Describe what you are looking for

##Summary

Reductions like Tensor.sum and nk.sum only accept axis=None (global) or a single int. There is no NumPy-style axis=(0, 1, …) for collapsing several dimensions in one call.

Current behavior

• Stubs: axis: Optional[int] only (numkong/__init__.pyi — Tensor.sum, nk.sum).
• Runtime: non-int axis raises, e.g. TypeError: an integer is required.

Motivation

For arrays shaped like (H, W, C) or (N, H, W, C), per-channel sums require reducing all spatial axes while keeping the channel axis. In NumPy that is one reduction, e.g. np.sum(x, axis=tuple(range(x.ndim - 1))). With NumKong, callers must chain several sum(axis=…) calls (and handle keepdims carefully) or reshape, which is awkward and error-prone.

Request
Either:

1. Extend axis to int | tuple[int, ...] | None with NumPy-compatible semantics (and keepdims), at least for sum, or
2. Add a documented, first-class API for “sum over all axes except last” / multi-axis reductions if a tuple axis is intentionally out of scope.

import numpy as np import numkong as nk t = nk.Tensor(np.ones((2, 3, 4), dtype=np.float32)) t.sum(axis=(0, 1)) # TypeError: an integer is required

Can you contribute to the implementation?

• I can contribute

Is your feature request specific to a certain interface?

It applies to everything

Contact Details

No response

Is there an existing issue for this?

• I have searched the existing issues

Code of Conduct

• I agree to follow this project's Code of Conduct