How to compute actions with RLlib and Tune after training

I encountered the same issue. You can see my post on it here. Unfortunately, I don’t have a solution yet. Hopefully this will get attention soon - it seems like such a basic capability.