diff --git a/demo/elmo/predict.py b/demo/elmo/predict.py
index be056b732c11cb42fd62d3e2a140a24b891c30b0..3fb9b7140a06e16437f022e1e91d844a35769e05 100644
--- a/demo/elmo/predict.py
+++ b/demo/elmo/predict.py
@@ -168,7 +168,8 @@ if __name__ == '__main__':
     ]
 
     index = 0
-    results = elmo_task.predict(data=data)
+    run_states = elmo_task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
     for batch_result in results:
         # get predict index
         batch_result = np.argmax(batch_result, axis=2)[0]
diff --git a/demo/image-classification/predict.py b/demo/image-classification/predict.py
index 532d813f8673781b14c1501f983723fc5bcbaa9f..ab7d34d94c2c542946a957ad65ede63f71606adf 100644
--- a/demo/image-classification/predict.py
+++ b/demo/image-classification/predict.py
@@ -76,7 +76,8 @@ def predict(args):
     label_map = dataset.label_dict()
     index = 0
     # get classification result
-    results = task.predict(data=data)
+    run_states = task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
     for batch_result in results:
         # get predict index
         batch_result = np.argmax(batch_result, axis=2)[0]
diff --git a/demo/multi-label-classification/predict.py b/demo/multi-label-classification/predict.py
index 5e7caa997b5edc89769dce0e0f530a7d0bbeb2d6..138c5ade3108e72faafeac5d1dfafe45eb2353ab 100644
--- a/demo/multi-label-classification/predict.py
+++ b/demo/multi-label-classification/predict.py
@@ -99,7 +99,8 @@ if __name__ == '__main__':
     ]
 
     index = 0
-    results = multi_label_cls_task.predict(data=data)
+    run_states = multi_label_cls_task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
     for result in results:
         # get predict index
         label_ids = []
diff --git a/demo/senta/predict.py b/demo/senta/predict.py
index 4d22b7bb38e1105c9bc5462f000c5268991530df..1c258343a286d124a4fb70c551168fb1d4f0be22 100644
--- a/demo/senta/predict.py
+++ b/demo/senta/predict.py
@@ -60,7 +60,8 @@ if __name__ == '__main__':
 
     data = ["这家餐厅很好吃", "这部电影真的很差劲"]
 
-    results = cls_task.predict(data=data)
+    run_states = cls_task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
     index = 0
     for batch_result in results:
         batch_result = np.argmax(batch_result, axis=2)[0]
diff --git a/demo/sequence-labeling/predict.py b/demo/sequence-labeling/predict.py
index fb26b86a3035f2ec61f79e83a50f4bc14354f7e3..01fa42807d157d868ce1d0a8b3fe1fd597abe91a 100644
--- a/demo/sequence-labeling/predict.py
+++ b/demo/sequence-labeling/predict.py
@@ -96,7 +96,8 @@ if __name__ == '__main__':
         ["不过重在晋趣，略增明人气息，妙在集古有道、不露痕迹罢了。"],
     ]
 
-    results = seq_label_task.predict(data=data)
+    run_states = seq_label_task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
 
     for num_batch, batch_results in enumerate(results):
         infers = batch_results[0].reshape([-1]).astype(np.int32).tolist()
diff --git a/demo/text-classification/predict.py b/demo/text-classification/predict.py
index 1465d0d62077d924086eb55fd2ed8b5a1f02ca64..e0088dba7d04dec3d338047178820d6c6e3450bd 100644
--- a/demo/text-classification/predict.py
+++ b/demo/text-classification/predict.py
@@ -97,7 +97,8 @@ if __name__ == '__main__':
     ]
 
     index = 0
-    results = cls_task.predict(data=data)
+    run_states = cls_task.predict(data=data)
+    results = [run_state.run_results for run_state in run_states]
     for batch_result in results:
         # get predict index
         batch_result = np.argmax(batch_result, axis=2)[0]
diff --git a/paddlehub/finetune/task.py b/paddlehub/finetune/task.py
index 13c776d4dae9641fa57005fe895bb02405c2f094..d438eaf60e770188c3a78e14fb47116faeea4925 100644
--- a/paddlehub/finetune/task.py
+++ b/paddlehub/finetune/task.py
@@ -498,7 +498,7 @@ class BasicTask(object):
             self.exe, dirname=dirname, main_program=self.main_program)
 
     def finetune_and_eval(self):
-        self.finetune(do_eval=True)
+        return self.finetune(do_eval=True)
 
     def finetune(self, do_eval=False):
         # Start to finetune
@@ -519,6 +519,7 @@ class BasicTask(object):
                 self.eval(phase="test")
 
             self._finetune_end_event(run_states)
+            return run_states
 
     def eval(self, phase="dev"):
         with self.phase_guard(phase=phase):
@@ -526,6 +527,7 @@ class BasicTask(object):
             self._eval_start_event()
             run_states = self._run()
             self._eval_end_event(run_states)
+            return run_states
 
     def predict(self, data, load_best_model=True):
         with self.phase_guard(phase="predict"):
@@ -539,7 +541,7 @@ class BasicTask(object):
             run_states = self._run()
             self._predict_end_event(run_states)
             self._predict_data = None
-        return [run_state.run_results for run_state in run_states]
+        return run_states
 
     def _run(self, do_eval=False):
         with fluid.program_guard(self.main_program, self.startup_program):