Skip to content

simet.providers.cifar

simet.providers.cifar

CIFARProvider 

CIFARProvider(data_path)

Bases: Provider

Provider that exposes the CIFAR-10 training split via VisionDataset.

Wraps torchvision.datasets.CIFAR10 with train=True, download=True, and applies the project Transform. Data will be downloaded (if missing) into self.data_path.

Notes
  • This provider is fixed to the training split (train=True). If you need the test/validation split, create a separate provider or make the split configurable.
  • The given Transform should handle any normalization/augmentation appropriate for CIFAR-10 (e.g., mean/std, random crops/flips).
Example

provider = CIFARProvider(data_path=Path("./data/cifar10")) ds = provider.get_data(transform=SomeTransform()) len(ds) # 50,000 images in the training split 50000

Source code in simet/providers/base/provider.py 
46
47
def __init__(self, data_path: Path) -> None:
    self.data_path = data_path

get_data 

get_data(transform)

Return the CIFAR-10 training dataset with the provided transform.

Parameters:

Name Type Description Default
transform Transform

Project transform wrapper; transform.get_transform() is passed to the underlying CIFAR10 dataset.

 required

Returns:

Name Type Description
VisionDataset VisionDataset

A CIFAR-10 training dataset ready for a DataLoader.
Source code in simet/providers/cifar.py 
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
@override
def get_data(self, transform: Transform) -> VisionDataset:
    """Return the CIFAR-10 **training** dataset with the provided transform.

    Args:
        transform: Project transform wrapper; `transform.get_transform()` is
            passed to the underlying `CIFAR10` dataset.

    Returns:
        VisionDataset: A CIFAR-10 training dataset ready for a `DataLoader`.
    """
    return CIFAR10(
        root=self.data_path,
        train=True,
        download=True,
        transform=transform.get_transform(),
    )