FIX Removes warning in HGBT when fitting on dataframes (#22908)

thomasjpfan · web-flow · commit 4408dfb86bdf · 2022-03-25T10:56:49.000+01:00
diff --git a/doc/whats_new/v1.1.rst b/doc/whats_new/v1.1.rst
@@ -391,6 +391,11 @@ Changelog
   for instance using cgroups quota in a docker container. :pr:`22566` by
   :user:`Jérémie du Boisberranger <jeremiedbb>`.
 
+- |Fix| :class:`ensemble.HistGradientBoostingClassifier` and
+  :class:`ensemble.HistGradientBoostingRegressor` no longer warns when
+  fitting on a pandas DataFrame with a non-default `scoring` parameter and
+  early_stopping enabled. :pr:`22908` by `Thomas Fan`_.
+
 :mod:`sklearn.feature_extraction`
 .................................
 
diff --git a/sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py b/sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.py
@@ -30,7 +30,7 @@
 from ...model_selection import train_test_split
 from ...preprocessing import LabelEncoder
 from ._gradient_boosting import _update_raw_predictions
-from .common import Y_DTYPE, X_DTYPE, X_BINNED_DTYPE, G_H_DTYPE
+from .common import Y_DTYPE, X_DTYPE, G_H_DTYPE
 
 from .binning import _BinMapper
 from .grower import TreeGrower
@@ -936,8 +936,10 @@ def _raw_predict(self, X, n_threads=None):
             The raw predicted values.
         """
         is_binned = getattr(self, "_in_fit", False)
-        dtype = X_BINNED_DTYPE if is_binned else X_DTYPE
-        X = self._validate_data(X, dtype=dtype, force_all_finite=False, reset=False)
+        if not is_binned:
+            X = self._validate_data(
+                X, dtype=X_DTYPE, force_all_finite=False, reset=False
+            )
         check_is_fitted(self)
         if X.shape[1] != self._n_features:
             raise ValueError(
diff --git a/sklearn/ensemble/_hist_gradient_boosting/tests/test_gradient_boosting.py b/sklearn/ensemble/_hist_gradient_boosting/tests/test_gradient_boosting.py
@@ -1,3 +1,5 @@
+import warnings
+
 import numpy as np
 import pytest
 from numpy.testing import assert_allclose, assert_array_equal
@@ -1141,3 +1143,20 @@ def test_loss_deprecated(old_loss, new_loss):
     est2 = HistGradientBoostingRegressor(loss=new_loss, random_state=0)
     est2.fit(X, y)
     assert_allclose(est1.predict(X), est2.predict(X))
+
+
+def test_no_user_warning_with_scoring():
+    """Check that no UserWarning is raised when scoring is set.
+
+    Non-regression test for #22907.
+    """
+    pd = pytest.importorskip("pandas")
+    X, y = make_regression(n_samples=50, random_state=0)
+    X_df = pd.DataFrame(X, columns=[f"col{i}" for i in range(X.shape[1])])
+
+    est = HistGradientBoostingRegressor(
+        random_state=0, scoring="neg_mean_absolute_error", early_stopping=True
+    )
+    with warnings.catch_warnings():
+        warnings.simplefilter("error", UserWarning)
+        est.fit(X_df, y)
