I think it just comes down to the fact that discrete action spaces are easier to work with for most RL frameworks (not sure what exactly open AI is using here t

I think it just comes down to the fact that discrete action spaces are easier to work with for most RL frameworks (not sure what exactly open AI is using here though).