Attention

June 2024 Status Update: Removing DataPipes and DataLoader V2

We are re-focusing the torchdata repo to be an iterative enhancement of torch.utils.data.DataLoader. We do not plan on continuing development or maintaining the [DataPipes] and [DataLoaderV2] solutions, and they will be removed from the torchdata repo. We’ll also be revisiting the DataPipes references in pytorch/pytorch. In release torchdata==0.8.0 (July 2024) they will be marked as deprecated, and in 0.9.0 (Oct 2024) they will be deleted. Existing users are advised to pin to torchdata==0.8.0 or an older version until they are able to migrate away. Subsequent releases will not include DataPipes or DataLoaderV2. Please reach out if you suggestions or comments (please use this issue for feedback)

Forker

class torchdata.datapipes.iter.Forker(datapipe: IterDataPipe, num_instances: int, buffer_size: int = 1000, copy: Optional[Literal['shallow', 'deep']] = None)

Creates multiple instances of the same Iterable DataPipe (functional name: fork).

Parameters:

• datapipe – Iterable DataPipe being copied

• num_instances – number of instances of the datapipe to create

• buffer_size – this restricts how far ahead the leading child DataPipe can read relative to the slowest child DataPipe. Defaults to 1000. Use -1 for the unlimited buffer.

• copy – copy strategy to use for items yielded by each branch. Supported options are None for no copying, "shallow" for shallow object copies, and "deep" for deep object copies. Defaults to None.

Note

All branches of the forked pipeline return the identical object unless the copy parameter is supplied. If the object is mutable or contains mutable objects, changing them in one branch will affect all others.

Example

>>> # xdoctest: +REQUIRES(module:torchdata)
>>> from torchdata.datapipes.iter import IterableWrapper
>>> source_dp = IterableWrapper(range(5))
>>> dp1, dp2 = source_dp.fork(num_instances=2)
>>> list(dp1)
[0, 1, 2, 3, 4]
>>> list(dp2)
[0, 1, 2, 3, 4]