PERM MNIST

Classes

class datasets.perm_mnist.MyMNIST(root, train=True, transform=None, target_transform=None, download=False)

Bases: MNIST

Overrides the MNIST dataset to change the getitem function.

class datasets.perm_mnist.PermutedMNIST(args)

Bases: ContinualDataset

Permuted MNIST Dataset.

Creates a dataset composed by a sequence of tasks, each containing a different permutation of the pixels of the MNIST dataset.

Parameters:

• NAME (str) – name of the dataset

• SETTING (str) – setting of the experiment

• N_CLASSES_PER_TASK (int) – number of classes in each task

• N_TASKS (int) – number of tasks

• SIZE (tuple) – size of the images

NAME: str = 'perm-mnist'

N_CLASSES_PER_TASK: int = 10

N_TASKS: int = 20

SETTING: str = 'domain-il'

SIZE: Tuple[int] = (28, 28)

get_backbone()

get_batch_size()

Return type:

int

get_class_names()

get_data_loaders()

Return type:

Tuple[DataLoader, DataLoader]

static get_denormalization_transform()

get_epochs()

static get_loss()

static get_normalization_transform()

static get_transform()