Cycler

class torchdata.datapipes.iter.Cycler(source_datapipe: IterDataPipe[T_co], count: int | None = None)

Cycles the specified input in perpetuity by default, or for the specified number of times (functional name: cycle).

If the ordering does not matter (e.g. because you plan to shuffle later) or if you would like to repeat an element multiple times before moving onto the next element, use Repeater.

Parameters:

• source_datapipe – source DataPipe that will be cycled through

• count – the number of times to read through source_datapipe` (if ``None, it will cycle in perpetuity)

Example

>>> from torchdata.datapipes.iter import IterableWrapper
>>> dp = IterableWrapper(range(3))
>>> dp = dp.cycle(2)
>>> list(dp)
[0, 1, 2, 0, 1, 2]