Ubuntu16.04.5 configuration NVIDIA GPU-accelerated NVIDIA graphics driver to achieve

Ubuntu16.04.5 configuration NVIDIA GPU-accelerated NVIDIA graphics driver to achieve

Tags (separated by spaces): operation and maintenance series

---

• One: the system environment and system initialization packet ready
• II: Installation Test Procedure

---

One: the system environment and system initialization packet ready

apt-get update 
apt-get install vim openssh-server

---

准备系统所需要的安装包

NVIDIA-Linux-x86_64-440.44.run

cuda_10.2.89_440.33.01_linux.run 

II: Installation Test Procedure

1.1 install Nvidia graphics driver

1. 到官网上下载自己GPU对应版本的显卡驱动。

下载地址：https://www.nvidia.cn/Download/index.aspx?lang=cn

选择你的显卡驱动版本 点击搜索下载即可

---

1.2 Installation NVIDIA-Linux-x86_64-440.44.run

屏蔽自带的显卡驱动

1） vim /etc/modprobe.d/blacklist.conf

2） 在最后一行加上：blacklist nouveau  ，这里是将Ubuntu自带的显卡驱动加入黑名单

3） 在终端输入：update-initramfs –u，使修改生效

4 )　从新启动系统：　reboot　

５）打开终端输入lsmod | grep nouveau，没有输出，则屏蔽成功

6 ) service lightdm stop 

安装 NVIDIA-Linux-x86_64-440.44.run

./NVIDIA-Linux-x86_64-440.44.run

1.3 Installation cuda_10.2.89_440.33.01_linux.run

1. 下载CUDA

下载地址：https://developer.nvidia.com/cuda-downloads

cuda_10.2.89_440.33.01_linux.run 

./cuda_10.2.89_440.33.01_linux.run 

配置环境变量

vim /etc/profile

----
到最后加上

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$PATH
----
source /etc/profile

测试是否安装成功

cd /usr/local/cuda/samples/1_Utilities/deviceQuery
make
./deviceQuery

1.4 Installation CuDNN

1. 下载

网址：https://developer.nvidia.com/rdp/cudnn-download

需要自己注册用户名与密码登录　才能下载
cudnn-10.2-linux-x64-v7.6.5.32.tgz
测试所需包

 tar -zxvf cudnn-10.2-linux-x64-v7.6.5.32.tgz

cd cuda/
cp include/cudnn.h /usr/local/cuda/include/
cp lib64/lib* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/cuda/include/cudnn.h
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

验证是否安装成功

网址：https://developer.nvidia.com/rdp/cudnn-download
下载
libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb

---

cd /usr/local/cuda/lib64/
sudo rm -rf libcudnn.so libcudnn.so.7
sudo ln -s libcudnn.so.7.0.5 libcudnn.so.7
sudo ln -s libcudnn.so.7 libcudnn.so 
sudo ldconfig 

dpkg -i libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb

dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb

dpkg -i libcudnn7-doc_7.6.5.32-1+cuda10.2_amd64.deb

cp -r /usr/src/cudnn_samples_v7/ /home/el/
cd /home/el/cudnn_samples_v7/mnistCUDNN
make clean && make
./mnistCUDNN

---

1.5 Installation anaconda3　

Anaconda3-2019.10-Linux-x86_64.sh 

chmod +x Anaconda3-2019.10-Linux-x86_64.sh 

vim /etc/profile

------
增加

export PATH=/opt/anaconda3/bin:$PATH
------

conda -V 

1.6 installed opencv

1. 下载

网址：https://pypi.org/project/opencv-python/#files

因为安装的python是3.7的，所以opencv名字中要是"cp37"的。
想要安装opencv3，所以名字中要为opencv_python-3.****
我的系统是linux 64位的的，所以名字要是***linux1_x86_64**

软件：
opencv_python-4.1.2.30-cp37-cp37m-manylinux1_x86_64.whl

pip install opencv_python-4.1.2.30-cp37-cp37m-manylinux1_x86_64.whl

conda list |grep opencv

Cmake version 1.7 update system

apt-get install cmake
cmake --version 

在Ubuntu16.04默认安装的cmake版本为3.5.x，可通过一下命令，查看版本。

cmake --version

有时需要安装高版本的cmake。

1.卸载旧版本

apt-get autoremove cmake

2.以安装3.12.3版本为例

$ sudo apt-get install build-essential
$ wget http://www.cmake.org/files/v3.12/cmake-3.12.3.tar.gz

3.解压、安装
$ tar xf cmake-3.12.3.tar.gz
$ cd cmake-3.11.3
$ ./configure
$ make
$ sudo make install

4.解决路径问题

export PATH=/usr/local/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
cmake

5.查看，安装成功

cmake --version

1.8 Configuration xgboost support

1. 下载源代码

apt-get install git

git clone --recursive https://github.com/dmlc/xgboost

2. 编译GPU共享库

cd xgboost
mkdir build
cd build
cmake .. -DUSE_CUDA=ON
make -j

3. 安装Python包
在xgboost根目录下

cd python-package
sudo python3 setup.py install

测试GPU加速

python3 tests/benchmark/benchmark.py