centos7下mxnet0.11.0源码安装

mxnet依赖项如下：

（1）nvidia 驱动，cuda8.0，cudnn5.1

（2）OpenBlas-0.2.20

（3）Opencv2.4.12

（4）Python-2.7.8  or higher

（5）mxnet安装

 

（1）nvidia 驱动，cuda8.0，cudnn5.1

去官网，http://www.nvidia.cn/Download/index.aspx?lang=cn下载驱动，

去官网，https://developer.nvidia.com/cuda-downloads下载cuda-8.0

去官网，https://developer.nvidia.com/rdp/form/cudnn-download-survey下载cudnn-5.1

yum  -y   install  gcc  kernel-devel  kernel-headers              #编译驱动需要

./NVIDIA-Linux-x86_64-384.59.run

./cuda_8.0.61_375.26_linux.run

./cuda_8.0.61.2_linux.run             #更新包

tar -xvf  cudnn-8.0-linux-x64-v5.1.tgz

cp  cuda/include/*  /usr/local/cuda-8.0/include/

cp  cuda/lib/* /usr/local/ cuda-8.0/lib

#添加环境变量

vim ~/.bashrc

export CUDA_HOME=/usr/local/cuda-8.0

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

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

export LD_LIBRARY_PATH="/usr/local/cuda-8.0/lib:${LD_LIBRARY_PATH}"

source  ~/.bashrc

 

（2）OpenBlas-0.2.20

这里也可以选择使用MKL（收费，学生可以免费申请），atlas，实际加速效果，MKL>OpenBlas>atlas

下载链接：https://sourceforge.net/projects/openblas/

tar OpenBLAS-0.2.20.tar.gz

make -j8

make prefix=/usr/local/openblas install

vim ~/.bashrc

加入下面的路径，

export LD_LIBRARY_PATH="/usr/local/openblas/lib:${LD_LIBRARY_PATH}"

source  ~/.bashrc

cp /usr/local/openblas/lib/*  /usr/local/lib/

cp /usr/local/openblas/include/*  /usr/local/include/

cp /usr/local/openblas/bin/*  /usr/bin/

 

（3）Opencv2.4.12

官网下载，http://opencv.org/

unzip  opencv-2.4.12.zip

cd  opencv-2.4.12

mkdir  build

cd  build

cmake..

make  -j8

make  install

vim ~/.bashrc

添加export  PKG_CONFIG_PATH=/usr/local/lib/pkgconfig

Source  ~/.bashrc

 

（4）Python-2.7.8  or higher

默认centos7自带python2.7.5，如果想安装别的版本或者3.2以上的版本，可以参考下面的操作，个人不建议换，里面牵扯的东西太多。

官网下载https://www.python.org/ftp/python/

tar -xvf  Python-2.7.8.tgz

./configure--prefix=/usr/local/python-2.7.8

make  -j8

make  install

mv  /usr/bin/python  python2.7.5

ln -s  /usr/local/python-2.7.8/bin/python   /usr/bin/python

vim /usr/bin/yum

将第一行#!/usr/bin/python改为 #!/usr/bin/python2.7.5，从而保证yum可以使用

vim /usr/libexec/urlgrabber-ext-down

将第一行#!/usr/bin/python改为 #!/usr/bin/python2.7.5，从而保证yum可以使用

 

保证ibus可以使用

vim  /usr/bin/ibus-setup

将最后一行exec /usr/bin/python改为exec /usr/bin/python2.7.5

 

安装numpy

wget http://jaist.dl.sourceforge.NET/project/numpy/NumPy/1.9.0/numpy-1.9.0.zip

unzip numpy-1.9.0.zip

cd numpy-1.9.0

 Python  setup.py  install

reboot

 

（5）mxnet安装

git  clone  --recursive  https://github.com/apache/incubator-mxnet.git   --branch0.11.0.rc3

cd  incubator-mxnet

cp   make/config.mk   ./

做如下的修改，

#-------------------------------------------------------------------------------

#  Template configuration for compiling mxnet

#

#  If you want to change the configuration,please use the following

#  steps. Assume you are on the root directoryof mxnet. First copy the this

#  file so that any local changes will beignored by git

#

#  $ cp make/config.mk .

#

#  Next modify the according entries, and thencompile by

#

#  $ make

#

#  or build in parallel with 8 threads

#

#  $ make -j8

#-------------------------------------------------------------------------------

 

#---------------------

#choice of compiler

#--------------------

 

exportCC = gcc

exportCXX = g++

exportNVCC = nvcc

 

#whether compile with options for MXNet developer

DEV= 0

 

#whether compile with debug

DEBUG= 0

 

#whether compiler with profiler

USE_PROFILER=

 

#the additional link flags you want to add

ADD_LDFLAGS=

 

#the additional compile flags you want to add

ADD_CFLAGS=

 

#---------------------------------------------

#matrix computation libraries for CPU/GPU

#---------------------------------------------

 

#whether use CUDA during compile

USE_CUDA = 1

 

#add the path to CUDA library to link and compile flag

#if you have already add them to environment variable, leave it as NONE

USE_CUDA_PATH = /usr/local/cuda

#USE_CUDA_PATH= NONE

 

#whether use CuDNN R3 library

USE_CUDNN = 1

 

#whether use cuda runtime compiling for writing kernels in native language (i.e.Python)

USE_NVRTC= 0

 

#whether use opencv during compilation

#you can disable it, however, you will not able to use

#imbin iterator

USE_OPENCV= 1

 

#use openmp for parallelization

USE_OPENMP= 1

 

#MKL ML Library for Intel CPU/Xeon Phi

#Please refer to MKL_README.md for details

 

#MKL ML Library folder, need to be root for /usr/local

#Change to User Home directory for standard user

#For USE_BLAS!=mkl only

MKLML_ROOT=/usr/local

 

#whether use MKL2017 library

USE_MKL2017= 0

 

#whether use MKL2017 experimental feature for high performance

#Prerequisite USE_MKL2017=1

USE_MKL2017_EXPERIMENTAL= 0

 

#whether use NNPACK library

USE_NNPACK= 0

 

#choose the version of blas you want to use

#can be: mkl, blas, atlas, openblas

#in default use atlas for linux while apple for osx

UNAME_S:= $(shell uname -s)

ifeq($(UNAME_S), Darwin)

USE_BLAS= apple

else

USE_BLAS = openblas

endif

 

#whether use lapack during compilation

#only effective when compiled with blas versions openblas/apple/atlas/mkl

USE_LAPACK= 1

 

#path to lapack library in case of a non-standard installation

USE_LAPACK_PATH=

 

#add path to intel library, you may need it for MKL, if you did not add the path

#to environment variable

USE_INTEL_PATH= NONE

 

#If use MKL only for BLAS, choose static link automatically to allow pythonwrapper

ifeq($(USE_MKL2017), 0)

ifeq($(USE_BLAS), mkl)

USE_STATIC_MKL= 1

endif

else

USE_STATIC_MKL= NONE

endif

 

#----------------------------

#Settings for power and arm arch

#----------------------------

ARCH:= $(shell uname -a)

ifneq(,$(filter $(ARCH), armv6l armv7l powerpc64le ppc64le aarch64))

         USE_SSE=0

else

         USE_SSE=1

endif

 

#----------------------------

#distributed computing

#----------------------------

 

#whether or not to enable multi-machine supporting

USE_DIST_KVSTORE= 0

 

#whether or not allow to read and write HDFS directly. If yes, then hadoop is

#required

USE_HDFS= 0

 

#path to libjvm.so. required if USE_HDFS=1

LIBJVM=$(JAVA_HOME)/jre/lib/amd64/server

 

#whether or not allow to read and write AWS S3 directly. If yes, then

#libcurl4-openssl-dev is required, it can be installed on Ubuntu by

#sudo apt-get install -y libcurl4-openssl-dev

USE_S3= 0

 

#----------------------------

#additional operators

#----------------------------

 

#path to folders containing projects specific operators that you don't want toput in src/operators

EXTRA_OPERATORS=

 

#----------------------------

#other features

#----------------------------

 

#Create C++ interface package

USE_CPP_PACKAGE= 0

 

#----------------------------

#plugins

#----------------------------

 

#whether to use caffe integration. This requires installing caffe.

#You also need to add CAFFE_PATH/build/lib to your LD_LIBRARY_PATH

#CAFFE_PATH = $(HOME)/caffe

#MXNET_PLUGINS += plugin/caffe/caffe.mk

 

#whether to use torch integration. This requires installing torch.

#You also need to add TORCH_PATH/install/lib to your LD_LIBRARY_PATH

#TORCH_PATH = $(HOME)/torch

#MXNET_PLUGINS += plugin/torch/torch.mk

 

#WARPCTC_PATH = $(HOME)/warp-ctc

#MXNET_PLUGINS += plugin/warpctc/warpctc.mk

 

#whether to use sframe integration. This requires build sframe

#git@github.com:dato-code/SFrame.git

#SFRAME_PATH = $(HOME)/SFrame

#MXNET_PLUGINS += plugin/sframe/plugin.mk

然后执行，

Make  -j8

yum  install  graphviz

pip  install  graphviz

完毕后，mxnet就安装完毕，使用官网的一个矩阵运算进行测试，

$ python

>>> import  mxnet  as mx

>>> a = mx.nd.ones((2, 3), mx.gpu())

>>> b = a * 2 + 1

>>> b.asnumpy()

array([[ 3.,  3., 3.],

      [ 3.,  3.,  3.]], dtype=float32)

References：

https://mxnet.incubator.apache.org/get_started/install.html

https://github.com/apache/incubator-mxnet