cfnet 0.0.12

A counterfactual explanation library using Jax

Welcome to cfnet

Key Features

• fast: code runs significantly faster than existing CF explanation libraries.
• scalable: code can be accelerated over CPU, GPU, and TPU
• flexible: we provide flexible API for researchers to allow full customization.

TODO: - implement various methods of CF explanations

Install

cfnet is built on top of Jax. It also uses Pytorch to load data.

Running on CPU

If you only need to run cfnet on CPU, you can simply install via pip or clone the GitHub project.

Installation via PyPI:

pip install cfnet

Editable Install:

git clone https://github.com/BirkhoffG/cfnet.git
pip install -e cfnet

Running on GPU or TPU

If you wish to run cfnet on GPU or TPU, please first install this library via pip install cfnet.

Then, you should install the right GPU or TPU version of Jax by following steps in the install guidelines.

An Example of using cfnet

See How to use cfnet.