RL-Image

[DeepCowProductions]

Describe the solution you'd like
An Image for specialized reinforcement Learning.
This would need things like OpenAI gym and most likely some OpenGL rendering or widgets to display a variety of RL environments.

[DeepCowProductions]

Looks like opengl aplications are possible.
Here are some ressources to go on:

https://gitlab.com/nvidia/container-images/samples/-/blob/master/opengl/ubuntu16.04/turbovnc-virtualgl/Dockerfile
https://hub.docker.com/r/nvidia/opengl
https://virtualgl.org/vgldoc/2_1_1/#hd002
https://turbovnc.org/About/TigerVNC
https://gursimarsm.medium.com/run-gui-applications-in-a-docker-container-ca625bad4638

This all relies on some form of vnc and server-side rendering, hence this is also a requirement for the hub:
https://github.com/jupyterhub/jupyter-remote-desktop-proxy

[DeepCowProductions]

Some desktop trials were successful and the opengl emulation works (although to a limited degree)
As resources are the limiting factor, anything opengl that requires 3d rendering most likely will be very laggy.