fedden / poker_ai

🤖 An Open Source Texas Hold'em AI

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Deepcopy very slows down with evaluator instance in PokerEngine

mamdf opened this issue · comments

commented

It slows down too much to apply action, I tried to train my own pseudo game but my old pc was burned out.

I look for the faulty and unless with a big abstraction it was the deep copy of the environment, however, the main reason is the unnecessary copy of evaluator instance into PokerEnigne, so as a naive fix I move out this to global module and improve the speed train X10, the same result testing the deepcopy with timeit

commented

@kimphilby , thanks for raising an issue!

We'll check it out and perhaps sneak it into our first release if viable.

I'm not sure what you are trying to run on your old pc, but we need 30+ gb of memory and 32+ cores to make the training times reasonable, FYI.