trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1995,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.75,
      "learning_rate": 3.7468671679198e-05,
      "loss": 1.049,
      "step": 500
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.493734335839599e-05,
      "loss": 0.8325,
      "step": 1000
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.2406015037593984e-05,
      "loss": 0.6958,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1995,
      "total_flos": 2694600320090112.0,
      "train_loss": 0.7912993390458568,
      "train_runtime": 812.2771,
      "train_samples_per_second": 24.538,
      "train_steps_per_second": 2.456
    }
  ],
  "max_steps": 1995,
  "num_train_epochs": 3,
  "total_flos": 2694600320090112.0,
  "trial_name": null,
  "trial_params": null
}