torchtree-scipy

About torchtree-scypi

torchtree-scipy is a package that enhances the functionalities of torchtree by incorporating additional features from SciPy.

Dependencies

• SciPy
• torchtree

Installation

Installing from source

git clone https://github.com/4ment/torchtree-scipy
pip install torchtree-scipy/

Installing latest stable version

You can install torchtree-scipy and its dependencies from PyPI with:

pip install torchtree-scipy

Check install

If the installation was succesfull, this command should print the version of the torchtree_scipy library

python -c "import torchtree_scipy;print(torchtree_scipy.__version__)"

Command line arguments

The torchtree-scipy plugin adds these arguments to the torchtree CLI:

torchtree-cli advi --help
  ...
  --scipy_gamma_site    use the GammaSiteModel implemented with SciPy

Features

Discrete gamma site model

This model implements the discretized gamma distribution to model rate heterogeity accross sites. The gradient of this model with respect to the shape parameter is calculated numerically using finite differences. The easiest way to use this model is to use the argument --scipy_gamma_site. This sets the type of the site model to torchtree_scipy.GammaSiteModel in the JSON configuration file. The default step size is epsilon=1.0e-6 but it can be changed manually in the JSON file.

License

Distributed under the GPLv3 License. See LICENSE for more information.

Acknowledgements

torchtree-scipy makes use of the following libraries and tools, which are under their own respective licenses:

• PyTorch
• SciPy
• torchtree