tensorflow-install

Docker + tensorflow

1. Base image tensorflow/tensorflow

   $ docker pull tensorflow/tensorflow
   

2. extension FROM tensorflow/tensorflow

   ==> Dockerfile
        FROM tensorflow/tensorflow:latest
        MAINTAINER suker <suker@nexell.co.kr>
        COPY run_tensorboard.sh /         # run_tensorboard.sh 의 내용 =>  tensorboard --logdir=/notebooks/board
   ==> build
        $ docker build -t suker/tensorflow .
   

3. 실행

   ==> run.sh
   #!/bin/bash
   docker run -d --name tenf -p 8888:8888 -p 6006:6006 -v ~/MachineLearning/shared:/notebooks -e PASSWORD=nexell suker/tensorflow
   sleep 3
   docker exec -d tenf /run_tensorboard.sh
   

8888 port는 jupyter notebook, 6006 port는 tensorboard 이다. 각각 locahost:8888 및 localhost:6006 에 접속하면 사용 가능하다.