trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 2500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 17.461204528808594,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.5563,
      "step": 500
    },
    {
      "epoch": 2.0,
      "grad_norm": 9.144333839416504,
      "learning_rate": 1.2e-05,
      "loss": 0.3244,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "grad_norm": 26.32133674621582,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.1664,
      "step": 1500
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.11387253552675247,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0827,
      "step": 2000
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.02141467295587063,
      "learning_rate": 0.0,
      "loss": 0.0431,
      "step": 2500
    },
    {
      "epoch": 5.0,
      "step": 2500,
      "total_flos": 1.863396760974336e+16,
      "train_loss": 0.2345771484375,
      "train_runtime": 1455.9557,
      "train_samples_per_second": 13.733,
      "train_steps_per_second": 1.717
    }
  ],
  "logging_steps": 500,
  "max_steps": 2500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 1.863396760974336e+16,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}