "step": 1310
    },
    {
      "epoch": 5.39,
      "grad_norm": 0.009542102925479412,
      "learning_rate": 5.473684210526316e-07,
      "logits/chosen": -1.7713199853897095,
      "logits/rejected": -1.8372914791107178,
      "logps/chosen": -130.9491729736328,
      "logps/rejected": -201.64227294921875,
      "loss": 0.0316,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.3896652460098267,
      "rewards/margins": 9.91728687286377,
      "rewards/rejected": -8.527620315551758,
      "step": 1320
    },
    {
      "epoch": 5.39,
      "eval_logits/chosen": -1.812363624572754,
      "eval_logits/rejected": -1.8517435789108276,
      "eval_logps/chosen": -189.71922302246094,
      "eval_logps/rejected": -264.22357177734375,
      "eval_loss": 0.09649749845266342,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.4471155405044556,
      "eval_rewards/margins": 7.758502006530762,
      "eval_rewards/rejected": -6.3113861083984375,
      "eval_runtime": 26.23,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1320
    },
    {
      "epoch": 5.43,
      "grad_norm": 25.36566162109375,
      "learning_rate": 5.12280701754386e-07,
      "logits/chosen": -1.800421118736267,
      "logits/rejected": -1.853594183921814,
      "logps/chosen": -154.73907470703125,
      "logps/rejected": -204.53807067871094,
      "loss": 0.015,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.8498456478118896,
      "rewards/margins": 9.068418502807617,
      "rewards/rejected": -6.218573570251465,
      "step": 1330
    },
    {
      "epoch": 5.47,
      "grad_norm": 0.0038690047804266214,
      "learning_rate": 4.771929824561404e-07,
      "logits/chosen": -1.7142912149429321,
      "logits/rejected": -1.7878398895263672,
      "logps/chosen": -123.2061996459961,
      "logps/rejected": -255.41015625,
      "loss": 0.0066,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.37266087532043457,
      "rewards/margins": 11.119722366333008,
      "rewards/rejected": -10.747061729431152,
      "step": 1340
    },
    {
      "epoch": 5.51,
      "grad_norm": 0.021649232134222984,
      "learning_rate": 4.421052631578947e-07,
      "logits/chosen": -1.895521879196167,
      "logits/rejected": -1.9479408264160156,
      "logps/chosen": -160.8917999267578,
      "logps/rejected": -227.413818359375,
      "loss": 0.0249,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.1835147887468338,
      "rewards/margins": 7.590100288391113,
      "rewards/rejected": -7.406586647033691,
      "step": 1350
    },
    {
      "epoch": 5.51,
      "eval_logits/chosen": -1.811531662940979,
      "eval_logits/rejected": -1.850866436958313,
      "eval_logps/chosen": -189.75291442871094,
      "eval_logps/rejected": -264.3137512207031,
      "eval_loss": 0.09495646506547928,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.4370089769363403,
      "eval_rewards/margins": 7.775452136993408,
      "eval_rewards/rejected": -6.338443279266357,
      "eval_runtime": 26.2475,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1350
    },
    {
      "epoch": 5.55,
      "grad_norm": 3.246389627456665,
      "learning_rate": 4.0701754385964917e-07,
      "logits/chosen": -1.8035504817962646,
      "logits/rejected": -1.8503916263580322,
      "logps/chosen": -155.70687866210938,
      "logps/rejected": -203.13211059570312,
      "loss": 0.0039,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 2.7533974647521973,
      "rewards/margins": 11.575220108032227,
      "rewards/rejected": -8.821822166442871,
      "step": 1360
    },
    {
      "epoch": 5.59,
      "grad_norm": 25.977346420288086,
      "learning_rate": 3.7192982456140356e-07,
      "logits/chosen": -1.7662136554718018,
      "logits/rejected": -1.8182075023651123,
      "logps/chosen": -159.5939178466797,
      "logps/rejected": -207.60775756835938,
      "loss": 0.0309,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.3097236752510071,
      "rewards/margins": 9.610837936401367,
      "rewards/rejected": -9.301115036010742,
      "step": 1370
    },
    {
      "epoch": 5.63,
      "grad_norm": 0.6150230169296265,
      "learning_rate": 3.368421052631579e-07,
      "logits/chosen": -1.8924853801727295,
      "logits/rejected": -1.8853254318237305,
      "logps/chosen": -134.17984008789062,
      "logps/rejected": -141.0009307861328,
      "loss": 0.2078,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.9433987736701965,
      "rewards/margins": 6.605888366699219,
      "rewards/rejected": -5.662489891052246,
      "step": 1380
    },
    {
      "epoch": 5.63,
      "eval_logits/chosen": -1.811058759689331,
      "eval_logits/rejected": -1.8503836393356323,
      "eval_logps/chosen": -189.82925415039062,
      "eval_logps/rejected": -264.44891357421875,
      "eval_loss": 0.09370919317007065,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.4141037464141846,
      "eval_rewards/margins": 7.793092250823975,
      "eval_rewards/rejected": -6.378988742828369,
      "eval_runtime": 26.2336,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1380
    },
    {
      "epoch": 5.67,
      "grad_norm": 2.6434385290485807e-05,
      "learning_rate": 3.017543859649123e-07,
      "logits/chosen": -1.8447850942611694,
      "logits/rejected": -1.852958083152771,
      "logps/chosen": -161.8138885498047,
      "logps/rejected": -138.2427978515625,
      "loss": 0.0033,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 4.558165073394775,
      "rewards/margins": 11.715656280517578,
      "rewards/rejected": -7.1574907302856445,
      "step": 1390
    },
    {
      "epoch": 5.71,
      "grad_norm": 0.035506412386894226,
      "learning_rate": 2.666666666666667e-07,
      "logits/chosen": -1.826308250427246,
      "logits/rejected": -1.8440215587615967,
      "logps/chosen": -192.70401000976562,
      "logps/rejected": -216.74386596679688,
      "loss": 0.0099,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.0126938819885254,
      "rewards/margins": 10.826337814331055,
      "rewards/rejected": -7.813642978668213,
      "step": 1400
    },
    {
      "epoch": 5.76,
      "grad_norm": 3.0411527156829834,
      "learning_rate": 2.315789473684211e-07,
      "logits/chosen": -1.835890769958496,
      "logits/rejected": -1.8625280857086182,
      "logps/chosen": -96.525390625,
      "logps/rejected": -223.0682373046875,
      "loss": 0.013,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 3.298449754714966,
      "rewards/margins": 11.045391082763672,
      "rewards/rejected": -7.746940612792969,
      "step": 1410
    },
    {
      "epoch": 5.76,
      "eval_logits/chosen": -1.8103302717208862,
      "eval_logits/rejected": -1.8497931957244873,
      "eval_logps/chosen": -189.79739379882812,
      "eval_logps/rejected": -264.4075622558594,
      "eval_loss": 0.09261856973171234,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.423659086227417,
      "eval_rewards/margins": 7.790247440338135,
      "eval_rewards/rejected": -6.366588115692139,
      "eval_runtime": 26.2481,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1410
    },
    {
      "epoch": 5.8,
      "grad_norm": 0.16603805124759674,
      "learning_rate": 1.9649122807017545e-07,
      "logits/chosen": -1.7975696325302124,
      "logits/rejected": -1.8887662887573242,
      "logps/chosen": -141.9566192626953,
      "logps/rejected": -302.58746337890625,
      "loss": 0.0015,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.0443235635757446,
      "rewards/margins": 11.86727523803711,
      "rewards/rejected": -10.822951316833496,
      "step": 1420
    },
    {
      "epoch": 5.84,
      "grad_norm": 9.943257331848145,
      "learning_rate": 1.6140350877192984e-07,
      "logits/chosen": -1.7811765670776367,
      "logits/rejected": -1.8620269298553467,
      "logps/chosen": -167.00860595703125,
      "logps/rejected": -238.9532470703125,
      "loss": 0.0021,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.4416310787200928,
      "rewards/margins": 11.767573356628418,
      "rewards/rejected": -10.32594108581543,
      "step": 1430
    },
    {
      "epoch": 5.88,
      "grad_norm": 0.0007007242529653013,
      "learning_rate": 1.2631578947368423e-07,
      "logits/chosen": -1.7201411724090576,
      "logits/rejected": -1.8131204843521118,
      "logps/chosen": -149.80282592773438,
      "logps/rejected": -283.00653076171875,
      "loss": 0.0194,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.16927139461040497,
      "rewards/margins": 11.347020149230957,
      "rewards/rejected": -11.17774772644043,
      "step": 1440
    },
    {
      "epoch": 5.88,
      "eval_logits/chosen": -1.810078740119934,
      "eval_logits/rejected": -1.8495510816574097,
      "eval_logps/chosen": -189.88156127929688,
      "eval_logps/rejected": -264.57855224609375,
      "eval_loss": 0.09226309508085251,
      "eval_rewards/accuracies": 0.9642857313156128,
      "eval_rewards/chosen": 1.398409128189087,
      "eval_rewards/margins": 7.816287040710449,
      "eval_rewards/rejected": -6.417877197265625,
      "eval_runtime": 26.2354,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 1440
    }
  ],
  "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
}