tf_docker

Docker container for tutorials on tensorflow.

Based on https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/docker

Building the container

This repository is linked to dockerhub, however sometimes during debugging it might be faster to localy build the container. You can do this via

docker build https://raw.githubusercontent.com/oduerr/tf_docker/cpu_r/Dockerfile

You can replace cpu_r with the branch you want to build e.g. master

To run a single instance locally do:

docker run -p 8080:8888 -p 8081:6006 -it oduerr/tf_docker

and open the browser pointing at http://localhost:8080/.

To run many instances

with different ports on a server (e.g. on the Amazon Web Server). Start multiple containers mapping to port 8080.

docker run  -p 8081:8888 oduerr/tf_docker&
docker run  -p 8082:8888 oduerr/tf_docker&

Or better use a loop

for i in `seq 8800 8850`; do docker run -d -p $i:8888 oduerr/tf_docker;  done;

Further information

More information on the docker container, can be found at https://tensorchiefs.github.io/dl_course_2018/docker.html

Notes using AWS

I used the provided Amazon AMI (Amazon Linux AMI 2016.03.3 (HVM), SSD Volume Type - ami-31490d51). However it does not come with docker installed. I installed it as descripbed in http://docs.aws.amazon.com/AmazonECS/latest/developerguide/docker-basics.html#install_docker

tmpnb

Alternative integrate the images in https://github.com/jupyter/tmpnb. Code not working prop yet

docker pull jupyter/minimal-notebook
export TOKEN=$( head -c 30 /dev/urandom | xxd -p )
docker run --net=host -d -e CONFIGPROXY_AUTH_TOKEN=$TOKEN --name=proxy jupyter/configurable-http-proxy --default-target http://127.0.0.1:8080
docker run --net=host -d -e CONFIGPROXY_AUTH_TOKEN=$TOKEN \
           -v /var/run/docker.sock:/docker.sock \
           jupyter/tmpnb python orchestrate.py --image='oduerr/tf_docker'