trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2625,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.57,
      "learning_rate": 4.047619047619048e-05,
      "loss": 2.5903,
      "step": 500
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.095238095238095e-05,
      "loss": 2.489,
      "step": 1000
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.1428571428571428e-05,
      "loss": 2.3848,
      "step": 1500
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.1904761904761905e-05,
      "loss": 2.3436,
      "step": 2000
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.3809523809523808e-06,
      "loss": 2.3011,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2625,
      "total_flos": 1.278595580424192e+16,
      "train_loss": 2.4165630696614584,
      "train_runtime": 3505.1584,
      "train_samples_per_second": 5.991,
      "train_steps_per_second": 0.749
    }
  ],
  "max_steps": 2625,
  "num_train_epochs": 3,
  "total_flos": 1.278595580424192e+16,
  "trial_name": null,
  "trial_params": null
}