Merge pull request #20 from basf/fix-scorer-sign-error

Fix scorer sign error

molpipeline/metrics/ignore_error_scorer.py

@@ -39,6 +39,8 @@ def ignored_value_scorer(
     score_func = scorer._score_func  # pylint: disable=protected-access
     response_method = scorer._response_method  # pylint: disable=protected-access
     scorer_kwargs = scorer._kwargs  # pylint: disable=protected-access
+    if scorer._sign < 0:  # pylint: disable=protected-access
+        scorer_kwargs["greater_is_better"] = False
 
     def newscore(
         y_true: npt.NDArray[np.float_ | np.int_],

molpipeline/pipeline/_skl_pipeline.py

@@ -4,7 +4,6 @@
 
 from typing import Any, Iterable, List, Literal, Optional, Tuple, TypeVar, Union
 
-
 try:
     from typing import Self  # type: ignore[attr-defined]
 except ImportError:

tests/test_metrics/test_ignore_error_scorer.py

@@ -3,6 +3,8 @@
 import unittest
 
 import numpy as np
+from sklearn import linear_model
+from sklearn.metrics import get_scorer
 
 from molpipeline.metrics import ignored_value_scorer
 
@@ -45,3 +47,47 @@ def test_filter_none_with_nan(self) -> None:
             ba_score._score_func(y_true, y_pred),  # pylint: disable=protected-access
             1.0,
         )
+
+    def test_correct_init_mse(self) -> None:
+        """Test that initialization is correct as we access via protected vars."""
+        x_train = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1]).reshape(
+            -1, 1
+        )
+        y_train = np.array([0.1, 0.3, 0.3, 0.4, 0.5, 0.5, 0.7, 0.88, 0.9, 1])
+        regr = linear_model.LinearRegression()
+        regr.fit(x_train, y_train)
+        cix_scorer = ignored_value_scorer("neg_mean_squared_error", None)
+        scikit_scorer = get_scorer("neg_mean_squared_error")
+        self.assertEqual(
+            cix_scorer(regr, x_train, y_train), scikit_scorer(regr, x_train, y_train)
+        )
+
+    def test_correct_init_rmse(self) -> None:
+        """Test that initialization is correct as we access via protected vars."""
+        x_train = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1]).reshape(
+            -1, 1
+        )
+        y_train = np.array([0.1, 0.3, 0.3, 0.4, 0.5, 0.5, 0.7, 0.88, 0.9, 1])
+        regr = linear_model.LinearRegression()
+        regr.fit(x_train, y_train)
+        cix_scorer = ignored_value_scorer("neg_root_mean_squared_error", None)
+        scikit_scorer = get_scorer("neg_root_mean_squared_error")
+        self.assertEqual(
+            cix_scorer(regr, x_train, y_train), scikit_scorer(regr, x_train, y_train)
+        )
+
+    def test_correct_init_inheritance(self) -> None:
+        """Test that initialization is correct if we pass an initialized scorer."""
+        x_train = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1]).reshape(
+            -1, 1
+        )
+        y_train = np.array([0.1, 0.3, 0.3, 0.4, 0.5, 0.5, 0.7, 0.88, 0.9, 1])
+        regr = linear_model.LinearRegression()
+        regr.fit(x_train, y_train)
+        scikit_scorer = get_scorer("neg_root_mean_squared_error")
+        cix_scorer = ignored_value_scorer(
+            get_scorer("neg_root_mean_squared_error"), None
+        )
+        self.assertEqual(
+            cix_scorer(regr, x_train, y_train), scikit_scorer(regr, x_train, y_train)
+        )