深度学习21天实战caffe学习笔记《3 ：准备Caffe环境》

准备Caffe环境

【如果是其他环境下的配置就请绕道喽，我也没有专门去试一试各个环境下的配置，请谅解~】

官网 http://caffe.berkeleyvision.org/installation.html；

首先在这里介绍一下我的硬件环境：

Ubuntu 14.04 ---------[ win10远程连接ssh（putty）+VNC ]：

• http://www.ubuntu.com/download/desktop 

gcc 4.8.4   || cuda 7.5.18  ||  

1、安装ssh、vnc；

win10 :  putty   、  VNC

Unbuntu : sudo apt-get install ssh

 

cuda7.5安装

1. cuda7.5下载：地址 https://developer.nvidia.com/cuda-downloads 
   文件： cuda_7.5.18_linux.run
2. 登录界面前按Ctrl+Alt+F1进入命令提示符 【禁用nouveau驱动】
3. 执行命令： sudo vi /etc/modprobe.d/blacklist-nouveau.conf
4. 输入以下内容

blacklist nouveau
options nouveau modset=0

最后保存退出（:wq)

1. 执行命令： sudo update-initramfs -u 
   再执行命令： lspci | grep nouveau 查看是否有内容 
   如果没有内容 ，说明禁用成功，如果有内容，就重启一下再查看 
   sudo reboot 
   重启后，进入登录界面的时候，不要登录进入桌面，直接按Ctrl+Alt+F1进入命令提示符。
2. 重启后，登录界面时直接按Ctrl+Alt+F1进入命令提示符
3. 安装依赖项： 
   sudo service lightdm stop 
   sudo apt-get install g++ 
   sudo apt-get install Git 
   sudo apt-get install freeglut3-dev
4. 假设cuda_7.5.18_linux.run位于 ~ 目录，切换到~目录： cd ~
5. 执行命令： sudo sh cuda_7.5.18_linux.run
6. 安装的时候，要让你先看一堆文字（EULA)，我们直接不停的按空格键到100%，然后输入一堆accept,yes,yes或回车进行安装。安装完成后，重启，然后用ls查看一下： 
   ls /dev/nvidia* 
   会看到/dev目录下生成多个nvidia开头文件(夹) 
   或者输入命令： sudo nvcc –version 会显示类似以下信息

dl@dl-Z170X-Gaming-3:~$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2015 NVIDIA Corporation
Built on Tue_Aug_11_14:27:32_CDT_2015
Cuda compilation tools, release 7.5, V7.5.17

• 
  

1. 配置环境变量 
   执行命令： sudo vi /etc/profile 
   文件底部添加以下内容：

export PATH=/usr/local/cuda-7.5/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64:$LD_LIBRARY_PATH

• 
  

1. 编译samples 
   安装成功后在～目录下可以看到一个NVIDIA_CUDA-7.5_Samples文件夹，切换到目录 
   输入sudo make， 大概等个十多分钟后就会把全部的samples编译完毕。生成的可执行文件位于 
   NVIDIA_CUDA-7.5_Samples/bin/x86_64/Linux/release 目录下 
   比如运行 ./nbody可以看到以下demo 

---

cuda安装过程中遇到的问题

Q1

–

1. 在执行命令： sudo apt-get install g++ 时出现以下错误 
   g++ : Depends: g++-4.8 (>= 4.8.2-5~) but it is not going to be installed
2. 是因为ubuntu 14.04的源过旧或不可访问导致，可以通过更新源解决。
3. 首先，备份原始源文件source.list 
   sudo cp /etc/apt/sources.list /etc/apt/sources.list_backup
4. 然后 
   sudo gedit /etc/apt/source.list 
   在文件尾部添加以下内容

deb http://archive.ubuntu.com/ubuntu/ trusty main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu/ trusty-security main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu/ trusty-updates main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu/ trusty-backports main restricted universe multiverse
deb-src http://archive.ubuntu.com/ubuntu/ trusty main restricted universe multiverse
deb-src http://archive.ubuntu.com/ubuntu/ trusty-security main restricted universe multiverse
deb-src http://archive.ubuntu.com/ubuntu/ trusty-updates main restricted universe multiverse
deb-src http://archive.ubuntu.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb-src http://archive.ubuntu.com/ubuntu/ trusty-backports main restricted universe multivers

1. 最后 sudo apt-get update

Q2

W: GPG 错误：http://archive.ubuntukylin.com:10006/ubuntukylin xenial InRelease: 由于没有公钥，无法验证下列签名： NO_PUBKEY 8D5A09DC9B929006
W: 仓库 “http://archive.ubuntukylin.com:10006/ubuntukylin xenial InRelease” 没有数字签名。
N: 无法认证来自该源的数据，所以使用它会带来潜在风险。
N: 参见 apt-secure(8) 手册以了解仓库创建和用户配置方面的细节。
W: 以下 ID 的密钥没有可用的公钥：
8D5A09DC9B929006

• 
  

solution:

sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 8D5A09DC9B929006

• 
  

注意最后的一串密钥就是报错信息里的, 每个人的不一样

Q3

1080Ti显卡安装完CUDA之后要安装显卡驱动，否则提示找不到显卡设备

sudo service lightdm stop  
sudo chmod 777 NVIDIA-Linux-x86_64-378.13.run   
sudo ./NVIDIA-Linux-x86_64-367.27.run  
sudo service lightdm start

• 
  

安装caffe

1. 下载caffe：执行命令： sudo git clone https://github.com/BVLC/caffe.git
2. 安装依赖项：
   sudo apt-get install libatlas-base-dev 
   sudo apt-get install libprotobuf-dev 
   sudo apt-get install libleveldb-dev 
   sudo apt-get install libsnappy-dev 
   sudo apt-get install libopencv-dev 
   sudo apt-get install libboost-all-dev 
   sudo apt-get install libhdf5-serial-dev 
   sudo apt-get install libgflags-dev 
   sudo apt-get install libgoogle-glog-dev 
   sudo apt-get install liblmdb-dev 
   sudo apt-get install protobuf-compiler
3. 编译caffe 
   cd ~/caffe 
   sudo cp Makefile.config.example Makefile.config 
   make all
4. 配置运行环境 
   sudo vi /etc/ld.so.conf.d/caffe.conf 
   添加内容： 
   /usr/local/cuda/lib64
5. 更新配置 
   sudo ldconfig
6. caffe测试，执行以下命令： 
   cd ~/caffe 
   sudo sh data/mnist/get_mnist.sh 
   sudo sh examples/mnist/create_mnist.sh 
   最后测试： 
   sudo sh examples/mnist/train_lenet.sh

运行结果如下： 

其他依赖项

我们查看caffe目录下 Makefile.config 内容如下：

## Refer to http://caffe.berkeleyvision.org/installation.html
# Contributions simplifying and improving our build system are welcome!

# cuDNN acceleration switch (uncomment to build with cuDNN).
 USE_CUDNN := 1

# CPU-only switch (uncomment to build without GPU support).
# CPU_ONLY := 1

# uncomment to disable IO dependencies and corresponding data layers
# USE_OPENCV := 0
# USE_LEVELDB := 0
# USE_LMDB := 0

# uncomment to allow MDB_NOLOCK when reading LMDB files (only if necessary)
#   You should not set this flag if you will be reading LMDBs with any
#   possibility of simultaneous read and write
# ALLOW_LMDB_NOLOCK := 1

# Uncomment if you're using OpenCV 3
 OPENCV_VERSION := 3

# To customize your choice of compiler, uncomment and set the following.
# N.B. the default for Linux is g++ and the default for OSX is clang++
# CUSTOM_CXX := g++

# CUDA directory contains bin/ and lib/ directories that we need.
CUDA_DIR := /usr/local/cuda
# On Ubuntu 14.04, if cuda tools are installed via
# "sudo apt-get install nvidia-cuda-toolkit" then use this instead:
# CUDA_DIR := /usr

# CUDA architecture setting: going with all of them.
# For CUDA < 6.0, comment the *_50 lines for compatibility.
CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
        -gencode arch=compute_20,code=sm_21 \
        -gencode arch=compute_30,code=sm_30 \
        -gencode arch=compute_35,code=sm_35 \
        -gencode arch=compute_50,code=sm_50 \
        -gencode arch=compute_50,code=compute_50

# BLAS choice:
# atlas for ATLAS (default)
# mkl for MKL
# open for OpenBlas
BLAS := mkl
# Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
# Leave commented to accept the defaults for your choice of BLAS
# (which should work)!
# BLAS_INCLUDE := /path/to/your/blas
# BLAS_LIB := /path/to/your/blas

# Homebrew puts openblas in a directory that is not on the standard search path
# BLAS_INCLUDE := $(shell brew --prefix openblas)/include
# BLAS_LIB := $(shell brew --prefix openblas)/lib

# This is required only if you will compile the matlab interface.
# MATLAB directory should contain the mex binary in /bin.
# MATLAB_DIR := /usr/local
# MATLAB_DIR := /Applications/MATLAB_R2012b.app

# NOTE: this is required only if you will compile the python interface.
# We need to be able to find Python.h and numpy/arrayobject.h.
PYTHON_INCLUDE := /usr/include/python2.7 \
        /usr/lib/python2.7/dist-packages/numpy/core/include
# Anaconda Python distribution is quite popular. Include path:
# Verify anaconda location, sometimes it's in root.
# ANACONDA_HOME := $(HOME)/anaconda
# PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
        # $(ANACONDA_HOME)/include/python2.7 \
        # $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \

# Uncomment to use Python 3 (default is Python 2)
# PYTHON_LIBRARIES := boost_python3 python3.5m
# PYTHON_INCLUDE := /usr/include/python3.5m \
#                 /usr/lib/python3.5/dist-packages/numpy/core/include

# We need to be able to find libpythonX.X.so or .dylib.
PYTHON_LIB := /usr/lib
# PYTHON_LIB := $(ANACONDA_HOME)/lib

# Homebrew installs numpy in a non standard path (keg only)
# PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; print(numpy.core.__file__)'))/include
# PYTHON_LIB += $(shell brew --prefix numpy)/lib

# Uncomment to support layers written in Python (will link against Python libs)
# WITH_PYTHON_LAYER := 1

# Whatever else you find you need goes here.
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib

# If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies
# INCLUDE_DIRS += $(shell brew --prefix)/include
# LIBRARY_DIRS += $(shell brew --prefix)/lib

# Uncomment to use `pkg-config` to specify OpenCV library paths.
# (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)
# USE_PKG_CONFIG := 1

BUILD_DIR := build
DISTRIBUTE_DIR := distribute

# Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171
# DEBUG := 1

# The ID of the GPU that 'make runtest' will use to run unit tests.
TEST_GPUID := 0

# enable pretty build (comment to see full commands)
Q ?= @

• 
  

可以看到诸如

# cuDNN acceleration switch (uncomment to build with cuDNN).
# USE_CUDNN := 1

# Uncomment if you're using OpenCV 3
# OPENCV_VERSION := 3

# BLAS choice:
# atlas for ATLAS (default)
# mkl for MKL
# open for OpenBlas
BLAS := atlas

都是使用默认的设置，我们可以安装其他依赖项提高caffe运行效率

opencv3.0安装

1. github上有人写好完整的运行脚本自动下载OpenCV，编译，安装，配置等
2. Caffe + Ubuntu 15.04 + CUDA 7.5 新手安装配置指南 作者 Xin-Yu Ou（欧新宇） 可以到他的网盘中下载 
   PS：为了方便大家使用，我提供一个百度云盘，用于分享部分安装过程中需要用到的软件包和链接地址（所有软件包仅供学术交流使用，请大家尽量去官网下载。）。百度云盘链接:http://pan.baidu.com/s/1qX1uFHa 密码:wysa
3. 在Install-opencv-master文件夹中包含安装各个版本opencv脚本
4. 切换到目录执行： 
   sudo sh Ubuntu/dependencies.sh 
   安装依赖项
5. 执行opencv3.0安装脚本 
   sudo sh Ubuntu/3.0/opencv3_0_0.sh 
   等待安装完成即可
6. 修改Makefile.config

# Uncomment if you're using OpenCV 3

OPENCV_VERSION := 3

1. (可选）opencv3.1已经发布，如果要安装最新的opencv3.1，我们可以先执行 
   sudo sh get_latest_version_download_file.sh 
   获取最新的地址，然后更新opencv3_0_0.sh中的下载地址，同时需要修正文件名等

arch=$(uname -m)
if [ "$arch" == "i686" -o "$arch" == "i386" -o "$arch" == "i486" -o "$arch" == "i586" ]; then
flag=1
else
flag=0
fi
echo "Installing OpenCV 3.0.0"
mkdir OpenCV
cd OpenCV
echo "Removing any pre-installed ffmpeg and x264"
sudo apt-get -y remove ffmpeg x264 libx264-dev
echo "Installing Dependenices"
sudo apt-get -y install libopencv-dev
sudo apt-get -y install build-essential checkinstall cmake pkg-config yasm
sudo apt-get -y install libtiff4-dev libjpeg-dev libjasper-dev
sudo apt-get -y install libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev libxine-dev libgstreamer0.10-dev libgstreamer-plugins-base0.10-dev libv4l-dev
sudo apt-get -y install python-dev python-numpy
sudo apt-get -y install libtbb-dev
sudo apt-get -y install libqt4-dev libgtk2.0-dev
sudo apt-get -y install libfaac-dev libmp3lame-dev libopencore-amrnb-dev libopencore-amrwb-dev libtheora-dev libvorbis-dev libxvidcore-dev
sudo apt-get -y install x264 v4l-utils ffmpeg
sudo apt-get -y install libgtk2.0-dev
echo "Downloading OpenCV 3.0.0"
wget -O opencv-3.0.0.zip http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/3.0.0/opencv-3.0.0.zip/download
echo "Installing OpenCV 3.0.0"
unzip opencv-3.0.0.zip
cd opencv-3.0.0
mkdir build
cd build
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D WITH_TBB=ON -D BUILD_NEW_PYTHON_SUPPORT=ON -D WITH_V4L=ON -D INSTALL_C_EXAMPLES=ON -D INSTALL_PYTHON_EXAMPLES=ON -D BUILD_EXAMPLES=ON -D WITH_QT=ON -D WITH_OPENGL=ON ..
make -j8
sudo make install
sudo sh -c 'echo "/usr/local/lib" > /etc/ld.so.conf.d/opencv.conf'
sudo ldconfig
echo "OpenCV 3.0.0 ready to be used"

安装opencv3遇到的问题

1. 在执行 
   sudo sh Ubuntu/3.0/opencv3_0_0.sh 
   出现有个地方一直卡住了，显示在下载一个文件： ippicv_linux_20141027.tgz 
   因为墙的原因，这个文件无法下载下来
2. [其他文档] ippicv_linux_20141027.tgz 处下载文件 ippicv_linux_20141027.tgz
3. 下载后拷贝到opencv/3rdparty/ippicv/downloads/linux-8b449a536a2157bcad08a2b9f266828b/ 目录下即
4. http://stackoverflow.com/questions/25726768/opencv-3-0-trouble-with-installation

---

安装BLAS——选择MKL

1. 首先下载 MKL(Intel(R) Parallel Studio XE Cluster Edition for Linux 2016) 
   网址： https://software.intel.com/en-us/intel-education-offerings 
   Caffe + Ubuntu 15.04 + CUDA 7.5 新手安装配置指南 作者 Xin-Yu Ou（欧新宇） 可以到他的网盘中下载， 需要自己申请序列号
2. 下载完成后： parallel_studio_xe_2016.tgz
3. 执行以下命令： 
   $ tar zxvf parallel_studio_xe_2016.tar.gz
   $ chmod a+x parallel_studio_xe_2016 -R
   $ sh install_GUI.sh
4. 环境配置： 
   $ sudo gedit /etc/ld.so.conf.d/intel_mkl.conf 
   然后添加以下内容

/opt/intel/lib/intel64
/opt/intel/mkl/lib/intel64

配置生效： sudo ldconfig -v 
安装MKL完成

1. 修改Makefile.config

# BLAS choice:
    # atlas for ATLAS (default)
    # mkl for MKL
    # open for OpenBlas
    BLAS := mkl

cuDNN安装

1. cudnn下载 
   下载地址：https://developer.nvidia.com/cudnn 
   或者到网盘： http://pan.baidu.com/s/1bnOKBO 下载 
   下载相应文件cudnn-7.0-linux-x64-v4.0-rc.tgz, 放到~根目录下
2. 切换到～目录，执行命令

sudo tar xvf cudnn-7.0-linux-x64-v4.0-rc.tgz
cd cuda/include
sudo cp *.h /usr/local/include/
cd ../lib64
sudo cp lib* /usr/local/lib/
cd /usr/local/lib
sudo chmod +r libcudnn.so.4.0.4
sudo ln -sf libcudnn.so.4.0.4 libcudnn.so.4
sudo ln -sf libcudnn.so.4 libcudnn.so
sudo ldconfig

• 
  

1. 修改Makefile.config

# cuDNN acceleration switch (uncomment to build with cuDNN).
USE_CUDNN := 1

---

cudnn版本问题

在make工程的时候出现以下错误：

...

NVCC src/caffe/layers/deconv_layer.cu
NVCC src/caffe/layers/cudnn_conv_layer.cu
src/caffe/layers/cudnn_conv_layer.cu(81): error: argument of type "cudnnAddMode_t" is incompatible with parameter of type "const void *"
          detected during instantiation of "void caffe::CuDNNConvolutionLayer<Dtype>::Forward_gpu(const std::vector<caffe::Blob<Dtype> *, std::allocator<caffe::Blob<Dtype> *>> &, const std::vector<caffe::Blob<Dtype> *, std::allocator<caffe::Blob<Dtype> *>> &) [with Dtype=float]" 
(157): here

...

20 errors detected in the compilation of "/tmp/tmpxft_00002703_00000000-16_cudnn_conv_layer.compute_50.cpp1.ii".
make: *** [.build_release/cuda/src/caffe/layers/cudnn_conv_layer.o] Error 1
make: *** Waiting for unfinished jobs....

解决方案：

更换V3版本cudnn Caffe 工程的一些编译错误以及解决方案

$ cd lib64/
$ sudo cp lib* /usr/local/cuda/lib64/
$ cd ../include/
$ sudo cp cudnn.h /usr/local/cuda/include/
$ cd /usr/local/cuda/lib64/
$ sudo rm -r libcudnn.so libcudnn.so.7.0
$ sudo ln -sf libcudnn.so.7.0.64 libcudnn.so.7.0
$ sudo ln -sf libcudnn.so.7.0 libcudnn.so
$ sudo ldconfig

重新编译测试caffe

1. 编译
   sudo make clean 
   sudo make all
2. sample测试： （ 比不使用cudnn快很多） 
   sh data/mnist/get_mnist.sh 
   sh examples/mnist/create_mnist.sh
3. 我们可以将迭代次数增加到50000次 
   sudo gedit examples/mnist/lenet_solver.prototxt 
   修改max_iter: 50000 
   最后： 
   sh examples/mnist/train_lenet.sh 

编译Python接口

依赖项

sudo apt-get install -y python-numpy python-scipy python-matplotlib python-sklearn python-skimage python-h5py python-protobuf python-leveldb python-networkx python-nose python-pandas python-gflags cython ipython

• 
  

sudo vi ~/.bashrc
添加： export PYTHONPATH=/home/dl/caffe/python:$PYTHONPATH
sudo ldconfig
sudo make pycaffe -j8

• 
  

---

编译matlab接口

1. 安装matlab2014 
   sh /usr/local/MATLAB/R2014a/bin/matlab
2. Makefile.config 中修改 ： MATLAB_DIR := /usr/local/MATLAB/R2014a
3. sudo make matcaffe -j8

---

其他

1. Vi编辑命令 常用vi编辑器命令行

A:当前行的尾部追加内容

i:游标前插入内容

I:游标后插入内容

o:在鼠标所在行的下面添加内容

O:在鼠标所在行的上面添加内容

ESC:退出编辑模式

Ctrl-T:移动到下一个tab

Backspace:向后移动一个字符

Ctrl-U:删除当前

cw:删除游标所在的字符，然后进入编辑模式

cc:删除游标所在的行，然后进入编辑模式

C:删除从游标所在的位置到行尾的字符，然后进入编辑模式

dd:删除当前行

ndd:删除第n行

D:删除当前行游标所在的位置后面的字符

dw:删除邮编所在的字符

d}:删除当前段剩余的字符

d^:删除游标前到行首的字符

c/pat:删除游标后面到第一次匹配字符间的内容

dn:删除游标后面到下一个匹配字符间的内容

dfa:删除当前行游标到匹配字符间的内容（匹配的字符也将被删）

dta:删除当前行游标到匹配字符间的内容（匹配的字符不被删）

dL:删除从游标到屏幕的最后一行之间的内容

dG:删除从游标到文件末尾之间的内容
J:连结上下两行的内容

p:在游标后面插入buffer中的内容

P:在游标前面插入buffer中的内容

rx:用x替换字符

Rtext:用text从游标开始处进行替换

u:撤销最后的改变

U:还原当前行的内容

x:向后删除游标所在位置的字符

X:向前删除游标前面的字符

nX:删除前面的n个字符，游标所在的字符将不会被删

.:还原最后的改变

~:反转字母的大小写

y:拷贝当前行到新的buffer

yy:拷贝当前行

"xyy:拷贝当前行的buffer名为x的buffer

ye:拷贝当单词的末尾

1. 搜狗输入法安装 
   Ubuntu14.04安装搜狗输入法
2. im-config 然后 ibus选取fcitx
3. fcitx-config-gtk3

参考资料

1. Caffe学习系列(1)：安装配置ubuntu14.04+cuda7.5+caffe+cudnn
2. Caffe + Ubuntu 15.04 + CUDA 7.5 新手安装配置指南
3. ubuntu 14.04 install g++问题＂g++:Depends:g++-4.8(>= 4.8.2-5
4. ippicv_linux_20141027.tgz
5. http://stackoverflow.com/questions/25726768/opencv-3-0-trouble-with-installation