trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 0,
  "global_step": 2532,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.79,
      "learning_rate": 8.025276461295419e-06,
      "loss": 0.5594,
      "step": 500
    },
    {
      "epoch": 1.58,
      "learning_rate": 6.050552922590838e-06,
      "loss": 0.4529,
      "step": 1000
    },
    {
      "epoch": 2.37,
      "learning_rate": 4.075829383886256e-06,
      "loss": 0.3755,
      "step": 1500
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.101105845181675e-06,
      "loss": 0.2917,
      "step": 2000
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.263823064770932e-07,
      "loss": 0.1953,
      "step": 2500
    },
    {
      "epoch": 4.0,
      "step": 2532,
      "total_flos": 5252556024446976.0,
      "train_loss": 0.37261163435087763,
      "train_runtime": 593262.1709,
      "train_samples_per_second": 0.273,
      "train_steps_per_second": 0.004
    }
  ],
  "logging_steps": 500,
  "max_steps": 2532,
  "num_train_epochs": 4,
  "save_steps": 1.0,
  "total_flos": 5252556024446976.0,
  "trial_name": null,
  "trial_params": null
}