remove routing from non fit and score methods

adrinjalali · adrinjalali · commit 26a98f7d2b5e · 2023-08-15T11:21:45.000+02:00
diff --git a/sklearn/model_selection/_search.py b/sklearn/model_selection/_search.py
@@ -555,7 +555,7 @@ def score_samples(self, X):
         return self.best_estimator_.score_samples(X)
 
     @available_if(_estimator_has("predict"))
-    def predict(self, X, **params):
+    def predict(self, X):
         """Call predict on the estimator with the best found parameters.
 
         Only available if ``refit=True`` and the underlying estimator supports
@@ -567,27 +567,17 @@ def predict(self, X, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``predict``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         y_pred : ndarray of shape (n_samples,)
             The predicted labels or values for `X` based on the estimator with
             the best found parameters.
         """
         check_is_fitted(self)
-        return self.best_estimator_.predict(
-            X, **_get_params_for_method(self, "predict", params)
-        )
+        return self.best_estimator_.predict(X)
 
     @available_if(_estimator_has("predict_proba"))
-    def predict_proba(self, X, **params):
+    def predict_proba(self, X):
         """Call predict_proba on the estimator with the best found parameters.
 
         Only available if ``refit=True`` and the underlying estimator supports
@@ -599,14 +589,6 @@ def predict_proba(self, X, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``predict_proba``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         y_pred : ndarray of shape (n_samples,) or (n_samples, n_classes)
@@ -615,12 +597,10 @@ def predict_proba(self, X, **params):
             to that in the fitted attribute :term:`classes_`.
         """
         check_is_fitted(self)
-        return self.best_estimator_.predict_proba(
-            X, **_get_params_for_method(self, "predict_proba", params)
-        )
+        return self.best_estimator_.predict_proba(X)
 
     @available_if(_estimator_has("predict_log_proba"))
-    def predict_log_proba(self, X, **params):
+    def predict_log_proba(self, X):
         """Call predict_log_proba on the estimator with the best found parameters.
 
         Only available if ``refit=True`` and the underlying estimator supports
@@ -632,14 +612,6 @@ def predict_log_proba(self, X, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``predict_log_proba``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         y_pred : ndarray of shape (n_samples,) or (n_samples, n_classes)
@@ -648,12 +620,10 @@ def predict_log_proba(self, X, **params):
             corresponds to that in the fitted attribute :term:`classes_`.
         """
         check_is_fitted(self)
-        return self.best_estimator_.predict_log_proba(
-            X, **_get_params_for_method(self, "predict_log_proba", params)
-        )
+        return self.best_estimator_.predict_log_proba(X)
 
     @available_if(_estimator_has("decision_function"))
-    def decision_function(self, X, **params):
+    def decision_function(self, X):
         """Call decision_function on the estimator with the best found parameters.
 
         Only available if ``refit=True`` and the underlying estimator supports
@@ -665,14 +635,6 @@ def decision_function(self, X, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``decision_function``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         y_score : ndarray of shape (n_samples,) or (n_samples, n_classes) \
@@ -681,12 +643,10 @@ def decision_function(self, X, **params):
             the best found parameters.
         """
         check_is_fitted(self)
-        return self.best_estimator_.decision_function(
-            X, **_get_params_for_method(self, "decision_function", params)
-        )
+        return self.best_estimator_.decision_function(X)
 
     @available_if(_estimator_has("transform"))
-    def transform(self, X, **params):
+    def transform(self, X):
         """Call transform on the estimator with the best found parameters.
 
         Only available if the underlying estimator supports ``transform`` and
@@ -698,27 +658,17 @@ def transform(self, X, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``transform``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         Xt : {ndarray, sparse matrix} of shape (n_samples, n_features)
             `X` transformed in the new space based on the estimator with
             the best found parameters.
         """
         check_is_fitted(self)
-        return self.best_estimator_.transform(
-            X, **_get_params_for_method(self, "transform", params)
-        )
+        return self.best_estimator_.transform(X)
 
     @available_if(_estimator_has("inverse_transform"))
-    def inverse_transform(self, Xt, **params):
+    def inverse_transform(self, Xt):
         """Call inverse_transform on the estimator with the best found params.
 
         Only available if the underlying estimator implements
@@ -730,24 +680,14 @@ def inverse_transform(self, Xt, **params):
             Must fulfill the input assumptions of the
             underlying estimator.
 
-        **params : dict
-            Parameters to be passed the underlying estimator's ``inverse_transform``.
-
-            Only available if `enable_metadata_routing=True`. See the
-            :ref:`User Guide <metadata_routing>`.
-
-            ..versionadded:: 1.4
-
         Returns
         -------
         X : {ndarray, sparse matrix} of shape (n_samples, n_features)
             Result of the `inverse_transform` function for `Xt` based on the
             estimator with the best found parameters.
         """
         check_is_fitted(self)
-        return self.best_estimator_.inverse_transform(
-            Xt, **_get_params_for_method(self, "inverse_transform", params)
-        )
+        return self.best_estimator_.inverse_transform(Xt)
 
     @property
     def n_features_in_(self):
diff --git a/sklearn/tests/metadata_routing_common.py b/sklearn/tests/metadata_routing_common.py
@@ -49,7 +49,6 @@ def check_recorded_metadata(obj, method, split_params=tuple(), **kwargs):
     split_params : tuple, default=empty
         specifies any parameters which are to be checked as being a subset
         of the original values.
-
     """
     records = getattr(obj, "_records", dict()).get(method, dict())
     assert set(kwargs.keys()) == set(records.keys())
diff --git a/sklearn/tests/test_metaestimators_metadata_routing.py b/sklearn/tests/test_metaestimators_metadata_routing.py
@@ -107,14 +107,8 @@ def enable_slep006():
         "init_args": {"param_grid": {"alpha": [0.1, 0.2]}},
         "X": X,
         "y": y,
-        "estimator_routing_methods": [
-            "fit",
-            "predict",
-            "predict_proba",
-            "predict_log_proba",
-            "decision_function",
-        ],
-        "preserves_metadata": False,
+        "estimator_routing_methods": ["fit"],
+        "preserves_metadata": "subset",
         "scorer_name": "scoring",
         "scorer_routing_methods": ["fit", "score"],
         "cv_name": "cv",
@@ -127,14 +121,8 @@ def enable_slep006():
         "init_args": {"param_distributions": {"alpha": [0.1, 0.2]}},
         "X": X,
         "y": y,
-        "estimator_routing_methods": [
-            "fit",
-            "predict",
-            "predict_proba",
-            "predict_log_proba",
-            "decision_function",
-        ],
-        "preserves_metadata": False,
+        "estimator_routing_methods": ["fit"],
+        "preserves_metadata": "subset",
         "scorer_name": "scoring",
         "scorer_routing_methods": ["fit", "score"],
         "cv_name": "cv",
@@ -147,14 +135,8 @@ def enable_slep006():
         "init_args": {"param_grid": {"alpha": [0.1, 0.2]}},
         "X": X,
         "y": y,
-        "estimator_routing_methods": [
-            "fit",
-            "predict",
-            "predict_proba",
-            "predict_log_proba",
-            "decision_function",
-        ],
-        "preserves_metadata": False,
+        "estimator_routing_methods": ["fit"],
+        "preserves_metadata": "subset",
         "scorer_name": "scoring",
         "scorer_routing_methods": ["fit", "score"],
         "cv_name": "cv",
@@ -167,14 +149,8 @@ def enable_slep006():
         "init_args": {"param_distributions": {"alpha": [0.1, 0.2]}},
         "X": X,
         "y": y,
-        "estimator_routing_methods": [
-            "fit",
-            "predict",
-            "predict_proba",
-            "predict_log_proba",
-            "decision_function",
-        ],
-        "preserves_metadata": False,
+        "estimator_routing_methods": ["fit"],
+        "preserves_metadata": "subset",
         "scorer_name": "scoring",
         "scorer_routing_methods": ["fit", "score"],
         "cv_name": "cv",
@@ -193,10 +169,15 @@ def enable_slep006():
 - y: y-data to fit
 - estimator_routing_methods: list of all methods to check for routing metadata
   to the sub-estimator
-- preserves_metadata: Whether the metaestimator passes the metadata to the
-  sub-estimator without modification or not. If it does, we check that the
-  values are identical. If it doesn't, no check is performed. TODO Maybe
-  something smarter could be done if the data is modified.
+- preserves_metadata:
+    - True (default): the metaestimator passes the metadata to the
+      sub-estimator without modification. We check that the values recorded by
+      the sub-estimator are identical to what we've passed to the
+      metaestimator.
+    - False: no check is performed regarding values, we only check that a
+      metadata with the expected names/keys are passed.
+    - "subset": we check that the recorded metadata by the sub-estimator is a
+      subset of what is passed to the metaestimator.
 - scorer_name: The name of the argument for the scorer
 - scorer_routing_methods: list of all methods to check for routing metadata
   to the scorer
@@ -341,6 +322,8 @@ def set_request(estimator, method_name):
 
     for method_name in routing_methods:
         for key in ["sample_weight", "metadata"]:
+            if method_name == "predict":
+                pass
             val = {"sample_weight": sample_weight, "metadata": metadata}[key]
             method_kwargs = {key: val}
 
@@ -360,12 +343,20 @@ def set_request(estimator, method_name):
                 instance.fit(X, y)
                 method(X, **method_kwargs)
 
-            if preserves_metadata:
-                # sanity check that registry is not empty, or else the test
-                # passes trivially
-                assert registry
+            # sanity check that registry is not empty, or else the test passes
+            # trivially
+            assert registry
+            if preserves_metadata is True:
                 for estimator in registry:
                     check_recorded_metadata(estimator, method_name, **method_kwargs)
+            elif preserves_metadata == "subset":
+                for estimator in registry:
+                    check_recorded_metadata(
+                        estimator,
+                        method_name,
+                        split_params=method_kwargs.keys(),
+                        **method_kwargs,
+                    )
 
 
 @pytest.mark.parametrize("metaestimator", METAESTIMATORS, ids=METAESTIMATOR_IDS)
