diff --git a/docs/demo_guides/cuda.md b/docs/demo_guides/cuda.md
index 8b3e76acef590bda19a59388017added6a0b8d52..f863fd86864194c6d022e4cf1fc75eb46725cc2c 100644
--- a/docs/demo_guides/cuda.md
+++ b/docs/demo_guides/cuda.md
@@ -48,7 +48,7 @@ cuda的编译结果位于 `build_cuda/inference_lite_lib`
 
 4、 `demo` 文件夹：c++ demo.
 
-如果编译打开了python选项，则会在 `build_cuda/inference_lite_lib/python/lib/` 目录下生成 `lite_core.so`。
+如果编译打开了python选项，则会在 `build_cuda/inference_lite_lib/python/lib/` 目录下生成 `lite.so`。
 
 ## 运行
 
@@ -66,7 +66,7 @@ wget https://paddle-inference-dist.cdn.bcebos.com/PaddleLite/kite.jpg
 
 二： 运行   
 
-**NOTE:**此处示例使用的是python接口。
+**NOTE：** 此处示例使用的是python接口。
 
 ``` python
 #-*- coding: utf-8 -*-
@@ -75,7 +75,7 @@ import sys
 import numpy as np
 import cv2
 sys.path.append('build_cuda/inference_lite_lib/python/lib')
-from lite_core import *
+from lite import *
 
 def read_img(im_path, resize_h, resize_w):
   im = cv2.imread(im_path).astype('float32')
diff --git a/lite/api/paddle_api.h b/lite/api/paddle_api.h
index dfb0a7fa68579e24eac22a7edee89a8cf9e12d5c..b08f2f5c745f87cda2be181bdea2444b2c11313c 100644
--- a/lite/api/paddle_api.h
+++ b/lite/api/paddle_api.h
@@ -181,7 +181,7 @@ class LITE_API CxxConfig : public ConfigBase {
 #endif
 #ifdef LITE_WITH_CUDA
   void set_multi_stream(bool multi_stream) { multi_stream_ = multi_stream; }
-  int multi_stream() const { return multi_stream_; }
+  bool multi_stream() const { return multi_stream_; }
 #endif
 
 #ifdef LITE_WITH_MLU
diff --git a/lite/demo/cxx/cuda_demo/CMakeLists.txt b/lite/demo/cxx/cuda_demo/CMakeLists.txt
index e27548b4e56ce03098c5c82b3eee49add62cc0a4..f057a1f189fdb92ff33f00d5ceacc83f7fc28c5d 100644
--- a/lite/demo/cxx/cuda_demo/CMakeLists.txt
+++ b/lite/demo/cxx/cuda_demo/CMakeLists.txt
@@ -1,20 +1,24 @@
-project(demo CXX C)
 cmake_minimum_required(VERSION 2.8)
+project(demo CXX C)
+
+add_definitions(-DLITE_WITH_CUDA)
 
 set(TARGET demo)
 set(CMAKE_CXX_FLAGS "-std=c++11 -O3")
 
-set(LITE_LIB "${PROJECT_SOURCE_DIR}/../../cxx")
-set(PROTOBUF_LIB "${PROJECT_SOURCE_DIR}/../../third_party/protobuf")
+set(LITE_ROOT "${PROJECT_SOURCE_DIR}/../../cxx")
+set(PROTOBUF_ROOT "${PROJECT_SOURCE_DIR}/../../third_party/protobuf")
 
-include_directories("${LITE_LIB}/include")
-link_directories("${LITE_LIB}/lib")
-link_directories("${PROTOBUF_LIB}/lib")
+include_directories("${LITE_ROOT}/include")
+link_directories("${LITE_ROOT}/lib")
+link_directories("${PROTOBUF_ROOT}/lib")
+# cuda lib
+link_directories("/usr/local/cuda/lib64/")
 
 add_executable(${TARGET} ${TARGET}.cc)
 
-set(DEPS ${LITE_LIB}/lib/libpaddle_full_api_shared.so)
+set(DEPS ${LITE_ROOT}/lib/libpaddle_full_api_shared.so)
 set(DEPS ${DEPS} protobuf-lite)
-set(DEPS ${DEPS} "-lrt -lpthread -ldl")
+set(DEPS ${DEPS} "-lrt -lpthread -ldl -lcudart")
 
 target_link_libraries(${TARGET} ${DEPS})
diff --git a/lite/tools/build.sh b/lite/tools/build.sh
index 632da81cb12d5a7533252afe5feb8d938fc96cb2..115834460765945f16881f0c5999411b485a1b31 100755
--- a/lite/tools/build.sh
+++ b/lite/tools/build.sh
@@ -350,6 +350,7 @@ function make_cuda {
             -DLITE_WITH_LIGHT_WEIGHT_FRAMEWORK=OFF \
             -DWITH_TESTING=OFF \
             -DLITE_WITH_ARM=OFF \
+            -DLITE_WITH_STATIC_CUDA=OFF \
             -DLITE_WITH_PYTHON=${BUILD_PYTHON} \
             -DLITE_BUILD_EXTRA=ON \
             -DLITE_WITH_XPU=$BUILD_XPU \