e3nn/e3nn: 2021-06-21

Mario Geiger, Tess Smidt, M Alby, B. Miller, Wouter Boomsma, Bradley Dice, Kostiantyn Lapchevskyi, Maurice Weiler, Michał Tyszkiewicz, Simon Batzner, Jes Frellsen, Nuri Hyun Jung, Sophia Sanborn, Josh Rackers, Michael BaileyZenodo 2021 · DOI

Abstract

[0.3.3] - 2021-06-21 Changed <code>FullyConnectedNet</code> is now a <code>torch.nn.Sequential</code> Fixed <code>BatchNorm</code> was not equivariant for pseudo-scalars Added <code>biases</code> argument to <code>o3.Linear</code> <code>nn.models.v2106</code>: <code>MessagePassing</code> takes a sequence of irreps <code>nn.models.v2106</code>: <code>Convolution</code> inpired from Batch Normalization Biases Residual Blocks Towards the Identity Function in Deep Networks
BibTeX
@inproceedings{geiger2021e,
  author = {Mario Geiger and Tess Smidt and M Alby and B. Miller and Wouter Boomsma and Bradley Dice and Kostiantyn Lapchevskyi and Maurice Weiler and Michał Tyszkiewicz and Simon Batzner and Jes Frellsen and Nuri Hyun Jung and Sophia Sanborn and Josh Rackers and Michael Bailey},
  title = {{e3nn/e3nn: 2021-06-21}},
  booktitle = {Zenodo},
  year = {2021},
  doi = {10.5281/zenodo.5006322},
}

Loading related content...

Figure