Faster Text Generation with TensorFlow and XLA

This notebook is a companion to the 🤗 blog post with the same title. It is meant to illustrate how to use XLA with TensorFlow text generation.

It contains two stand-alone examples, one for encoder-decoder models and another for decoder-only models.

⚠️ If you are running this on colab, you might not have access to a GPU. The benefits of XLA are best observed with a GPU!