trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1671,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.9,
      "learning_rate": 3.5038898862956315e-05,
      "loss": 3.0168,
      "step": 500
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.0077797725912628e-05,
      "loss": 2.2677,
      "step": 1000
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.116696588868941e-06,
      "loss": 1.9156,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1671,
      "total_flos": 401468825419776.0,
      "train_loss": 2.3393769826381097,
      "train_runtime": 272.9817,
      "train_samples_per_second": 24.463,
      "train_steps_per_second": 6.121
    }
  ],
  "max_steps": 1671,
  "num_train_epochs": 3,
  "total_flos": 401468825419776.0,
  "trial_name": null,
  "trial_params": null
}