Skip to content

ci: test with torch 2.13 and numpy 1.26 - #9

Merged
maxg0811 merged 1 commit into
mainfrom
ci/test-torch-213-numpy-matrix
Nov 16, 2026
Merged

maxg0811 merged 1 commit into
mainfrom
ci/test-torch-213-numpy-matrix

Conversation

@maxg0811

@maxg0811 maxg0811 commented Nov 16, 2026 •

Copy link
Copy Markdown

Why

The Test workflow fails on main (run). The setup action installed torch 2.0.0, which is built against numpy 1. The unpinned pip install scipy then pulled in numpy 2.2.6, which broke torch and numpy interop (Failed to initialize NumPy: _ARRAY_API not found). 4 tests failed with Numpy is not available and Could not infer dtype of numpy.float32.

Changes

  • Bump the setup action's default torch version from 2.0.0 to 2.13.0.
  • Replace the unpinned scipy install with pyg-lib==0.8.0. InducedNeighborSampler needs pyg-lib or torch-sparse.
  • Remove the unused full_install input.
  • Pin numpy 1.26.4 and scipy 1.10.1 in the test install.

Testing

Local run on Python 3.10 with torch 2.13.0, numpy 1.26.4, scipy 1.10.1 and pyg-lib: 9 passed, 1 skipped. Without pyg-lib, the sampler test fails, which is why the install stays.

🤖 Generated with Claude Code

@ox-security

ox-security Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

OX Security Logo

OX Security reviewed this pull request — nothing to fix.

✔ No issues found

Branch ci/test-torch-213-numpy-matrix → main

View scan in OX Security →

@maxg0811 maxg0811 changed the title ci: test with torch 2.13 against numpy 1 and numpy 2 ci: test with torch 2.13 and numpy 1.26.4 Nov 16, 2026
@codecov

codecov Bot commented Nov 16, 2026

Copy link
Copy Markdown

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

ℹ️ You can also turn on project coverage checks and project coverage reporting on Pull Request comment

Thanks for integrating Codecov - We've got you covered ☂️

CI installed torch 2.0.0, which is built against numpy 1, but an unpinned
scipy install pulled in numpy 2.2.6. That broke torch and numpy interop
and failed 4 tests.

- Bump the setup action's default torch version to 2.13.0.
- Install pyg-lib 0.8.0 instead of unpinned scipy, since
  InducedNeighborSampler needs pyg-lib or torch-sparse.
- Remove the unused full_install input.
- Pin numpy 1.26.4 and scipy 1.10.1 in the test install.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@maxg0811
maxg0811 force-pushed the ci/test-torch-213-numpy-matrix branch from dc4c06b to 5bd5342 Compare November 16, 2026 10:52
@maxg0811 maxg0811 changed the title ci: test with torch 2.13 and numpy 1.26.4 ci: test with torch 2.13 and numpy 1.26 Nov 16, 2026

@philippj2 philippj2 left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@maxg0811
maxg0811 merged commit 257d768 into main Nov 16, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants