Introduction to the TensorFlow seq2seq library
We used the raw TensorFlow API for all our implementations in this book for better transparency of the actual functionality of the models and for a better learning experience. However, TensorFlow has various libraries that hide all the fine-grained details of the implementations. This allows users to implement sequence-to-sequence models like the Neural Machine Translation (NMT) model we saw in Chapter 10, Sequence-to-Sequence Learning – Neural Machine Translation with fewer lines of code and without worrying about more specific technical details about how they work. Knowledge about these libraries is important as they provide a much cleaner way of using these models in production code or researching beyond the existing methods. Therefore, we will go through a quick introduction of how to use the TensorFlow seq2seq library. This code is available as an exercise in the seq2seq_nmt.ipynb file.
