Refer to this paper: https://arxiv.org/pdf/1904.09925.pdf
This is an (ongoing) implementation of the paper "Attention Augmented Convolutional Networks" by Bello et al, written using Pytorch. The code has been adapted from the Tensorflow code provided in the original paper.
- Layer Implementation
- Training models
- Tests and comparisons
The layer implementation can be found in the file named AAConv.ipynb.