MAINT compatibility with sklearn 1.6 #1104
Conversation
imblearn/base.py
Outdated
| return parse_version( | ||
| parse_version(sklearn.__version__).base_version | ||
| ) >= parse_version("1.6") |
There was a problem hiding this comment.
I'd probably do a try / except with import Tags kinda thing though. With a note that "remove this after 1.6 becomes minimum"
There was a problem hiding this comment.
This is not the pattern that we document. But actually, I don't know why we are telling to use available_if. Right now, we could have both _more_tags and __sklearn_tags__ define.
The available_if would make sense if you want to avoid a deprecation warning of an error that we could raise in the future if you define both _more_tags and __sklearn_tags__. This is something that I mentioned here: scikit-learn/scikit-learn#30257
imblearn/ensemble/_bagging.py
Outdated
| if sklearn_version < parse_version("1.6"): | ||
| kwargs = {"force_all_finite": False} | ||
| else: | ||
| kwargs = {"ensure_all_finite": False} |
There was a problem hiding this comment.
this logic can go in your fixes.validate_data
imblearn/utils/_tags.py
Outdated
| if sklearn_version >= parse_version("1.6"): | ||
| from sklearn.utils._tags import InputTags |
There was a problem hiding this comment.
A try / except around the import (from the public path) makes more sense I think, which shows the exact thing you care about, instead of using the version as a proxy.
There was a problem hiding this comment.
The Tags are not available in a public path if I'm not wrong.
There was a problem hiding this comment.
Sorry I check the wrong __init__.py :)
Working on the compatibility with scikit-learn 1.6