[python] potential fix for #325

python/dalex/NEWS.md

@@ -20,7 +20,7 @@ dalex (development)
 ### defaults
 
 * wrong parameter name `title_x` changed to `y_title` in `CeterisParibus.plot` and `AggregatedProfiles.plot` ([#317](https://github.com/ModelOriented/DALEX/issues/317))
-
+* now warning the user in `Explainer` when `predict_function` returns an error or doesn't return `numpy.ndarray (1d)` ([#325](https://github.com/ModelOriented/DALEX/issues/325))
 
 dalex 0.2.1
 ----------------------------------------------------------------

python/dalex/dalex/_explainer/checks.py

@@ -1,6 +1,7 @@
 # check functions for Explainer.__init__
 import pandas as pd
 from copy import deepcopy
+from warnings import warn
 
 from .helper import *
 from .yhat import *
@@ -189,9 +190,14 @@ def check_predict_function_and_model_type(predict_function, model_type,
                                        np.min(y_hat), np.mean(y_hat), np.max(y_hat)), verbose=verbose)
 
             except (Exception, ValueError, TypeError) as error:
-                verbose_cat("  -> predicted values  :  the predict_function returns an error when executed \n",
-                            verbose=verbose)
-                print(error)
+                # verbose_cat("  -> predicted values  : the predict_function returns an error when executed \n",
+                #             verbose=verbose)
+
+                warn("\n  -> predicted values  : the predict_function returns an error when executed \n" +
+                     str(error), stacklevel=2)
+
+        if not isinstance(y_hat, np.ndarray) or y_hat.shape != (data.shape[0], ):
+            warn("\n  -> predicted values  : predict_function must return numpy.ndarray (1d)", stacklevel=2)
 
         # check if predict_function accepts arrays
         try:

python/dalex/test/test_explainer.py

@@ -7,6 +7,9 @@
 from sklearn.preprocessing import StandardScaler, LabelEncoder, OneHotEncoder
 
 import dalex as dx
+import numpy as np
+
+import warnings
 
 
 class ExplainerTest(unittest.TestCase):
@@ -63,5 +66,54 @@ def test(self):
         self.assertIsInstance(case5, dx.instance_level.BreakDown)
         self.assertIsInstance(case6, dx.instance_level.CeterisParibus)
 
-        case5 = dx.Explainer(self.model, self.X, self.y, predict_function=1, verbose=False)
+        with warnings.catch_warnings(record=True) as w:
+            # Cause all warnings to always be triggered.
+            warnings.simplefilter("always")
+            # Trigger a warning.
+            case5 = dx.Explainer(self.model, self.X, self.y, predict_function=1, verbose=False)
+            assert issubclass(w[-1].category, UserWarning)
+
         self.assertIsInstance(case5, dx.Explainer)
+
+    def test_errors(self):
+
+        from sklearn.ensemble import RandomForestRegressor
+
+        data = dx.datasets.load_fifa()
+        X = data.drop(columns=['nationality', 'value_eur']).iloc[1:100, :]
+        y = data['value_eur'][1:100]
+
+        model = RandomForestRegressor()
+        model.fit(X, y)
+
+        def predict_function_return_2d(m, d):
+            n_rows = d.shape[0]
+            prediction = m.predict(d)
+            return prediction.reshape((n_rows, 1))
+
+        def predict_function_return_3d(m, d):
+            n_rows = d.shape[0]
+            prediction = m.predict(d)
+            return prediction.reshape((n_rows, 1, 1))
+
+        def predict_function_return_one_element_array(m, d):
+            return np.array(0.2)
+
+        warnings.simplefilter("always")
+        with warnings.catch_warnings(record=True) as w:
+            # Trigger a warning.
+            dx.Explainer(model, X, y, verbose=False, model_type='regression',
+                         predict_function=predict_function_return_2d)
+            assert issubclass(w[-1].category, UserWarning)
+
+        with warnings.catch_warnings(record=True) as w:
+            # Trigger a warning.
+            dx.Explainer(model, X, y, verbose=False, model_type='regression',
+                         predict_function=predict_function_return_3d)
+            assert issubclass(w[-1].category, UserWarning)
+
+        with warnings.catch_warnings(record=True) as w:
+            # Trigger a warning.
+            dx.Explainer(model, X, y, verbose=False, model_type='regression',
+                         predict_function=predict_function_return_one_element_array)
+            assert issubclass(w[-1].category, UserWarning)