class anndata.experimental.AnnLoader(adatas, batch_size=1, shuffle=False, use_default_converter=True, use_cuda=False, **kwargs)[source]#

PyTorch DataLoader for AnnData objects.

Builds DataLoader from a sequence of AnnData objects, from an AnnCollection object or from an AnnCollectionView object. Takes care of the required conversions.

adatas Sequence[AnnData] | dict[str, AnnData]

AnnData objects or an AnnCollection object from which to load the data.

batch_size int (default: 1)

How many samples per batch to load.

shuffle bool (default: False)

Set to True to have the data reshuffled at every epoch.

use_default_converter bool (default: True)

Use the default converter to convert arrays to pytorch tensors, transfer to the default cuda device (if use_cuda=True), do memory pinning (if pin_memory=True). If you pass an AnnCollection object with prespecified converters, the default converter won’t overwrite these converters but will be applied on top of them.

use_cuda bool (default: False)

Transfer pytorch tensors to the default cuda device after conversion. Only works if use_default_converter=True


Arguments for PyTorch DataLoader. If adatas is not an AnnCollection object, then also arguments for AnnCollection initialization.