Home

SpikingRL Lab

SpikingRL Lab is an experimental and extensible framework for reinforcement learning with configurable artificial and spiking policy networks. Algorithms, environment backends, policies, and network nodes are selected through typed registries and Hydra configuration rather than being fixed in the experiment runner. The project combines Gymnasium environments, skrl training loops, PyTorch and Norse neurons, Hydra configuration, and MLflow experiment tracking.

The initial set of implementations provides:

• a REINFORCE agent for discrete and continuous action spaces
• feed-forward and stateful node-based policy networks
• standard PyTorch layers and Norse LIF/LI nodes
• parallel Gymnasium environments through skrl
• reproducible Hydra configurations, checkpoints, and MLflow metadata.

Warning

The project is under active development.

Where to start

Page	Use it to
Getting started	Install the project, run training, evaluate a checkpoint, and inspect results
Project architecture	Understand execution, packages, registries, and state
Configuration and experiments	Compose configs and retain experiments
Algorithms	Open per-algorithm notes with equations and spiking-network details
Development	Add components, follow project conventions, and run quality checks

For the shortest installation and launch instructions, see the repository README.