Outdated code in scikit-learn-extra

[TimotheeMathieu]

As of now, there is not really anybody maintaining this library. I tried to do a few things but I can't really debug code that I did not write and I don't understand (I am talking in particular about eigenpro that gives weird dimensionality error in the CI, see #175 ).

This means I cannot really do a new release and this seems to be problematic because there are a few people that want to install the library but do not seem to manage to do it with the current release (see #164).

For now, I still manage to install from the source, so anybody having installation problem can try to install scikit-learn-extra with the following command:

pip install git+https://github.com/scikit-learn-contrib/scikit-learn-extra

EDIT: If this does not work, use the following fork:

pip install git+https://github.com/TimotheeMathieu/scikit-learn-extra

[handwerkerd]

Hello. I was able to install from source based on your command above, but when I tried to import in an environment with numpy>2.0 I get the error:

>>> import sklearn_extra

A module that was compiled using NumPy 1.x cannot be run in
NumPy 2.1.1 as it may crash. To support both 1.x and 2.x
versions of NumPy, modules must be compiled with NumPy 2.0.
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.

If you are a user of the module, the easiest solution will be to
downgrade to 'numpy<2' or try to upgrade the affected module.
We expect that some modules will need time to support NumPy 2.

I'm not sure if this is a build issue that's specific to sklearn_extra if if you have another dependency that has the same numpy version issue.

[TimotheeMathieu]

Ah yes, sorry I fixed that in the PR but could not merge due to failing CI. You can try installing the version from the PR (which comes from my fork):

pip install git+https://github.com/TimotheeMathieu/scikit-learn-extra

[handwerkerd]

That works. Thank you!

[kubaziz1]

@TimotheeMathieu Hello, I just wanted to ask if there is some alternative install of scikit-learn-extra for version 13?

[TimotheeMathieu]

Hello. Nothing for python 3.13 yet, it has not been a long time since the release and python 3.13 is not available on major package managers yet so it is not tested in scikit-learn-extra.

[AVHopp]

@TimotheeMathieu first thanks for your support here :) I run in similar issues when trying to install on an M1 Mac and tried all of your suggestions, but always get the following error:

(.......)

I just wanted to ask for your expertise - is this also related to the same issues discussed here and in #164 or is this some other error that I am running into? Thanks in advance :)

[TimotheeMathieu]

@AVHopp: I am not sure about your error, it seems linked to MacOS somehow but I don't have this in the MacOS CI (i.e. the installation work just fine for python 3.10 using the command I mention in the first post of this issue). Maybe try to install using the fork mentionned in the first post and be careful about using a virtual environment in order to isolate the installation from the system python install.

[AVHopp]

Already did that, and this did unfortunately not solve the problem for me :/ I will try to investigate further and see if I can find out more

[MarkcoHK]

Hi, are there anyone still facing the same problem? I got the same problem as @AVHopp , and I have no cue how to fix it. I even created new environnment for python ver.3.10, but I still got the ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (scikit-learn-extra).
Anyone faced this problem and solve it successfully?

[AVHopp]

I was able to fix it by explicitly using a x86 environment. I did this by creating a micromamba environment and specifying the "CONDA_SUBDIR" environment variable. This then creates an environment using this architecture for all packages, so depending on other packages you want to use, this might not be a solution.

[TimotheeMathieu]

In order to do a release in spite of the aforementionned problem with some outdated code, I removed the outdated code from the PR and then the binaries build alright. If somebody wants to help repair eigenpro they are welcome to open a PR, otherwise I will soon do the release with the updated code from PR #175 and doing this release should solve most of the installation problems.

[MarkcoHK]

@AVHopp it works for me, appreciate that !! I am using M-chips macbook, and I directly create new env. with x64 as the code below.

CONDA_SUBDIR=osx-64 conda create -n (ENV. NAME) python
conda activate (ENV. NAME)
conda config --env --set subdir osx-64

[MarkcoHK]

@TimotheeMathieu First thanks your maintenance and support here!!!
I still got some issues by downloading the new release, but I solve by the method mentioned.

[AVHopp]

@MarkcoHK glad that I could help :)