Skip to content

Gym environment for training agents to play Ratchet & Clank

Notifications You must be signed in to change notification settings

bordplate/Helga

Repository files navigation

Helga

Helga is a reinforcement learning gym environment for the original Ratchet & Clank HD trilogy on PS3.

Supported games

  • Ratchet & Clank 1 (PAL) - NPEA00385
  • Ratchet & Clank 3: Up Your Arsenal (PAL) - NPEA00387

Demo

Ratchet-RL-Eudora.mp4

Requirements

  • Windows
  • RPCS3
  • A legal copy of the PAL version of the game already installed in RPCS3
  • At least Python 3.8
  • Redis installed and running

How to start training your agent

Install the Python requirements:

pip install -r requirements.txt

Redis

Helga uses Redis to communicate between workers and the main learner process. The way the code is structured, Redis is a hard requirement right now.

You can have Redis installed locally on your machine, or you can use a different machine on the network for it. It's probably best not to use a cloud service for this, as the learner and worker communicate a lot, and the transfers might be too slow.

Redis on Windows

Official Redis documentation

Redis' official documentation on how to install Redis on Windows involves using WSL or Docker.

Unofficial Redis Windows port

If you don't want to use WSL or Docker, you can use an unofficial Windows port of Redis. This is the one that I use, and it has been working fine.

WandB

You can use Weights and Biases to track your training progress. Just follow their quickstart guide here until you're logged in.

Add --wandb to the command line arguments to enable Weights and Biases tracking.

Contributors

@davubnub on Discord for the Ratchet faces

About

Gym environment for training agents to play Ratchet & Clank

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published