8000 CI/FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat by ogrisel · Pull Request #27899 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

CI/FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat #27899

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
lesteve merged 5 commits into from
Dec 4, 2023
Merged

CI/FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat #27899

lesteve merged 5 commits into from
Dec 4, 2023

Conversation

@ogrisel
Copy link
Member
@ogrisel ogrisel commented Dec 4, 2023 •
edited
Loading

While reviewing #27075, I realized that the current state of our Linux_Nightly pylatest_pip_scipy_dev does not actually test against numpy dev without us upper-bounding the version of numpy explicitly (yet):

https://dev.azure.com/scikit-learn/scikit-learn/_build/results?buildId=61365&view=logs&j=dfe99b15-50db-5d7b-b1e9-4105c42527cf&t=eb5122d5-ab7e-5479-a8ce-245b4d64938b&l=442

The goal of this PR is to trigger a run to:

  • check if scipy and pandas actually break at import time when running against numpy 2 dev,
  • if not, check if scikit-learn itself can import successfully and then run its own tests,
  • fix scikit-learn itself in case of small problems revealed by the test results.

@github-actions
Copy link
github-actions bot commented Dec 4, 2023 •
edited
Loading

✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

Generated for commit: ed58999. Link to the linter CI: here

@ogrisel
Copy link
Member Author
ogrisel commented Dec 4, 2023

Good news. Most scikit-learn tests pass (thanks to the fixes contributed earlier by @seberg). We just need to fix a test with unexpected dtype values. I suspect the changes in dtype promotion rules: https://numpy.org/neps/nep-0050-scalar-promotion.html.

I will have a look.

glemaitre
glemaitre reviewed Dec 4, 2023
View reviewed changes
Copy link
Member
@glemaitre glemaitre left a comment

Choose a reason for hiding this comment

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

LGTM. If the fix is simply in a test, I could imagine to add it inside this PR as well.

@ogrisel
Copy link
Member Author
ogrisel commented Dec 4, 2023

LGTM. If the fix is simply in a test, I could imagine to add it inside this PR as well.

Not it's not just the test. The dtype of some fitted attribute is upcasted. I have a fix, I will push to the PR and update the title.

@ogrisel ogrisel changed the title CI: force install numpy dev in [scipy-dev] build Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat Dec 4, 2023
@ogrisel ogrisel added the Bug label Dec 4, 2023
@ogrisel ogrisel added this to the 1.4 milestone Dec 4, 2023
@ogrisel
Copy link
Member Author
ogrisel commented Dec 4, 2023

I milestoned this PR 1.4 so that scikit-learn would not have any numpy 2 incompatible code in it.

@ogrisel ogrisel mentioned this pull request Dec 4, 2023
Closed
4 tasks
@glemaitre glemaitre changed the title Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat Dec 4, 2023
@glemaitre glemaitre changed the title FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat CI/FIX Enable numpy dev in [scipy-dev] build and fix Bayesian linear models for numpy 2 compat Dec 4, 2023
glemaitre
glemaitre approved these changes Dec 4, 2023
View reviewed changes
Copy link
Member
@glemaitre glemaitre left a comment

Choose a reason for hiding this comment

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

LGTM.

lesteve
lesteve approved these changes Dec 4, 2023
View reviewed changes
Copy link
Member
@lesteve lesteve left a comment

Choose a reason for hiding this comment

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

LGTM

It seems like float32< 8000 /code> dtype was maintained in numpy<2 by relying on float64_scalar * float32_array => float32_array (try executing np.float64(1.) * np.ones(2, dtype=np.float32) if you want to double-check)

The code in this PR is more careful about dtype.

@lesteve
Copy link
Member
lesteve commented Dec 4, 2023

And also it seems like pandas dev wheel is the one that wants numpy<2 ...

@lesteve lesteve enabled auto-merge (squash) December 4, 2023 15:44
@lesteve lesteve merged commit 87e6908 into scikit-learn:main Dec 4, 2023
@thomasjpfan thomasjpfan mentioned this pull request Dec 4, 2023
Closed
@ogrisel ogrisel deleted the test-against-numpy-dev-in-scipy-dev-build branch December 8, 2023 10:21
@lesteve lesteve mentioned this pull request Dec 19, 2023
Closed
3 tasks
@lesteve lesteve mentioned this pull request Jan 18, 2024
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@lesteve lesteve lesteve approved these changes

@glemaitre glemaitre glemaitre approved these changes

Assignees

No one assigned

Labels

Bug Build / CI

Projects

None yet

Milestone

1.4

Development

Successfully merging this pull request may close these issues.

3 participants

@ogrisel @lesteve @glemaitre
0