While implementing shap.PermutationExplainer, I used a simple numpy.array as a masker (as is suggested in the documentation). However, when I tried to implement the shap_values method for this explainer I got the error:

TypeError: ‘numpy.ndarray’ object is not callable

… which is justified because internally, while computing the shap values, a masked model is set up which has the following commands, which assumes self.masker to be a function only (I think) where as my self.masker is an numpy.array

delta_mask = mask ^ last_mask
                if do_delta_masking and delta_mask.sum() == 1:
                    delta_ind = np.nonzero(delta_mask)[0][0]
                    masked_inputs = self.masker(delta_ind, *self.args).copy()
                else:
                    masked_inputs = self.masker(mask, *self.args)