{
  "best_metric": 0.30117151141166687,
  "best_model_checkpoint": "./mistral/21-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.11-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-366_batch_4_2024-04-21_ppid_9/checkpoint-270",
  "epoch": 4.390243902439025,
  "eval_steps": 30,
  "global_step": 270,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.16260162601626016,
      "grad_norm": 11.074726104736328,
      "learning_rate": 3.6363636363636366e-06,
      "logits/chosen": -1.7639459371566772,
      "logits/rejected": -1.8396574258804321,
      "logps/chosen": -186.35623168945312,
      "logps/rejected": -242.7454376220703,
      "loss": 0.6923,
      "rewards/accuracies": 0.550000011920929,
      "rewards/chosen": 0.0025298120453953743,
      "rewards/margins": 0.002917728153988719,
      "rewards/rejected": -0.000387916574254632,
      "step": 10
    },
    {
      "epoch": 0.3252032520325203,
      "grad_norm": 13.123202323913574,
      "learning_rate": 4.9154929577464795e-06,
      "logits/chosen": -1.8020210266113281,
      "logits/rejected": -1.827330231666565,
      "logps/chosen": -179.5525665283203,
      "logps/rejected": -159.48426818847656,
      "loss": 0.6821,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.0269151721149683,
      "rewards/margins": 0.020867865532636642,
      "rewards/rejected": 0.00604730611667037,
      "step": 20
    },
    {
      "epoch": 0.4878048780487805,
      "grad_norm": 14.401883125305176,
      "learning_rate": 4.774647887323944e-06,
      "logits/chosen": -1.7544463872909546,
      "logits/rejected": -1.8119304180145264,
      "logps/chosen": -131.93418884277344,
      "logps/rejected": -186.62869262695312,
      "loss": 0.6531,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.0737195536494255,
      "rewards/margins": 0.08270301669836044,
      "rewards/rejected": -0.008983459323644638,
      "step": 30
    },
    {
      "epoch": 0.4878048780487805,
      "eval_logits/chosen": -1.7289739847183228,
      "eval_logits/rejected": -1.7717361450195312,
      "eval_logps/chosen": -135.4998321533203,
      "eval_logps/rejected": -185.3151397705078,
      "eval_loss": 0.6393889784812927,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.12710317969322205,
      "eval_rewards/margins": 0.1346290558576584,
      "eval_rewards/rejected": -0.007525864988565445,
      "eval_runtime": 17.3728,
      "eval_samples_per_second": 1.612,
      "eval_steps_per_second": 0.806,
      "step": 30
    },
    {
      "epoch": 0.6504065040650406,
      "grad_norm": 15.026944160461426,
      "learning_rate": 4.647887323943662e-06,
      "logits/chosen": -1.8521251678466797,
      "logits/rejected": -1.9095609188079834,
      "logps/chosen": -167.38348388671875,
      "logps/rejected": -215.0012969970703,
      "loss": 0.6343,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.11476834118366241,
      "rewards/margins": 0.10640791803598404,
      "rewards/rejected": 0.008360415697097778,
      "step": 40
    },
    {
      "epoch": 0.8130081300813008,
      "grad_norm": 15.622106552124023,
      "learning_rate": 4.507042253521127e-06,
      "logits/chosen": -1.7336807250976562,
      "logits/rejected": -1.7874727249145508,
      "logps/chosen": -151.20423889160156,
      "logps/rejected": -212.6553192138672,
      "loss": 0.5966,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.16852515935897827,
      "rewards/margins": 0.19663922488689423,
      "rewards/rejected": -0.0281140748411417,
      "step": 50
    },
    {
      "epoch": 0.975609756097561,
      "grad_norm": 15.992025375366211,
      "learning_rate": 4.3661971830985915e-06,
      "logits/chosen": -1.8077377080917358,
      "logits/rejected": -1.8177680969238281,
      "logps/chosen": -166.0524139404297,
      "logps/rejected": -166.77749633789062,
      "loss": 0.5732,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.28099626302719116,
      "rewards/margins": 0.25574439764022827,
      "rewards/rejected": 0.0252518467605114,
      "step": 60
    },
    {
      "epoch": 0.975609756097561,
      "eval_logits/chosen": -1.734244465827942,
      "eval_logits/rejected": -1.7778174877166748,
      "eval_logps/chosen": -133.12557983398438,
      "eval_logps/rejected": -185.54177856445312,
      "eval_loss": 0.5475863218307495,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.36452820897102356,
      "eval_rewards/margins": 0.39471906423568726,
      "eval_rewards/rejected": -0.030190791934728622,
      "eval_runtime": 17.4075,
      "eval_samples_per_second": 1.609,
      "eval_steps_per_second": 0.804,
      "step": 60
    },
    {
      "epoch": 1.1382113821138211,
      "grad_norm": 13.867135047912598,
      "learning_rate": 4.225352112676057e-06,
      "logits/chosen": -1.8027187585830688,
      "logits/rejected": -1.882067084312439,
      "logps/chosen": -174.7197723388672,
      "logps/rejected": -223.87844848632812,
      "loss": 0.5589,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.2828216254711151,
      "rewards/margins": 0.361534059047699,
      "rewards/rejected": -0.07871242612600327,
      "step": 70
    },
    {
      "epoch": 1.3008130081300813,
      "grad_norm": 14.660954475402832,
      "learning_rate": 4.0845070422535216e-06,
      "logits/chosen": -1.8306058645248413,
      "logits/rejected": -1.8653805255889893,
      "logps/chosen": -170.78506469726562,
      "logps/rejected": -210.1864013671875,
      "loss": 0.4755,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.4126508831977844,
      "rewards/margins": 0.48294001817703247,
      "rewards/rejected": -0.07028912007808685,
      "step": 80
    },
    {
      "epoch": 1.4634146341463414,
      "grad_norm": 15.84911060333252,
      "learning_rate": 3.943661971830986e-06,
      "logits/chosen": -1.7707087993621826,
      "logits/rejected": -1.81671142578125,
      "logps/chosen": -143.22067260742188,
      "logps/rejected": -205.97433471679688,
      "loss": 0.5252,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.4095473885536194,
      "rewards/margins": 0.5323301553726196,
      "rewards/rejected": -0.12278278172016144,
      "step": 90
    },
    {
      "epoch": 1.4634146341463414,
      "eval_logits/chosen": -1.74051034450531,
      "eval_logits/rejected": -1.7852736711502075,
      "eval_logps/chosen": -130.74990844726562,
      "eval_logps/rejected": -186.0896759033203,
      "eval_loss": 0.464372456073761,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.6020953059196472,
      "eval_rewards/margins": 0.6870760321617126,
      "eval_rewards/rejected": -0.08498072624206543,
      "eval_runtime": 17.3793,
      "eval_samples_per_second": 1.611,
      "eval_steps_per_second": 0.806,
      "step": 90
    },
    {
      "epoch": 1.6260162601626016,
      "grad_norm": 15.015800476074219,
      "learning_rate": 3.8028169014084508e-06,
      "logits/chosen": -1.795932412147522,
      "logits/rejected": -1.8505065441131592,
      "logps/chosen": -152.42788696289062,
      "logps/rejected": -204.91549682617188,
      "loss": 0.4682,
      "rewards/accuracies": 0.8500000238418579,
      "rewards/chosen": 0.4214203953742981,
      "rewards/margins": 0.5449811816215515,
      "rewards/rejected": -0.12356076389551163,
      "step": 100
    },
    {
      "epoch": 1.7886178861788617,
      "grad_norm": 14.755510330200195,
      "learning_rate": 3.6619718309859158e-06,
      "logits/chosen": -1.7966296672821045,
      "logits/rejected": -1.860388159751892,
      "logps/chosen": -146.94325256347656,
      "logps/rejected": -258.38568115234375,
      "loss": 0.429,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.4777925908565521,
      "rewards/margins": 0.659986138343811,
      "rewards/rejected": -0.18219345808029175,
      "step": 110
    },
    {
      "epoch": 1.951219512195122,
      "grad_norm": 14.414965629577637,
      "learning_rate": 3.5211267605633804e-06,
      "logits/chosen": -1.7987533807754517,
      "logits/rejected": -1.8348747491836548,
      "logps/chosen": -159.26161193847656,
      "logps/rejected": -196.20147705078125,
      "loss": 0.4402,
      "rewards/accuracies": 0.8500000238418579,
      "rewards/chosen": 0.5999616384506226,
      "rewards/margins": 0.764100193977356,
      "rewards/rejected": -0.1641385406255722,
      "step": 120
    },
    {
      "epoch": 1.951219512195122,
      "eval_logits/chosen": -1.7491638660430908,
      "eval_logits/rejected": -1.794945478439331,
      "eval_logps/chosen": -128.98277282714844,
      "eval_logps/rejected": -186.99697875976562,
      "eval_loss": 0.40887242555618286,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.7788085341453552,
      "eval_rewards/margins": 0.9545186161994934,
      "eval_rewards/rejected": -0.17571023106575012,
      "eval_runtime": 17.373,
      "eval_samples_per_second": 1.612,
      "eval_steps_per_second": 0.806,
      "step": 120
    },
    {
      "epoch": 2.113821138211382,
      "grad_norm": 11.437810897827148,
      "learning_rate": 3.3802816901408454e-06,
      "logits/chosen": -1.7318217754364014,
      "logits/rejected": -1.7831417322158813,
      "logps/chosen": -181.61026000976562,
      "logps/rejected": -181.42465209960938,
      "loss": 0.3371,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.925055205821991,
      "rewards/margins": 1.1250288486480713,
      "rewards/rejected": -0.19997362792491913,
      "step": 130
    },
    {
      "epoch": 2.2764227642276422,
      "grad_norm": 9.344759941101074,
      "learning_rate": 3.2394366197183104e-06,
      "logits/chosen": -1.8318933248519897,
      "logits/rejected": -1.8962070941925049,
      "logps/chosen": -149.0543975830078,
      "logps/rejected": -251.8573455810547,
      "loss": 0.356,
      "rewards/accuracies": 0.8500000238418579,
      "rewards/chosen": 0.8429641723632812,
      "rewards/margins": 0.9255543947219849,
      "rewards/rejected": -0.08259038627147675,
      "step": 140
    },
    {
      "epoch": 2.4390243902439024,
      "grad_norm": 13.787588119506836,
      "learning_rate": 3.0985915492957746e-06,
      "logits/chosen": -1.7744439840316772,
      "logits/rejected": -1.814591646194458,
      "logps/chosen": -148.4331512451172,
      "logps/rejected": -177.21852111816406,
      "loss": 0.4138,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.8420242071151733,
      "rewards/margins": 0.8596542477607727,
      "rewards/rejected": -0.017630038782954216,
      "step": 150
    },
    {
      "epoch": 2.4390243902439024,
      "eval_logits/chosen": -1.7541459798812866,
      "eval_logits/rejected": -1.8013830184936523,
      "eval_logps/chosen": -127.49993133544922,
      "eval_logps/rejected": -188.32179260253906,
      "eval_loss": 0.3651679456233978,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.9270926713943481,
      "eval_rewards/margins": 1.2352840900421143,
      "eval_rewards/rejected": -0.30819153785705566,
      "eval_runtime": 17.3736,
      "eval_samples_per_second": 1.612,
      "eval_steps_per_second": 0.806,
      "step": 150
    },
    {
      "epoch": 2.6016260162601625,
      "grad_norm": 6.115133762359619,
      "learning_rate": 2.9577464788732396e-06,
      "logits/chosen": -1.7456220388412476,
      "logits/rejected": -1.834590196609497,
      "logps/chosen": -118.06221008300781,
      "logps/rejected": -232.33853149414062,
      "loss": 0.325,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.0931917428970337,
      "rewards/margins": 1.5432584285736084,
      "rewards/rejected": -0.4500666558742523,
      "step": 160
    },
    {
      "epoch": 2.7642276422764227,
      "grad_norm": 18.476879119873047,
      "learning_rate": 2.8169014084507046e-06,
      "logits/chosen": -1.7815649509429932,
      "logits/rejected": -1.8541799783706665,
      "logps/chosen": -177.2996826171875,
      "logps/rejected": -254.4600067138672,
      "loss": 0.3698,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.6835691928863525,
      "rewards/margins": 1.2215800285339355,
      "rewards/rejected": -0.5380108952522278,
      "step": 170
    },
    {
      "epoch": 2.926829268292683,
      "grad_norm": 16.214574813842773,
      "learning_rate": 2.676056338028169e-06,
      "logits/chosen": -1.8389543294906616,
      "logits/rejected": -1.853319525718689,
      "logps/chosen": -160.57228088378906,
      "logps/rejected": -136.90098571777344,
      "loss": 0.3863,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.666906476020813,
      "rewards/margins": 0.767067015171051,
      "rewards/rejected": -0.10016050189733505,
      "step": 180
    },
    {
      "epoch": 2.926829268292683,
      "eval_logits/chosen": -1.7583568096160889,
      "eval_logits/rejected": -1.8064650297164917,
      "eval_logps/chosen": -126.48298645019531,
      "eval_logps/rejected": -189.43382263183594,
      "eval_loss": 0.3402670919895172,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.0287870168685913,
      "eval_rewards/margins": 1.4481836557388306,
      "eval_rewards/rejected": -0.4193967878818512,
      "eval_runtime": 17.3755,
      "eval_samples_per_second": 1.611,
      "eval_steps_per_second": 0.806,
      "step": 180
    },
    {
      "epoch": 3.089430894308943,
      "grad_norm": 7.2306013107299805,
      "learning_rate": 2.535211267605634e-06,
      "logits/chosen": -1.8473081588745117,
      "logits/rejected": -1.9084949493408203,
      "logps/chosen": -179.919189453125,
      "logps/rejected": -254.76321411132812,
      "loss": 0.2924,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.619147002696991,
      "rewards/margins": 1.209044098854065,
      "rewards/rejected": -0.589897096157074,
      "step": 190
    },
    {
      "epoch": 3.252032520325203,
      "grad_norm": 9.260942459106445,
      "learning_rate": 2.3943661971830984e-06,
      "logits/chosen": -1.8063428401947021,
      "logits/rejected": -1.8332942724227905,
      "logps/chosen": -133.8800048828125,
      "logps/rejected": -181.50469970703125,
      "loss": 0.3248,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 1.0484669208526611,
      "rewards/margins": 1.2968560457229614,
      "rewards/rejected": -0.24838905036449432,
      "step": 200
    },
    {
      "epoch": 3.4146341463414633,
      "grad_norm": 17.881528854370117,
      "learning_rate": 2.2535211267605635e-06,
      "logits/chosen": -1.8409783840179443,
      "logits/rejected": -1.8688993453979492,
      "logps/chosen": -188.42379760742188,
      "logps/rejected": -208.37612915039062,
      "loss": 0.2999,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.6589190363883972,
      "rewards/margins": 1.1798341274261475,
      "rewards/rejected": -0.5209150910377502,
      "step": 210
    },
    {
      "epoch": 3.4146341463414633,
      "eval_logits/chosen": -1.762878179550171,
      "eval_logits/rejected": -1.8114169836044312,
      "eval_logps/chosen": -125.83088684082031,
      "eval_logps/rejected": -190.80966186523438,
      "eval_loss": 0.3209560513496399,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.093997597694397,
      "eval_rewards/margins": 1.6509748697280884,
      "eval_rewards/rejected": -0.5569776296615601,
      "eval_runtime": 17.3805,
      "eval_samples_per_second": 1.611,
      "eval_steps_per_second": 0.806,
      "step": 210
    },
    {
      "epoch": 3.5772357723577235,
      "grad_norm": 6.462300777435303,
      "learning_rate": 2.1126760563380285e-06,
      "logits/chosen": -1.9442710876464844,
      "logits/rejected": -1.963648796081543,
      "logps/chosen": -184.78109741210938,
      "logps/rejected": -209.25723266601562,
      "loss": 0.3841,
      "rewards/accuracies": 0.8500000238418579,
      "rewards/chosen": 0.808915913105011,
      "rewards/margins": 1.145354986190796,
      "rewards/rejected": -0.3364391326904297,
      "step": 220
    },
    {
      "epoch": 3.7398373983739837,
      "grad_norm": 8.52739143371582,
      "learning_rate": 1.971830985915493e-06,
      "logits/chosen": -1.7989740371704102,
      "logits/rejected": -1.8550697565078735,
      "logps/chosen": -118.730712890625,
      "logps/rejected": -222.0437469482422,
      "loss": 0.3073,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.988745391368866,
      "rewards/margins": 1.832834005355835,
      "rewards/rejected": -0.8440886735916138,
      "step": 230
    },
    {
      "epoch": 3.902439024390244,
      "grad_norm": 6.697244644165039,
      "learning_rate": 1.8309859154929579e-06,
      "logits/chosen": -1.8378446102142334,
      "logits/rejected": -1.8886947631835938,
      "logps/chosen": -150.81031799316406,
      "logps/rejected": -189.8251495361328,
      "loss": 0.2311,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 1.1466209888458252,
      "rewards/margins": 1.7799263000488281,
      "rewards/rejected": -0.6333051919937134,
      "step": 240
    },
    {
      "epoch": 3.902439024390244,
      "eval_logits/chosen": -1.7646329402923584,
      "eval_logits/rejected": -1.8135557174682617,
      "eval_logps/chosen": -125.47990417480469,
      "eval_logps/rejected": -191.78990173339844,
      "eval_loss": 0.30853360891342163,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.129095196723938,
      "eval_rewards/margins": 1.7840988636016846,
      "eval_rewards/rejected": -0.655003547668457,
      "eval_runtime": 17.3741,
      "eval_samples_per_second": 1.612,
      "eval_steps_per_second": 0.806,
      "step": 240
    },
    {
      "epoch": 4.065040650406504,
      "grad_norm": 14.190593719482422,
      "learning_rate": 1.6901408450704227e-06,
      "logits/chosen": -1.8366336822509766,
      "logits/rejected": -1.8869634866714478,
      "logps/chosen": -172.4488525390625,
      "logps/rejected": -221.21475219726562,
      "loss": 0.3199,
      "rewards/accuracies": 0.8500000238418579,
      "rewards/chosen": 0.31344354152679443,
      "rewards/margins": 0.9199511408805847,
      "rewards/rejected": -0.6065076589584351,
      "step": 250
    },
    {
      "epoch": 4.227642276422764,
      "grad_norm": 10.718254089355469,
      "learning_rate": 1.5492957746478873e-06,
      "logits/chosen": -1.6653333902359009,
      "logits/rejected": -1.769636869430542,
      "logps/chosen": -112.0279312133789,
      "logps/rejected": -213.3590850830078,
      "loss": 0.2095,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 1.3316088914871216,
      "rewards/margins": 2.201448440551758,
      "rewards/rejected": -0.8698397874832153,
      "step": 260
    },
    {
      "epoch": 4.390243902439025,
      "grad_norm": 9.33919620513916,
      "learning_rate": 1.4084507042253523e-06,
      "logits/chosen": -1.8406511545181274,
      "logits/rejected": -1.914043664932251,
      "logps/chosen": -146.75958251953125,
      "logps/rejected": -264.20391845703125,
      "loss": 0.2992,
      "rewards/accuracies": 0.949999988079071,
      "rewards/chosen": 0.8267571330070496,
      "rewards/margins": 1.8074918985366821,
      "rewards/rejected": -0.9807347059249878,
      "step": 270
    },
    {
      "epoch": 4.390243902439025,
      "eval_logits/chosen": -1.7653776407241821,
      "eval_logits/rejected": -1.8147268295288086,
      "eval_logps/chosen": -125.27703857421875,
      "eval_logps/rejected": -192.666748046875,
      "eval_loss": 0.30117151141166687,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 1.149381399154663,
      "eval_rewards/margins": 1.8920701742172241,
      "eval_rewards/rejected": -0.7426888346672058,
      "eval_runtime": 17.3768,
      "eval_samples_per_second": 1.611,
      "eval_steps_per_second": 0.806,
      "step": 270
    }
  ],
  "logging_steps": 10,
  "max_steps": 366,
  "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
}