LaDyS
LaDyS (Latent Dynamical Systems) provides PyTorch benchmark scaffolding for latent variable models of neural dynamics.
Documentation
CLI
Run a full LaDyS experiment with:
ladys run -d lorenz -m cassm
The public registries include synthetic datasets such as lorenz and
chaotic_rnn, plus model entries such as ilqr_vae, lfads, and ndt.
This builds the dataset, model, trainer, evaluation metrics, and a self-contained
run folder through the public ladys.Experiment orchestration API.
Model pages are generated from model class docstrings and config defaults:
python scripts/generate_model_docs.py
Images referenced from model docstrings should live under website/docs/assets/.
Use links like  in docstrings; generated
model pages rewrite those links to the correct relative path.
Serve the documentation site locally with:
mkdocs serve