inferno.io.core package

Submodules

inferno.io.core.base module

class inferno.io.core.base.IndexSpec(index=None, base_sequence_at_index=None)

Bases: object

Class to wrap any extra index information a Dataset object might want to send back. This could be useful in (say) inference, where we would wish to (asynchronously) know more about the current input.

class inferno.io.core.base.SyncableDataset

Bases: torch.utils.data.dataset.Dataset

sync_with(dataset)

inferno.io.core.concatenate module

class inferno.io.core.concatenate.Concatenate(*datasets, transforms=None)

Bases: torch.utils.data.dataset.Dataset

Concatenates mutliple datasets to one. This class does not implement synchronization primitives.

map_index(index)

inferno.io.core.data_utils module

inferno.io.core.data_utils.defines_base_sequence(dataset)

inferno.io.core.data_utils.implements_sync_primitives(dataset)

inferno.io.core.zip module

class inferno.io.core.zip.Zip(*datasets, sync=False, transforms=None)

Bases: inferno.io.core.base.SyncableDataset

Zip two or more datasets to one dataset. If the datasets implement synchronization primitives, they are all synchronized with the first dataset.

sync_datasets()

sync_with(dataset)

class inferno.io.core.zip.ZipReject(*datasets, sync=False, transforms=None, rejection_dataset_indices, rejection_criterion)

Bases: inferno.io.core.zip.Zip

Extends Zip by the functionality of rejecting samples that don’t fulfill a specified rejection criterion.

fetch_from_rejection_datasets(index)

Module contents