A Dota 2 bot to be trained with deep reinforcement learning using RLlib and run in conjunction with the 5v5dota2ai-addon.
This project is built as part of a dissertation submitted for a Master's degree in Artificial Intelligence.
Clone the repo, install dependencies with poetry install, and run either
rampagebot/rl/main_dqn.py or rampagebot/rl/main_ppo.py depending on which
algorithm you want to train. Then launch Dota 2 Workshop Tools, and run the
dota2ai custom game.
Curious about this project, its background, or anything else? Feel free to submit an issue in this repo and I will try to make time to answer there.
This project is released to the public under the GNU GPL-3.0 license, in the hopes that it may be useful to as many people as possible.