A small library for conditional density estimation using masked autoregressive flows, with Python & C implementations and a small CLI.
The CLI/C implementation, is usable from source via clone repo & cmake . && make, otherwise download from releases on GitHub, where binaries for each platform are available.
The Python implementation depends on numpy, scipy and autograd only, making it easy to just drop the file cde.py into your project.
You can run the tests with Python
uv venv env; . env/bin/activate
uv pip install pytest scipy autograd scikit_learn tqdm
pytest
Please see README.ipynb for info on usage & examples.