This code library contains functions to use pyRDDLGym with the Vejde library for relational reinforcement learning. 

It wraps pyRDDLGym, and provides an child class of BaseModel that pulls the required fields from the pyRDDLGym simulator.
This lets you experiment and train Vejde agents with many of the problems in the library of RDDL problems.

How do I run it?

1. Install the package.

This project uses uv for dependency management. Run uv sync --extra cu128 for PyTorch compiled with CUDA 12.8 and uv sync --extra cpu for CPU only.

2. Run one of the included scripts

If you want a simple sanity check, run test/test_imitation_mdp.py which runs a round of supervised learning on a very simple environment using a set policy as the expert.

For a quick reinforcement learning setup, you can use the train_agent script:

uv run train --domain SysAdmin_MDP_ippc2011 --instances 1 2 3