diff --git a/python/paddle/static/__init__.py b/python/paddle/static/__init__.py
index bca045852fd06e65cd03841b27d257c5864f5dca..7c9c034e8f9740a403d11c64c60577ae08ace618 100644
--- a/python/paddle/static/__init__.py
+++ b/python/paddle/static/__init__.py
@@ -23,8 +23,8 @@ __all__ = [
 ]
 
 from . import nn
-from .io import save_inference_model
-from .io import load_inference_model
+from .io import save_inference_model #DEFINE_ALIAS
+from .io import load_inference_model #DEFINE_ALIAS
 from ..fluid import Scope  #DEFINE_ALIAS
 from .input import data  #DEFINE_ALIAS
 from .input import InputSpec  #DEFINE_ALIAS
diff --git a/python/paddle/static/io.py b/python/paddle/static/io.py
index 14536b880f585a4e302d18dee35148a9c0037437..a25a8fb191bb2db45746ef3c027fc2993fe78ef4 100644
--- a/python/paddle/static/io.py
+++ b/python/paddle/static/io.py
@@ -97,7 +97,7 @@ def save_inference_model(path_prefix, feed_vars, fetch_vars, executor):
             # Feed data and train process
 
             # Save inference model. Note we don't save label and loss in this example
-            paddle.static.io.save_inference_model(path_prefix, [image], [predict], exe)
+            paddle.static.save_inference_model(path_prefix, [image], [predict], exe)
 
             # In this example, the save_inference_mode inference will prune the default
             # main program according to the network's input node (img) and output node(predict).
@@ -239,10 +239,10 @@ def load_inference_model(path_prefix, executor, **configs):
 
             # Save the inference model
             path_prefix = "./infer_model"
-            paddle.static.io.save_inference_model(path_prefix, [image], [hidden_b], exe)
+            paddle.static.save_inference_model(path_prefix, [image], [hidden_b], exe)
 
             [inference_program, feed_target_names, fetch_targets] = (
-                paddle.static.io.load_inference_model(path_prefix, exe))
+                paddle.static.load_inference_model(path_prefix, exe))
             tensor_img = np.array(np.random.random((64, 784)), dtype=np.float32)
             results = exe.run(inference_program,
                           feed={feed_target_names[0]: tensor_img},