trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2049,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.73,
      "learning_rate": 3.779892630551489e-05,
      "loss": 2.8291,
      "step": 500
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.559785261102977e-05,
      "loss": 2.0769,
      "step": 1000
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.3396778916544655e-05,
      "loss": 1.8033,
      "step": 1500
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.1957052220595413e-06,
      "loss": 1.5679,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2049,
      "total_flos": 587090646368256.0,
      "train_loss": 2.057774809525035,
      "train_runtime": 368.2343,
      "train_samples_per_second": 22.241,
      "train_steps_per_second": 5.564
    }
  ],
  "max_steps": 2049,
  "num_train_epochs": 3,
  "total_flos": 587090646368256.0,
  "trial_name": null,
  "trial_params": null
}