Getting started

Requirements

Use Python 3.13, 3.14 or 3.15 and the .NET 10 runtime. The package includes its OakEmu assemblies and depends on the OakEmu Python wrappers.

Training uses PyTorch. A supported NVIDIA GPU is useful for larger experiments; environments and tests also run on a CPU. The included experiment configurations select devices explicitly, so check them before starting a run.

Install the package

Add the package to a uv project:

uv add mrkwatkins-oakemu-rl

Or use pip:

python -m pip install mrkwatkins-oakemu-rl

Work from the repository

The repository includes game experiments, sample configuration files and tests. Clone it and install its locked dependencies:

git clone https://github.com/MrKWatkins/OakEmu-RL.git
cd OakEmu-RL
uv sync --locked

The development environment selects a CUDA build of PyTorch. The published package leaves PyTorch selection to the installer. See the PyTorch installation instructions for a build that suits your machine.

Game data and snapshots referenced by configuration files must exist before you run an environment. Check the paths in the selected game's YAML files.

Continue with game environments for a small example, or training experiments to use the included framework.