More information regarding KataGo and its installation can be found here:
In this instructions/tutorial, I will be using the same steps as the channel:
Jim Plays Go - Be sure to check out his video if there are any issues with this tutorial.
1) I will be using Sabaki to run KataGo. To install Sabaki, go to: https://github.com/SabakiHQ/Sabaki/releases and install the latest version of Sabaki best fit for your computer. (Different installations depend on what type of computer you have; Windows, Mac, Linux).
2) After you've installed Sabaki, head to: https://github.com/lightvector/KataGo/releases to download its latest releases. We recommend using the opencl version because your computer may not have the Cuda library.
3) Extract or unzip the file that you've just downloaded into a specific location onto your PC. I've personally extracted the file onto my desktop screen so it looks like this:
4) You will need to then download a neural network for the AI system to run on. To find the current best Neural Network to install, go to https://d3dndmfyhecmj0.cloudfront.net/g170/neuralnets/index.html and download the latest .bin.gz on the website. I'm going to download this file directly into my KataGo folder. After doing so, you must rename the file to: default_model.bin
Do not add a .gz to the end of the name! It should look like this:
5) You will then need to access the command prompt into the directory of your folder. To first access the command prompt, just type command prompt into the search bar of your PC. Once you have your command prompt open, type: cd <location_of_folder> to access that specific directory. An easy way to find the directory name is to just double-click the search bar in the files navigation system containing your name and to copy and paste that direct folder name into the command prompt.
6) Next, you will want to run a benchmark in the directory, so type the command: .\katago benchmark and run the command. It will do the tuning for your computer and will then do the benchmarking for your file. 6 search threads should be the default number of search threads. If you want to alter the amount of search threads, go to your KataGo folder and go to default_gtp. Open the file as a notepad file and alter numSearchThreads.
7) Now to configure Sabaki to run KataGo. Launch Sabaki and on the top left corner, you should see 'Engines'. Click it and there will be a drop-down menu. Hit 'Show Engines Sidebar'. Then press the play button in the corner and click 'Manage Engines...'. That should open up a bar at the bottom with all the Ai engines you currently have.
8)Click the 'Add' button on the bottom left of the screen and it will give you a space to put the name of you engine. Input any name you'd like. You will then need to input the katago's full file location under the Path section. To find the full file name, go to the KataGo Folder. Right click the KataGo application and click 'properties'. Then go to 'security' and copy and paste the full file name. Put this under the Path section. The only argument it will need is: gtp and for initial commands, input: time_settings 0 5 1
It should look something like this:
10) Enjoy running KataGo! You can set it as an analyzer by going back to the 'Engines Sidebar' and by clicking the triangle button again. This time you should see KataGo(whatever you named your engine) as an option to select. Select the engine and it should take some time to configure. Once the line stops spinning, right click KataGo and set as analyzer. You can use this on all your game records!
11) You can all set KataGo as your opponent by starting a new game and enabling KataGo as your opponent like so:
Hopefully this was helpful! If there were any questions, please ask below!