simet.providers.subsampled_provider¶

simet.providers.subsampled_provider ¶

SubsampledProvider ¶

SubsampledProvider(data_path=Path(), dataset=None)

Bases: Provider

Provider that wraps an already-built VisionDataset.

This provider is a thin adapter: instead of discovering data on disk, it simply returns a preconstructed VisionDataset (e.g., a subsampled or filtered view) when get_data(...) is called. The transform argument is ignored on purpose—assume the wrapped dataset already applies the desired preprocessing.

Parameters:

Name	Type	Description	Default
`data_path`	`Path`	Kept for API parity with `Provider`; not used by this implementation.	`Path()`
`dataset`	`VisionDataset`	The dataset instance to expose (e.g., a `Subset`, custom dataset, or any `VisionDataset` you prepared externally).	`None`

Attributes:

Name	Type	Description
`dataset`	`VisionDataset`	The wrapped dataset returned by `get_data()`.

Notes

Because transform is ignored, make sure the dataset you pass is already configured with the correct transforms.
data_path is retained for uniformity but is not consulted.
If dataset is omitted, the default VisionDataset() placeholder will likely raise at runtime; in practice you should always pass a concrete dataset.

Example

from torch.utils.data import Subset from torchvision.datasets import ImageFolder import torchvision.transforms as T

base = ImageFolder("data/train", transform=T.ToTensor()) small = Subset(base, indices=list(range(1000))) # subsampled view provider = SubsampledProvider(dataset=small) ds = provider.get_data(transform=... ) # transform ignored len(ds) 1000

Source code in simet/providers/subsampled_provider.py

def __init__(
    self,
    data_path: Path = Path(),
    dataset: VisionDataset = None,  # type: ignore
) -> None:
    super().__init__(data_path)
    self.dataset = dataset if dataset else VisionDataset()

get_data ¶

get_data(transform)

Return the wrapped dataset; transform is ignored.

Parameters:

Name	Type	Description	Default
`transform`	`Transform`	Unused. Present for interface compatibility.	required

Returns:

Name	Type	Description
`VisionDataset`	`VisionDataset`	The dataset provided at construction time.

Source code in simet/providers/subsampled_provider.py

def get_data(self, transform: Transform) -> VisionDataset:
    """Return the wrapped dataset; `transform` is ignored.

    Args:
        transform (Transform): Unused. Present for interface compatibility.

    Returns:
        VisionDataset: The dataset provided at construction time.
    """
    return self.dataset