RLGym-Learn

A flexible framework for efficiently using RLGym v2 to train models.

Features

• Full support for all generics of the RLGym v2 API
• Full support for all functionality of RLGym v2 across multiple environments
• Fast parallelization of environments using Rust and shared memory
• Support for metrics gathering from environments
• Detailed checkpointing system
• File-based configuration
• Provided optimized PPO implementation
• Allows multiple learning algorithms to provide actions for agents within an environment
• Multi-platform (Windows, Linux)

Installation

1. install RLGym via pip install rlgym. If you're here for Rocket League, you can use pip install rlgym[rl-rlviser] instead to get the RLGym API as well as the Rocket League / Sim submodules and rlviser support.
2. If you would like to use a GPU install PyTorch with CUDA
3. Install this project via pip install rlgym-learn
4. Install rlgym-learn-algos via pip install rlgym-learn-algos
5. If pip installing fails at first, install Rust by following the instructions here

Usage

See the RLGym website for complete documentation and demonstration of functionality [COMING SOON]. For now, you can take a look at quick_start_guide.py and speed_test.py to get a sense of what's going on.

Credits

This project was built using Matthew Allen's wonderful RLGym-PPO as a starting point. Although this project has grown to share almost no code with its predecessor, I couldn't have done this without his support in talking through the design of abstractions and without RLGym-PPO to reference.

All of his files which remain similar have been refactored out to rlgym-learn-algos, although there is still util/KBHit.py contributed by Ian Cunnyngham which comes from RLGym-ppo.

Disclaimer

This framework is designed to be usable in every situation you might use the RLGym API in. However, there are a couple assumptions on the usage of RLGym which are baked into the functionality of this framework. These are pretty niche, but are listed below just in case:

1. The AgentID hash must fit into a signed 64 bit integer.
2. The obs space type and action space type should not change after the associated configuration objects' associated get_x_type functions have been called, and they should be the same across all agents and all envs.