trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 20.0,
  "eval_steps": 500,
  "global_step": 3140,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.18,
      "learning_rate": 4.2038216560509556e-05,
      "loss": 1.2813,
      "step": 500
    },
    {
      "epoch": 6.37,
      "learning_rate": 3.407643312101911e-05,
      "loss": 0.2307,
      "step": 1000
    },
    {
      "epoch": 9.55,
      "learning_rate": 2.6114649681528662e-05,
      "loss": 0.1029,
      "step": 1500
    },
    {
      "epoch": 12.74,
      "learning_rate": 1.8152866242038215e-05,
      "loss": 0.0472,
      "step": 2000
    },
    {
      "epoch": 15.92,
      "learning_rate": 1.0191082802547772e-05,
      "loss": 0.0256,
      "step": 2500
    },
    {
      "epoch": 19.11,
      "learning_rate": 2.229299363057325e-06,
      "loss": 0.0184,
      "step": 3000
    },
    {
      "epoch": 20.0,
      "step": 3140,
      "total_flos": 1.95972567552e+16,
      "train_loss": 0.27299969211505476,
      "train_runtime": 864.6725,
      "train_samples_per_second": 115.651,
      "train_steps_per_second": 3.631
    }
  ],
  "logging_steps": 500,
  "max_steps": 3140,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 1.95972567552e+16,
  "trial_name": null,
  "trial_params": null
}