ECS Cuda Environments Setup Step

ECS Cuda Environments Setup

1.Anaconda

• version: Py 3.5.2
• download url: https://repo.continuum.io/archive/
• https://repo.continuum.io/archive/Anaconda3-4.2.0-Linux-x86_64.sh
• install: sudo bash Anaconda3-4.2.0-Linux-x86_64.sh
• addition:
  • prepending PATH=/home/ming_test/anaconda3/bin to PATH in /home/ming_test/.bashrc
  • source ~/.bashrc

2.Cuda、Cudnn

• version: Cuda9.0 + Cudnn7.0.5 + Tensorflow-gpu1.8.0
• download url:
  • cudnn: https://developer.nvidia.com/rdp/cudnn-archive
  • cuda: https://developer.nvidia.com/cuda-toolkit-archive
• install: https://blog.youkuaiyun.com/qq_34296627/article/details/82833172

Cuda9.0 downloads:

Cuda install:
sudo dpkg -i cuda-repo-ubuntu1604-9-0-local_9.0.176-1_amd64.deb
sudo apt-get update
sudo apt-get install aptitude
sudo aptitude install cuda
sudo gedit ~/.bashrc
export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
export CUDA_HOME=/usr/local/cuda

Cudnn7.0.5 for linux Cuda9.0 downloads:
https://developer.download.nvidia.com/compute/machine-learning/cudnn/secure/v7.0.5/prod/9.0_20171129/cudnn-9.0-linux-x64-v7.tgz?wr6dJhFFBEX1TVnQoqn1LIsgBbICaCVMJmDhkzUsQvDPHsVWEJUHsqxfBOoXKvsincrgtiT1Mwn7zM0HABfCmaZykOMRP7FmQo7XZITlEgr-m5hoOndg9jwdHl2dMjtvcm3KUNJTKCwCx0eLyiwdUSLUxNHfJ5Cswu46xDNVS2FhVPg-Qh8Mb1pz70XomEpcnnZok-MOb5w
Cudnn install:
tar -zxvf cudnn-9.0-linux-x64-v7.tgz 
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/ 
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/ -d 
sudo chmod a+r /usr/local/cuda/include/cudnn.h 
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

Tensorflow-Gpu install
sudo pip install -i https://pypi.tuna.tsinghua.edu.cn/simple tensorflow-gpu==1.8

3.conda virtual

• New: conda create --name fashiongpu --file requirments.txt
• enter: source activate fashiongpu
• exit: source deactivate fashiongpu
• look info: conda info --envs

Google cloud platform Environments Setup

GCP（Google Cloud Platform）入门

1.设置root密码
sudo passwd root

2.查看显卡、内存、硬盘空间
lspci |grep -i vga
lspci |grep -i nvidia
https://blog.youkuaiyun.com/dashan8608/article/details/86242754

top

df -h1

Ubuntu 查看磁盘空间大小命令
ubuntu sda是什么

3.安装Anaconda

4.sudo pip设置
修改sudoers的读写权限

nschineseart@instance-1:~$ ll /etc/sudoers
-r--r----- 1 root root 755 Jun 10 22:53 /etc/sudoers
nschineseart@instance-1:~$ sudo chmod 777 /etc/sudoers
nschineseart@instance-1:~$ ll /etc/sudoers
-rwxrwxrwx 1 root root 755 Jun 10 22:53 /etc/sudoers*
nschineseart@instance-1:~$ sudo chmod 400 /etc/sudoers       再把权限修改回来

Ref
为用户增加sudo权限(修改sudoers文件)
~/.bashrc = /home/user_neme/.bashrc

Failed building wheel for uWSGI，没有安装C编译器
sudo apt-get install gcc
https://stackoverflow.com/questions/47868161/failed-building-wheel-for-uwsgi?rq=1

pip install -r requirements.txt

5.测试Django、uwsgi、nginx

• 入站、出站防火墙设置
  http://zhuanlan.zhihu.com/p/40983101

• 静态文件链接

在服务器上创建目录结构“/var/www/artwork_creation/”
修改目录权限
sudo chmod 777 /var/www/artwork_creation
创建 static 目录，注意顺序是先分配权限，再创建目录
mkdir static

修改settings.py文件
STATIC_ROOT='/var/www/artwork_creation/static/'
STATIC_URL='/static/'
收集所有静态文件到 static_root 指定目录：python manage.py collectstatic

• 服务器方式

1.Django自带的服务器
python manage.py runserver 0:8088
http://外网ip:8088/chinaink/index

2.uwsgi 协议服务器
修改uwsgi.ini文件，打开http，关闭socket，IP为内网ip

[uwsgi]
#socket=10.128.0.2:8099
http=10.128.0.2:8080
#chdir=/Users/xufeng/Code/Demo/Django/artwork_creation
chdir=/home/nschineseart/Code/artwork_creation
wsgi-file=artwork_creation/wsgi.py
processes=1
threads=1
master=True
pidfile=uwsgi.pid
daemonize=uswgi.log

启动：uwsgi --ini uwsgi.ini
停止：uwsgi --stop uwsgi.pid
重启：uwsgi --reload uwsgi.pid
http://外网ip:8080/chinaink/index

3.Nginx服务器
sudo apt-get install nginx
修改/etc/nginx/nginx.conf 如下

user root;
worker_processes auto;
pid /run/nginx.pid;

http{
		...
        server {
                listen      8000;
                server_name  localhost;
                
                #charset koi8-r;
                #access_log  logs/host.access.log  main;
                location / {
                        include uwsgi_params;
                        uwsgi_pass 172.27.0.11:8099;  # 将请求转发到uwsgi开放的socket端口
                 }

                 location /static {
                        alias /var/www/artwork_creation/static/;  # 链接静态资源请求
                 }
        }

}

启动：

cd /usr/sbin/
sudo nginx

停止、重启

nginx -s stop
nginx -s reload

80端口被占用，修改nginx默认端口

nschineseart@instance-vcpu1:/usr/sbin$ sudo nginx 
nginx: [emerg] bind() to 0.0.0.0:80 failed (98: Address already in use)
nginx: [emerg] bind() to [::]:80 failed (98: Address already in use)

解决方案是编辑nginx的配置文件

nschineseart@instance-vcpu1:/usr/sbin$ sudo vi /etc/nginx/sites-available/default
修改这一段：

listen 80;
listen [::]:80 default_server;
为

listen 8008;
listen [::]:8008 ipv6only=on default_server;

访问：
http://外网ip:8000/chinaink/index

linux 修改文件用户组和所有者
sudo chown -R nschineseart artwork_creation/

6.安装cuda、cudnn
新增或移除 GPU，Cuda的安装

服务器没有图像界面因此和笔记本上的配置方法不同