Caffe + Ubuntu 16.04 + GTK780 + GIGABYTE-uefi DualBIOS 配置

1.安装Ubuntu16.04

进入BIOS功能，将启动优先权各项全部关闭，F10保存，进入U盘启动模式，安装Ubuntu16.04

1.1 EFI引导分区 200M

1.2 swap area(交换空间）大约与本机内存相等

1.3 剩余空间全部用作Ext4日志文件系统， 挂在点选“/”

1.4 Device foor boot loader installation 选择在EFI引导分区下

2.预检

# lspci | grep -i nvidia

01:00.0 VGA compatible controller: NVIDIA Corporation GK110 [Geforce GTX 780] (rev a1)

01:00.1 Audio device:NVIDIA Corporation GK110 HDMI Audio (rev a1)

#uname -m && cat /etc/*release

x86_64

DISTRIB_ID=Ubuntu

DISTRIB_RELEASE=16.04

DISTRIB_CODENAME=xenial
DISTRIB_DESCRIPTION="Ubuntu 16.04.1 LTS"

NAME="Ubuntu"

VERSION="16.04.1 LTS (Xenial Xerus)"

ID=ubuntu

ID_LIKE=debian

PRETTY_NAME="Ubuntu 16.04.1 LTS"

VERSION_ID="16.04"

HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"

BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"

UBUNTU_CODENAME=xenial

#gcc --version

gcc (Ubuntu 5.4.0-6ubuntu1~16.04.02) 5.4.0 20160609
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO

warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

3.更新 /etc/apt/sources.list

deb http://mirrors.163.com/ubuntu/ raring main restricted universe multiverse

deb http://mirrors.163.com/ubuntu/ raring-security main restricted universe multiverse
deb http://mirrors.163.com/ubuntu/ raring-updates main restricted universe multiverse
deb http://mirrors.163.com/ubuntu/ raring-proposed main restricted universe multiverse
deb http://mirrors.163.com/ubuntu/ raring-backports main restricted universe multiverse
deb-src http://mirrors.163.com/ubuntu/ raring main restricted universe multiverse
deb-src http://mirrors.163.com/ubuntu/ raring-security main restricted universe multiverse
deb-src http://mirrors.163.com/ubuntu/ raring-updates main restricted universe multiverse
deb-src http://mirrors.163.com/ubuntu/ raring-proposed main restricted universe multiverse

deb-src http://mirrors.163.com/ubuntu/ raring-backports main restricted universe multiverse

deb http://ubuntu.cn99.com/ubuntu/ precise main restricted universe multiverse
deb http://ubuntu.cn99.com/ubuntu/ precise-updates main restricted universe multiverse
deb http://ubuntu.cn99.com/ubuntu/ precise-security main restricted universe multiverse
deb http://ubuntu.cn99.com/ubuntu/ precise-backports main restricted universe multiverse

deb http://ubuntu.cn99.com/ubuntu-cn/ precise main restricted universe multiverse

#apt-get update 

4. 安装cmake-2.8.12.1 

# sudo wget  http://www.cmake.org/files/v2.8/cmake-2.8.12.1-Linux-i386.tar.gz

#tar xzvfcmake-2.8.11.tar.gz

# cd cmake-2.8.11

# ./configure;  make;  make install

5. 安装依赖

# sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev    protobuf-compiler

# sudo apt-get install --no-install-recommends libboost-all-dev

# sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev

6. 安装cuda

因为需要GPU，故安装CUDA

6.1 安装NVIDIA驱动

Nouveau是由第三方为NVIDIA显卡开发的一个开源3D驱动，没能得到NVIDIA的认可与支持，不过确让Linux更容易的应对各种复杂的NVIDIA显卡环境，让用户安装完系统即可进入桌面并且有不错的显示效果，故很多Linux发行版默认集成了Nouveau驱动，在遇到NVIDIA显卡时默认安装。企业版的Linux更是如此，几乎所有支持图形界面的企业Linux发行版都将Nouveau收入其中。

关闭Nouveau

#sudo gedit /etc/modprobe.d/blacklist.conf  在文件后面加入blacklist nouveau 

#mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r).img.bak (无效)

#dracut -v /boot/initramfs-$(uname -r).img $(uname -r) （sudo apt-get install dracut）

下载驱动程序NVIDIA-Linux-x86_64-367.44.run

# sudo service lightdm stop   进入命令行

# sudo ./NVIDIA-Linux-x86_64-367.44.run

You appear to be running an X server; please exit X ...

# ps -e | grep X

# sudo kill Xorg  kill掉X进程

# sudo ./NVIDIA-Linux-x86_64-367.44.run

中途会报一些错误与警告，不用搭理

# sudo service lightdm start   返回图像界面

进入图像界面后报错：System program problem detected (未解决)

# nvidia-smi  检查驱动

6.2 安装CUDA

# sudo ./cuda_7.5.18_linux.run --override

# sudo gedit ~/.bashrc 修改环境变量

export PATH=$PATH:/usr/local/cuda-7.5/bin

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64

#source~/.bashrc

# sudo gedit /etc/ld.so.conf   修改动态链接

/usr/local/cuda/lib64

# sudo ldconfig

# nvcc -V  检查CUDA

# cd /usr/local/cuda-7.5/samples

# sudo make 测试CUDA

报错：unsupported GNU version! gcc versions later than 4.9 are not supported

gcc版本过高，修改CUDA配置文档，没有选择去降低gcc版本

# cd /usr/local/cuda/include/

# sudo cp host_config.h host_config.h.bak

# sudo gedit host_config.h 

if _GNUC_>4 || (_GNUC_ == 4 && _GNUC_MINOR_ > 9)  将两个4改为5

# cd /usr/local/cuda-7.5/samples

# sudo make  测试CUDA，通过

6.3 安装CUDNN

下载cudnn7.5-linux-x64-v5.1.tgz

# sudo cpinclude/cudnn.h /usr/local/cuda/include

# sudo cp lib64/libcudnn.*  /usr/local/cuda/lib64

# cd /usr/local/cuda/lib 

需要对cudnn进行修改，添加软链接

# sudo rm -rf libcudnn.so libcudnn.so.5

# sudo ln -s libcudnn.5.1.3 libcudnn.so.5

# sudo ln -s libcudnn.so libcudnn.so

7. 安装 MKL   （需要序列号，现有文件包安装）

免费获取网址 https://software.intel.com/zh-cn/intel-mkl/

# tar xzvf l_mkl2017.0.098.tgz

# cd l_mkl2017.0.098

# ./install.sh

# serial number xxxx-xxxxxxxx

# sudo vim ~/.bashrc

export LD_LIBRARY_PATH=/opt/intel/mkl/lib/intel64:$LD_LIBRARY_PATH

# source ~/.bashrc

8. 安装Anaconda

下载Anaconda2-4.1.1-Linux-x86_64

# sudo bash Anaconda2-4.1.1-Linux-x86_64

PREFIX=/usr/anaconda2

installing: hdf5-1.8.16-0 ...

installing: mkl-11.3.3-0 ...

anaconda 自动添加环境变量

export PATH="/root/anaconda/bin:$PATH"  

9. 安装opencv

9.1 opencv2.4.13安装

下载opencv2.4.13.zip

# unzip opencv2.4.13.zip  && cd opencv2.4.13

# mkdir release

# cd release

# cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local/opencv2 -D CUDA_GENERATION=Kepler ..  （GTX780显卡，Kelper架构）

# make 

报错

error:/usr/include/string.h:652:42:error: ‘memcpy’ wasnotdeclaredinthis scope

原因g++版本太新了，兼容一下，在出现上面错误时，在CMakeLists.txt中前几行添加

set(CMAKE_CXX_FLAGS"${CMAKE_CXX_FLAGS} -D_FORCE_INLINES")

参考：http://blog.csdn.net/Swearos/article/details/51307304

# sudo make install

set runtime path of "/usr/local/opencv2/bin/opencv_visualisation" to "/usr/local/opencv2/lib:/usr/local/cuda/lib64" 

... 

9.2  配置环境变量

# sudo gedit ~/.bashrc

# exportPKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/opencv2/lib/pkgconfig

# exportLD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/opencv2/lib

# source ~/.bashrc

9.3 opencv测试

# cd  /usr/opencv-2.4.13/samples/c

# sudo ./build_all.sh

报错： coutours.c:1:39 fatal error: opencv2/imgproc/imgproc_c.h: No such file or direction

第7步已经配置了环境变量，大费周章后

# sudo cp /usr/local/lib/pkgconfig/opencv.pc /usr/lib/pkgconfig 

参考：http://blog.sina.com.cn/s/blog_77ed43e301018iuu.html

#/usr/opencv-2.4.13/samples/c下     ./find_obj 测试通过

10. 安装 caffe

10.1 下载caffe-master.zip 进行安装

# unzip caffe-master.zip 

# cd caffe-master

# sudo cp Makefile.config.example Makefile.config

# sudo vim Makefile.config

###############[Makefile.config]

USE_CUDNN = 1

BLAS := mkl

PYTHON_INCLUDE := /usr/anaconda2/include/python2.7 \

                                   /usr/anaconda2/lib/python2.7/site-packages/numpy/core/include

PYTHON_LIB := /usr/anaconda2/lib

INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include \

                             /opt/intel/mkl/include 

LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib \

                            /opt/intel/mkl/lib 

###############[Makefile.config]

# make all -j8

报错：fatal error: hdf5.h: No such file or direction

进入 Makefile.config文件

INCLUDE_DIRS := $(PYTHON_INCLUDE) ... \

/usr/include/hdf5/serial        将该路径添加至INCLUDE_DIRS中

进入 Makefile 181行

LIBRARIES += ... hdf5_hl hdf5

改为

LIBRARIES += ... hdf5_serial_hl hdf5_serial

参考：http://blog.csdn.net/xue_wenyuan/article/details/52037121

报错： ./include/caffe/util/cudnn.hpp: In function 'void caffe::cudnn::createPoolingDesc（...'

         ./include/caffe/util/cudnn.hpp:127:41: error: too few arguments to function ‘cudnnStatus_t

caffe版本与cudnn版本不匹配,下载最新版caffe安装 

参考：http://blog.csdn.net/u011070171/article/details/52292680

报错： NVCC src/caffe/layers/silence_layer.cu

         /usr/include/string.h:652:42 error:'mempy' was not declared in this scope

在Makefile文件609行，添加 -D_FORCE_INLINES

NVCCFLAGS += -D_FORCE_INLINES -ccbin=$(CXX) -Xcompiler -fPIC $(COMMON_FLAGS)

参考：http://blog.csdn.net/eagelangel/article/details/51531801

# make test -j8

# make runtest -j8

报错：symbol lookup error: /opt/intel/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: 

libmkl_intel_thread未找到libomp5

#sudo cp /opt/intel/lib/intel64_lin/libomp5.so /usr/local/lib

# make pycaffe

export PYTHONPATH=$PYTHONPATH: .../caffe/python

进入python环境，测试import caffe通过

10.2 测试MNIST

进入到caffe根目录下

./data/mnist/get_mnist.sh                       下载minst数据库

一定要在caffe的根目录下运行./examples/mnist/create_mnist.sh

否则会有错误“ build/examples/mnist/convert_mnist_data.bin: not found”

./examples/mnist/create_mnist.sh        数据转成lmdb格式

./examples/mnist/train_lenet.sh            训练mnist

输出正确率为99.05%

至此，caffe安装完成