If you would like to contribute to the LCZero project in any way, it would be appreciated!
Here is a brief list of ideas what you can do:
Contributing without coding
If you are not a developer, there is still plenty of ways to contribute:
- Contribute training games. Self-play training game generation is vital part of the LCZero project.
- Help on #help channel in our Discord.
- Contributing to improving this site.
- Maintaining benchmark tool and submitting your benchmark data.
- Testing and tuning of the engine (join #test-discuss in Discord).
- Run Lc0 Lichess bots.
- Maintain or moderate one of our social surfaces (forum, twitch, blog).
- Do Lc0 games streaming or youtube videos with analysis.
- Create and maintain Lc0 packages in various Linux distros.
- Be a point of contact with competition administration or coordinate with GUI authors.
Helping with development
We maintain several codebases and it’s not only the engine itself. It’s possible to find codebases for a wide variety of tastes:
- Lc0, the engine itself. Written in C++.
- Example improvement: Revive tensorflow backend and look into support of TPUs.
- Another idea: Investigate what is needed for Lc0 to be IDeA compatible.
- Training code. Written in Python.
- Training client. Written in Go.
- For example, make screensaver version of the client, to make it use GPU only when computer is not in use and display generated games as they are being played.
- Training server. Written in Go.
- Example improvement: support competitive contribution (API, teams, more often stats update)
- Our Discord bot Little Teapot. Written in Python, being rewritten in Go.
New project ideas
For people who prefer to code from scratch, there is also plenty of ideas to try. Some ideas for inspiration:
- Network evaluation framework as a centralized server (e.g. position-tester-as-a-service).
- Visualization of networks (structure and activation).
- Come up with more stats, and run them.
- Web site for gathering stats for people who wants to provide more stats.
- Tools to be able to run chess engine remotely (e.g. on colab while doing local analysis)
- GUI for engines running remotely (e.g. like TUI we used for WCCC but through web).
- Better monitoring for server infrastructure.
- Automate backup management.
- Improve detection of malicious training games.
- Maintain a webassembly version of Lc0 (Lichess wants to have that!)
- Maintain a version of Lc0 for mobile platforms.
- Maintain Linux / MacOS compilation.
- Write a “hub” website for competitions (so that people won’t have to stream their competitions through twitch). Basically, do a distributed CCCC.