Ubuntu 14.04 源码安装Tensorflow 1.0 (CUDA8.0 or CUDA７.5 or CUDA7.0 版本)

大家在安装Tensorflow时，一个简单的安装方法就是类似：

    pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/mac/tensorflow-0.8.0rc0-py2-none-any.whl

    但是这个安装方法需要提前安装相应版本的CUDA和CUDNN，若有一个版本不对，就会出现问题。尤其是最新的Tensorflow 1.0 需要CUDA 8.0版本，但是一般情况下电脑都装的CUDA 7.0 或7.5。因此，本片博客就是介绍如何使用之前安装的CUDA 7.5和CUDNN 5，使用源码编译Tensorflow 1.0。

    参考链接：https://github.com/tensorflow/tensorflow/blob/r0.12/tensorflow/g3doc/get_started/os_setup.md
    1、首先下载源码：

        $git clone https://github.com/tensorflow/tensorflow

    2、其次，安装编译工具bazel: https://bazel.build/versions/master/docs/install.html

        安装JDK 8

        $ sudo add-apt-repository ppa:webupd8team/java
        $ sudo apt-get update
        $ sudo apt-get install oracle-java8-installer

        把bazel加入到源

        $ 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 -

        安装bazel

        $ sudo apt-get update && sudo apt-get install bazel

        在终端输入bazel测试是否安装成功，效果如下：

    3、配置Tensorflow，如果编译过程中出现问题，多重复几次

        $cd tensorflow
        $ ./configure
        Please specify the location of python. [Default is /usr/bin/python]:
        Do you wish to build TensorFlow with Google Cloud Platform support? [y/N] N
        No Google Cloud Platform support will be enabled for TensorFlow
        Do you wish to build TensorFlow with GPU support? [y/N] y
        GPU support will be enabled for TensorFlow
        Please specify which gcc nvcc should use as the host compiler. [Default is /usr/bin/gcc]:
        Please specify the Cuda SDK version you want to use, e.g. 7.0. [Leave empty to use system default]: 7.5
        Please specify the location where CUDA 7.5 toolkit is installed. Refer to README.md for more details. [Default is /usr/local/cuda]:
        Please specify the cuDNN version you want to use. [Leave empty to use system default]: 5
        Please specify the location where cuDNN 5 library is installed. Refer to README.md for more details. [Default is /usr/local/cuda]:
        Please specify a list of comma-separated Cuda compute capabilities you want to build with.
        You can find the compute capability of your device at: https://developer.nvidia.com/cuda-gpus.
        Please note that each additional compute capability significantly increases your build time and binary size.
        [Default is: "3.5,5.2"]: 3.0
        Setting up Cuda include
        Setting up Cuda lib
        Setting up Cuda bin
        Setting up Cuda nvvm
        Setting up CUPTI include
        Setting up CUPTI lib64
        Configuration finished

    ４、产生pip包，并安装,如果编译过程中出现问题，多重复几次

        # To build with GPU support:
        $ bazel build -c opt --config=cuda //tensorflow/tools/pip_package:build_pip_package

        $ bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg

        # 在/tmp/tensorflow_pkg/文件夹下查找.whl文件，并ｐｉｐ安装，如名字为：tensorflow-1.0.0-cp27-cp27mu-linux_x86_64.whl
        $ sudo pip install /tmp/tensorflow_pkg/tensorflow-1.0.0-cp27-cp27mu-linux_x86_64.whl

        此时，终端会出现错误，如下：

        tensorflow-1.0.0-cp27-cp27mu-linux_x86_64.whl is not a supported wheel on this platform.
        Storing debug log for failure in /home/jiaqi/.pip/pip.log
        处理方法为：进入/tmp/tensorflow_pkg/文件夹下，将tensorflow-1.0.0-cp27-cp27mu-linux_x86_64.whl文件名修改为tensorflow-1.0.0-cp27-ｎｏｎｅ-linux_x86_64.whl即可。

        $sudo mv tensorflow-1.0.0-cp27-cp27mu-linux_x86_64.whl tensorflow-1.0.0-cp27-none-linux_x86_64.whl

    5、最后，设置Tensorflow，并安装到python中

        # To build with GPU support:

        $bazel build -c opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
        $mkdir _python_build
        $cd _python_build
        $ln -s ../bazel-bin/tensorflow/tools/pip_package/build_pip_package.runfiles/org_tensorflow/* .
        $ln -s ../tensorflow/tools/pip_package/* .

        $sudo python setup.py develop

        此时，安装成功，可以在python 中加载一下。

    ６、此时Tensorflow 源码编译成功。

    7、若出现如下问题：

        AttributeError: type object 'NewBase' has no attribute 'is_abstract'

        出现这个问题，应该是six包安装有问题，可以卸载原有版本，重新安装：参照:http://www.cnblogs.com/xiaodi914/p/5687477.html

        $ sudo pip uninstall six
        $ sudo pip install six --upgrade

        一般来说six包的安装位置是/usr/lib/python2.7/dist-packages，建议先试前者（工作站上也是前者），如果six版本还是没有改变，则指定安装位置，如下：

        $ sudo pip install six --upgrade --target="/usr/lib/python2.7/dist-packages"