Sequence-to-sequence using Tensorflow 2

This project is partly derived from the official tutorial, an example which trains a sequence-to-sequence (seq2seq) model for Spanish to English translation.

Installation

Please follow the official guide to install Tensorflow. In addition, you need to install scikit-learn.

Getting Started

Preparing a Dataset

prepare_dataset.py downloads a parallel corpus with some preprocessing. The resultant files will consume about 8.1MB.

$ python prepare_dataset.py <directory-to-save-dataset>

Training a Seq2seq Model

train.py starts training a seq2seq model on the dataset.

$ python train.py --dataset <directory-to-save-dataset> --device 0