adds set_shap_values methods

RELEASE_NOTES.md

@@ -9,9 +9,11 @@
 ### New Features
 - Export your ExplainerHub to static html with `hub.to_html()` and `hub.save_html()` methods
 - Export your ExplainerHub to a zip file with static html exports with `to_zip()` method
+- Manually add pre-calculated shap values with `explainer.set_shap_values()`
+- Manually add pre-calculated shap interaction values with `explainer.set_shap_interaction_values()`
 
 ### Bug Fixes
--
+- Fixed bug with What if tab components static html export (missing `</div>`)
 -
 
 ### Improvements

explainerdashboard/explainers.py

@@ -962,6 +962,24 @@ def get_shap_values_df(self, pos_label=None):
                     self._shap_values_df, self.onehot_dict, self.merged_cols).astype(self.precision)
         return self._shap_values_df
 
+    def set_shap_values(self, base_value:float, shap_values:np.ndarray):
+        """Set shap values manually. This is useful if you already have 
+        shap values calculated, and do not want to calculate them again inside
+        the explainer instance. Especially for large models and large datasets
+        you may want to calculate shap values on specialized hardware, and then
+        add them to the explainer manually.
+
+        Args:
+            base_value (float): the shap intercept generated by e.g. 
+                base_value = shap.TreeExplainer(model).shap_values(X_test).expected_value
+            shap_values (np.ndarray]): Generated by e.g. 
+                shap_values = shap.TreeExplainer(model).shap_values(X_test)
+        """
+        self._shap_base_value = base_value
+        self._shap_values_df = pd.DataFrame(shap_values, columns=self.columns)
+        self._shap_values_df = merge_categorical_shap_values(
+                    self._shap_values_df, self.onehot_dict, self.merged_cols).astype(self.precision)
+
     @insert_pos_label
     def get_shap_row(self, index=None, X_row=None, pos_label=None):
         if index is not None:
@@ -1018,6 +1036,27 @@ def shap_interaction_values(self, pos_label=None):
                     self.onehot_dict).astype(self.precision)
         return self._shap_interaction_values
 
+    def set_shap_interaction_values(self, shap_interaction_values:np.ndarray):
+        """Manually set shap interaction values in case you have already pre-computed
+        these elsewhere and do not want to re-calculate them again inside the
+        explainer instance.
+
+        Args:
+            shap_interaction_values (np.ndarray): shap interactions values of shape (n, m, m)
+
+        """
+        if not isinstance(shap_interaction_values, np.ndarray):
+            raise ValueError("shap_interaction_values should be a numpy array")
+        if not shap_interaction_values.shape == (len(self.X), len(self.original_cols), len(self.original_cols)):
+            raise ValueError("shap interaction_values should be of shape "
+                        f"({len(self.X)}, {len(self.original_cols)}, {len(self.original_cols)})!")
+
+        self._shap_interaction_values = \
+            merge_categorical_shap_interaction_values(
+                shap_interaction_values, self.columns, self.merged_cols, 
+                self.onehot_dict).astype(self.precision)
+
+
     @insert_pos_label
     def mean_abs_shap_df(self, pos_label=None):
         """Mean absolute SHAP values per feature."""
@@ -2328,6 +2367,58 @@ def get_shap_values_df(self, pos_label=None):
             else:
                 raise ValueError(f"pos_label={pos_label}, but should be either 1 or 0!")
 
+    def set_shap_values(self, base_value:List[float], shap_values:List):
+        """Set shap values manually. This is useful if you already have 
+        shap values calculated, and do not want to calculate them again inside
+        the explainer instance. Especially for large models and large datasets
+        you may want to calculate shap values on specialized hardware, and then
+        add them to the explainer manually.
+
+        Args:
+            base_value (list[float]): list of shap intercept generated by e.g. 
+                base_value = shap.TreeExplainer(model).shap_values(X_test).expected_value.
+                Should be a list with a float for each class. For binary classification
+                and some models shap only provides the base value for the positive class,
+                in which case you need to provide [1-base_value, base_value] or [-base_value, base_value]
+                depending on whether the shap values are for probabilities or logodds. 
+            shap_values (list[np.ndarray]): Generated by e.g. 
+                shap_values = shap.TreeExplainer(model).shap_values(X_test)
+                For binary classification
+                and some models shap only provides the shap values for the positive class,
+                in which case you need to provide [1-shap_values, shap_values] or [-shap_values, shap_values]
+                depending on whether the shap values are for probabilities or logodds. 
+        """
+        if isinstance(base_value, np.ndarray) and base_value.shape == (len(self.labels),):
+            base_value = list(base_value)
+        if not isinstance(base_value, list):
+            raise ValueError("base_value should be a list of floats with an expected value for each class")
+        if not len(base_value) == len(self.labels):
+            raise ValueError("base value should be a list with an expected value "
+                f"for each class, so should be length {len(self.labels)}")
+        self._shap_base_value = base_value
+
+        self._shap_values_df = []
+        if not isinstance(shap_values, list):
+            raise ValueError("shap_values should be a list of np.ndarray with shap values for each class")
+        if len(shap_values) != len(self.labels):
+            raise ValueError("shap_values be a list with a np.ndarray of shap values "
+                f"for each class, so should be length {len(self.labels)}")
+        for sv in shap_values:
+            if not isinstance(sv, np.ndarray):
+                raise ValueError("each element of shap_values should be an np.ndarray!")
+            if sv.shape[0] != len(self.X):
+                raise ValueError(f"Expected shap values to have {len(self.X)} rows!")
+            if sv.shape[1] != len(self.original_cols):
+                raise ValueError(f"Expected shap values to have {len(self.original_columns)} columns!")
+            self._shap_values_df.append(
+                merge_categorical_shap_values(
+                    pd.DataFrame(sv, columns=self.columns), 
+                    self.onehot_dict, self.merged_cols).astype(self.precision)
+            )
+        if len(self.labels) == 2:
+            self._shap_values_df = self._shap_values_df[1]
+
+
     @insert_pos_label
     def get_shap_row(self, index=None, X_row=None, pos_label=None):
         def X_row_to_shap_row(X_row):
@@ -2418,6 +2509,33 @@ def shap_interaction_values(self, pos_label=None):
             else:
                 raise ValueError(f"pos_label={pos_label}, but should be either 1 or 0!")
 
+    def set_shap_interaction_values(self, shap_interaction_values:List[np.ndarray]):
+        """Manually set shap interaction values in case you have already pre-computed
+        these elsewhere and do not want to re-calculate them again inside the
+        explainer instance.
+
+        Args:
+            shap_interaction_values (np.ndarray): shap interactions values of shape (n, m, m)
+
+        """
+        self._shap_interaction_values = []
+        if not isinstance(shap_interaction_values, list):
+            raise ValueError("shap_interaction_values should be a list of np.ndarray with shap interaction values for each class")
+        if len(shap_interaction_values) != len(self.labels):
+            raise ValueError("shap_interaction_values should be a list with a np.ndarray of shap interaction values "
+                f"for each class, so should be length {len(self.labels)}")
+        for siv in shap_interaction_values:
+            if not isinstance(siv, np.ndarray):
+                raise ValueError("each element of shap_values should be an np.ndarray!")
+            if siv.shape != (len(self.X), len(self.original_cols), len(self.original_cols)):
+                raise ValueError(f"Expected shap interaction values to have shape of "
+                            f"({len(self.X)}, {len(self.original_cols)}, {len(self.original_cols)})")
+            self._shap_interaction_values.append(
+                merge_categorical_shap_interaction_values(
+                    siv, self.columns, self.merged_cols, self.onehot_dict).astype(self.precision))
+        if len(self.labels) == 2:
+            self._shap_interaction_values = self._shap_interaction_values[1]
+
     @insert_pos_label
     def mean_abs_shap_df(self, pos_label=None):
         """mean absolute SHAP values"""