BUG: Avoid NumPy 2 deprecation warning in __array_wrap__

@mruberry

🐛 Describe the bug

In NumPy 2, NumPy added a DeprecationWarning on __array_wrap__ to extend the information passed in and avoid an indefinite try/except chain.

This means that the e.g. the following gives a warning:

import warnings

import numpy as np
import torch

warnings.filterwarnings("error")  # with this an error

np.delete(torch.arange(10), 0)

Now, the solution is straight forward: Just add the two new arguments and ignore them.
(torch has no scalars, and the context is info that was always there and you don't need). That is __array_wrap__(obj, context=None, return_scalar=None).

I'll guess end-users use this path but because of how DeprecationWarnings work they won't actually see them normally, so it makes sense it wasn't spotted.
It would be nice to escalate the warning to something visible, but I would prefer if by that time torch has the fix.

Versions

current main.

cc @mruberry @rgommers

I'd like to work on this. The fix is straightforward: update __array_wrap__ in torch/_tensor.py to accept context=None, return_scalar=None and ignore them, suppressing the NumPy 2 deprecation warning.

Fix NumPy 2 DeprecationWarning in Tensor.__array_wrap__

Fix NumPy 2 DeprecationWarning in Tensor.__array_wrap__ (pytorch#187190)

Fix NumPy 2 DeprecationWarning in Tensor.__array_wrap__ (pytorch#187190…

Navigation Menu

Repository navigation

BUG: Avoid NumPy 2 deprecation warning in `__array_wrap__` #180657

🐛 Describe the bug

Versions

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

BUG: Avoid NumPy 2 deprecation warning in __array_wrap__ #180657

Description

🐛 Describe the bug

Versions

Activity

ogulcanaydogan commented on Apr 28, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

BUG: Avoid NumPy 2 deprecation warning in `__array_wrap__` #180657