MinimalDataset

class axtreme.data.dataset.MinimalDataset(data: Sequence[T] | T)

Bases: Dataset[T]

Creates a Dataset over a list-like data source.

Note

list, np.array or tensor forfill the __getitem__ and __len__ requirement directly - but this makes them conform with the __add__ behaviour defined for datasets.

__init__(data: Sequence[T] | T) → None

Creates a Dataset over a list-like data source.

Parameters:

• data – Supports being indexed (supports __getitem__) and len, where index values are considered datapoints.

• e.g (-)

• DataLoader. (- Datapoints should be compatible with) –

  • General work with numerics and matrixes

  • Specifics of what DataLoader consume from datasets see torch._utils.collate.default_collate

Examples

>>> data = [1, 2, 3]
>>> ds = MinimalDataset(data)
>>> ds[0]
1

>>> data = np.array([[1, 2, 3], [4, 5, 6]])
>>> ds = MinimalDataset(data)
>>> ds[0]
1p.array([1,2,3])

>>> data = torch.tensor([[1], [2]])
>>> ds = MinimalDataset(data)
>>> ds[0]
torch.tensor([1])

Methods

__init__(data)

Creates a Dataset over a list-like data source.