Ubuntu16.04安装cuda8.0+cudnn6.0与tensorflow源代码安装

系统：Ubuntu16.04

一、安装Nvidia驱动

如果是笔记本的话，一般是有两块驱动，我们可以使用集成显卡用于显示，用Nvidia显卡计算（AMD目前不支持，可以跳过）

第1步：检查你的电脑有哪些显卡

打开终端，粘贴下面的命令

lspci -k | grep -A 2 -i "VGA"

可以看见，我们笔记本有Intel显卡和Nvidia显卡，如果你的电脑也有这两种显卡，那么继续阅读。

第2步：检查你的电脑Nvidia显卡是否支持cuda以及你的显卡的计算能力

网址为

https://developer.nvidia.com/cuda-gpus

或者

http://blog.csdn.net/real_myth/article/details/44308169

可以看见我们的显卡GT 640M LE的计算能力为3.0，满足要求

第3步：安装合适的Nvidia显卡

cuda自身是包含驱动的，但是它自带的显卡并不一定能够适合我们的笔记本（我的就是如此，显示分辨率始终有问题）

所以在终端里输入下面的命令来查看哪一个专有驱动是推荐安装的

sudo ubuntu-drivers devices

可以看到推荐我们安装的驱动为384，而cuda8.0要求我们的的显卡驱动版本>=375.26，所以我们安装推荐的驱动版本

在官网上下载该驱动，注意下载*.run文件，因为我们需要做一些配置，官网为

http://www.nvidia.cn/Download/index.aspx?lang=cn

下载完成后需要禁用nouveau（Ubuntu16.04系统自带的Nvidia驱动）

打开编辑配置文件

sudo vim /etc/modprobe.d/blacklist.conf

在最后一行添加

blacklist nouveau

保存退出后执行

sudo update-initramfs -u

重启后执行

lsmod | grep nouveau

没有输出即屏蔽好了，进入命令行界面

Ctrl-Alt+F1

禁用X服务

sudo service lightdm stop

切换到你下载驱动的目录执行

sudo ./NVIDIA-Linux-x86_64-384.22.run –no-opengl-files

–no-opengl-files 只安装驱动文件，不安装OpenGL文件，这个参数非常重要

安装完成后重新启动x服务

sudo service lightdm start

运行

nvidia-smi

出现以下类似内容就说明驱动安装完成了

 二、安装cuda

和安装nvidia驱动一样，我们采用*.run文件安装cuda，文件下载网址为

https://developer.nvidia.com/cuda-downloads

第1步：安装依赖

sudo apt-get install freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa libglu1-mesa-dev

第2步：安装cuda

在cuda文件目录下执行安装

sudo ./cuda_8.0.61_375.26_linux.run

除了驱动选择不安装，其他都选择yes

第3步：配置环境

编辑 /etc/profile，在末尾添加

export PATH=/usr/local/cuda-8.0/bin:$PATH

export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64

第4步：检测安装是否完成

在命令行输入

nvcc -V

显示以上信息说明安装完成了，如果还不放心，切换到NVIDIA_CUDA-8.0_Samples/1_Utilities/deviceQuery目录下，执行make

运行产生的deviceQuery

这样说明我们的安装真的已经完成了

三、导入cudnn

下载cudnn6.0，网址为

https://developer.nvidia.com/cudnn

将头文件.h导入/usr/local/cuda-8.0/include

将库文件.so和.a导入/usr/local/cuda-8.0/lib64

四、从源代码安装tensorflow

良心建议：最好参照官网教程https://www.tensorflow.org/install/install_sources

主要是因为软件一直在更新，本教程只适合版本tensorflow-1.4.0rc1，从源代码安装主要是因为我这里网络有问题，直接pip安装老是会中断

第1步：下载源代码

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

cd tensorflow

git checkout r1.4

第2步：配置环境

安装Bazel(个人理解类似于cxx的make)，官网网址https://docs.bazel.build/versions/master/install.html

bazel0.9.0会报错，安装0.8.1即可

安装python2.7相关依赖项

sudo apt-get install python-numpy python-dev python-pip python-wheel

最后安装libcupti-dev

sudo apt-get install libcupti-dev

第3步：配置TensorFlow依赖项所在路径

主要是cuda的环境，我们前面已经配置好了

cd tensorflow # cd to the top-level directory created

./configure

之后会弹出

You have bazel 0.7.0 installed.

Please specify the location of python. [Default is /usr/bin/python]:

直接回车

Found possible Python library paths:

/usr/local/lib/python2.7/dist-packages /usr/lib/python2.7/dist-packages Please input the desired Python library path to use. Default is [/usr/lib/python2.7/dist-packages]

输入/usr/local/lib/python2.7/dist-packages

Do you wish to build TensorFlow with jemalloc as malloc support? [Y/n]: n
No jemalloc as malloc support will be enabled for TensorFlow.

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 Hadoop File System support? [Y/n]: n
No Hadoop File System support will be enabled for TensorFlow.

Do you wish to build TensorFlow with Amazon S3 File System support? [Y/n]: n
No Amazon S3 File System support will be enabled for TensorFlow.

Do you wish to build TensorFlow with XLA JIT support? [y/N]: n
No XLA JIT support will be enabled for TensorFlow.

Do you wish to build TensorFlow with GDR support? [y/N]: n
No GDR support will be enabled for TensorFlow.

Do you wish to build TensorFlow with VERBS support? [y/N]: n
No VERBS support will be enabled for TensorFlow.

Do you wish to build TensorFlow with OpenCL support? [y/N]: n
No OpenCL support will be enabled for TensorFlow.

这些都选择否，暂时用不上，之后配置cuda

Do you wish to build TensorFlow with CUDA support? [y/N]: y
CUDA support will be enabled for TensorFlow.

Please specify the CUDA SDK version you want to use, e.g. 7.0. [Leave empty to default to CUDA 8.0]:

Please specify the location where CUDA 8.0 toolkit is installed. Refer to README.md for more details. [Default is /usr/local/cuda]: /usr/local/cuda-8.0

Please specify the cuDNN version you want to use. [Leave empty to default to cuDNN 6.0]:

Please specify the location where cuDNN 6 library is installed. Refer to README.md for more details. [Default is /usr/local/cuda-8.0]:

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: 3.0]

后面的大同小异，如下：

Do you want to use clang as CUDA compiler? [y/N]: n
nvcc will be used as CUDA compiler.

Please specify which gcc should be used by nvcc as the host compiler. [Default is /usr/bin/gcc]:

Do you wish to build TensorFlow with MPI support? [y/N]: n
No MPI support will be enabled for TensorFlow.

Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]:

Add "--config=mkl" to your bazel command to build with MKL support.
Please note that MKL on MacOS or windows is still not supported.
If you would like to use a local MKL instead of downloading, please set the environment variable "TF_MKL_ROOT" every time before build.
Configuration finished

第4步：编译生成*.whl文件（pip package）

bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package

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

第5步：安装编译生成*.whl文件

sudo pip install /tmp/tensorflow_pkg/tensorflow-1.4.0rc1-cp27-cp27mu-linux_x86_64.whl

第6步：做一个简单的测试

注意：在非Tensorflow文件目录下（比如/home下）打开python shell环境

python

然后输入：

import tensorflow as tf

hello = tf.constant('Hello, TensorFlow!')

sess = tf.Session()

print(sess.run(hello))

如果结果显示：

Hello, TensorFlow!

表示安装完成

 截图纪念！！！

五、编译tensorflow的cxx api接口

由于后期程序采用的cxx编写的，所以这里顺便编译了cxx接口

第1步：编译接口

进入TensorFlow/tensorflow文件执行

bazel build :libtensorflow_cc.so

这样在bazel-bin/tensorflow/目录下会出现libtensorflow.so/libtensorflow_cc.so文件

第2步：配置其他依赖

在使用tensorflow c/c++接口时，会有很多头文件依赖、protobuf版本依赖等问题

tensorflow/contrib/makefile目录下，找到build_all_xxx.sh文件并执行，例如准备在linux上使用，就执行build_all_linux.sh文件

./tensorflow/contrib/makefile/build_all_linux.sh

成功后会在makefile目录下出现一个gen文件夹

在这一步我遇到一个问题

./autogen.sh: 48: ./autogen.sh: autoreconf: not found

原因是因为因为没有安装automake工具

执行sudo apt-get install autoconf automake libtool

然后再次执行build_all_linux.sh

第3步：简单测试

#include <iostream>
#include <tensorflow/core/platform/env.h>
#include <tensorflow/core/public/session.h>

using namespace std;
using namespace tensorflow;

int main(int argc, char **argv) {
  Session* session;
  Status status = NewSession(SessionOptions(), &session);
  if(!status.ok()){
    cout << status.ToString() << endl;
  }
  cout << "Session successfully created." << endl;
  
  return 0;
}

对应的CMakeLists.txt为

cmake_minimum_required(VERSION 2.6)
project(DeeperDepthPrediction)
set( CMAKE_CXX_FLAGS "-std=c++11" )

add_executable(FirstTry  FirstTry.cpp)

include_directories( /home/ai/Programs/TensorFlow
  /home/ai/Programs/TensorFlow/bazel-genfiles  
  /home/ai/Programs/TensorFlow/tensorflow/contrib/makefile/gen/protobuf/include
  /home/ai/Programs/TensorFlow/tensorflow/contrib/makefile/downloads/nsync/public
  /home/ai/Programs/TensorFlow/tensorflow/contrib/makefile/downloads/eigen
)

target_link_libraries( FirstTry 
  /home/ai/Programs/TensorFlow/bazel-bin/tensorflow/libtensorflow_cc.so
  /home/ai/Programs/TensorFlow/bazel-bin/tensorflow/libtensorflow_framework.so

)

 对应的输出结果为

Starting: /home/ai/Projects/DeeperDepthPrediction/DeeperDepthPrediction/build/FirstTry
2017-11-04 19:09:06.960681: 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
2017-11-04 19:09:07.134301: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:892] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2017-11-04 19:09:07.134625: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1031] Found device 0 with properties: 
name: GeForce GT 640M LE major: 3 minor: 0 memoryClockRate(GHz): 0.7085
pciBusID: 0000:02:00.0
totalMemory: 1.95GiB freeMemory: 1.94GiB
2017-11-04 19:09:07.134663: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1121] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GT 640M LE, pci bus id: 0000:02:00.0, compute capability: 3.0)
Session successfully created.
*** Exited normally ***