{
  "best_metric": 0.4368760883808136,
  "best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.6-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-90_batch_16_2024-04-20_ppid_9/checkpoint-90",
  "epoch": 5.806451612903226,
  "eval_steps": 30,
  "global_step": 90,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.65,
      "grad_norm": 16.76080894470215,
      "learning_rate": 4.770114942528735e-06,
      "logits/chosen": -1.8055893182754517,
      "logits/rejected": -1.828260064125061,
      "logps/chosen": -46.100120544433594,
      "logps/rejected": -82.48685455322266,
      "loss": 0.6795,
      "rewards/accuracies": 0.25,
      "rewards/chosen": 0.04209427908062935,
      "rewards/margins": 0.04848780483007431,
      "rewards/rejected": -0.006393528077751398,
      "step": 10
    },
    {
      "epoch": 1.29,
      "grad_norm": 11.445436477661133,
      "learning_rate": 4.1954022988505746e-06,
      "logits/chosen": -1.8350013494491577,
      "logits/rejected": -1.8602994680404663,
      "logps/chosen": -37.913516998291016,
      "logps/rejected": -67.64692687988281,
      "loss": 0.6251,
      "rewards/accuracies": 0.3499999940395355,
      "rewards/chosen": 0.15703639388084412,
      "rewards/margins": 0.1911666840314865,
      "rewards/rejected": -0.0341302864253521,
      "step": 20
    },
    {
      "epoch": 1.94,
      "grad_norm": 11.545478820800781,
      "learning_rate": 3.620689655172414e-06,
      "logits/chosen": -1.8373997211456299,
      "logits/rejected": -1.847825050354004,
      "logps/chosen": -29.99410057067871,
      "logps/rejected": -59.1256217956543,
      "loss": 0.5772,
      "rewards/accuracies": 0.30000001192092896,
      "rewards/chosen": 0.27005138993263245,
      "rewards/margins": 0.30424270033836365,
      "rewards/rejected": -0.0341913141310215,
      "step": 30
    },
    {
      "epoch": 1.94,
      "eval_logits/chosen": -1.81316077709198,
      "eval_logits/rejected": -1.8455785512924194,
      "eval_logps/chosen": -36.48888397216797,
      "eval_logps/rejected": -55.34504699707031,
      "eval_loss": 0.5171023011207581,
      "eval_rewards/accuracies": 0.5,
      "eval_rewards/chosen": 0.5191236138343811,
      "eval_rewards/margins": 0.5845221877098083,
      "eval_rewards/rejected": -0.06539855152368546,
      "eval_runtime": 2.4917,
      "eval_samples_per_second": 11.237,
      "eval_steps_per_second": 1.605,
      "step": 30
    },
    {
      "epoch": 2.58,
      "grad_norm": 7.317459583282471,
      "learning_rate": 3.0459770114942533e-06,
      "logits/chosen": -1.7586491107940674,
      "logits/rejected": -1.7903881072998047,
      "logps/chosen": -52.9969367980957,
      "logps/rejected": -87.54074096679688,
      "loss": 0.5181,
      "rewards/accuracies": 0.5,
      "rewards/chosen": 0.6470302939414978,
      "rewards/margins": 0.7092825174331665,
      "rewards/rejected": -0.06225220113992691,
      "step": 40
    },
    {
      "epoch": 3.23,
      "grad_norm": 5.132155418395996,
      "learning_rate": 2.471264367816092e-06,
      "logits/chosen": -1.8951135873794556,
      "logits/rejected": -1.9210799932479858,
      "logps/chosen": -34.33287811279297,
      "logps/rejected": -68.03052520751953,
      "loss": 0.5415,
      "rewards/accuracies": 0.32499998807907104,
      "rewards/chosen": 0.5781847238540649,
      "rewards/margins": 0.6501666903495789,
      "rewards/rejected": -0.07198190689086914,
      "step": 50
    },
    {
      "epoch": 3.87,
      "grad_norm": 7.053220748901367,
      "learning_rate": 1.896551724137931e-06,
      "logits/chosen": -1.9114513397216797,
      "logits/rejected": -1.9271589517593384,
      "logps/chosen": -25.962200164794922,
      "logps/rejected": -48.20077133178711,
      "loss": 0.5125,
      "rewards/accuracies": 0.25,
      "rewards/chosen": 0.47203439474105835,
      "rewards/margins": 0.5603113174438477,
      "rewards/rejected": -0.08827687054872513,
      "step": 60
    },
    {
      "epoch": 3.87,
      "eval_logits/chosen": -1.817034363746643,
      "eval_logits/rejected": -1.8500827550888062,
      "eval_logps/chosen": -35.225791931152344,
      "eval_logps/rejected": -55.629276275634766,
      "eval_loss": 0.4517291486263275,
      "eval_rewards/accuracies": 0.5,
      "eval_rewards/chosen": 0.8980507254600525,
      "eval_rewards/margins": 1.048719048500061,
      "eval_rewards/rejected": -0.15066833794116974,
      "eval_runtime": 2.4913,
      "eval_samples_per_second": 11.239,
      "eval_steps_per_second": 1.606,
      "step": 60
    },
    {
      "epoch": 4.52,
      "grad_norm": 3.6152443885803223,
      "learning_rate": 1.3218390804597702e-06,
      "logits/chosen": -1.8332040309906006,
      "logits/rejected": -1.8538074493408203,
      "logps/chosen": -34.457115173339844,
      "logps/rejected": -72.48670959472656,
      "loss": 0.4838,
      "rewards/accuracies": 0.375,
      "rewards/chosen": 0.6265449523925781,
      "rewards/margins": 0.7487497329711914,
      "rewards/rejected": -0.12220476567745209,
      "step": 70
    },
    {
      "epoch": 5.16,
      "grad_norm": 3.271343231201172,
      "learning_rate": 7.471264367816093e-07,
      "logits/chosen": -1.7745481729507446,
      "logits/rejected": -1.784841537475586,
      "logps/chosen": -52.409339904785156,
      "logps/rejected": -76.24700164794922,
      "loss": 0.4665,
      "rewards/accuracies": 0.4749999940395355,
      "rewards/chosen": 0.939862847328186,
      "rewards/margins": 1.0275475978851318,
      "rewards/rejected": -0.08768495172262192,
      "step": 80
    },
    {
      "epoch": 5.81,
      "grad_norm": 2.297419786453247,
      "learning_rate": 1.7241379310344828e-07,
      "logits/chosen": -1.8448705673217773,
      "logits/rejected": -1.8637123107910156,
      "logps/chosen": -33.81181335449219,
      "logps/rejected": -63.00860595703125,
      "loss": 0.491,
      "rewards/accuracies": 0.3499999940395355,
      "rewards/chosen": 0.6832905411720276,
      "rewards/margins": 0.8434805870056152,
      "rewards/rejected": -0.1601901352405548,
      "step": 90
    },
    {
      "epoch": 5.81,
      "eval_logits/chosen": -1.8184903860092163,
      "eval_logits/rejected": -1.8517436981201172,
      "eval_logps/chosen": -34.81117630004883,
      "eval_logps/rejected": -55.731048583984375,
      "eval_loss": 0.4368760883808136,
      "eval_rewards/accuracies": 0.5,
      "eval_rewards/chosen": 1.0224344730377197,
      "eval_rewards/margins": 1.2036339044570923,
      "eval_rewards/rejected": -0.18119940161705017,
      "eval_runtime": 2.4928,
      "eval_samples_per_second": 11.232,
      "eval_steps_per_second": 1.605,
      "step": 90
    }
  ],
  "logging_steps": 10,
  "max_steps": 90,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 90,
  "total_flos": 0.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}