LucasAlegre / morl-baselines

Multi-Objective Reinforcement Learning algorithms implementations.
https://lucasalegre.github.io/morl-baselines
MIT License
271 stars 44 forks source link

Implement a fix to #69 #70

Closed wilrop closed 11 months ago

wilrop commented 11 months ago

This pull request implements a fix for issue #69 . I also went further and implemented support for all discretizable action and observation spaces.

Note that it may still throw exceptions when the discretized space is too large. For example, in the four-rooms environment, numpy complains that it has to allocate way too much memory. This is not a "bug" but rather a limitation of PQL which requires you to establish Q sets for all state-action pairs.

wilrop commented 11 months ago

I just realised that I'm a git noob and didn't give the correct commit message. I thought if your commit is stopped by the pre-commit hook, your commit goes through but you have to make a new commit before you are allowed to push. I now learned that it simply cancels your commit. I'm not sure if this matters, but I originally wanted to call the commit "Extend PQL to allow for all discretizable action and observation spaces".