trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 45,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.67,
      "grad_norm": 0.612777054309845,
      "learning_rate": 8.000000000000001e-07,
      "loss": 0.4081,
      "step": 10
    },
    {
      "epoch": 1.33,
      "grad_norm": 0.4069637358188629,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 0.3882,
      "step": 20
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.8410348892211914,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.4163,
      "step": 30
    },
    {
      "epoch": 2.67,
      "grad_norm": 0.7110201716423035,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 0.376,
      "step": 40
    }
  ],
  "logging_steps": 10,
  "max_steps": 45,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 210090472243200.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}