This is a reimplementation of the Transformer architecture for NMT (Vaswani et al. 2017, https://arxiv.org/abs/1706.03762), reimplemented using the NamedTensor library (https://github.com/harvardnlp/namedtensor).
littell/NamedTransformer
An implementation of the Transformer network for NMT, implemented in NamedTensor
Jupyter NotebookMIT