last-checkpoint/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.691164972634871,
  "eval_steps": 500,
  "global_step": 12000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.17,
      "learning_rate": 0.0001,
      "loss": 0.204,
      "step": 3000
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0001,
      "loss": 0.1084,
      "step": 6000
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0001,
      "loss": 0.0949,
      "step": 9000
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0001,
      "loss": 0.0866,
      "step": 12000
    }
  ],
  "logging_steps": 3000,
  "max_steps": 12790,
  "num_train_epochs": 5,
  "save_steps": 3000,
  "total_flos": 158256349577216.0,
  "trial_name": null,
  "trial_params": null
}