CIFAR10#

class dmme.data.CIFAR10(data_dir: str = '.', batch_size: int = 128, augs: Optional[List[Callable]] = None)[source]#

CIFAR10 Dataset scaled to \([-1, 1]\)

Download CIFAR10 using torchvision and apply augmentations from argument and finally scale images to \([-1, 1]\)

Parameters:

  • data_dir (str) – path to cifar10

  • batch_size (int) – batch size

  • augs (List[Transform]) – augmentations on PIL images only from torchvision

prepare_data()[source]#

Download dataset

dataset(augs=[])[source]#

Dataset builder

setup_train()[source]#

Prepare training set

setup_test()[source]#

Prepare test set