trainer_state.json

{
  "best_metric": 0.2942250669002533,
  "best_model_checkpoint": "vit-large-patch16-224-dungeon-geo-morphs-001/checkpoint-10",
  "epoch": 10.0,
  "eval_steps": 10,
  "global_step": 15,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 6.666666666666667,
      "grad_norm": 2.7785141468048096,
      "learning_rate": 2e-05,
      "loss": 0.9027,
      "step": 10
    },
    {
      "epoch": 6.666666666666667,
      "eval_accuracy": 1.0,
      "eval_loss": 0.2942250669002533,
      "eval_runtime": 0.409,
      "eval_samples_per_second": 73.344,
      "eval_steps_per_second": 9.779,
      "step": 10
    },
    {
      "epoch": 10.0,
      "step": 15,
      "total_flos": 1.2874949963753472e+17,
      "train_loss": 0.6321113010247549,
      "train_runtime": 81.0916,
      "train_samples_per_second": 8.694,
      "train_steps_per_second": 0.185
    }
  ],
  "logging_steps": 10,
  "max_steps": 15,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 15,
  "save_steps": 10,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.2874949963753472e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}