trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 1355,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.85,
      "learning_rate": 3.1549815498154986e-05,
      "loss": 2.9715,
      "step": 500
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.3099630996309964e-05,
      "loss": 2.7861,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1355,
      "total_flos": 2829792706560000.0,
      "train_loss": 2.838261835865429,
      "train_runtime": 580.6833,
      "train_samples_per_second": 9.325,
      "train_steps_per_second": 2.333
    }
  ],
  "logging_steps": 500,
  "max_steps": 1355,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 2829792706560000.0,
  "trial_name": null,
  "trial_params": null
}