trainer_state.json

{
  "best_metric": 0.8784,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0220/checkpoints/checkpoint-666",
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 666,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 22.285633087158203,
      "learning_rate": 5e-05,
      "loss": 0.7794,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8434666666666667,
      "eval_loss": 0.5113060474395752,
      "eval_runtime": 6.2861,
      "eval_samples_per_second": 596.554,
      "eval_steps_per_second": 9.386,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": NaN,
      "learning_rate": 5e-05,
      "loss": 0.2127,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8784,
      "eval_loss": 0.4164274036884308,
      "eval_runtime": 6.5792,
      "eval_samples_per_second": 569.979,
      "eval_steps_per_second": 8.968,
      "step": 666
    }
  ],
  "logging_steps": 500,
  "max_steps": 2664,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 8,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 3.29482641245184e+18,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}