tf.data is the recommended API for creating TensorFlow input pipelines and is relied upon by countless external and internal Google users. The API enables you to build complex input pipelines from simple, reusable pieces and makes it possible to handle large amounts of data, different data formats, and perform complex transformations. In this talk, I will present an overview of the project and highlight best practices for creating performant input pipelines.
Jiri is a tech lead of the tf.data project and a software engineer at Google. He holds a PhD from Carnegie Mellon University and throughout his career he has worked on distributed systems and performance, most recently TensorFlow.