Fix model info to work with pipeline (#405)

* Fix model info to work with pipeline

* Organize imports

* Fix pylint blank line

deepchecks/checks/overview/model_info.py

@@ -14,6 +14,7 @@
 
 from deepchecks import ModelOnlyBaseCheck, CheckResult
 from deepchecks.utils.validation import model_type_validation
+from deepchecks.utils.model import get_model_of_pipeline
 
 
 __all__ = ['ModelInfo']
@@ -31,13 +32,11 @@ def run(self, model: BaseEstimator) -> CheckResult:
         Returns:
             CheckResult: value is dictionary in format {type: <model_type>, params: <model_params_dict>}
         """
-        return self._model_info(model)
-
-    def _model_info(self, model: BaseEstimator):
         model_type_validation(model)
-        model_type = type(model).__name__
-        model_params = model.get_params()
-        default_params = type(model)().get_params()
+        estimator = get_model_of_pipeline(model)
+        model_type = type(estimator).__name__
+        model_params = estimator.get_params()
+        default_params = type(estimator)().get_params()
 
         # Create dataframe to show
         model_param_df = pd.DataFrame(model_params.items(), columns=['Parameter', 'Value'])

tests/checks/overview/model_info_test.py

@@ -10,6 +10,9 @@
 #
 """Tests for Model Info."""
 from hamcrest import assert_that, has_entries, calling, raises
+from sklearn.impute import SimpleImputer
+from sklearn.pipeline import Pipeline
+
 from deepchecks.checks.overview.model_info import ModelInfo
 from deepchecks.errors import DeepchecksValueError
 
@@ -38,6 +41,16 @@ def test_model_info_object(iris_adaboost):
     assert_model_result(result)
 
 
+def test_model_info_pipeline(iris_adaboost):
+    # Arrange
+    simple_pipeline = Pipeline([('nan_handling', SimpleImputer(strategy='most_frequent')),
+                                ('adaboost', iris_adaboost)])
+    # Act
+    result = ModelInfo().run(simple_pipeline)
+    # Assert
+    assert_model_result(result)
+
+
 def test_model_info_wrong_input():
     # Act
     assert_that(calling(ModelInfo().run).with_args('some string'),