Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/lichess-org/kaladin

Machine learning tool aimed at automating cheat detection using insights data.
https://github.com/lichess-org/kaladin

Last synced: about 2 months ago
JSON representation

Machine learning tool aimed at automating cheat detection using insights data.

Awesome Lists containing this project

README

        

# Kaladin
This machine learning tool is aimed at automating cheat detection on Lichess using insights ([example](https://lichess.org/insights/thibault)).

It is built using CNNs on Keras/TensorFlow.

## Setup
You will need:
- Linux OS (tested on Ubuntu 20.04 LTS)
- Docker
- MongoDB

## Docker container setup for Tensorflow with CPU or GPU

### Pre-requisites
Install Docker using your favorite package manager, or for example you can [follow this guide](https://www.tensorflow.org/install/docker).

### Create custom image and container

Run `$./docker.sh gpu|cpu [dev|prod]` with the needed target, it will create/update the image and start the container. `dev` (default) will open `bash`, while `prod` will directly launch the queue manager: `python3 queue_manager.py`

### Useful commands

To restart the container: `docker restart kaladin`

To view the logs: `docker logs -f kaladin`

## Configuration

For the list of options and default values used by Kaladin, see `src/.env.base`. You can override these either by setting environmental variables or create a `src/.env` file.

## Acknowledgments

The Kaladin repository was re-created when transitioning to open source to ensure that user data was not made public. Git history was expunged during that transition. A record of the commits prior to the transition can be found [here](doc/lost_commits.txt):
Special thanks to:
* [kraktus](https://github.com/kraktus) for your work on the queue manager, Docker config, error handling, lila integration, and integration testing.
* [michael1241](https://github.com/michael1241) for your domain expertise, design discussions, initial queue manager and mongo and deployment support.
* [ornicar](https://github.com/ornicar) for your support, your mongo wizardry, and your lila integration work.
* the others around the globe who helped by validating the model output, generating ideas, and providing valuable feedback.