torch_concepts.data.datasets.mnist.MNISTAddition¶

class MNISTAddition(root, train, target_transform=None, download=True)[source]¶

The MNIST addition dataset is a modified version of the MNIST dataset where each image is a concatenation of two MNIST images and the target label is the sum of the two digits. The concept label is a one-hot encoding of the two digits.

concept_names¶: The names of the concept labels.

task_names¶: The names of the task labels.

root¶: The root directory where the dataset is stored.

train¶: Whether to load the training or test split. Default is False.

transform¶: The transformations to apply to the images. Default is None.

target_transform¶: The transformations to apply to the target labels. Default is None.

download¶: Whether to download the dataset if it does not exist. Default is False.

__init__(root, train, target_transform=None, download=True)[source]¶

Methods

`__init__`(root, train[, target_transform, ...])
`download`()	Download the MNIST data if it doesn't exist already.
`extra_repr`()

Attributes

`class_to_idx`
`classes`
`concept_names`
`input_dim`
`input_shape`
`mirrors`
`n_concepts`
`n_tasks`
`name`
`processed_folder`
`raw_folder`
`resources`
`task_names`
`test_data`
`test_file`
`test_labels`
`train_data`
`train_labels`
`training_file`
`transform`