“KataGo”的版本间的差异

来自Gea-Suan Lin's Wiki
跳到导航 跳到搜索
第37行: 第37行:
 
== 快速安裝 ==
 
== 快速安裝 ==
  
<synaxhighlight lang="bash">
+
<syntaxhighlight lang="bash">
 
sudo apt update; sudo apt install -y build-essential cmake git libboost-filesystem-dev libgoogle-perftools-dev libzip-dev nvidia-435 ocl-icd-opencl-dev zlib1g-dev; git clone https://github.com/lightvector/KataGo.git; cd KataGo/cpp; cmake . -DBUILD_MCTS=1 -DUSE_BACKEND=OPENCL -DUSE_TCMALLOC=1; make -j8; cd ~/KataGo/cpp; cp configs/gtp_example.cfg ~/Lizzie/; cd ~/Lizzie/; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b40c256x2-s1349368064-d524332537.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b30c320x2-s1287828224-d525929064.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170e-b20c256x2-s2430231552-d525879064.zip; echo *.zip | xargs -n1 unzip
 
sudo apt update; sudo apt install -y build-essential cmake git libboost-filesystem-dev libgoogle-perftools-dev libzip-dev nvidia-435 ocl-icd-opencl-dev zlib1g-dev; git clone https://github.com/lightvector/KataGo.git; cd KataGo/cpp; cmake . -DBUILD_MCTS=1 -DUSE_BACKEND=OPENCL -DUSE_TCMALLOC=1; make -j8; cd ~/KataGo/cpp; cp configs/gtp_example.cfg ~/Lizzie/; cd ~/Lizzie/; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b40c256x2-s1349368064-d524332537.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b30c320x2-s1287828224-d525929064.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170e-b20c256x2-s2430231552-d525879064.zip; echo *.zip | xargs -n1 unzip
 
</synaxhighlight>
 
</synaxhighlight>

2020年3月7日 (六) 20:45的版本

KataGo是一套电脑围棋软件。

基本环境

这边是用AWS上的Ubuntu 18.04,开g4dn.xlarge测试的。

编译

先安装需要的套件:

apt install -y build-essential cmake git libboost-filesystem-dev libgoogle-perftools-dev libzip-dev nvidia-435 ocl-icd-opencl-dev zlib1g-dev

再来编译:

git clone https://github.com/lightvector/KataGo.git; cd KataGo/cpp; cmake . -DBUILD_MCTS=1 -DUSE_BACKEND=OPENCL -DUSE_TCMALLOC=1; make -j8

设定

等下会使用Lizzie当作前端,所以这边会把KataGo的设定档丢到Lizzie的目录下:

cd KataGo/cpp; cp configs/gtp_example.cfg ~/Lizzie/

下载最新的训练资料

这边会下载目前最强的20 blocks、30 blocks与40 blocks版本到Lizzie的目录下并且解开:

cd ~/Lizzie/; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b40c256x2-s1349368064-d524332537.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b30c320x2-s1287828224-d525929064.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170e-b20c256x2-s2430231552-d525879064.zip; echo *.zip | xargs -n1 unzip

快速安装

<syntaxhighlight lang="bash"> sudo apt update; sudo apt install -y build-essential cmake git libboost-filesystem-dev libgoogle-perftools-dev libzip-dev nvidia-435 ocl-icd-opencl-dev zlib1g-dev; git clone https://github.com/lightvector/KataGo.git; cd KataGo/cpp; cmake . -DBUILD_MCTS=1 -DUSE_BACKEND=OPENCL -DUSE_TCMALLOC=1; make -j8; cd ~/KataGo/cpp; cp configs/gtp_example.cfg ~/Lizzie/; cd ~/Lizzie/; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b40c256x2-s1349368064-d524332537.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170-b30c320x2-s1287828224-d525929064.zip; wget https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/g170e-b20c256x2-s2430231552-d525879064.zip; echo *.zip | xargs -n1 unzip </synaxhighlight>

相关连结

外部链接