Please note that currently the installation and configuration is somewhat involved. This is expected to be simplified in the future.
- Windows 10 (a Linux version is under development)
- Processor based on Intel/AMD x86 architecture
- Processor support for AVX instruction set
- At least one NVIDIA GPU with CUDA drivers installed
- Minimum 8GB of memory (recommended)
Please follow these preliminary steps (as needed) for installing the Ceres development environment:
- Install Microsoft .NET 5.0 runtime or SDK from https://dotnet.microsoft.com/download/dotnet/5.0
- Download version v0.26.3 for CUDA backend of Leela Chess Zero (file lc0-v0.26.3-windows-gpu-nvidia-cuda.zip from https://github.com/LeelaChessZero/lc0/releases/tag/v0.26.3)
- Build or download a backend plug-in customized for Ceres (LC0.DLL) and place this file in the same directory as the location where LC0 was installed above (see below for more details)
- Use git to download the Ceres code from github at https://github.com/dje-dev/ceres
- Download and install Microsoft Visual Studio (free Community Edition) from https://visualstudio.microsoft.com/downloads/
If a pre-built LC0.DLL is not provided or available for the version of CUDA installed on your computer it will be necessary to build from source.
Now you can launch Visual Studio and open the solution file Ceres.sln under the src directory. For best performance, make sure to choose "Release" from the top toolbar (not "Debug"). You can then run Ceres by "Run without debugging" command (Ctrl-F5.).
The final required step is to make sure a required configuration file with the name Ceres.json is present in the working directory from which Ceres is launched (typically under a subdirectory with name of the form artifacts/release/net5.0). This file contains various configuration entries (some required, some optional).
If Ceres launched run without a Ceres.json file in the working directory, the user will be prompted to enter the values of 4 required configuration settings. These will then be used to initialize a functional Ceres.json file, which can then be customzied at will using a text editor (or the SETOPT command in Ceres).
The following is a complete example of this setup process, followed by a simple UCI search command using the newly configured settings.
