How to search for a subnet on cifar100 using Autoformer?

[callmewenhao]

Hi bro! I got a question: what happens when we run evolution.py using dataset as cifar100 or cifar10?
It seems that this will use the test data of cifar100 and cifar10 for search. Does this cause data leakage?

Related code

dataset_val, args.nb_classes = build_dataset(is_train=False, args=args, folder_name="subImageNet")
dataset_test, _ = build_dataset(is_train=False, args=args, folder_name="val")

and

if args.data_set == 'CIFAR10':
    dataset = datasets.CIFAR10(args.data_path, train=is_train, transform=transform, download=True)
    nb_classes = 10
elif args.data_set == 'CIFAR100':
    dataset = datasets.CIFAR100(args.data_path, train=is_train, transform=transform, download=True)
    nb_classes = 100

Looking forward to your answer！🤗