scikit-learn · OmarManzoor · Feb 6, 2024 · Feb 4, 2024 · Feb 4, 2024
diff --git a/sklearn/model_selection/_search.py b/sklearn/model_selection/_search.py
@@ -781,18 +781,11 @@ def _select_best_index(refit, refit_metric, results):
            best_index = results[f"rank_test_{refit_metric}"].argmin()
        return best_index

-    def _get_scorers(self, convert_multimetric):
+    def _get_scorers(self):
        """Get the scorer(s) to be used.

        This is used in ``fit`` and ``get_metadata_routing``.

-        Parameters
-        ----------
-        convert_multimetric : bool
-            Whether to convert a dict of scorers to a _MultimetricScorer. This
-            is used in ``get_metadata_routing`` to include the routing info for
-            multiple scorers.
-
        Returns
        -------
        scorers, refit_metric
@@ -807,10 +800,9 @@ def _get_scorers(self, convert_multimetric):
            scorers = _check_multimetric_scoring(self.estimator, self.scoring)
            self._check_refit_for_multimetric(scorers)
            refit_metric = self.refit
-            if convert_multimetric and isinstance(scorers, dict):
-                scorers = _MultimetricScorer(
-                    scorers=scorers, raise_exc=(self.error_score == "raise")
-                )
+            scorers = _MultimetricScorer(
+                scorers=scorers, raise_exc=(self.error_score == "raise")
+            )

        return scorers, refit_metric

@@ -866,10 +858,7 @@ def fit(self, X, y=None, **params):
            Instance of fitted estimator.
        """
        estimator = self.estimator
-        # Here we keep a dict of scorers as is, and only convert to a
-        # _MultimetricScorer at a later stage. Issue:
-        # https://github.com/scikit-learn/scikit-learn/issues/27001
-        scorers, refit_metric = self._get_scorers(convert_multimetric=False)
+        scorers, refit_metric = self._get_scorers()

        X, y = indexable(X, y)
        params = _check_method_params(X, params=params)
@@ -1015,7 +1004,10 @@ def evaluate_candidates(candidate_params, cv=None, more_results=None):
                self.feature_names_in_ = self.best_estimator_.feature_names_in_

        # Store the only scorer not as a dict for single metric evaluation
-        self.scorer_ = scorers
+        if isinstance(scorers, _MultimetricScorer):
+            self.scorer_ = scorers._scorers
+        else:
+            self.scorer_ = scorers

        self.cv_results_ = results
        self.n_splits_ = n_splits
@@ -1147,7 +1139,7 @@ def get_metadata_routing(self):
            method_mapping=MethodMapping().add(caller="fit", callee="fit"),
        )

-        scorer, _ = self._get_scorers(convert_multimetric=True)
+        scorer, _ = self._get_scorers()
        router.add(
            scorer=scorer,
            method_mapping=MethodMapping()

diff --git a/sklearn/model_selection/_validation.py b/sklearn/model_selection/_validation.py
@@ -358,18 +358,11 @@ def cross_validate(
        scorers = check_scoring(estimator, scoring)
    else:
        scorers = _check_multimetric_scoring(estimator, scoring)
+        scorers = _MultimetricScorer(
+            scorers=scorers, raise_exc=(error_score == "raise")
+        )

    if _routing_enabled():
-        # `cross_validate` will create a `_MultiMetricScorer` if `scoring` is a
-        # dict at a later stage. We need the same object for the purpose of
-        # routing. However, creating it here and passing it around would create
-        # a much larger diff since the dict is used in many places.
-        if isinstance(scorers, dict):
-            _scorer = _MultimetricScorer(
-                scorers=scorers, raise_exc=(error_score == "raise")
-            )
-        else:
-            _scorer = scorers
        # For estimators, a MetadataRouter is created in get_metadata_routing
        # methods. For these router methods, we create the router to use
        # `process_routing` on it.
@@ -386,7 +379,7 @@ def cross_validate(
                method_mapping=MethodMapping().add(caller="fit", callee="fit"),
            )
            .add(
-                scorer=_scorer,
+                scorer=scorers,
                method_mapping=MethodMapping().add(caller="fit", callee="score"),
            )
        )
@@ -901,8 +894,8 @@ def _fit_and_score(
        if error_score == "raise":
            raise
        elif isinstance(error_score, numbers.Number):
-            if isinstance(scorer, dict):
-                test_scores = {name: error_score for name in scorer}
+            if isinstance(scorer, _MultimetricScorer):
+                test_scores = {name: error_score for name in scorer._scorers}
                if return_train_score:
                    train_scores = test_scores.copy()
            else:
@@ -966,13 +959,9 @@ def _fit_and_score(
 def _score(estimator, X_test, y_test, scorer, score_params, error_score="raise"):
    """Compute the score(s) of an estimator on a given test set.

-    Will return a dict of floats if `scorer` is a dict, otherwise a single
+    Will return a dict of floats if `scorer` is a _MultiMetricScorer, otherwise a single
    float is returned.
    """
-    if isinstance(scorer, dict):
-        # will cache method calls if needed. scorer() returns a dict
-        scorer = _MultimetricScorer(scorers=scorer, raise_exc=(error_score == "raise"))
-
    score_params = {} if score_params is None else score_params

    try:

diff --git a/sklearn/model_selection/tests/test_validation.py b/sklearn/model_selection/tests/test_validation.py
@@ -43,6 +43,7 @@
    precision_score,
    r2_score,
 )
+from sklearn.metrics._scorer import _MultimetricScorer
 from sklearn.model_selection import (
    GridSearchCV,
    GroupKFold,
@@ -2323,7 +2324,9 @@ def three_params_scorer(i, j, k):
        ),
        (
            True,
-            {"sc1": three_params_scorer, "sc2": three_params_scorer},
+            _MultimetricScorer(
+                scorers={"sc1": three_params_scorer, "sc2": three_params_scorer}
+            ),
            3,
            (1, 3),
            (0, 1),
@@ -2332,7 +2335,9 @@ def three_params_scorer(i, j, k):
        ),
        (
            False,
-            {"sc1": three_params_scorer, "sc2": three_params_scorer},
+            _MultimetricScorer(
+                scorers={"sc1": three_params_scorer, "sc2": three_params_scorer}
+            ),
            10,
            (1, 3),
            (0, 1),