Fix #982: update CI and environment dependencies with proper formatting

[mariam851]

Hi @rabst,

Description

This PR addresses the issue #982 where test failures occurred in the feature_selection module due to a dependency conflict between NumPy 2.4.0 and Numba.

Changes

• Updated environment.yml to pin NumPy to >=1.21.0,<2.4.0.
• Modified .github/workflows/python-package-pip.yml and python-package-conda.yml to ensure stable NumPy versions during CI runs.
• Verified that all 95 tests in mlxtend/feature_selection/ pass successfully in a local environment.

[rasbt]

Thanks for the PR! Maybe I missed (or forgot) something, but why is numba added?

[mariam851]

Hi @rasbt, you're right. numba isn't a direct dependency of mlxtend. I added it to the Conda workflow because of the compatibility issues between NumPy 2.x and older Numba versions which were causing failures during environment setup in the CI.

However, since NumPy is now pinned to <2.4.0, we probably don't need to install numba explicitly anymore. I can remove it from the workflow to keep the dependencies minimal. Would you like me to do that?

[rasbt]

Hi @rasbt, you're right. numba isn't a direct dependency of mlxtend. I added it to the Conda workflow because of the compatibility issues between NumPy 2.x and older Numba versions which were causing failures during environment setup in the CI.

I see, thanks for explaining, that makes sense. I removed it here in the PR to see if it works, and if not, we can add it.

[rasbt]

Actually, I am just looking at #982 and it's from 2022. This one has been addressed a long time ago. I don't think we need to change anything because there was no failure in main.