datamllab / rlcard

Reinforcement Learning / AI Bots in Card (Poker) Games - Blackjack, Leduc, Texas, DouDizhu, Mahjong, UNO.
http://www.rlcard.org
MIT License
2.85k stars 614 forks source link

Action Space setting of No-Limit Texas Holdem is Unreasonable #59

Open lujiayou123 opened 4 years ago

lujiayou123 commented 4 years ago

When I set self.init_chips=2000 (1000bbs) in game.py, the code runs extremly slow. I found out it's because of the action space setting. The num of leagl actions is too large(from 2 to 2000).

hsywhu commented 4 years ago

@lujiayou123 Hi, thanks for reporting the issue. The No-Limit Texas Holdem game is implemented just following the original rule so the large action space is an inevitable problem. That's also the reason why we want to implement some simplified version of the games like Leduc Holdem (more specific introduction can be found in this issue

However, you can still set up your own rules to reduce the action space like limiting the chips you can put when raising.