{
  "best_metric": 0.4401554763317108,
  "best_model_checkpoint": "./mistral/19-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.3-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-366_batch_4_2024-04-19_ppid_9/checkpoint-90",
  "epoch": 1.4634146341463414,
  "eval_steps": 30,
  "global_step": 90,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.16,
      "grad_norm": 29.063692092895508,
      "learning_rate": 3.181818181818182e-06,
      "logits/chosen": -1.7455532550811768,
      "logits/rejected": -1.7793405055999756,
      "logps/chosen": -54.72136306762695,
      "logps/rejected": -97.28812408447266,
      "loss": 0.6924,
      "rewards/accuracies": 0.15000000596046448,
      "rewards/chosen": 0.003122234484180808,
      "rewards/margins": 0.0015356539515778422,
      "rewards/rejected": 0.0015865802997723222,
      "step": 10
    },
    {
      "epoch": 0.33,
      "grad_norm": 15.157337188720703,
      "learning_rate": 4.9154929577464795e-06,
      "logits/chosen": -1.8641210794448853,
      "logits/rejected": -1.8705828189849854,
      "logps/chosen": -40.3618049621582,
      "logps/rejected": -55.96257400512695,
      "loss": 0.6737,
      "rewards/accuracies": 0.32499998807907104,
      "rewards/chosen": 0.04275550693273544,
      "rewards/margins": 0.040880247950553894,
      "rewards/rejected": 0.0018752537434920669,
      "step": 20
    },
    {
      "epoch": 0.49,
      "grad_norm": 13.148392677307129,
      "learning_rate": 4.774647887323944e-06,
      "logits/chosen": -1.8284633159637451,
      "logits/rejected": -1.8519256114959717,
      "logps/chosen": -44.69489288330078,
      "logps/rejected": -76.79936981201172,
      "loss": 0.6371,
      "rewards/accuracies": 0.42500001192092896,
      "rewards/chosen": 0.12410124391317368,
      "rewards/margins": 0.1248731017112732,
      "rewards/rejected": -0.0007718756678514183,
      "step": 30
    },
    {
      "epoch": 0.49,
      "eval_logits/chosen": -1.7379764318466187,
      "eval_logits/rejected": -1.7726188898086548,
      "eval_logps/chosen": -50.49373245239258,
      "eval_logps/rejected": -83.54340362548828,
      "eval_loss": 0.5864880681037903,
      "eval_rewards/accuracies": 0.4642857015132904,
      "eval_rewards/chosen": 0.24716782569885254,
      "eval_rewards/margins": 0.25581416487693787,
      "eval_rewards/rejected": -0.008646338246762753,
      "eval_runtime": 8.1282,
      "eval_samples_per_second": 3.445,
      "eval_steps_per_second": 1.722,
      "step": 30
    },
    {
      "epoch": 0.65,
      "grad_norm": 24.143388748168945,
      "learning_rate": 4.633802816901409e-06,
      "logits/chosen": -1.8080145120620728,
      "logits/rejected": -1.8319604396820068,
      "logps/chosen": -46.848201751708984,
      "logps/rejected": -81.42877960205078,
      "loss": 0.5649,
      "rewards/accuracies": 0.42500001192092896,
      "rewards/chosen": 0.3360167443752289,
      "rewards/margins": 0.34645381569862366,
      "rewards/rejected": -0.010437069460749626,
      "step": 40
    },
    {
      "epoch": 0.81,
      "grad_norm": 0.0,
      "learning_rate": 4.492957746478874e-06,
      "logits/chosen": -1.8638660907745361,
      "logits/rejected": -1.8754875659942627,
      "logps/chosen": -22.06886863708496,
      "logps/rejected": -41.17422103881836,
      "loss": 0.6377,
      "rewards/accuracies": 0.20000000298023224,
      "rewards/chosen": 0.25195080041885376,
      "rewards/margins": 0.16642150282859802,
      "rewards/rejected": 0.08552931249141693,
      "step": 50
    },
    {
      "epoch": 0.98,
      "grad_norm": 7.51575231552124,
      "learning_rate": 4.352112676056338e-06,
      "logits/chosen": -1.8215672969818115,
      "logits/rejected": -1.8419084548950195,
      "logps/chosen": -44.14409255981445,
      "logps/rejected": -77.38871002197266,
      "loss": 0.5496,
      "rewards/accuracies": 0.3499999940395355,
      "rewards/chosen": 0.4203863739967346,
      "rewards/margins": 0.458610475063324,
      "rewards/rejected": -0.038224123418331146,
      "step": 60
    },
    {
      "epoch": 0.98,
      "eval_logits/chosen": -1.7419661283493042,
      "eval_logits/rejected": -1.7773857116699219,
      "eval_logps/chosen": -49.36268997192383,
      "eval_logps/rejected": -83.60606384277344,
      "eval_loss": 0.4963766038417816,
      "eval_rewards/accuracies": 0.4642857015132904,
      "eval_rewards/chosen": 0.5864797234535217,
      "eval_rewards/margins": 0.613922655582428,
      "eval_rewards/rejected": -0.02744293212890625,
      "eval_runtime": 8.1315,
      "eval_samples_per_second": 3.443,
      "eval_steps_per_second": 1.722,
      "step": 60
    },
    {
      "epoch": 1.14,
      "grad_norm": 3.454394817352295,
      "learning_rate": 4.211267605633803e-06,
      "logits/chosen": -1.7844343185424805,
      "logits/rejected": -1.8126220703125,
      "logps/chosen": -39.737998962402344,
      "logps/rejected": -78.76033782958984,
      "loss": 0.4925,
      "rewards/accuracies": 0.4000000059604645,
      "rewards/chosen": 0.6417733430862427,
      "rewards/margins": 0.7573493123054504,
      "rewards/rejected": -0.11557593196630478,
      "step": 70
    },
    {
      "epoch": 1.3,
      "grad_norm": 14.345088005065918,
      "learning_rate": 4.070422535211268e-06,
      "logits/chosen": -1.822148084640503,
      "logits/rejected": -1.8463541269302368,
      "logps/chosen": -35.735836029052734,
      "logps/rejected": -75.4683837890625,
      "loss": 0.4965,
      "rewards/accuracies": 0.375,
      "rewards/chosen": 0.6439998745918274,
      "rewards/margins": 0.8037986755371094,
      "rewards/rejected": -0.1597987860441208,
      "step": 80
    },
    {
      "epoch": 1.46,
      "grad_norm": 12.988151550292969,
      "learning_rate": 3.9295774647887325e-06,
      "logits/chosen": -1.858496069908142,
      "logits/rejected": -1.8683990240097046,
      "logps/chosen": -30.166736602783203,
      "logps/rejected": -50.970542907714844,
      "loss": 0.5185,
      "rewards/accuracies": 0.30000001192092896,
      "rewards/chosen": 0.6545813679695129,
      "rewards/margins": 0.7556509971618652,
      "rewards/rejected": -0.10106959193944931,
      "step": 90
    },
    {
      "epoch": 1.46,
      "eval_logits/chosen": -1.7461166381835938,
      "eval_logits/rejected": -1.7827314138412476,
      "eval_logps/chosen": -47.9538688659668,
      "eval_logps/rejected": -83.84149169921875,
      "eval_loss": 0.4401554763317108,
      "eval_rewards/accuracies": 0.4642857015132904,
      "eval_rewards/chosen": 1.0091263055801392,
      "eval_rewards/margins": 1.1071958541870117,
      "eval_rewards/rejected": -0.09806957095861435,
      "eval_runtime": 8.1286,
      "eval_samples_per_second": 3.445,
      "eval_steps_per_second": 1.722,
      "step": 90
    }
  ],
  "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": 2,
  "trial_name": null,
  "trial_params": null
}