A lightweight docker image with ludwig and nvidia gpu support for further customization using poetry. The compressed size is about 6GB compared to the 9GB provided by ludwigai, see here.
The motivation for this docker image lies in the Os4ML project in order to save cluster resources. Internal developers started to use this image as starting point for customized jupyterhub images providing a pre-installed individual set of python package.
KNOWN ISSUE: This is a pytorch-optimized version regarding the gpu supprt. Tensorflow won't work out of the box. This is because Tensorflow depends on the system cuda libraries - not the python packages - which are not installed.