Version 1.0 release

[adrinjalali]

This is the issue to cover things we need to do to finish the release.

@scikit-learn/core-devs please tag this issue when you want a PR to be backported to 1.0 from now :)

Question: the travis job triggered by [cd build] succeeded, but we don't have any ARM wheels on the anaconda repo, I'm puzzled.

• update news and what's new date in release branch
• update news and what's new date and sklearn dev0 version in main branch
• check that the for the release wheels can be built successfully
• merge the PR with [cd build] commit message to upload wheels to the staging repo
• upload the wheels and source tarball to https://test.pypi.org
• create tag on the main github repo
• upload the wheels and source tarball to PyPI
• https://github.com/scikit-learn/scikit-learn/releases draft
• confirm bot detected at
  https://github.com/conda-forge/scikit-learn-feedstock and wait for merge
• https://github.com/scikit-learn/scikit-learn/releases publish
• fix the binder release version in .binder/requirement.txt (see
  [MRG] More explicit binder requirements for 0.22 #15847)
• announce on mailing list, LinkedIn, and on twitter

[glemaitre]

The ARM are located there: https://pypi.anaconda.org/scikit-learn-wheels-staging/simple/scikit-learn/

scikit_learn-1.0rc1-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
scikit_learn-1.0rc1-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
scikit_learn-1.0rc1-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl

[glemaitre]

However, something is weird with the naming (basically for all manylinux that has 2 dots). I don't know if this would be an issue when installing.

[adrinjalali]

what about the M1 chip wheels? are we doing them?

[glemaitre]

Not yet. We are awaiting for SciPy to provide a wheel. It is almost there but @ogrisel just discovered the following bug a couple of days ago: scipy/scipy#14688

[adrinjalali]

The release candidate is uploaded, and the tag is created on the release branch: https://pypi.org/project/scikit-learn/1.0rc1/

[adrinjalali]

Documentation: https://scikit-learn.org/1.0/

[tupui]

I see you dropped Python 3.6 for this release. Why not dropping Python 3.7? Both NumPy and SciPy dropped it on master (https://scipy.github.io/devdocs/dev/toolchain.html and xref scipy/scipy#14655 ). And it would be closer to NEP29.

[adrinjalali]

We probably would for next release. But I'm happy to have 3 python release support for our release.

[reshamas]

@adrinjalali
Can we update top note to include: share on LinkedIn page:
https://www.linkedin.com/company/scikit-learn/

[adrinjalali]

@reshamas which note do you mean?

[reshamas]

@adrinjalali

@reshamas which note do you mean?

The checklist at very top of this page has an item for "announce on mailing list and on twitter".
Can we add an item for: "share on scikit-learn LinkedIn page"?

[ogrisel]

I see you dropped Python 3.6 for this release. Why not dropping Python 3.7? Both NumPy and SciPy dropped it on master (https://scipy.github.io/devdocs/dev/toolchain.html and xref scipy/scipy#14655 ). And it would be closer to NEP29.

I would take NEP29 as a minimal requirement for scikit-learn. If it's easy for us to support a little more, then let's do it. I like the last 3 Python versions rule of thumb. It's still manageable from a CI maintenance point of view.

[tupui]

I see you dropped Python 3.6 for this release. Why not dropping Python 3.7? Both NumPy and SciPy dropped it on master (https://scipy.github.io/devdocs/dev/toolchain.html and xref scipy/scipy#14655 ). And it would be closer to NEP29.

I would take NEP29 as a minimal requirement for scikit-learn. If it's easy for us to support a little more, then let's do it. I like the last 3 Python versions rule of thumb. It's still manageable from a CI maintenance point of view.

Python 3.10 being out next month basically, I would read this as: let's drop Python 3.7 to support 3.8, 3.9 and 3.10.

[ogrisel]

We are about to release 1.0.0 today hopefully, so let's keep 3.7 for this one :)

[tupui]

Oh great! 😃 Sure then.

[adrinjalali]

@tupui does conda-forge hope for libraries to drop 3.7 for CPU cycles or maintainance?

Also, dropping 3.7 means probably using 3.8 syntax, which means no pypy support. I'm not sure what we think about that.

pinging @rth on this

[ogrisel]

@tupui does conda-forge hope for libraries to drop 3.7 for CPU cycles or maintainance?

Even if that's the case, it's would still be ok to support 3.7 for sdist and wheels on pypi.org.

[ogrisel]

For information, the conda-forge packages are being prepared here: conda-forge/scikit-learn-feedstock#166

[tupui]

@ogrisel I would be happy to try a PR to remove 3.7 once v1 is out. Let me know if this is wanted, trying to get on board in some ways 😅

[adrinjalali]

we probably would keep 3.7 on main for at least the first two minor releases maybe. But once it's time to do so, then you're more than welcome to remove all the 3.7 stuff 😄

[ogrisel]

@adrinjalali can you please update the todo list of this issue? It seems that most of it is done and I am not sure what is left to do besides general announcements and updating the scikit-learn.org/stable synlink.

As @thomasjpfan said on gitter.im, we still need to update var latestStable = '1.0'; in versionwarning.js:

scikit-learn/scikit-learn.github.io@3d1baaa

[adrinjalali]

the change is done now, was waiting for the builds to finish before pushing it

[adrinjalali]

The fix the binder release version in .binder/requirement.txt doesn't seem to be required, right?

[thomasjpfan]

From #14832 I think we do not need to update it anymore and it could be removed from the checklist.

[ogrisel]

I think we can close this one. Thanks everyone and @adrinjalali in particular!

[GaelVaroquaux]

Thanks everyone and @adrinjalali in particular!

Yes!!

[reshamas]

@adrinjalali
In the checklist above, can we add an item: "share announcement with NumFOCUS for newsletter"

(I assume this checklist will be used for the next release.)

[adrinjalali]

Arliss from NumFocus sends reminders before each newsletter, we just need to respond to those.