trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 40,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.3892,
      "step": 10
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001,
      "loss": 0.136,
      "step": 20
    },
    {
      "epoch": 3.0,
      "learning_rate": 5e-05,
      "loss": 0.0401,
      "step": 30
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0,
      "loss": 0.0266,
      "step": 40
    },
    {
      "epoch": 4.0,
      "step": 40,
      "total_flos": 4.959487335333888e+16,
      "train_loss": 0.14795949459075927,
      "train_runtime": 31.341,
      "train_samples_per_second": 20.421,
      "train_steps_per_second": 1.276
    }
  ],
  "max_steps": 40,
  "num_train_epochs": 4,
  "total_flos": 4.959487335333888e+16,
  "trial_name": null,
  "trial_params": null
}