beobench
beobench copied to clipboard
A toolkit providing easy and unified access to building control environments for reinforcement learning (RL).
This issue describes the roadmap towards splitting up the agent and environment side of Beobench experiment containers. This improvement was first suggested in #86 by @david-woelfle, big thanks! The goal...
### Problem When defaulting to buildx there can be issues with accessing the different build stages. Thus, in the current dev version, only builds via a container registry (e.g. GHCR,...
### Problem Currently the agent and the environment are executed in the same container. This makes installing dependencies for new research algorithms (for agents) potentially complicated, especially as some of...
### Problem Currently the agent script data saving is not very well described in the docs ### Potential Solution Add more information on this. See #88.
### Problem Build times of containers take aaaaaaaaages. ### Potential Solution Consider uploading build containers to dockerhub. That would furthermore also make results computed which beobench easier to reproduce as...
### Problem If the Beobench repo is mounted inside a docker-inside-docker/docker in docker/dind container, and then a `beobench run` command is run inside this dind container, the agent script is...
### Problem The documentation can always be improved. ### Solution Add the following parts and improvements to the docs: - Env list - #38 (priority) - #42 - #45 -...
### Problem Beobench is able to use a lot of disk space. We should add a section in the docs that elaborates how to view the docker related disk usage...
### Problem Wrapper usage not described in docs. ### Potential Solution Add example of wrapper usage to advanced usage guides.