sensitivity_torch is a package designed to allow taking first- and second-order derivatives through optimization or any other fixed-point process.

This package builds on top of PyTorch. We also maintain an implementation in JAX here.

Documentation can be found here.

Install using pip

$ pip install git+https://github.com/rdyro/sensitivity_torch.git

or from source

$ git clone git@github.com:rdyro/sensitivity_torch.git
$ cd sensitivity_torch
$ python3 setup.py install --user

Run all unit tests using

$ python3 setup.py test