combinations

→Tensor

combinations(input: Tensor, r: int = 2, with_replacement: bool = False)

source edit

All r-length combinations of the elements of a 1-D input.

Composite over Python itertools — non-differentiable (treats the values as opaque scalars). Use only for low-rate utility code; don't put on a hot path.

Parameters

inputTensor

1-D source tensor of length

n

. Raises on higher rank.

rint= 2

Combination length. Default 2.

with_replacementbool= False

When True the output enumerates multisets (an element can appear multiple times in a single row), giving

\binom{n + r - 1}{r}

rows. Default False (strict combinations,

\binom{n}{r}

rows).

Returns

Tensor

Shape (C, r) where $C$ is the combination count. Dtype and device follow input.