Black reformat

python/interpret-core/interpret/api/templates.py

@@ -135,7 +135,7 @@ def visualize(self, key=None):
                 is_multiclass = is_multiclass_local_data_dict(data_dict)
                 if is_multiclass:
                     # Sort by predicted class' abs feature values
-                    pred_idx = data_dict['perf']['predicted']
+                    pred_idx = data_dict["perf"]["predicted"]
                     sort_fn = lambda x: -abs(x[pred_idx])
                 else:
                     # Sort by abs feature values

python/interpret-core/interpret/blackbox/__init__.py

@@ -5,4 +5,5 @@
 from .shap import ShapKernel  # noqa: F401
 from .sensitivity import MorrisSensitivity  # noqa: F401
 from .partialdependence import PartialDependence  # noqa: F401
+
 # from .permutationimportance import PermutationImportance  # noqa: F401

python/interpret-core/interpret/glassbox/decisiontree.py

@@ -3,7 +3,12 @@
 
 from ..api.base import ExplainerMixin, ExplanationMixin
 from ..utils import unify_data
-from ..utils import gen_name_from_class, gen_local_selector, gen_global_selector, gen_perf_dicts
+from ..utils import (
+    gen_name_from_class,
+    gen_local_selector,
+    gen_global_selector,
+    gen_perf_dicts,
+)
 
 from sklearn.base import ClassifierMixin, RegressorMixin
 from sklearn.tree import DecisionTreeClassifier as SKDT
@@ -365,7 +370,7 @@ def explain_local(self, X, y=None, name=None):
                 "nodes": nodes,
                 "edges": edges,
                 "decision": decision,
-                "perf": None if perf_dicts is None else perf_dicts[i]
+                "perf": None if perf_dicts is None else perf_dicts[i],
             }
             for i, decision in enumerate(decisions)
         ]

python/interpret-core/interpret/glassbox/ebm/ebm.py

@@ -77,13 +77,17 @@ def visualize(self, key=None):
         Returns:
             A Plotly figure.
         """
-        from ...visual.plot import plot_continuous_bar, plot_horizontal_bar, sort_take, is_multiclass_global_data_dict
+        from ...visual.plot import (
+            plot_continuous_bar,
+            plot_horizontal_bar,
+            sort_take,
+            is_multiclass_global_data_dict,
+        )
 
         data_dict = self.data(key)
         if data_dict is None:
             return None
 
-
         # Overall graph
         if self.explanation_type == "global" and key is None:
             data_dict = sort_take(
@@ -121,11 +125,7 @@ class EBMPreprocessor(BaseEstimator, TransformerMixin):
     """ Transformer that preprocesses data to be ready before EBM. """
 
     def __init__(
-        self,
-        feature_names=None,
-        feature_types=None,
-        max_bins=255,
-        binning="quantile",
+        self, feature_names=None, feature_types=None, max_bins=255, binning="quantile",
     ):
         """ Initializes EBM preprocessor.
 
@@ -165,7 +165,9 @@ def fit(self, X):
 
         self.has_fitted_ = False
 
-        schema = autogen_schema(X, feature_names=self.feature_names, feature_types=self.feature_types)
+        schema = autogen_schema(
+            X, feature_names=self.feature_names, feature_types=self.feature_types
+        )
 
         for col_idx in range(X.shape[1]):
             col_name = list(schema.keys())[col_idx]
@@ -192,11 +194,7 @@ def fit(self, X):
                             )
                         )
                     else:  # pragma: no cover
-                        raise ValueError(
-                            "Unknown binning: '{}'.".format(
-                                self.binning
-                            )
-                        )
+                        raise ValueError("Unknown binning: '{}'.".format(self.binning))
 
                 _, bin_edges = np.histogram(col_data, bins=bins)
 
@@ -587,16 +585,15 @@ def __init__(
         #   models without fitting them ourselves.  To conform to a common explaination API we get the feature_names
         #   here.
         feature_names,
-
-        # other packages LightGBM, CatBoost, Scikit-Learn (future) are using categorical specific ways to indicate 
+        # other packages LightGBM, CatBoost, Scikit-Learn (future) are using categorical specific ways to indicate
         #   feature_types.  The benefit to them is that they can accept multiple ways of specifying categoricals like:
         #   categorical = [true, false, true, true] OR categorical = [1, 4, 8] OR categorical = 'all'/'auto'/'none'
         #   We're choosing a different route because for visualization we want to be able to express multiple
         #   different types of data.  For example, if the user has data with strings of "low", "medium", "high"
         #   We want to keep both the ordinal nature of this feature and we wish to preserve the text for visualization
-        #   scikit-learn callers can pre-convert these things to [0, 1, 2] in the correct order because they don't 
-        #   need to worry about visualizing the data afterwards, but for us we  need a way to specify the strings 
-        #   back anyways.  So we need some way to express both the categorical nature of features and the order 
+        #   scikit-learn callers can pre-convert these things to [0, 1, 2] in the correct order because they don't
+        #   need to worry about visualizing the data afterwards, but for us we  need a way to specify the strings
+        #   back anyways.  So we need some way to express both the categorical nature of features and the order
         #   mapping.  We can do this and more complicated conversions via:
         #   feature_types = ["categorical", ["low", "medium", "high"], "continuous", "time", "bool"]
         feature_types,
@@ -629,9 +626,9 @@ def __init__(
         # Native
         learning_rate,
         max_leaves,
-        # Holte, R. C. (1993) "Very simple classification rules perform well on most commonly used datasets" 
+        # Holte, R. C. (1993) "Very simple classification rules perform well on most commonly used datasets"
         # says use 6 as the minimum instances https://link.springer.com/content/pdf/10.1023/A:1022631118932.pdf
-        # TODO PK try setting this (not here, but in our caller) to 6 and run tests to verify the best value.  
+        # TODO PK try setting this (not here, but in our caller) to 6 and run tests to verify the best value.
         min_samples_leaf,
         # Overall
         n_jobs,
@@ -782,9 +779,7 @@ def fit(self, X, y):  # noqa: C901
         # a single float64 for regression, so we do the same
         if is_classifier(self):
             self.intercept_ = np.zeros(
-                EBMUtils.get_count_scores_c(n_classes),
-                dtype=np.float64,
-                order="C",
+                EBMUtils.get_count_scores_c(n_classes), dtype=np.float64, order="C",
             )
         else:
             self.intercept_ = np.float64(0)
@@ -808,9 +803,7 @@ def train_model(estimator, X, y, n_classes):
                 # Discard initial interactions
                 new_model = []
                 new_feature_combinations = []
-                for i, feature_combination in enumerate(
-                    estimator.feature_groups_
-                ):
+                for i, feature_combination in enumerate(estimator.feature_groups_):
                     if len(feature_combination) != 1:
                         continue
                     new_model.append(estimator.model_[i])
@@ -827,8 +820,7 @@ def staged_fit_fn(estimator, X, y, inter_indices=[]):
                     )
 
                 staged_fit_args_iter = (
-                    (estimators[i], X, y, pair_indices)
-                    for i in range(self.outer_bags)
+                    (estimators[i], X, y, pair_indices) for i in range(self.outer_bags)
                 )
 
                 estimators = provider.parallel(staged_fit_fn, staged_fit_args_iter)
@@ -1110,9 +1102,7 @@ def explain_global(self, name=None):
         # Obtain min/max for model scores
         lower_bound = np.inf
         upper_bound = -np.inf
-        for feature_combination_index, _ in enumerate(
-            self.feature_groups_
-        ):
+        for feature_combination_index, _ in enumerate(self.feature_groups_):
             errors = self.term_standard_deviations_[feature_combination_index]
             scores = self.additive_terms_[feature_combination_index]
 
@@ -1302,17 +1292,11 @@ def explain_local(self, X, y=None, name=None):
         is_classification = is_classifier(self)
         if is_classification:
             scores = EBMUtils.classifier_predict_proba(
-                instances,
-                self.feature_groups_,
-                self.additive_terms_,
-                self.intercept_,
+                instances, self.feature_groups_, self.additive_terms_, self.intercept_,
             )
         else:
             scores = EBMUtils.regressor_predict(
-                instances,
-                self.feature_groups_,
-                self.additive_terms_,
-                self.intercept_,
+                instances, self.feature_groups_, self.additive_terms_, self.intercept_,
             )
 
         perf_list = []

python/interpret-core/interpret/glassbox/ebm/internal.py

@@ -645,11 +645,7 @@ def close(self):
         log.info("Deallocation boosting end")
 
     def boosting_step(
-        self,
-        feature_combination_index,
-        learning_rate,
-        max_leaves,
-        min_samples_leaf,
+        self, feature_combination_index, learning_rate, max_leaves, min_samples_leaf,
     ):
 
         """ Conducts a boosting step per feature
@@ -699,9 +695,7 @@ def boosting_step(
                 ct.byref(metric_output),
             )
             if return_code != 0:  # pragma: no cover
-                raise Exception(
-                    "Out of memory in ApplyModelFeatureCombinationUpdate"
-                )
+                raise Exception("Out of memory in ApplyModelFeatureCombinationUpdate")
 
         # log.debug("Boosting step end")
         return metric_output.value
@@ -1094,8 +1088,7 @@ def get_interactions(
         ) as native_ebm_interactions:
             for feature_combination in iter_feature_combinations:
                 score = native_ebm_interactions.get_interaction_score(
-                    feature_combination,
-                    min_samples_leaf,
+                    feature_combination, min_samples_leaf,
                 )
                 interaction_scores.append((feature_combination, score))
 

python/interpret-core/interpret/glassbox/ebm/utils.py

@@ -20,7 +20,7 @@
 class EBMUtils:
     @staticmethod
     def convert_to_intervals(cuts):
-        cuts = np.array(cuts, dtype = np.float64)
+        cuts = np.array(cuts, dtype=np.float64)
 
         if np.isnan(cuts).any():
             raise Exception("cuts cannot contain nan")
@@ -87,7 +87,9 @@ def ebm_train_test_split(
                 y_uniq = len(set(y))
                 n_test_samples = test_size if test_size >= 1 else len(y) * test_size
                 if n_test_samples < y_uniq:  # pragma: no cover
-                    warnings.warn("Too few samples per class, adapting test size to guarantee 1 sample per class.")
+                    warnings.warn(
+                        "Too few samples per class, adapting test size to guarantee 1 sample per class."
+                    )
                     test_size = y_uniq
 
             X_train, X_val, y_train, y_val = train_test_split(
@@ -103,7 +105,6 @@ def ebm_train_test_split(
         if not is_train:
             X_train, y_train = None, None
 
-
         # TODO PK doing a fortran re-ordering here (and an extra copy) isn't the most efficient way
         #         push the re-ordering right to our first call to fit(..) AND stripe convert
         #         groups of rows at once and they process them in fortran order after that

python/interpret-core/interpret/provider/visualize.py

@@ -8,7 +8,10 @@
 from warnings import warn
 
 from ..version import __version__
-JS_URL = "https://unpkg.com/@interpretml/interpret-inline@{}/dist/interpret-inline.js".format(__version__)
+
+JS_URL = "https://unpkg.com/@interpretml/interpret-inline@{}/dist/interpret-inline.js".format(
+    __version__
+)
 
 log = logging.getLogger(__name__)
 
@@ -190,4 +193,9 @@ def __init__(self, detected_envs=None, js_url=None):
     def render(self, explanation, key=-1, **kwargs):
         from ..visual.inline import render
 
-        render(explanation, default_key=key, detected_envs=self.detected_envs, js_url=self.js_url)
+        render(
+            explanation,
+            default_key=key,
+            detected_envs=self.detected_envs,
+            js_url=self.js_url,
+        )

python/interpret-core/interpret/test/test_explainers.py

@@ -5,6 +5,7 @@
 
 from .utils import synthetic_classification, get_all_explainers
 from .utils import assert_valid_explanation, assert_valid_model_explainer
+
 # from ..blackbox import PermutationImportance
 
 from ..glassbox import LogisticRegression

python/interpret-core/interpret/test/test_selenium.py

@@ -4,6 +4,7 @@
 from .utils import synthetic_classification, get_all_explainers
 from ..glassbox import LogisticRegression
 from ..glassbox.decisiontree import TreeExplanation
+
 # from ..blackbox import PermutationImportance
 from ..visual.interactive import set_show_addr, shutdown_show_server, show_link
 from copy import deepcopy

python/interpret-core/interpret/test/utils.py

@@ -8,6 +8,7 @@
 from ..blackbox import ShapKernel
 from ..blackbox import MorrisSensitivity
 from ..blackbox import PartialDependence
+
 # from ..blackbox import PermutationImportance
 
 from ..greybox import TreeInterpreter

python/interpret-core/interpret/utils/all.py

@@ -128,14 +128,9 @@ def gen_local_selector(data_dicts, round=3, is_classification=True):
         records.append(record)
 
     if is_classification:
-        columns = [
-            "Predicted", "PrScore", "Actual", "AcScore",
-            "Resid", "AbsResid"
-        ]
+        columns = ["Predicted", "PrScore", "Actual", "AcScore", "Resid", "AbsResid"]
     else:
-        columns = [
-            "Predicted", "Actual", "Resid", "AbsResid"
-        ]
+        columns = ["Predicted", "Actual", "Resid", "AbsResid"]
 
     df = pd.DataFrame.from_records(records, columns=columns)
     if round is not None:

python/interpret-core/interpret/utils/environment.py

@@ -79,7 +79,7 @@ def _detect_azure_notebook():
 
 def _detect_azureml_notebook_vm():
     nbvm_file_path = "/mnt/azmnt/.nbvm"
-    return (os.path.exists(nbvm_file_path) and os.path.isfile(nbvm_file_path))
+    return os.path.exists(nbvm_file_path) and os.path.isfile(nbvm_file_path)
 
 
 def _detect_vscode():

python/interpret-core/interpret/utils/test/test_all.py

@@ -20,12 +20,7 @@ def test_gen_perf_dicts_regression():
 
 def test_gen_perf_dicts_classification():
     y = np.array([0, 2])
-    scores = np.array(
-        [
-            [0.9, 0.06, 0.04],
-            [0.1, 0.5, 0.4],
-        ]
-    )
+    scores = np.array([[0.9, 0.06, 0.04], [0.1, 0.5, 0.4],])
     expected_predicted = np.array([0, 1])
     expected_actual_score = np.array([0.9, 0.4])
     expected_predicted_score = np.array([0.9, 0.5])

python/interpret-core/interpret/visual/inline.py

@@ -120,7 +120,9 @@ def _build_javascript(viz_obj, id_str=None, default_key=-1, js_url=None):
     else:
         init_js = """
         <script type="text/javascript" src="{0}"></script>
-        """.format(js_url)
+        """.format(
+            js_url
+        )
 
     if id_str is None:
         div_id = "_interpret-viz-{0}".format(uuid.uuid4())
@@ -200,7 +202,9 @@ def render(explanation, id_str=None, default_key=-1, detected_envs=None, js_url=
     else:
         viz_obj = _build_viz_obj(explanation)
 
-    init_js, body_js = _build_javascript(viz_obj, id_str, default_key=default_key, js_url=js_url)
+    init_js, body_js = _build_javascript(
+        viz_obj, id_str, default_key=default_key, js_url=js_url
+    )
 
     if "databricks" in detected_envs:
         _render_databricks(init_js + body_js)