From: Colin Watson Date: Sun, 19 Oct 2025 02:17:51 +0100 Subject: Handle estimator checking changes in scikit-learn 1.6 Forwarded: https://github.com/trevorstephens/gplearn/pull/305 Bug-Debian: https://bugs.debian.org/1117991 Last-Update: 2025-10-19 --- .github/workflows/build.yml | 2 +- doc/rtd-pip-requirements | 2 +- gplearn/genetic.py | 71 +++++++++++++++++++++++++++------- gplearn/tests/test_estimator_checks.py | 11 +++++- gplearn/utils.py | 8 ++++ setup.py | 3 +- 6 files changed, 80 insertions(+), 17 deletions(-) diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml index 3b198b8..5df8e5b 100644 --- a/.github/workflows/build.yml +++ b/.github/workflows/build.yml @@ -41,7 +41,7 @@ jobs: python -m pip install pandas - name: Install minimal dependencies if: ${{ matrix.python_version == '3.8' }} - run: python -m pip install scikit-learn==1.0.2 joblib==1.0.0 + run: python -m pip install scikit-learn==1.1.0 joblib==1.0.0 - name: Install gplearn run: python -m pip install . - name: Describe Python environment diff --git a/doc/rtd-pip-requirements b/doc/rtd-pip-requirements index 947fdee..fbf725e 100644 --- a/doc/rtd-pip-requirements +++ b/doc/rtd-pip-requirements @@ -1,5 +1,5 @@ numpy>=1.8.1 numpydoc>=0.5 scipy>=0.13 -scikit-learn>=0.22.1 +scikit-learn>=1.1.0 joblib>=0.13.0 diff --git a/gplearn/genetic.py b/gplearn/genetic.py index 157bc92..d64dedb 100644 --- a/gplearn/genetic.py +++ b/gplearn/genetic.py @@ -23,11 +23,13 @@ from sklearn.exceptions import NotFittedError from sklearn.utils import compute_sample_weight from sklearn.utils.validation import check_array, _check_sample_weight from sklearn.utils.multiclass import check_classification_targets +from sklearn.utils.multiclass import type_of_target from ._program import _Program from .fitness import _fitness_map, _Fitness from .functions import _function_map, _Function, sig1 as sigmoid from .utils import _partition_estimators +from .utils import _sklearn_version_ge from .utils import check_random_state __all__ = ['SymbolicRegressor', 'SymbolicClassifier', 'SymbolicTransformer'] @@ -301,6 +303,18 @@ class BaseSymbolic(BaseEstimator, metaclass=ABCMeta): if isinstance(self, ClassifierMixin): X, y = self._validate_data(X, y, y_numeric=False) check_classification_targets(y) + # Once we require scikit-learn >= 1.6, this should pass + # raise_unknown=True rather than checking for "unknown" + # manually. + y_type = type_of_target(y, input_name="y") + if y_type == "unknown": + raise ValueError("Unknown label type for y: %r" % y) + elif y_type != "binary": + raise ValueError( + "Only binary classification is supported. The type of the " + "target is %s." + % y_type + ) if self.class_weight: if sample_weight is None: @@ -599,7 +613,7 @@ class BaseSymbolic(BaseEstimator, metaclass=ABCMeta): return self -class SymbolicRegressor(BaseSymbolic, RegressorMixin): +class SymbolicRegressor(RegressorMixin, BaseSymbolic): """A Genetic Programming symbolic regressor. @@ -868,7 +882,15 @@ class SymbolicRegressor(BaseSymbolic, RegressorMixin): if not hasattr(self, '_program'): raise NotFittedError('SymbolicRegressor not fitted.') - X = check_array(X) + try: + # scikit-learn >= 1.6 + from sklearn.utils.validation import validate_data + + X = validate_data(self, X, reset=False) + except ImportError: + # scikit-learn < 1.6 + X = check_array(X) + _, n_features = X.shape if self.n_features_in_ != n_features: raise ValueError('Number of features of the model must match the ' @@ -881,7 +903,7 @@ class SymbolicRegressor(BaseSymbolic, RegressorMixin): return y -class SymbolicClassifier(BaseSymbolic, ClassifierMixin): +class SymbolicClassifier(ClassifierMixin, BaseSymbolic): """A Genetic Programming symbolic classifier. @@ -1142,6 +1164,11 @@ class SymbolicClassifier(BaseSymbolic, ClassifierMixin): return self.__repr__() return self._program.__str__() + def __sklearn_tags__(self): + tags = super().__sklearn_tags__() + tags.classifier_tags.multi_class = False + return tags + def _more_tags(self): return {'binary_only': True} @@ -1164,7 +1191,15 @@ class SymbolicClassifier(BaseSymbolic, ClassifierMixin): if not hasattr(self, '_program'): raise NotFittedError('SymbolicClassifier not fitted.') - X = check_array(X) + try: + # scikit-learn >= 1.6 + from sklearn.utils.validation import validate_data + + X = validate_data(self, X, reset=False) + except ImportError: + # scikit-learn < 1.6 + X = check_array(X) + _, n_features = X.shape if self.n_features_in_ != n_features: raise ValueError('Number of features of the model must match the ' @@ -1196,7 +1231,7 @@ class SymbolicClassifier(BaseSymbolic, ClassifierMixin): return self.classes_.take(np.argmax(proba, axis=1), axis=0) -class SymbolicTransformer(BaseSymbolic, TransformerMixin): +class SymbolicTransformer(TransformerMixin, BaseSymbolic): """A Genetic Programming symbolic transformer. @@ -1467,14 +1502,16 @@ class SymbolicTransformer(BaseSymbolic, TransformerMixin): output = str([gp.__str__() for gp in self]) return output.replace("',", ",\n").replace("'", "") - def _more_tags(self): - return { - "_xfail_checks": { - "check_sample_weights_invariance": ( - "zero sample_weight is not equivalent to removing samples" - ), + if not _sklearn_version_ge("1.6"): + def _more_tags(self): + return { + "_xfail_checks": { + "check_sample_weights_invariance": ( + "zero sample_weight is not equivalent to removing " + "samples" + ), + } } - } def transform(self, X): """Transform X according to the fitted transformer. @@ -1494,7 +1531,15 @@ class SymbolicTransformer(BaseSymbolic, TransformerMixin): if not hasattr(self, '_best_programs'): raise NotFittedError('SymbolicTransformer not fitted.') - X = check_array(X) + try: + # scikit-learn >= 1.6 + from sklearn.utils.validation import validate_data + + X = validate_data(self, X, reset=False) + except ImportError: + # scikit-learn < 1.6 + X = check_array(X) + _, n_features = X.shape if self.n_features_in_ != n_features: raise ValueError('Number of features of the model must match the ' diff --git a/gplearn/tests/test_estimator_checks.py b/gplearn/tests/test_estimator_checks.py index af57fb3..534b50e 100644 --- a/gplearn/tests/test_estimator_checks.py +++ b/gplearn/tests/test_estimator_checks.py @@ -10,6 +10,7 @@ from sklearn.utils.estimator_checks import check_estimator from gplearn.genetic import SymbolicClassifier, SymbolicRegressor from gplearn.genetic import SymbolicTransformer +from gplearn.utils import _sklearn_version_ge def test_sklearn_regressor_checks(): @@ -29,6 +30,14 @@ def test_sklearn_classifier_checks(): def test_sklearn_transformer_checks(): """Run the sklearn estimator validation checks on SymbolicTransformer""" + kwargs = {} + if _sklearn_version_ge("1.6"): + kwargs["expected_failed_checks"] = { + "check_sample_weights_invariance": ( + "zero sample_weight is not equivalent to removing samples" + ), + } check_estimator(SymbolicTransformer(population_size=50, hall_of_fame=10, - generations=5)) + generations=5), + **kwargs) diff --git a/gplearn/utils.py b/gplearn/utils.py index 7eee1bd..a210dec 100644 --- a/gplearn/utils.py +++ b/gplearn/utils.py @@ -6,10 +6,18 @@ order to maintain compatibility across different versions of scikit-learn. """ +import importlib.metadata import numbers import numpy as np from joblib import cpu_count +from packaging.version import Version + + +def _sklearn_version_ge(min_version): + """Check whether we have at least min_version of scikit-learn.""" + sklearn_version = importlib.metadata.version("scikit-learn") + return Version(sklearn_version) >= Version(min_version) def check_random_state(seed): diff --git a/setup.py b/setup.py index 1eef653..72b13fd 100644 --- a/setup.py +++ b/setup.py @@ -34,5 +34,6 @@ setup(name='gplearn', '*.tests.*']), zip_safe=False, package_data={'': ['LICENSE']}, - install_requires=['scikit-learn>=1.0.2', + install_requires=['packaging', + 'scikit-learn>=1.1.0', 'joblib>=1.0.0'])