Sb3

[zarns]

This isn't intended to merge, as this PR is a total mess. @bcollazo, you mentioned in the catanatron_gym README.md that the repo works well with sb3_contrib and I've been trying to set up a MaskablePPO player without too much luck. Do you have any history on trying to integrate the sb3_contrib models into catanatron? I've been struggling to get this working. I can get the model to train (sort of) using, python catanatron_experimental/catanatron_experimental/train_model.py --episodes 10 --model my_model and I can get it to play against the random player using, catanatron-play --players=R,MP --num=100 --code=catanatron_experimental/catanatron_experimental/my_player.py

I'm guessing you experimented some with sb3_contrib but must've excluded the changes, do you have any insight/tips to help get sb3_contrib working properly?