The new version that allows for mixed numerical precision has been released (1.0.1) here and a working example/tutorial can be found full dimensional and non-full dimensional This package constains one of the fastest polytope samplers known to date. It only works with linear inequality restrictions, it does not accept V-representation.

A paper with the asymptotic complexity and tests for MHAR is in:

arXiv: https://arxiv.org/abs/2104.07097

LACSC2021 conference: https://www.youtube.com/watch?v=o2CxnI6onts

Original code for the paper: https://github.com/uumami/mhar_pytorch

A Jupyter notebook with examples of how to use it is in (works in Colab):
https://github.com/uumami/mhar/blob/beta/tutorial/turorial.ipynb

PyPi: https://pypi.org/project/mhar/

You need to install Pytorch, then you can use the following command to install the MHAR package:
pip install mhar

Contributors to the code: uumami (Mario Vazquez Corte)