trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 3068,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.33,
      "learning_rate": 4.185136897001304e-05,
      "loss": 0.7763,
      "step": 500
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.370273794002607e-05,
      "loss": 0.6385,
      "step": 1000
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5554106910039117e-05,
      "loss": 0.6041,
      "step": 1500
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.740547588005215e-05,
      "loss": 0.5271,
      "step": 2000
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.25684485006519e-06,
      "loss": 0.515,
      "step": 2500
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.108213820078227e-06,
      "loss": 0.5088,
      "step": 3000
    },
    {
      "epoch": 2.0,
      "step": 3068,
      "total_flos": 5.166258268431053e+16,
      "train_loss": 0.592957808732054,
      "train_runtime": 5580.5786,
      "train_samples_per_second": 140.739,
      "train_steps_per_second": 0.55
    }
  ],
  "max_steps": 3068,
  "num_train_epochs": 2,
  "total_flos": 5.166258268431053e+16,
  "trial_name": null,
  "trial_params": null
}