Transformer architecture in Keras tensorflow 2.0

Introduction

This project is about using new Tensorflow 2.0 and Keras to construct your own neural machine translation with transformer architecture (Scientific paper here.
Tensorflow 2.0 is the brend new version of Tensorflow that is more user-friendly than the previous version. It also incorporate a Keras high Level API. In this project, there is two way of training your own neural machine translation:

1. Using tensorflow tf.data API
2. Using the old fashion way with lists of inputs/outputs

With tensorflow data API

The tensorflow data API allow you to process data (image, text etc.) properly with a simple API.
It also allow to use generator in order to save RAM (very usefull when you are dealing with a lot of data).
You can find the documentation here.

With a list of inputs

WIP...