hepengli
Results
3
comments of
hepengli
> Just to be clear, the problem you mention happens when `delta_enemy` is negative with a larger magnitude than `delta_deaths`, which would mean that agents are rewarded by injuring enemies,...
Thanks! I think this will do. I will check this again and get back to you soon.
Yes! I've also noticed this situation happens in MMM and MMM2. But your solution by changing the only positive reward to "reward = max(0, delta_enemy + delta_deaths)" is able to...