新围棋AI miniGo

Minigo是一套依照AlphaGo Zero所发表的论文所实做出的开源计算机围棋程序,也就是不使用人类棋谱与累积的围棋知识,仅实做围棋规则,使用单一类神经网络从自我对弈中学习。

百度网盘:
链接: https://pan.baidu.com/s/1lfriWi4B3TsTKziKZX8SVQ
提取码: tg1u

https://github.com/JYPark09/leela-zero/releases/tag/minigo-v17

https://github.com/tensorflow/minigo

https://go.ctolib.com/tensorflow-minigo.html

主要的项目贡献者Andrew Jackson是Google员工,虽然Google与DeepMind没有正式参与Minigo项目,但Andrew Jackson使用的是Google所提供的20%时间,并且得到Google赞助提供硬件资源进行运算,供Minigo团队确认程序正确性。目前为止Minigo跑谱1千1百多万,LeelaZero跑谱一千二百多万。因为TPU的加持,minigo跑谱速度远超LZ。也许不久将来,Minigo就会坐上开源狗的头把交椅。
目前放出的权重来看,minigo20B和LZ40B,ELF20B对战胜负基本55开。但是高配置高POminigo胜率高出上面两个权重。

https://userscloud.com/wd0tqdqkqvia

D:\>d:\leela-zero-0.16-win64\leelaz.exe -w 939-heron.gz
Using 2 thread(s).
RNG seed: 4878283789657530464
Leela Zero 0.16  Copyright (C) 2017-2018  Gian-Carlo Pascutto and contributors
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under certain conditions; see the COPYING file for details.

BLAS Core: Haswell
Detecting residual layers…v2…256 channels…19 blocks.
Initializing OpenCL (autodetecting precision).

Wavefront/Warp size: 32
Max workgroup size: 1024
Max workgroup dimensions: 1024 1024 64
Using OpenCL half precision (at least 5% faster than single).
Setting max tree size to 4077 MiB and cache size to 453 MiB.

Passes: 0            Black (X) Prisoners: 0
Black (X) to move    White (O) Prisoners: 0

   a b c d e f g h j k l m n o p q r s t
19 . . . . . . . . . . . . . . . . . . . 19
18 . . . . . . . . . . . . . . . . . . . 18

https://userscloud.com/cat842csy8es

990权重

1005权重
https://userscloud.com/w598ji58xolm

作者:

喜欢围棋和编程。

发表回复

您的电子邮箱地址不会被公开。 必填项已用 * 标注