Ubuntu14.04下配置Keras2.0.4+Tensorflow1.1.0+CUDA7.5+cuDNN5.1.10

Tensorflow1.0以后的版本对于CUDA的默认要求是8.0，所以如果打算直接使用pip命令安装Tensorflow，需要检查CUDA是否匹配，如果匹配，可以使用命令进行安装：

CPU版本：sudo pip install --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.1.0-cp27-none-linux_x86_64.whlGPU版本：sudo pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.1.0-cp27-none-linux_x86_64.whl  

本文在机器已经配置好了CUDA7.5的前提下进行环境配置，已经安装好了NVIDIA显卡驱动，为了避免更换CUDA8.0带来的麻烦，需要通过Tensorflow源码在CUDA7.5下安装Tensorflow1.1.0，关于CUDA的安装，可以参考以下文章：

Ubuntu 安装 tensorflow-gpu + keras

ubuntu16.04下安装TensorFlow(GPU加速)----详细图文教程

以上两篇都要通过deb安装，但是本人用runfile进行安装比较顺利，参考：

深度学习（TensorFlow）环境搭建：（二）Ubuntu16.04+1080Ti显卡驱动

深度学习（TensorFlow）环境搭建：（三）Ubuntu16.04+CUDA8.0+cuDNN7+Anaconda4.4+Python3.6+TensorFlow1.3

Ubuntu 14.04 安装 CUDA 问题及解决

1. cuDNN 5.1下载及环境配置

参考上述文章，选择CUDA和Linux对应的版本下载cuDNN 5.1并解压，将头文件和lib库复制到cuda的目录下，本文定义为/usr/local/cuda

sudo cp cuda/include/cudnn.h /usr/local/cuda/include # 复制到 include 中sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64 # 复制到 lib64 中sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn* # 将头文件复制进去

配置CUDA的环境变量，可以配置为用户环境变量或者全局环境变量，分别在～/bash_profile和/etc/profile中进行配置，配置完后source令其生效：

#set cuda environmentexport CUDA_HOME=/usr/local/cudaexport PATH=/usr/local/cuda/bin:/usr/local/bin:$PATHexport LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$CUDA_HOME/lib64:$CUDA_HOME/extras/CUPTI/lib64"

运行以下命令检查CUDA配置是否生效

nvidia-settings # 打开 NVIDIA 设置界面nvidia-smi

2. 源码安装Tensorflow 1.1.0

2.1 安装编译工具bazel

首先从github下载相应版本的源码，然后安装编译工具bazel：https://bazel.build/versions/master/docs/install.html

安装JDK8

sudo add-apt-repository ppa:webupd8team/javasudo apt-get updatesudo 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.listcurl 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

2.2 配置Tensorflow

解压源码，进入解压后目录，运行：./configure进行配置，配置选项如下：

Please specify the location of python. [Default is /usr/bin/python]: Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]: Do you wish to use jemalloc as the malloc implementation? [Y/n] jemalloc enabledDo you wish to build TensorFlow with Google Cloud Platform support? [y/N] NNo Google Cloud Platform support will be enabled for TensorFlowDo you wish to build TensorFlow with Hadoop File System support? [y/N] NNo Hadoop File System support will be enabled for TensorFlowDo you wish to build TensorFlow with the XLA just-in-time compiler (experimental)? [y/N] No XLA support will be enabled for TensorFlowFound possible Python library paths:  /usr/local/lib/python2.7/dist-packages  /usr/lib/python2.7/dist-packagesPlease input the desired Python library path to use.  Default is [/usr/local/lib/python2.7/dist-packages]Using python library path: /usr/local/lib/python2.7/dist-packagesDo you wish to build TensorFlow with OpenCL support? [y/N] NNo OpenCL support will be enabled for TensorFlowDo you wish to build TensorFlow with CUDA support? [y/N] yCUDA support will be enabled for TensorFlowPlease specify which gcc should be used by nvcc 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]: 7.5Please 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]: 5.1.10Please specify the location where cuDNN 5.1.10 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

特别需要注意选择CUDA版本只要到7.5就好，但是选择cdDNN版本需要能找到对应的libcudnn.so.5.1.10这样的文件，在/usr/local/cuda/lib64中，所以版本选择5.1.10，其他默认即可

2.3 产生pip包并安装

# To build with GPU support:bazel build -c opt --config=cuda //tensorflow/tools/pip_package:build_pip_packagebazel-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-1.1.0-cp27-cp27mu-linux_x86_64.whl sudo pip install /tmp/tensorflow_pkg/tensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl

如果出现如下错误：

tensorflow-1.1.0-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

则进入/tmp/tensorflow_pkg目录，更改tensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl为tensorflow-1.1.0-cp27-none-linux_x86_64.whl

再次pip即可

2.4 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_packagemkdir _python_buildcd _python_buildln -s ../bazel-bin/tensorflow/tools/pip_package/build_pip_package.runfiles/org_tensorflow/* .ln -s ../tensorflow/tools/pip_package/* .sudo python setup.py develop

安装成功后，运行 python -c "import tensorflow;print(tensorflow.__version__)"测试是否可行

运行以下测试用例测试GPU是否可用

import tensorflow as tfhello = tf.constant("hello TensorFlow!")  sess=tf.Session()  print(sess.run(hello))    a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')  b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')  c = tf.matmul(a, b)  sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))  print(sess.run(c)) 

3. 安装Keras 2.0.4

安装之前需要一些依赖，否则在安装scipy时会出错，比如Atlas和gfortran，其他深度学习常用赖项可参考：2015.08.17 Ubuntu 14.04+cuda 7.5+caffe安装配置中关于caffe的依赖设置

#安装scipy所需依赖sudo apt-get install libatlas-base-dev  sudo apt-get install gfortran#安装kerassudo pip install keras==2.0.4

使用命令

python -c "import keras"

测试keras是否安装成功