Support for Discrete Action Spaces

[Maab-Ibrahim]

From the current implementation and policy modules, and correct me if I have missed something, it seems that FSRL is built primarily around continuous (Gaussian) action spaces. I’m working with a discrete environment (e.g., GridWorld), and I was wondering:

Is there currently any built-in support or recommended way to use FSRL with discrete action spaces?

Or would the intended approach be to fork and extend FSRL by implementing a discrete policy class (e.g., using categorical distributions) and adjusting relevant modules (e.g., ActorCritic)?

Thanks in advance for any guidance!