BACKpropagation PACKage - a backpack for PyTorch

branch	tests & examples	coverage
master
development

A backpack for PyTorch that extends the backward pass of feedforward networks to compute quantities beyond the gradient.

• Check out the cheatsheet for an overview of quantities.
• Check out the examples on how to use the code.

Installation

pip install backpack-for-pytorch

Alternative: Clone the repository and run setup.py

git clone https://github.com/f-dangel/backpack.git ~/backpack
cd ~/backpack
python setup.py install

How to cite

If you are using backpack for your research, consider citing the paper

@inproceedings{dangel2020backpack,
    title     = {Back{PACK}: Packing more into Backprop},
    author    = {Felix Dangel and Frederik Kunstner and Philipp Hennig},
    booktitle = {International Conference on Learning Representations},
    year      = {2020},
    url       = {https://openreview.net/forum?id=BJlrF24twB}
}