diff --git a/learnware/learnware/reuse.py b/learnware/learnware/reuse.py
index d3b0c54..1526190 100644
--- a/learnware/learnware/reuse.py
+++ b/learnware/learnware/reuse.py
@@ -1,6 +1,6 @@
 import torch
 import numpy as np
-import tensorflow as tf
+# import tensorflow as tf
 from typing import Tuple, Any, List, Union, Dict
 from cvxopt import matrix, solvers
 from lightgbm import LGBMClassifier
@@ -56,8 +56,11 @@ class JobSelectorReuser(BaseReuser):
                 pred_y = self.learnware_list[idx].predict(user_data[data_idx_list])
                 if isinstance(pred_y, torch.Tensor):
                     pred_y = pred_y.detach().cpu().numpy()
-                elif isinstance(pred_y, tf.Tensor):
-                    pred_y = pred_y.numpy()
+                # elif isinstance(pred_y, tf.Tensor):
+                #     pred_y = pred_y.numpy()
+                
+                if not isinstance(pred_y, np.ndarray):
+                    raise TypeError(f"Model output must be np.ndarray or torch.Tensor")
 
                 pred_y_list.append(pred_y)
                 data_idxs_list.append(data_idx_list)
@@ -292,9 +295,12 @@ class AveragingReuser(BaseReuser):
             pred_y = self.learnware_list[idx].predict(user_data)
             if isinstance(pred_y, torch.Tensor):
                 pred_y = pred_y.detach().cpu().numpy()
-            elif isinstance(pred_y, tf.Tensor):
-                pred_y = pred_y.numpy()
+            # elif isinstance(pred_y, tf.Tensor):
+            #     pred_y = pred_y.numpy()
 
+            if not isinstance(pred_y, np.ndarray):
+                raise TypeError(f"Model output must be np.ndarray or torch.Tensor")
+            
             if self.mode == "mean":
                 if mean_pred_y is None:
                     mean_pred_y = pred_y