trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1560,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.96,
      "learning_rate": 3.397435897435898e-05,
      "loss": 3.6448,
      "step": 500
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.794871794871795e-05,
      "loss": 2.7292,
      "step": 1000
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.9230769230769234e-06,
      "loss": 2.2904,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1560,
      "total_flos": 358897071341568.0,
      "train_loss": 2.865567085070488,
      "train_runtime": 244.3152,
      "train_samples_per_second": 25.528,
      "train_steps_per_second": 6.385
    }
  ],
  "max_steps": 1560,
  "num_train_epochs": 3,
  "total_flos": 358897071341568.0,
  "trial_name": null,
  "trial_params": null
}