0

I'm trying to have my RL model play a game, but I've encountered a peculiar problem.

I am kind of new to all this, so maybe it's stupid, but:

My environment and everything are set up nicely and when testing works like a charm. I can see the inputs the model makes and the data it receives.

The issue is that after model.learn() the script just freezes. Now I tried removing the model.train() line and in that case, it follows through and finishes, although of course untrained.

I don't see anyone talking about implementing the train() function, but when I look at it it seems empty and unimplemented. Is this normal? Do I have to build the train() function myself?

Thanks in advance, if you need any code let me know. But I think the problem lies in my understanding?

CHAHI Saad
  • 309
  • 4
  • 15
Temojikato
  • 1
  • 2
  • Does it freeze or is it training? Training takes time. Any way you can go to that function and print some outputs to check the training progress (if there is any at all)? – M.K Feb 04 '22 at 09:55
  • Well the function itself seems empty and unimplemented for some reason, but I guess I'll try. I've had it run for a while, but maybe I'll try it again and just leave it be for a way longer time. – Temojikato Feb 05 '22 at 10:20
  • 1
    Alright, the issue seems resolved and u did indeed point me in the right direction. Because of how I setup my script and environment I was accidentally starting multiple, leading to edging on my preformance, making the calculations extremely slow. It now finishes cleanly within appropriate time. – Temojikato Feb 07 '22 at 11:13
  • That sounds great! – M.K Feb 07 '22 at 11:41

1 Answers1

0

Alright, the issue seems resolved and u did indeed point me in the right direction. Because of how I setup my script and environment I was accidentally starting multiple, leading to edging on my preformance, making the calculations extremely slow. It now finishes cleanly within appropriate time.

Temojikato
  • 1
  • 2