trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 7.0,
  "global_step": 14,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7982832789421082,
      "eval_f1": 0.8459016393442623,
      "eval_loss": 0.8990994691848755,
      "eval_runtime": 9.7523,
      "eval_samples_per_second": 47.784,
      "step": 2
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8519313335418701,
      "eval_f1": 0.8824531516183987,
      "eval_loss": 0.5915661454200745,
      "eval_runtime": 9.8478,
      "eval_samples_per_second": 47.32,
      "step": 4
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8776823878288269,
      "eval_f1": 0.9001751313485113,
      "eval_loss": 0.46527373790740967,
      "eval_runtime": 9.8666,
      "eval_samples_per_second": 47.23,
      "step": 6
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8841201663017273,
      "eval_f1": 0.9039145907473309,
      "eval_loss": 0.4179508090019226,
      "eval_runtime": 9.8783,
      "eval_samples_per_second": 47.174,
      "step": 8
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8948497772216797,
      "eval_f1": 0.9129662522202486,
      "eval_loss": 0.385845422744751,
      "eval_runtime": 9.9037,
      "eval_samples_per_second": 47.053,
      "step": 10
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9034335017204285,
      "eval_f1": 0.9197860962566845,
      "eval_loss": 0.3674018383026123,
      "eval_runtime": 9.9401,
      "eval_samples_per_second": 46.881,
      "step": 12
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9034335017204285,
      "eval_f1": 0.9203539823008849,
      "eval_loss": 0.36275243759155273,
      "eval_runtime": 9.9586,
      "eval_samples_per_second": 46.794,
      "step": 14
    },
    {
      "epoch": 7.0,
      "step": 14,
      "total_flos": 29010408999936.0,
      "train_runtime": 87.5566,
      "train_samples_per_second": 0.16
    }
  ],
  "max_steps": 14,
  "num_train_epochs": 7,
  "total_flos": 29010408999936.0,
  "trial_name": null,
  "trial_params": null
}