Book Image

What's New in TensorFlow 2.0

By : Ajay Baranwal, Alizishaan Khatri, Tanish Baranwal
Book Image

What's New in TensorFlow 2.0

By: Ajay Baranwal, Alizishaan Khatri, Tanish Baranwal

Overview of this book

TensorFlow is an end-to-end machine learning platform for experts as well as beginners, and its new version, TensorFlow 2.0 (TF 2.0), improves its simplicity and ease of use. This book will help you understand and utilize the latest TensorFlow features. What's New in TensorFlow 2.0 starts by focusing on advanced concepts such as the new TensorFlow Keras APIs, eager execution, and efficient distribution strategies that help you to run your machine learning models on multiple GPUs and TPUs. The book then takes you through the process of building data ingestion and training pipelines, and it provides recommendations and best practices for feeding data to models created using the new tf.keras API. You'll explore the process of building an inference pipeline using TF Serving and other multi-platform deployments before moving on to explore the newly released AIY, which is essentially do-it-yourself AI. This book delves into the core APIs to help you build unified convolutional and recurrent layers and use TensorBoard to visualize deep learning models using what-if analysis. By the end of the book, you'll have learned about compatibility between TF 2.0 and TF 1.x and be able to migrate to TF 2.0 smoothly.
Table of Contents (13 chapters)
Title Page

Best practices and the performance optimization of a data pipeline in TF 2.0 

Here is a summary of the best practices to follow while building an efficient input data pipeline in TF 2.0:

  • It's recommended to use a shuffling (shuffle) API before repeating the transformation.
  • Use the prefetch transformation to overlap the work of a producer (fetching the next batch of data) and consumer (using the current batch of data for training). Also, it's extremely important to note that the prefetch transformation should be added to the end of your input pipeline after shuffling (shuffle), repeating (repeat), and batching (batch) the data pipeline. This should look something like this:
# buffer_size could be either 1 or 2 which represents 1 or 2 batches of data
dataset = dataset.shuffle(count).repeat().batch(batch_size).prefetch(buffer_size)
  • It's strongly recommended to...