Thanks Brent. Yes it's great just how easily accessible ML is now. I haven't really look at PyTorch though (I need to).
Actually stuck on something relating to that, perhaps worth asking you:
Do you know how to make a DQRN agent work if you have multiple action spaces, like over 100 different actions? Or possibly two sets of different action types. At the moment I have built things that can decide on 6 actions at most and only making one decision at a time. If not no probs, thanks for the welcome.
