class

SequentialSampler

extendsSampler

SequentialSampler(data_source: Dataset)

source

Yield indices in fixed order 0, 1, ..., len(data_source) - 1.

The default sampler used when shuffle=False and no explicit sampler is supplied to DataLoader.

Parameters

data_sourceDataset

Dataset whose __len__ determines the index range.

Notes

Order is fully deterministic — no RNG is consulted — so two passes over the sampler always yield the same index sequence. This is the right choice for evaluation / inference loops where ordering must line up with external bookkeeping (e.g. per-row metric arrays).

Examples

>>> sampler = SequentialSampler(my_dataset)
>>> list(sampler)[:5]
[0, 1, 2, 3, 4]

Methods (3)

dunder

init

→None

__init__(data_source: Dataset)

source

Store data_source for length introspection.

Parameters

data_sourceDataset

Dataset to be iterated sequentially.

dunder

iter

→Iterator[int]

__iter__()

source

Yield 0, 1, ..., len(data_source) - 1 in order.

dunder

len

→int

__len__()

source

Return len(data_source).