SAI is a variable-komi fork of leela-zero, a Go program with no human provided knowledge.
The relevant papers are SAI7x7 and SAI9x9.
The server is currently running here, on resources of Chieti-Pescara University.
The 19x19 run started just recently and its play is still quite random, but we have pretty strong 9x9 networks such as S1 from the 9x9 paper. (But you will need to compile the program with modified settings, for it to be able to play on 9x9 goban.)
You need a PC with a GPU, i.e. a discrete graphics card made by NVIDIA or AMD, preferably not too old, and with the most recent drivers installed.
It is possible to run the program without a GPU, but performance will be much lower. If your CPU is not very recent (Haswell or newer, Ryzen or newer), performance will be outright bad, and it's probably of no use trying to join the distributed effort. But you can still play, especially if you are patient.
Head to the Github releases page at https://github.com/sai-dev/sai/releases, download the latest release and unzip. (There is also a cpu-only version if you need it.)
Then you can use the main program SAI. You have to open a Windows command prompt in the directory with the program to run it. It will need a network to work and networks can be found on the server. But you can immediately launch
sai.exe --help
to see the options.
If you'd rather play with a strong SAI 9x9 network, we provide a compiled 9x9 executable sai9x9.exe but you can also simply run sai9x9.bat and you will play agains the S1 network from the 9x9 paper, which is included inside the zip file.
To help the collective effort with games and matches, you have to create credentials here. This way we can track and contact people that, for any reason, are uploading wrong data. (This happened sometimes for leela-zero project.)
Choose a password of no importance and that you don't use elsewhere, since it will be stored unencrypted in your local machine.
Then simply double-click on sai.hta – authorization for this operation may be requested, if so please grant it. The possible options can be found calling from the command prompt
autogtp.exe --help
We suggest to use the option -g2 (or -g3, -g4 or larger) if your computer is powerful enough to play more than one game at the same time.
Please note that sai.hta is simply a wrapper for autogtp.exe and launches
autogtp.exe --url http://sai.unich.it/ --username <your_computername> --password <your_password> <your_options>
The client autogtp will connect to the server automatically and do its work in the background, uploading results after each game. You can just close the autogtp window to stop it.
Follow the instructions given on leela-zero github to compile the sai and autogtp binaries in the build subdirectory.
Then you can use the main program SAI. You have to open a shell and to the directory with the program to run it. It will need a network to work and networks can be found on the server. But you can immediately launch
sai --help
to see the options.
If you'd rather play with a strong SAI 9x9 network, you can compile a SAI 9x9 executable by editing src/config.cpp. A pretty strong 9x9 network is S1 from the 9x9 paper, downloadable from the link.
To help the collective effort with games and matches, you have to create credentials here. This way we can track and contact people that, for any reason, are uploading wrong data. (This happened sometimes for leela-zero project.)
Choose a password of no importance and that you don't use elsewhere, since it will be sent unencrypted.
Then, be sure that sai and autogtp executables are in the $PATH and simply launch
autogtp --url http://sai.unich.it/ --username <your_username> --password <your_password> -g <n>
where n is the number of games you want to play at the same time (depending on your hardware).
The client autogtp will connect to the server automatically and do its work in the background, uploading results after each game. You can stop it with Ctrl-C.
The code is released under the GPLv3 or later, except for ThreadPool.h, cl2.hpp, half.hpp and the eigen and clblast_level3 subdirs, which have specific licenses (compatible with GPLv3) mentioned in those files.
Additional permission under GNU GPL version 3 section 7
If you modify this Program, or any covered work, by linking or combining it with NVIDIA Corporation's libraries from the NVIDIA CUDA Toolkit and/or the NVIDIA CUDA Deep Neural Network library and/or the NVIDIA TensorRT inference library (or a modified version of those libraries), containing parts covered by the terms of the respective license agreement, the licensors of this Program grant you additional permission to convey the resulting work.