trainer_state.json

{
  "best_metric": 0.8805333333333333,
  "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0156/checkpoints/checkpoint-999",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 999,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.28086981177330017,
      "learning_rate": 5e-05,
      "loss": 0.8076,
      "step": 333
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8504,
      "eval_loss": 0.5132982134819031,
      "eval_runtime": 7.0162,
      "eval_samples_per_second": 534.478,
      "eval_steps_per_second": 8.409,
      "step": 333
    },
    {
      "epoch": 2.0,
      "grad_norm": NaN,
      "learning_rate": 5e-05,
      "loss": 0.2578,
      "step": 666
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8717333333333334,
      "eval_loss": 0.44963639974594116,
      "eval_runtime": 7.105,
      "eval_samples_per_second": 527.798,
      "eval_steps_per_second": 8.304,
      "step": 666
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.023187046870589256,
      "learning_rate": 5e-05,
      "loss": 0.1366,
      "step": 999
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8805333333333333,
      "eval_loss": 0.4148768186569214,
      "eval_runtime": 6.8028,
      "eval_samples_per_second": 551.246,
      "eval_steps_per_second": 8.673,
      "step": 999
    }
  ],
  "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": 4.94223961867776e+18,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}