PyCharm + Docker + GPUs

1.docker运行命令：

docker run --runtime=nvidia -dit --name=my-develop --publish=39822:22 --volume=/home/my/remote_develop:/remote_develop --restart=always euleros-cuda-py373:0.1.0 /bin/bash
 
docker exec -it my-develop /bin/bash
#【ctrl】+【p】 【ctrl】+【q】 （皮球pi_qiu）退出

2.测试GPU：

# 进入镜像
nvidia-docker run -it mxnet/python:gpu bash

#执行python
python

import mxnet as mx
a = mx.nd.ones((2, 3), mx.gpu())
b = a * 2 + 1
b.asnumpy()

参考资料：

https://zhuanlan.zhihu.com/p/27114995

https://blog.51cto.com/5249302/2359420

深度学习环境配置docker+pycharm+GPU

ttps://blog.csdn.net/Ryanpinwei/article/details/78806052

https://cloud.tencent.com/developer/article/1422566

https://blog.51cto.com/5249302/2359420

https://cloud.tencent.com/developer/article/1422566

错误：

nvidia-docker run -it mxnet/python:gpu bash

描述：

docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused \"process_linux.go:432: running prestart hook 1 caused \\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: driver error: failed to process request\\\\n\\\"\"": unknown.
ERRO[0001] error waiting for container: context canceled

处理：

sudo su
echo /var/IEF/nvidia/lib64/ > /etc/ld.so.conf.d/nvidia_driver.conf
ldconfig -v
exit
 

sudo chmod 777 /var
sudo chmod 777 /var/IEF
sudo ls /var/IEF/nvidia -d|xargs -I{} sudo chmod 777 {}
sudo ls /var/IEF/* -ld|grep -E "^d"|awk '{print "sudo chmod 777 "$9}'
sudo ls /var/IEF/* -ld|grep -E "^d"|awk '{print "sudo chmod 777 "$9}'|sh
sudo ls /var/IEF/*/* -ld|grep -E "^d"|awk '{print "sudo chmod 777 "$9}'|sh
sudo ls /var/IEF/*/*/* -ld|grep -E "^d"|awk '{print "sudo chmod 777 "$9}'|sh
sudo ls /var/IEF/*/*/*/* -ld|grep -E "^d"|awk '{print "sudo chmod 777 "$9}'|sh

sudo ls -ld /var/IEF/nvidia/* |grep "\-rwxr\-x"|awk '{print "sudo chmod 755 " $9}'|sh
sudo ls -ld /var/IEF/nvidia/drivers/* |grep "\-rwxr\-x"|awk '{print "sudo chmod 755 " $9}'|sh
sudo ls -ld /var/IEF/nvidia/bin/* |grep "\-rwxr\-x"|awk '{print "sudo chmod 755 " $9}'|sh
sudo ls -ld /var/IEF/nvidia/lib64/* |grep "\-rwxr\-x"|awk '{print "sudo chmod 755 " $9}'|sh
sudo ls -ld /var/IEF/nvidia/lib64/vdpau/* |grep "\-rwxr\-x"|awk '{print "sudo chmod 755 " $9}'|sh
nvidia-container-cli -k -d /dev/tty info

描述：

docker run --runtime=nvidia nvidia/cuda:9.0-base nvidia-smi
docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "exec: \"nvidia-smi\": executable file not found in $PATH": unknown.
ERRO[0001] error waiting for container: context canceled

处理：

docker run --runtime=nvidia -v /var/IEF/nvidia:/usr/local/nvidia nvidia/cuda:9.0-base nvidia-smi

描述：

docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused \"process_linux.go:432: running prestart hook 1 caused \\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: mount error: stat failed: /dev/nvidia-modeset: no such file or directory\\\\n\\\"\"": unknown.
ERRO[0001] error waiting for container: context canceled

处理：

docker run是添加参数，不能包含display

NVIDIA_DRIVER_CAPABILITIES
此选项控制将在容器内部安装哪些驱动程序库/二进制文件。

可能的值
compute,video，graphics,utility…：以逗号分隔的驱动程序列表，列出了容器所需的功能。
all：启用所有可用的驱动程序功能。
空或未设置：使用默认驱动程序功能：utility。
支持的驱动程序功能
compute：对于CUDA和OpenCL应用程序是必需的。
compat32：运行32位应用程序所需。
graphics：运行OpenGL和Vulkan应用程序所需。
utility：使用nvidia-smi和NVML必需。
video：使用视频编解码器SDK所需。
display：使用X11显示器需要。

docker run --env NVIDIA_DRIVER_CAPABILITIES=compute,graphics,utility,video

来源：oschina

链接：https://my.oschina.net/mengyoufengyu/blog/3159287

标签

SMI

pycharm

Docker