Centos7下安裝部署MXNET

Centos下安裝MXNET，參考官方文檔http://mxnet.io/get_started/setup.html#prerequisites，

一、安裝python的mxnet步驟如下：

#!/usr/bin/env bash
######################################################################
# This script installs MXNet for Python along with all required dependencies on a Fedora Machine.
# Tested on Fedora 21.0 + distro.
######################################################################
set -e

MXNET_HOME="$HOME/mxnet/"
echo "MXNet root folder: $MXNET_HOME"

echo "Installing basic development tools, atlas, opencv, pip, graphviz ..."
sudo yum update
sudo yum groupinstall -y "Development Tools" "Development Libraries"
sudo yum install -y atlas atlas-devel opencv opencv-devel graphviz graphviz-devel

echo "Building MXNet core. This can take few minutes..."
cd "$MXNET_HOME"
make -j$(nproc)

echo "Installing Numpy..."
sudo yum install numpy

echo "Installing Python setuptools..."
sudo yum install -y python-setuptools python-pip

echo "Adding MXNet path to your ~/.bashrc file"         
echo "export PYTHONPATH=$MXNET_HOME/python:$PYTHONPATH" >> ~/.bashrc
source ~/.bashrc

echo "Install Graphviz for plotting MXNet network graph..."
sudo pip install graphviz

echo "Installing Jupyter notebook..."
sudo pip install jupyter

echo "Done! MXNet for Python installation is complete. Go ahead and explore MXNet with Python :-)"

測試下是否安裝成功：

>>> import mxnet as mx
>>> 

ok，成功！

1 error解決：

1.1 安裝opencv需要注意的地方：

1)、如果停在了下載ippicv的地方，可以自行下載ippicv_linux_20151201.tgz（鏈接：http://www.linuxfromscratch.org/blfs/view/7.9/general/opencv.html），

然后將剛才下載的ippicv文件直接拷貝進入opencv源碼的下面這個目錄：3rdparty/ippicv/downloads/linux-808b791a6eac9ed78d32a7666804320e

2)、錯誤：/usr/bin/ld: /usr/local/include/../lib/libswscale.a(swscale.o): relocation R_X86_64_PC32 against symbol `ff_M24A' can not be used when making a shared object; recompile with -fPIC

需要使用PIC選項重新編譯ffmpeg： CFLAGS="-O3 -fPIC" ./configure --enable-nonfree --enable-pic --enable-shared

若ffmpeg正常安裝后執行ffmpeg時出現如下錯誤：ffmpeg: error while loading shared libraries: libavdevice.so.57: cannot open shared object file: No such file or directory

則 vim /etc/ld.so.conf
加入：/usr/local/lib
執行 ldconfig

 如果上面的方法都失敗，那么需要如下處理：將FFmpeg靜態庫更改為使用動態：

sed -i -e 's/libavformat\.a/libavformat.so/g' \
      -e 's/libavutil\.a/libavutil.so/g' \
      -e 's/libswscale\.a/libswscale.so/g' \
      -e 's/libavresample\.a/libavresample.so/g' \
      -e 's/libavcodec\.a/libavcodec.so/g' \
      cmake/OpenCVFindLibsVideo.cmake

然后：

cmake -D BUILD_opencv_gpu=OFF -D WITH_EIGEN=ON -D WITH_TBB=ON -D WITH_CUDA=OFF -D WITH_TIFF=OFF -DWITH_IPP=OFF -D WITH_1394=OFF -D CMAKE_BUILD_TYPE=RELEASE -DWITH_FFMPEG=ON -DWITH_GSTREAMER=OFF -D CMAKE_INSTALL_PREFIX=/usr/local
 ..
sudo make PREFIX=/usr/local install

 1.2 安裝nnvm錯誤解決

1）如果出現這個錯誤：/usr/bin/ld: cannot find -lcblas

cd /usr/lib64
ln -s libcblas.so.3 libcblas.so

二、安裝R包mxnet

#!/usr/bin/env bash
######################################################################
# This script installs MXNet for R along with all required dependencies on a Ubuntu Machine.
# We recommend to install Microsoft RServer together with Intel MKL library for optimal performance
# More information can be found here:
# https://blogs.technet.microsoft.com/machinelearning/2016/09/15/building-deep-neural-networks-in-the-cloud-with-azure-gpu-vms-mxnet-and-microsoft-r-server/
# Tested on Ubuntu 14.04+ distro.
######################################################################
set -e

MXNET_HOME="$HOME/mxnet/"
echo "MXNet root folder: $MXNET_HOME"

echo "Building MXNet core. This can take few minutes..."
cd "$MXNET_HOME"
make -j$(nproc)

echo "Installing R dependencies. This can take few minutes..."

# make sure we have essential tools installed
is_rscript_installed=$(which Rscript | wc -l)
if [ "$is_rscript_installed" = "0" ]; then
        read -p "Seems like Rscript is not installed. Install Rscript? [Y/n]"
        if [ x"$REPLY" = x"" -o x"$REPLY" = x"y" -o x"$REPLY" = x"Y" ]; then
                sudo apt-get install -y r-base-core
        fi
fi

# libcurl4-openssl-dev and libssl-dev are needed for devtools.
sudo apt-get -y install libcurl4-openssl-dev libssl-dev

# Needed for R XML
sudo apt-get install libxml2-dev

sudo Rscript -e "install.packages('devtools', repo = 'https://cran.rstudio.com')"
cd R-package
sudo Rscript -e "library(devtools); library(methods); options(repos=c(CRAN='https://cran.rstudio.com')); install_deps(dependencies = TRUE)"
cd ..

echo "Compiling R package. This can take few minutes..."
sudo make rpkg

echo "Installing R package..."
sudo R CMD INSTALL mxnet_current_r.tar.gz

echo "Done! MXNet for R installation is complete. Go ahead and explore MXNet with R :-)"

測試下是否安裝成功：

> library(mxnet)
Init Rcpp> 
> 
> a <- mx.nd.ones(c(2,3))
> a
     [,1] [,2] [,3]
[1,]    1    1    1
[2,]    1    1    1

ok！