diff --git a/python/paddle_serving_server/serve.py b/python/paddle_serving_server/serve.py
index 3f7ea5c7557bb3c30167392bbe18c634bfae2386..248ff583d6f5e7fa3af9353d06abf07a6e416d8b 100644
--- a/python/paddle_serving_server/serve.py
+++ b/python/paddle_serving_server/serve.py
@@ -233,7 +233,7 @@ def start_multi_card(args, serving_port=None):  # pylint: disable=doc-string-mis
             env_gpus = []
     if args.use_lite:
         print("run using paddle-lite.")
-        start_gpu_card_model(-1, -1, args)
+        start_gpu_card_model(-1, -1, serving_port, args)
     elif len(gpus) <= 0:
         print("gpu_ids not set, going to run cpu service.")
         start_gpu_card_model(-1, -1, serving_port, args)
diff --git a/python/paddle_serving_server/server.py b/python/paddle_serving_server/server.py
index daa315854d96ad45644f79b536b3f298d1424e7a..7ed32fed1c93988b2cac3d02b281e1810b551cb1 100644
--- a/python/paddle_serving_server/server.py
+++ b/python/paddle_serving_server/server.py
@@ -364,9 +364,9 @@ class Server(object):
                        cube_conf=None):
         if workdir == None:
             workdir = "./tmp"
-            os.system("mkdir {}".format(workdir))
+            os.system("mkdir -p {}".format(workdir))
         else:
-            os.system("mkdir {}".format(workdir))
+            os.system("mkdir -p {}".format(workdir))
         os.system("touch {}/fluid_time_file".format(workdir))
 
         if not self.port_is_available(port):