expand

→Tensor

expand(input: Tensor, sizes: int | Sequence[int] = ())

source

Return a view with singleton dimensions broadcast to new sizes.

For each axis whose current size is 1, expand replaces it with the requested size by setting the stride along that axis to 0. Axes that already have non-1 sizes must be passed as -1 (or the same size). No data is copied.

Parameters

inputTensor

Source tensor.

sizesint or sequence of int= ()

Target sizes. -1 means "keep current size".

Returns

Tensor

A view sharing storage with input.

Notes

expand is the in-place-style cousin of broadcast_to; they produce equivalent views but expand requires the input to already have a 1-sized dim where the expansion happens. Writing into the result is undefined because multiple logical entries alias the same physical element.

Examples

>>> import lucid
>>> x = lucid.tensor([[1.0], [2.0], [3.0]])  # shape (3, 1)
>>> lucid.expand(x, 3, 4).shape
(3, 4)