Docker配置TensorFlow-CPU开发环境

查找TensorFlow镜像

seven@seven:~$ sudo docker search tensorflow  -s 10
NAME                                DESCRIPTION                                     STARS               OFFICIAL            AUTOMATED
tensorflow/tensorflow               Official docker images for deep learning fra…   1150                                    
jupyter/tensorflow-notebook         Jupyter Notebook Scientific Python Stack w/ …   86                                      
xblaster/tensorflow-jupyter         Dockerized Jupyter with tensorflow              50                                      [OK]
tensorflow/serving                  Official images for TensorFlow Serving (http…   22                                      
floydhub/tensorflow                 tensorflow                                      14                                      [OK]
bitnami/tensorflow-serving          Bitnami Docker Image for TensorFlow Serving     13                                      [OK]

PULL TensorFlow镜像

seven@seven:~$ sudo docker run -it tensorflow/tensorflow
Unable to find image 'tensorflow/tensorflow:latest' locally
latest: Pulling from tensorflow/tensorflow
3b37166ec614: Pull complete 
504facff238f: Pull complete 
ebbcacd28e10: Pull complete 
c7fb3351ecad: Pull complete 
2e3debadcbf7: Pull complete 
568ddecd541b: Pull complete 
cb5781b11958: Pull complete 
c6f383503f95: Pull complete 
71dcbb855dc9: Pull complete 
5cfd34784f24: Pull complete 
c26bfefb0572: Pull complete 
88302acc21c8: Pull complete 
4c2b848f6d49: Pull complete 
...
Status: Downloaded newer image for tensorflow/tensorflow:latest

创建TensorFlow容器–jupyter notebook

1	seven@seven:~$ docker run --name seven-tensorflow -it -p 8888:8888 -v ~/Docker-tensorflow:/demo tensorflow/tensorflow

–name：创建的容器名，即seven-tensorflow
-it：保留命令行运行
p 8888:8888：将本地的8888端口和http://localhost:8888/映射
-v ~/Docker-tensorflow:/demo:将本地的~/Docker-tensorflow挂载到容器内的/demo下
tensorflow/tensorflow ：默认是tensorflow/tensorflow:latest,指定使用的镜像

输入以上命令后，默认容器就被启动了，命令行显示

[I 13:16:07.331 NotebookApp] Writing notebook server cookie secret to /root/.local/share/jupyter/runtime/notebook_cookie_secret
[I 13:16:07.347 NotebookApp] Serving notebooks from local directory: /notebooks
[I 13:16:07.347 NotebookApp] The Jupyter Notebook is running at:
[I 13:16:07.347 NotebookApp] http://(7191961747da or 127.0.0.1):8888/?token=2734a3a619e376f876eb72ba562852fa79efd94a5f3f871a
[I 13:16:07.347 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[C 13:16:07.347 NotebookApp] 
    
    Copy/paste this URL into your browser when you connect for the first time,
    to login with a token:
        http://(7191961747da or 127.0.0.1):8888/?token=2734a3a619e376f876eb72ba562852fa79efd94a5f3f871a

拷贝带token的URL在浏览器打开

1	http://(7191961747da or 127.0.0.1):8888/?token=257ce32bf00cd16dee9019462f8753a3b06154618885d682

接下来，你就可以在jupyter notebook上运行你的TensorFlow代码啦

关闭容器

1 2	seven@seven:~$ sudo docker stop seven-tensorflow seven-tensorflow

再次打开容器

1	seven@seven:~$ sudo docker start seven-tensortflow

如果不喜欢用Jupyter Notebook，我们也可以创建基于命令行的容器

基于命令行的TensorFlow容器

1 2	seven@seven:~$ sudo docker run -it --name bash_tensorflow tensorflow/tensorflow /bin/bash root@cb158fab4040:/notebooks#

这样我们就创建了名为bash_tensorflow的容器，并已经进入到容器

验证TensorFlow环境

默认的是python2.x安装的TensorFlowCPU版

root@cb158fab4040:/notebooks# python
Python 2.7.12 (default, Dec  4 2017, 14:50:18) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow 
>>> import tensorflow as tf
>>> print tf.__version__
1.11.0

转载请注明：Seven的博客