总结:在安装protobuf,hdf5等的时候指定了安装路径,这导致在11、12两个步骤中要配置编译选项及链接库的位置,这些软件默认应该是安装在
/usr/local
下的,如果这个推断是正确的,那么只需要编译caffe之前在/etc/ld.so.conf
中添加如下内容:
/usr/local/lib
/usr/local/lib64
一、服务器配置
操作系统:centos 6.5
GPU:
[root@localhost ~]# lspci | grep -i nvidia
02:00.0 3D controller: NVIDIA Corporation GK110GL [Tesla K20c] (rev a1)
04:00.0 3D controller: NVIDIA Corporation GK110GL [Tesla K20c] (rev a1)
83:00.0 3D controller: NVIDIA Corporation GK110GL [Tesla K20c] (rev a1)
84:00.0 3D controller: NVIDIA Corporation GK110GL [Tesla K20c] (rev a1)
- 1
- 2
- 3
- 4
- 5
- 1
- 2
- 3
- 4
- 5
注意:这里使用的是命令行版的centos,如果是图形界面,需要另外配置。
二、安装过程1
在安装之前,建议建立一个文件件,将涉及到的安装包放入其内,便于管理:
$ mkdir caffe
$ cd caffe
- 1
- 2
- 1
- 2
另外,记得更新一下系统:
$ yum update -y
- 1
- 1
1.安装cuda
nvidia官网提供了yum源,因此只需安装yum源,便可轻松安装cuda,省去编译的复杂过程。需要注意的是一定要根据直接的操作系统选择合适的版本,详情点这里:
$ yum install epel-release
$ wget http://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-repo-rhel7-7.0-28.x86_64.rpm
$ rpm -iv cuda-repo-rhel7-7.0-28.x86_64.rpm
$ yum search cuda
$ yum install cuda
- 1
- 2
- 3
- 4
- 5
- 6
- 1
- 2
- 3
- 4
- 5
- 6
2.安装opencv
$ yum install opencv-devel
# 配置 OpenCV 环境
$ git clone https://github.com/jayrambhia/Install-OpenCV
$ cd Install-OpenCV/RedHat
$ ./opencv_latest.sh
- 1
- 2
- 3
- 4
- 5
- 1
- 2
- 3
- 4
- 5
在一些资料上看到这样的安装过程,有两个问题:(1)是否需要执行yum install OpenCV-devel
,根据Install-OpenCV/RedHat/opencv_install.sh
文件,应该是不需要的,有勇气的同学可以试一下!(2)执行./opencv_latest.sh
后发现安装不成功,原因可能是centos一些软件版本过低造成的。没办法选择手动安装,记得参考Install-OpenCV/RedHat/opencv_install.sh
2:
$ yum -y groupinstall "Development Tools"
$ yum -y install wget unzip opencv opencv-devel gtk2-devel cmake
$ wget -O opencv-2.4.9.tar.gz
$ http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/2.4.9/opencv-2.4.9.tar.gz/download
$ tar -zxf opencv-2.4.9.tar.gz
$ cd opencv-2.4.9
$ mkdir build
$ cd build
$ cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D CUDA_GENERATION=Kepler ..
$ make -j8
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
cmake需要指定GPU架构3 :-D CUDA_GENERATION=Kepler,否则报错:Unsupported gpu architecture ‘compute_11’
安装的过程可能没那么顺利,经常会遇到各种错误。例如生成lib/libopencv_highgui.so.2.4.9找不到png_set_longjmp_fn
,可以这样解决:
#google之后发现这个函数定义在libpng中
#查看libopencv_highgui.so.2.4.9引用了哪个libpng
$ ldd lib/libopencv_highgui.so.2.4.9 | grep libpng
#查看该lipng中是否定义了png_set_longjmp_fn,注意根据自己的环境替换掉/usr/lib64/libpng12.so.0.49.0
$ readelf -s /usr/lib64/libpng12.so.0.49.0 | grep png_set_longjmp_fn
#没有任何输出,说明没有定义,则需要安装更新的版本
$ wget http://github.com/glennrp/libpng-releases/raw/master/libpng-1.5.23.tar.xz
$ tar -xvf libpng-1.5.23.tar.xz
$ cd libpng-1.5.23
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
安装后发现仍然找不到,这是因为链接时查找路径的问题4:
$ cd /usr/local/lib
$ cp -d libpng15.a libpng15.la libpng15.so* libpng.a libpng.* /usr/lib64
- 1
- 2
- 1
- 2
另外可能遇到的问题是编译过程中提示:
opencv-2.4.9/modules/gpu/src/nvidia/core/NCVPixelOperations.hpp(51): error: a storage class is not allowed in an explicit specialization
- 1
- 1
解决方法是下载NCVPixelOperations.hpp,替换原来的文件重新编译即可5.
3.安装atlas、snappy、boost
$ yum install atlas-devel snappy-devel boost-devel
- 1
- 1
4.安装protobuf
$ cd ~/caffe
$ tar –xvf protobuf-2.5.0.tar.gz
$ cd protobuf-2.5.0
$ ./configure --prefix=/opt/protobuf
$ make
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 1
- 2
- 3
- 4
- 5
- 6
5.安装leveldb
$ cd ~/caffe
$ tar –xvf leveldb-1.7.0.tar.gz
$ cd leveldb-1.7.0
$ make
$ cp libleveldb* /usr/lib/
$ cp –r include/leveldb /usr/local/include
- 1
- 2
- 3
- 4
- 5
- 6
- 1
- 2
- 3
- 4
- 5
- 6
6.安装hdf5
$ cd ~/caffe
$ tar –xvf hdf5-1.8.8.tar.bz2
$ cd hdf5-1.8.8
$ ./configure --prefix=/opt/hdf5
$ make
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 1
- 2
- 3
- 4
- 5
- 6
7.安装glog
$ cd ~/caffe
$ tar –xvf glog-0.3.3.tar.gz
$ cd glog-0.3.3
$ ./configure
$ make
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 1
- 2
- 3
- 4
- 5
- 6
8.安装gflags
$ cd ~/caffe
$ unzip master.zip
$ cd gflags-master
$ mkdir build
$ cd build
$ export CXXFLAGS=”-fPIC”
$ cmake ..
$ make
$ make install
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
9.安装lmdb
$ wget https://codeload.github.com/LMDB/lmdb/tar.gz/LMDB_0.9.15
$ tar -xvf LMDB_0.9.15.tar.gz
$ make
$ make install
- 1
- 2
- 3
- 4
- 1
- 2
- 3
- 4
10.将 /opt/protobuf/bin 加入到 PATH
echo 'export PATH=$PATH:/opt/protobuf/bin' >> ~/.bashrc
- 1
- 1
11.安装caffe
$ cd ~/caffe
$ unzip caffe-master.zip
$ cd caffe-master
$ cp Makefile.config.example Makefile.config
$ vi Makefile
# 修改内容为:
# COMMON_FLAGS 加上 –I/opt/protobuf/include –I/opt/hdf5/include
# LDFLAGS 加上 –L/opt/protobuf/lib –L/opt/hdf5/lib
# LIBRARIES += boost_thread 改为 LIBRARIES += boost_thread-mt
$ vi Makefile.config
# 修改内容为:
# LIBRARY_DIRS 加上 /usr/lib64/atlas
$ make all
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
12. 配置运行环境
$ vi /etc/ld.so.conf.d/caffe.conf
# 增加内容
# /usr/local/cuda/lib64
# /opt/protobuf/lib
# /opt/hdf5/lib
# /usr/local/lib
$ ldconfig
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 1
- 2
- 3
- 4
- 5
- 6
- 7
13. 运行
$ reboot
$ sh data/mnist/get_mnist.sh
$ sh examples/mnist/create_mnist.sh
$ vi examples/mnist/lenet_solver.prototxt
# 修改 ~/caffe/caffe-master/examples/mnist/lenet_solver.prototxt 文件设定运行 CPU 版本或者 GPU 版本。
# 修改最后一行: solver_mode: CPU 或者 solver_mode: GPU
$ time sh examples/mnist/train_lenet.sh
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 1
- 2
- 3
- 4
- 5
- 6
- 7