Feature: Support for Scikit-Learn v1.7+ Compatibility #422
Description
It seems that for scikit-learn>=1.7 there is a new requirement to implement __sklearn_tags__. I came across this version pinning workaround from an XGBoost issue for continuing to use scikit-learn and PyGAM together in the meantime. The change to scikit-learn means that any model training functionality cannot be run with PyGAM models for the latest releases of scikit-learn. I don't believe this is an immediate need, but it would be great if it could be addressed at some point.
Setup specifics
Failing Environment settings
- Python 3.14
- scikit-learn>1.6.1
- pygam>=0.11
Successful workaround environment settings
- Python 3.14
- scikit-learn<1.7
- pygam>=0.11
In-code call
The failing setup can boiled down to the following, but comes from OpenOA.
import numpy as np
from pygam import GAM
from sklearn.metrics import r2_score, make_scorer
from sklearn.model_selection import KFold, RandomizedSearchCV
scorer = make_scorer(r2_score, greater_is_better=True)
random_search = RandomizedSearchCV(
GAM,
cv=KFold,
param_distributions={"n_splines": np.arange(5, 40)},
n_iter=20,
scoring=scorer,
verbose=0,
return_train_score=True,
)
random_search.fit(X, y) # X and y Relevant traceback
estimator = GAM(callbacks=['deviance', 'diffs'], distribution='normal',
fit_intercept=True, link='identity', max_iter=100, terms='auto',
tol=0.0001, verbose=False)
def get_tags(estimator) -> Tags:
"""Get estimator tags.
:class:`~sklearn.BaseEstimator` provides the estimator tags machinery.
For scikit-learn built-in estimators, we should still rely on
`self.__sklearn_tags__()`. `get_tags(est)` should be used when we
are not sure where `est` comes from: typically
`get_tags(self.estimator)` where `self` is a meta-estimator, or in
the common checks.
.. versionadded:: 1.6
Parameters
----------
estimator : estimator object
The estimator from which to get the tag.
Returns
-------
tags : :class:`~.sklearn.utils.Tags`
The estimator tags.
"""
try:
tags = estimator.__sklearn_tags__()
except AttributeError as exc:
if "object has no attribute '__sklearn_tags__'" in str(exc):
# Happens when `__sklearn_tags__` is implemented by calling
# `super().__sklearn_tags__()` but there is no `__sklearn_tags__`
# method in the base class. Typically happens when only inheriting
# from Mixins.
> raise AttributeError(
f"The following error was raised: {exc}. It seems that "
"there are no classes that implement `__sklearn_tags__` "
"in the MRO and/or all classes in the MRO call "
"`super().__sklearn_tags__()`. Make sure to inherit from "
"`BaseEstimator` which implements `__sklearn_tags__` (or "
"alternatively define `__sklearn_tags__` but we don't recommend "
"this approach). Note that `BaseEstimator` needs to be on the "
"right side of other Mixins in the inheritance order."
)
E AttributeError: The following error was raised: 'GAM' object has no attribute '__sklearn_tags__'. It seems that there are no classes that implement `__sklearn_tags__` in the MRO and/or all classes in the MRO call `super().__sklearn_tags__()`. Make sure to inherit from `BaseEstimator` which implements `__sklearn_tags__` (or alternatively define `__sklearn_tags__` but we don't recommend this approach). Note that `BaseEstimator` needs to be on the right side of other Mixins in the inheritance order.
../../../miniconda3/envs/openoa-nrel/lib/python3.14/site-packages/sklearn/utils/_tags.py:283: AttributeError