fix: sklearn 1.6 compatibility - add __sklearn_tags__ and fix error msg

scikit-learn 1.6 removed _more_tags/_get_tags in favor of __sklearn_tags__.
Any estimator that defines _more_tags without __sklearn_tags__ raises
TypeError in check_estimator(). Add __sklearn_tags__ to all 9 affected
classes while keeping _more_tags for backward compatibility.

Also fix utilities.py error message format to match sklearn 1.6's
check_n_features_in_after_fitting regex:
  "X has N features, but ClassName is expecting M features as input"

Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>

Author: 3 people

python/abess/decomposition.py

@@ -135,6 +135,11 @@ def __init__(self, support_size=None, group=None,
     def _more_tags(self):
         return {'requires_y': False}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.requires_y = False
+        return tags
+
     def transform(self, X):
         r"""
         For PCA model, apply dimensionality reduction
@@ -574,6 +579,11 @@ def _more_tags(self):
         # (It just returns the transformation of `X`.)
         return {'_skip_test': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags._skip_test = True
+        return tags
+
     def fit(self, X, y=None, r=None, sparse_matrix=False):
         r"""
         The fit function is used to transfer the information of

python/abess/linear.py

@@ -94,6 +94,12 @@ def _more_tags(self):
         return {'binary_only': True,
                 'no_validation': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.classifier_tags.multi_class = False
+        tags.no_validation = True
+        return tags
+
     def predict_proba(self, X):
         r"""
         Give the probabilities of new sample
@@ -254,6 +260,10 @@ def __init__(self, path_type="seq", support_size=None,
     def _more_tags(self):
         return {'multioutput': False}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        return tags
+
     def predict(self, X):
         r"""
         Predict on given data.
@@ -384,6 +394,11 @@ def _more_tags(self):
         # 2-column `y` should be given (one for time, another for censoring).
         return {'_skip_test': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags._skip_test = True
+        return tags
+
     def predict(self, X):
         r"""
         Returns the time-independent part of hazard function,
@@ -540,6 +555,11 @@ def __init__(self, path_type="seq", support_size=None,
     def _more_tags(self):
         return {"poor_score": True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.regressor_tags.poor_score = True
+        return tags
+
     def predict(self, X):
         r"""
         Predict on given data.
@@ -688,6 +708,12 @@ def _more_tags(self):
         return {'multioutput': True,
                 'multioutput_only': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.target_tags.multi_output = True
+        tags.target_tags.single_output = False
+        return tags
+
     def predict(self, X):
         r"""
         Prediction of the mean of each response on given data.
@@ -825,6 +851,12 @@ def _more_tags(self):
                 'no_validation': True,
                 'poor_score': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.no_validation = True
+        tags.classifier_tags.poor_score = True
+        return tags
+
     def predict_proba(self, X):
         r"""
         Give the probabilities of new data being assigned to different classes.
@@ -996,6 +1028,12 @@ def _more_tags(self):
         return {'poor_score': True,
                 'requires_positive_y': True}
 
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.regressor_tags.poor_score = True
+        tags.target_tags.required_positive = True
+        return tags
+
     def predict(self, X):
         r"""
         Predict on given data.

python/abess/utilities.py

@@ -39,8 +39,10 @@ def new_data_check(self, X, y=None, weights=None):
     # Check2 : X validation
     X = check_array(X, accept_sparse=True)
     if X.shape[1] != self.n_features_in_:
-        raise ValueError("X.shape[1] should be " +
-                         str(self.n_features_in_))
+        raise ValueError(
+            f"X has {X.shape[1]} features, but {type(self).__name__} "
+            f"is expecting {self.n_features_in_} features as input"
+        )
 
     # Check3 : X, y validation
     if (y is not None) and (weights is None):