trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 20.0,
  "eval_steps": 500,
  "global_step": 1580,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 6.33,
      "learning_rate": 3.4177215189873416e-05,
      "loss": 0.9086,
      "step": 500
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.8354430379746836e-05,
      "loss": 0.0941,
      "step": 1000
    },
    {
      "epoch": 18.99,
      "learning_rate": 2.531645569620253e-06,
      "loss": 0.0241,
      "step": 1500
    },
    {
      "epoch": 20.0,
      "step": 1580,
      "total_flos": 1.95972567552e+16,
      "train_loss": 0.32572943477690974,
      "train_runtime": 684.9677,
      "train_samples_per_second": 145.992,
      "train_steps_per_second": 2.307
    }
  ],
  "logging_steps": 500,
  "max_steps": 1580,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 1.95972567552e+16,
  "trial_name": null,
  "trial_params": null
}