trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2994,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.5,
      "learning_rate": 4.1649966599866405e-05,
      "loss": 1.6003,
      "step": 500
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.32999331997328e-05,
      "loss": 1.0598,
      "step": 1000
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.49498997995992e-05,
      "loss": 0.3999,
      "step": 1500
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.65998663994656e-05,
      "loss": 0.3829,
      "step": 2000
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.249832999331998e-06,
      "loss": 0.131,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2994,
      "total_flos": 4688778409821792.0,
      "train_loss": 0.616782742018053,
      "train_runtime": 648.5631,
      "train_samples_per_second": 36.903,
      "train_steps_per_second": 4.616
    }
  ],
  "max_steps": 2994,
  "num_train_epochs": 3,
  "total_flos": 4688778409821792.0,
  "trial_name": null,
  "trial_params": null
}