Let's run latest releases of lc0, lczero-client and lichess-bot under NVIDIA docker.

Run Lc0 self-play training games

To download and run the prebuilt Docker image

docker run --runtime nvidia --rm -it vochicong/lc0-docker

otherwise, clone this repository then run

docker-compose up lc0

Kubernetes

See k8s.yaml for a sample configuration.

Version

The docker image is tagged by lc0 release version, see .github/main.workflow, k8s.yaml and docker-compose.yml.

Run Lc0 with lichess-bot

• Clone this repository, cd lc0-nvdia-docker
• Download and put syzygy tablebases into data/syzygy
• Download and put networks (weights) into data/networks
• Edit lc0.config and lcbot-config.yml to your preference
• Put your LICHESS_API_TOKEN in file .env (NEVER git commit this file)

Run

docker-compose up lcbot

Requirements

• Docker
• nvidia-docker
• docker-compose >= 1.23.2