Run docker container

docker run -u $(id -u):$(id -g) -v $(realpath ~/):/tf/notebooks -it -p 8888:8888 --gpus all balthasarschachtner/radler:v{x.y}

use password authentication instead of tokens:

• write password to ~/.jupyter/jupyter_server_config.json by executing the command docker run -u $(id -u):$(id -g) -v $(realpath ~/):/tf/notebooks -it balthasarschachtner/radler:v{x.y} jupyter server password
• this only works if you mount ~/ into the container (as done by the run command above)
• otherwise copy .jupyter/jupyter_server_config.json into the notebook directory (needs to appear to the container in /tf/notebooks/.jupyter/jupyter_server_config.json

Execute arbitrary commands within the docker environment:

• docker run -u $(id -u):$(id -g) -v $(realpath ~/):/tf/notebooks -it balthasarschachtner/radler:v{x.y} /bin/bash

mount additional directories

• add another volume with another option, e.g. -v /data:/data

Development resources:

• https://www.tensorflow.org/install/docker
• https://github.com/tensorflow/tensorflow/blob/master/tensorflow/tools/dockerfiles/dockerfiles/gpu-jupyter.Dockerfile
• https://hub.docker.com/r/tensorflow/tensorflow

Test setup

• Download notebook here: https://www.tensorflow.org/tutorials/quickstart/beginner