diff --git a/tools/paddle/prune_paddle_model.py b/tools/paddle/prune_paddle_model.py
index 8578b75ad..36cbbbe89 100755
--- a/tools/paddle/prune_paddle_model.py
+++ b/tools/paddle/prune_paddle_model.py
@@ -6,15 +6,13 @@
 import os
 
 
-def new_prepend_feed_ops(inference_program,
-                         feed_target_names,
-                         feed_holder_name='feed'):
+def prepend_feed_ops(program, feed_target_names):
     if len(feed_target_names) == 0:
         return
 
-    global_block = inference_program.global_block()
+    global_block = program.global_block()
     feed_var = global_block.create_var(
-        name=feed_holder_name,
+        name='feed',
         type=core.VarDesc.VarType.FEED_MINIBATCH,
         persistable=True)
 
@@ -33,13 +31,13 @@ def new_prepend_feed_ops(inference_program,
             attrs={'col': i})
 
 
-def append_fetch_ops(program, fetch_target_names, fetch_holder_name='fetch'):
+def append_fetch_ops(program, fetch_target_names):
     """
     In this palce, we will add the fetch op
     """
     global_block = program.global_block()
     fetch_var = global_block.create_var(
-        name=fetch_holder_name,
+        name='fetch',
         type=core.VarDesc.VarType.FETCH_LIST,
         persistable=True)
     print("the len of fetch_target_names:%d" % (len(fetch_target_names)))
@@ -51,16 +49,20 @@ def append_fetch_ops(program, fetch_target_names, fetch_holder_name='fetch'):
             attrs={'col': i})
 
 
-def insert_fetch(program, fetchs, fetch_holder_name="fetch"):
+def insert_by_op_type(program, op_names, op_type):
     global_block = program.global_block()
     need_to_remove_op_index = list()
     for i, op in enumerate(global_block.ops):
-        if op.type == 'fetch':
+        if op.type == op_type:
             need_to_remove_op_index.append(i)
     for index in need_to_remove_op_index[::-1]:
         global_block._remove_op(index)
     program.desc.flush()
-    append_fetch_ops(program, fetchs, fetch_holder_name)
+
+    if op_type == "feed":
+        prepend_feed_ops(program, op_names)
+    else:
+        append_fetch_ops(program, op_names)
 
 
 def parse_arguments():
@@ -73,6 +75,10 @@ def parse_arguments():
         '--model_filename', required=True, help='The input model file name.')
     parser.add_argument(
         '--params_filename', required=True, help='The parameters file name.')
+    parser.add_argument(
+        '--input_names',
+        nargs='+',
+        help='The inputs of pruned model.')
     parser.add_argument(
         '--output_names',
         required=True,
@@ -94,7 +100,6 @@ def parse_arguments():
         sys.exit(-1)
 
     paddle.enable_static()
-    paddle.static.io.prepend_feed_ops = new_prepend_feed_ops
     print("Start to load paddle model...")
     exe = static.Executor(paddle.CPUPlace())
     [program, feed_target_names, fetch_targets] = static.io.load_inference_model(
@@ -102,9 +107,18 @@ def parse_arguments():
         exe,
         model_filename=args.model_filename,
         params_filename=args.params_filename)
-    insert_fetch(program, args.output_names)
-    feed_vars = [program.global_block().var(name) for name in feed_target_names]
-    fetch_vars = [program.global_block().var(out_name) for out_name in args.output_names]
+
+    if args.input_names is not None:
+        insert_by_op_type(program, args.input_names, 'feed')
+        feed_vars = [program.global_block().var(name) for name in args.input_names]
+    else:
+        feed_vars = [program.global_block().var(name) for name in feed_target_names]
+
+    if args.output_names is not None:
+        insert_by_op_type(program, args.output_names, 'fetch')
+        fetch_vars = [program.global_block().var(out_name) for out_name in args.output_names]
+    else:
+        fetch_vars = [out_var for out_var in fetch_targets]
 
     model_name = args.model_filename.split(".")[0]
     path_prefix = os.path.join(args.save_dir, model_name)