ubuntu 16.04搭建doker环境下的深度学习环境

安装N卡驱动

tensorflow 与cuda版本的适配

版本不对应会导致tensorflow镜像无法运行
搭建TensorFlow的GPU版本，必备条件是一块能够支持CUDA的NVIDIA显卡，首先需要安装其基础支持平台CUDA和其机器学习库cuDNN，然后在此基础上搭建对应TensorFlow GPU版本

TensorFlow1.2~2.1各GPU版本CUDA和cuDNN对应版本如下：

安装docker

安装nvidia-docker

在docker环境下测试GPU是否可以正常运行

sudo nvidia-docker run --rm -it nvidia/cuda:9.0-base nvidia-smi

zkf@zkf-ThinkPad-T490:~$ sudo nvidia-docker run --rm  -it  nvidia/cuda:9.0-base nvidia-smi
Tue Aug 25 09:16:43 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.130                Driver Version: 384.130                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Graphics Device     Off  | 00000000:3C:00.0 Off |                  N/A |
| N/A   52C    P8    N/A /  N/A |    327MiB /  2001MiB |     21%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

拉取深度学习镜像

修改默认的pull路径

查看默认路径

docker info

zkf@zkf-ThinkPad-T490:~$ docker info
Client:
WARNING: No swap limit support
 Debug Mode: false
.
.
.
WARNING: No swap limit support
WARNING: No swap limit support
 Docker Root Dir: /var/lib/docker        #默认PULL路径
 .
 .
 .
 Registry Mirrors:
  https://zio7xszp.mirror.aliyuncs.com/
 Live Restore Enabled: false

查看当前默认路径下占用空间的大小

du -h --max-depth=1 /var/lib/docker

zkf@zkf-ThinkPad-T490:~$ sudo du -h --max-depth=1 /var/lib/docker
20K	/var/lib/docker/plugins
219M	/var/lib/docker/overlay2
20K	/var/lib/docker/builder
72K	/var/lib/docker/buildkit
4.0K	/var/lib/docker/swarm
52K	/var/lib/docker/network
4.0K	/var/lib/docker/containers
1.1M	/var/lib/docker/image
4.0K	/var/lib/docker/tmp
4.0K	/var/lib/docker/runtimes
4.0K	/var/lib/docker/trust
28K	/var/lib/docker/volumes
220M	/var/lib/docker                    #镜像当前占用的空间

修改docker中默认镜像和容器的保存位置。

sudo vim /etc/docker/daemon.json

将里面的data-root改为新的docker容器存储位置，添加 "data-root": "/data/docker",
修改完成后长这样

zkf@zkf-ThinkPad-T490:~$ cat /etc/docker/daemon.json
{
 "data-root": "/home/dockerimages",
 "registry-mirrors": ["https://zio7xszp.mirror.aliyuncs.com"],  
 "runtimes": {
        "nvidia": {
            "path": "/usr/bin/nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

重启docker守护进程和服务

sudo systemctl stop docker
sudo systemctl start docker
sudo systemctl status docker

检查修改是否成功

docker info

zkf@zkf-ThinkPad-T490:~$ docker info 
Client:
 Debug Mode: false

.
.
.

 Docker Root Dir: /home/dockerimages        #修改成功
 Debug Mode: false
 Registry: https://index.docker.io/v1/      
 Labels:
 Experimental: false
 Insecure Registries:
  127.0.0.0/8
 Registry Mirrors:
  https://zio7xszp.mirror.aliyuncs.com/
 Live Restore Enabled: false

开始拉取镜像

这里我们使用了deepo镜像：https://hub.docker.com/r/ufoym/deepo/ 其下的 ufoym/deepo:all-py36-jupyter，该镜像收集了大部分深度学习框架，运行在GPU环境，以及配有jupyter。

docker pull ufoym/deepo:all-py36-jupyter

运行镜像

nvidia-docker run -it -d -p 3888:8888 --ipc=host -v /data:/data --name deepo01 ufoym/deepo:all-py36-jupyter jupyter notebook --no-browser --ip=0.0.0.0 --allow-root --NotebookApp.token= --notebook-dir='/data'
参数说明

-v /data:/data：左边是外部路径，右边是内部路径，例如我的文件放在/home/ubuntu/data下，需要挂载到docker内部的路径是/data，则参数配置应该是-v /home/ubuntu/data:/data
–notebook-dir：jupyter工作目录的默认路径，推荐与上面的docker内部数据路径相同，即/data
-p 8888:8888：左边是外部端口，右边是docker镜像端口。如果想将jupyter应用挂载在8080端口，只需修改参数-p 8080:8888即可
–NotebookApp.token：进入jupyter的密码，这里设置的是空

运行成功

root@959f94717541:/# nvidia-smi
Tue Aug 25 10:28:04 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.130                Driver Version: 384.130                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Graphics Device     Off  | 00000000:3C:00.0 Off |                  N/A |
| N/A   47C    P8    N/A /  N/A |    293MiB /  2001MiB |     24%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

运行jupyterbook

http://localhost:3888/tree?