add docstring

hypergbm/experiment.py

@@ -333,6 +333,20 @@ def __init__(self, pipeline_model, data, **kwargs):
         self._explainer = KernelExplainer(pred_f, data=data, keep_index=True, **kwargs)
 
     def __call__(self, X, **kwargs):
+        """Calc explanation of X using shap kernel method.
+
+        Parameters
+        ----------
+        X
+        kwargs
+
+        Returns
+        -------
+            For classification task, output type is List[Explanation], length is `n_classes` in the model,
+                and shape of each element is equal to X.shape.
+            For regression task, output type is Explanation, shape is equal to X.shape
+        """
+
         explainer = self._explainer
         shap_values_data = explainer.shap_values(X, **kwargs)
         from shap._explanation import Explanation

hypergbm/hyper_gbm.py

@@ -696,6 +696,21 @@ def expected_values(self):
             return self._explainers[0].expected_value
 
     def __call__(self, X, transform_kwargs=None, **kwargs):
+        """Calc explanation of X using shap tree method.
+
+        Parameters
+        ----------
+        X
+        transform_kwargs
+        kwargs
+
+        Returns
+        -------
+            For cv training, output type is List[Explanation], length is num folds of CV.
+            For train-test split training, output type is Explanation. if it's a LightGBM training one
+                classification task the output shape is (Xt_n_rows, Xt_n_cols, n_classes), for other algorithms
+                output shape is (Xt_n_rows, Xt_n_cols)
+        """
 
         if transform_kwargs is None:
             transform_kwargs = {}