Apollo
Apollo
@gooooloo I see your player gametreee and see you subtract virtual loss after all simulations(800). Why do you not use parallel_search_num and is it ok?
@gooooloo Oh, I see you subtract virtual loss inside backup(). I am checking now why my model can't beat ntest and windows reversi though low policy and value loss. I'll...
Did you test with multithread search instead of coroutine?
https://github.com/Akababa/Chess-Zero/blob/nohistory/src/chess_zero/agent/player_chess.py I see @Akababa implement multi thread search tree, but I know python muti thread is not parallel task because GIL. OMG.
@Akababa I believe multithreading has same speed with asyncio in Python.
what is the best reversi game? I have not iPhone but have Mac. My model beats all of Android Reversi and Windows App Reversi.
My model (black) beats GRhino lv5 with open book variation "Low" and randomness 0 now. I make web player html, but I haven't any server to run tensorflow model.
I see "Online Reversi" on the Microsoft Store is very excellent. My model beats level 2 hardly. (2018/01/10) My model beats level 3 hardly now. (2018/01/11)
@gooooloo Um... Is really useful history? When use history, the player can not play on the one board state. I see some games as chess have must play from some...
@gooooloo No, I mean that some game can play from some board state that is not initial state, for example Chess Puzzles.