count_nonzero

→Tensor

count_nonzero(x: Tensor, dim: int | Sequence[int] | None = None)

source edit

Count non-zero elements along dim (or whole tensor when None).

Parameters

xTensor

Input. Non-zero is determined by element-wise comparison to zero (NaN counts as non-zero).

dimint, sequence of int, or None= None

Axis or axes to reduce. None (default) reduces the entire tensor to a 0-D scalar.

Returns

Tensor

int64 count(s). Shape mirrors lucid.sum over the same dim argument.

Notes

Composite implementation: the CPU backend lacks where(bool, i64, i64) and astype(bool → i64), so the op builds a float32 1.0/0.0 mask, reduces, and astype-casts the float scalar back to int64 (a path the backend does support). Slightly more allocations than a native int reduction but semantically identical.