update fix

            self.layers[k.name].out_tensor: v

            for k, v in six.iteritems(feed_dict)

        }

        # Recording the number of samples in the input batch

        n_samples = max(feed_dict[self.membership.out_tensor]) + 1

        feed_dict[self._training_placeholder] = 0.0

        feed_results = self.session.run(outputs, feed_dict=feed_dict)

          result = undo_transforms(feed_results[0], transformers)

          feed_results = [result]

        for ind, result in enumerate(feed_results):

          # GraphConvTensorGraph constantly outputs batch_size number of

          # results, only valid samples should be appended to final results

          results[ind].append(result[:n_samples])

      final_results = []