Skip to content

BUG: "divide by zero encountered in matmul" on MacOS M4 with numpy v2.3.3 #29820

@maresb

Description

@maresb

Describe the issue:

We are seeing an error that appears identical to #28687 in our environment for the project brendanjmeade/celeri#259 (comment).

Reproduce the code example:

import numpy as np
print(f"numpy version: {np.__version__}")
print(np.identity(n=15) @ np.identity(n=15)) # produces warning
print(np.identity(n=14) @ np.identity(n=14)) # no warning

Error message:

numpy version: 2.3.3
/Users/meade/Desktop/celeri-org/celeri/celeri/scripts/celeri_solve.py:15: RuntimeWarning: divide by zero encountered in matmul
  print(f"{np.identity(n=15) @ np.identity(n=15)=}")
/Users/meade/Desktop/celeri-org/celeri/celeri/scripts/celeri_solve.py:15: RuntimeWarning: overflow encountered in matmul
  print(f"{np.identity(n=15) @ np.identity(n=15)=}")
/Users/meade/Desktop/celeri-org/celeri/celeri/scripts/celeri_solve.py:15: RuntimeWarning: invalid value encountered in matmul
  print(f"{np.identity(n=15) @ np.identity(n=15)=}")
np.identity(n=15) @ np.identity(n=15)=array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])
np.identity(n=14) @ np.identity(n=14)=array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])

Python and NumPy Versions:

Python 3.13.7
NumPy 2.3.3
MacOS: M4 with Sequoia 15.6.1 (24G90)

I will provide the exact command output on Monday.

Runtime Environment:

Installed from conda-forge via pixi install. lockfile

I will provide the exact command output on Monday

Context for the issue:

This bug results in a lot of spurious, confusing, and scary warnings that distract from real problems.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions