OakEmu-RL

OakEmu-RL trains reinforcement learning agents to play ZX Spectrum games. It connects the OakEmu emulator to Gymnasium and provides training frameworks built with PyTorch.

An agent observes screenshots, chooses actions and learns from the rewards those actions produce. Game configuration files define the controls, rewards and episode boundaries.

• Use a Gymnasium environment with your own learning algorithm.
• Run the included training experiments with DQN or PPO.
• Reuse useful gameplay through trajectory training.
• Save checkpoints, inspect metrics and export models for inference.

Start with installation, or browse the generated Python API.