This project is the successor to Phillip. While the original Phillip used pure deep RL, this one starts with behavioral cloning on slippi replays, which makes it play a lot more like a human.
The bot is available to play via netplay on my twitch channel.
I am hesitant to release any trained agents as I don't want people using them on ranked/unranked, so at the moment the bot isn't available to play against locally.
My youtube channel has some recordings, mainly of the original phillip. There is also a video of Aklo playing the new bot.
The main entry points are scripts/train.py for imitation learning, scripts/eval_two.py for playing a trained agent, and slippi_ai/rl/run.py for self-play RL. Dataset creation is handled by slippi_db/parse_local.py.
- Huge thanks to Fizzi for writing the fast-forward gecko code that significantly speeds up RL training, for providing most of the imitation training data in the form of anonymized ranked collections (link in the Slippi discord), and of course for giving us Slippi in the first place. Even prior to rollback netcode, slippi replays were what rekindled my interest in melee AI, and are what gave name to this repo.
- Big thanks also to altf4 for creating the libmelee interface to slippi dolphin, making melee AI development accessible to everyone.
- Thank you to the many players who have generously shared their replays.