trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 3360,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.49,
      "learning_rate": 4.255952380952381e-05,
      "loss": 7.4464,
      "step": 500
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.511904761904762e-05,
      "loss": 4.3785,
      "step": 1000
    },
    {
      "epoch": 4.46,
      "learning_rate": 2.767857142857143e-05,
      "loss": 3.9619,
      "step": 1500
    },
    {
      "epoch": 5.95,
      "learning_rate": 2.023809523809524e-05,
      "loss": 3.7302,
      "step": 2000
    },
    {
      "epoch": 7.44,
      "learning_rate": 1.2797619047619047e-05,
      "loss": 3.5937,
      "step": 2500
    },
    {
      "epoch": 8.93,
      "learning_rate": 5.357142857142857e-06,
      "loss": 3.5297,
      "step": 3000
    },
    {
      "epoch": 10.0,
      "step": 3360,
      "total_flos": 1.59009668522496e+16,
      "train_loss": 4.338697015671503,
      "train_runtime": 1250.2518,
      "train_samples_per_second": 26.875,
      "train_steps_per_second": 2.687
    }
  ],
  "max_steps": 3360,
  "num_train_epochs": 10,
  "total_flos": 1.59009668522496e+16,
  "trial_name": null,
  "trial_params": null
}