rprishabh / tensorflow-cpp-vs2015

tensorflow c++ example for VS2015

tensorflow C++ VS2015

1.编译安装tensorflow

git clone https://github.com/tensorflow/tensorflow
cd tensorflow
mkdir build
cmake ../tensorflow/contrib/cmake

勾选tensorflow_BUILD_SHARED_LIB和tensorflow_ENABLE_GPU 去掉tensorflow_ENABLE_GRPC_SUPPORT

编译生成好的项目，最后别忘了在INSTALL工程上右键生成，以便把所需的文件拷贝至相应的路径。

2.编译本工程

打开tensorflow-cpp.sln编译即可

3.新建项目

1.注意在Inlcude文件夹选项下加入:

C:\Program Files\tensorflow\include
D:\Anaconda\envs\Python35\Lib\site-packages\tensorflow\include
D:\CNN\tensorflow\build\external\nsync\public

2.在Lib文件夹下选项加入：

D:\CNN\tensorflow\build\

3.连接器->输入：

png\install\lib\png12.lib
sqlite\install\lib\sqlite.lib
zlib\install\lib\zlibstatic.lib
gif\install\lib\giflib.lib
jpeg\install\lib\libjpeg.lib
lmdb\install\lib\lmdb.lib
farmhash\install\lib\farmhash.lib
fft2d\\src\lib\fft2d.lib
highwayhash\install\lib\highwayhash.lib
nsync\install\lib\nsync.lib
jsoncpp\src\jsoncpp\src\lib_json\$(Configuration)\jsoncpp.lib
protobuf\src\protobuf\$(Configuration)\libprotobuf.lib
snappy\src\snappy\$(Configuration)\snappy.lib
tf_cc_while_loop.dir\$(Configuration)\tf_cc_while_loop.lib
tf_stream_executor.dir\$(Configuration)\tf_stream_executor.lib
$(Configuration)\tf_protos_cc.lib
$(Configuration)\tf_core_gpu_kernels.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cudart_static.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cuda.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cublas.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cublas_device.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cufft.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\curand.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\extras\CUPTI\libx64\cupti.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cusolver.lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\cudnn.lib

4.连接器->命令行选项：

/WHOLEARCHIVE:tf_core_lib.dir\$(Configuration)\tf_core_lib.lib
/WHOLEARCHIVE:tf_core_cpu.dir\$(Configuration)\tf_core_cpu.lib
/WHOLEARCHIVE:tf_core_framework.dir\$(Configuration)\tf_core_framework.lib
/WHOLEARCHIVE:tf_core_kernels.dir\$(Configuration)\tf_core_kernels.lib
/WHOLEARCHIVE:tf_cc_framework.dir\$(Configuration)\tf_cc_framework.lib
/WHOLEARCHIVE:tf_cc.dir\$(Configuration)\tf_cc_ops.lib 
/WHOLEARCHIVE:tf_core_direct_session.dir\$(Configuration)\tf_core_direct_session.lib 
/WHOLEARCHIVE:tf_core_ops.dir\$(Configuration)\tf_core_ops.lib   
/WHOLEARCHIVE:tf_stream_executor.dir\$(Configuration)\tf_stream_executor.lib
/WHOLEARCHIVE:tf_cc.dir\$(Configuration)\tf_cc.lib 
/WHOLEARCHIVE:tf_cc_ops.dir\$(Configuration)\tf_cc_ops.lib

Note:

本机测试环境为VS2015、CUDA8.0、CUDNN7.0

参考：

• Building a standalone C++ Tensorflow program on Windows
• Tensorflow C++ 编译和调用图模型
• 使用TensorFlow C++ API构建线上预测服务