tensorflow-cpp

Compile TensorFlow to C++ library for CMake project.

NOTE :This mannual only test under ubuntu 14.04, so other platforms are not guaranteed.

In general, for compiling tensorflow library as a c++ shared library, we need two other libraries: protobuf and eigen, and a compiling tool bazel.

If you are quite familiar with docker, you can go ahead to look into the Dockerfile in this repository and just ignore all the following explanation.

Prerequisites

platform: ubuntu 14.04

Step1: Install protobuf (version 3.4.0)

1. Install protobuf dependencies

sudo apt-get update && sudo apt-get install -y \
    autoconf \
    automake \
    ccache \
    curl \
    g++ \
    libtool \
    make \
    unzip \
    wget

2: Install protobuf Be careful, we must install version 3.4.0 as current tensorflow(v1.4.0-rc1) need this version's protobuf.

git clone https://github.com/google/protobuf \
    && cd protobuf \
    && git checkout v3.4.0 \
    && ./autogen.sh \
    && ./configure --prefix=/usr \
    && sudo make install \
    && cd ..

Step2: Install Bazel

1. Install JDK 8

sudo apt-get update \
    && sudo apt-get install -y software-properties-common \
    && sudo add-apt-repository -y ppa:webupd8team/java \
    && sudo apt-get update \
    && echo debconf shared/accepted-oracle-license-v1-1 select true | sudo debconf-set-selections \
    && echo debconf shared/accepted-oracle-license-v1-1 seen   true | sudo debconf-set-selections \
    && sudo apt-get install -y oracle-java8-installer

2. Add Bazel distribution URI as a package source

echo "deb [arch=amd64] http://storage.googleapis.com/bazel-apt stable jdk1.8" | sudo tee /etc/apt/sources.list.d/bazel.list \
    && curl https://bazel.build/bazel-release.pub.gpg | sudo apt-key add -

3. Install and update Bazel

sudo apt-get update && sudo apt-get install -y bazel && sudo apt-get upgrade -y bazel

Step3: Install TensorFlow Python dependencies

sudo apt-get update && sudo apt-get install -y \
    python-numpy \
    python-dev \
    python-pip \
    python-wheel \
    swig

Step4: Install TensorFlow

git clone --recurse-submodules -b v1.4.0-rc1 https://github.com/tensorflow/tensorflow.git \
    && cd tensorflow \
    && printf '\n\n\n\n\n\n\n\n\n\n\n\n' | ./configure \
    && bazel build //tensorflow:libtensorflow_cc.so \
    && sudo mkdir -p                /usr/local/include/google/tensorflow \
    && sudo cp -r bazel-genfiles/*  /usr/local/include/google/tensorflow/ \
    && sudo cp -r tensorflow        /usr/local/include/google/tensorflow/ \
    && sudo find                    /usr/local/include/google/tensorflow -type f  ! -name "*.h" -delete \
    && sudo cp -r third_party       /usr/local/include/google/tensorflow/ \
    && sudo cp bazel-bin/tensorflow/libtensorflow_cc.so /usr/local/lib \
    && sudo cp bazel-bin/tensorflow/libtensorflow_framework.so /usr/local/lib \
    && cd ..

Step5: Install eigen header files

Execute the file eigen.sh in this repository as follow, Please specify <tensorflow-root> as the root directory of the TensorFlow repository which cloned in step4, and /usr/local is the directory to where you wish to install eigen.

cd tensorflow-cpp
sudo ./eigen.sh install <tensorflow-root> /usr/local

Step6: Install nsync

For my case(under ubuntu 14.04), i found that i also need another library nsync after preparing everything. If you came across some errors similar to the following, it aproves you also need this:

/usr/local/include/google/tensorflow/tensorflow/core/platform/default/mutex.h:25:22: fatal error: nsync_cv.h: No such file or directory
 #include "nsync_cv.h"

So, just install it as following:

git clone https://github.com/google/nsync.git \
    && cd nsync \
    && git checkout 839fcc53ff9be58218ed55397deb3f8376a1444e \
    && cmake -DCMAKE_INSTALL_PREFIX=/usr \
    && sudo make -j4 install

Testing

After finishing above steps, everything should be ok. Now let's build a small CMake project which run the official tensorflow image recognition demo to validate our work.

1. After clone this repository.
2. Go inside: cd tensorflow-cpp, and Download the model data to the data directory:

   curl -L "https://storage.googleapis.com/download.tensorflow.org/models/inception_v3_2016_08_28_frozen.pb.tar.gz" | tar -C ./data -xz
   

3. Build this project: cmake . && make -j4
4. Run the binary: ./hello_tensorflow
5. If everything is ok, the result should be similar to the following:

   2017-10-27 08:53:28.165985: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
   2017-10-27 08:53:28.779356: I /tensorflow-cpp/main.cc:250] military uniform (653): 0.834306
   2017-10-27 08:53:28.779484: I /tensorflow-cpp/main.cc:250] mortarboard (668): 0.0218692
   2017-10-27 08:53:28.779523: I /tensorflow-cpp/main.cc:250] academic gown (401): 0.0103579
   2017-10-27 08:53:28.779551: I /tensorflow-cpp/main.cc:250] pickelhaube (716): 0.00800814
   2017-10-27 08:53:28.779732: I /tensorflow-cpp/main.cc:250] bulletproof vest (466): 0.00535088

Conclusion

If you want to integrate tensorflow c++ library to your c++ project, after building tensorflow c++ library successfully according to the above instructions:

1. Copy all cmake mond tensorflow_framework.dules(Eigen_VERSION.cmake, FindEigen.cmake, FindTensorFlow.cmake) in this repository to your own project's cmake/Modules directory.
2. Add your project's cmake/Modules directory to CMAKE_MODULE_PATH variable.
3. Link your executable with two more libraries: tensorflow_cc and tensorflow_framework