trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 1000,
  "global_step": 1530,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.2679738562091503,
      "grad_norm": 9.848502159118652,
      "learning_rate": 2.148648648648649e-06,
      "loss": 1.5025,
      "step": 1000
    },
    {
      "epoch": 3.2679738562091503,
      "eval_loss": 1.684686303138733,
      "eval_mse": 1.684686455470108,
      "eval_runtime": 4.3805,
      "eval_samples_per_second": 57.071,
      "eval_steps_per_second": 3.653,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1530,
      "total_flos": 1.140221868071424e+16,
      "train_loss": 1.0542351367426854,
      "train_runtime": 1107.9245,
      "train_samples_per_second": 22.086,
      "train_steps_per_second": 1.381
    }
  ],
  "logging_steps": 1000,
  "max_steps": 1530,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 1.140221868071424e+16,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}