Book Image

Learn TensorFlow Enterprise

By : KC Tung
Book Image

Learn TensorFlow Enterprise

By: KC Tung

Overview of this book

TensorFlow as a machine learning (ML) library has matured into a production-ready ecosystem. This beginner’s book uses practical examples to enable you to build and deploy TensorFlow models using optimal settings that ensure long-term support without having to worry about library deprecation or being left behind when it comes to bug fixes or workarounds. The book begins by showing you how to refine your TensorFlow project and set it up for enterprise-level deployment. You’ll then learn how to choose a future-proof version of TensorFlow. As you advance, you’ll find out how to build and deploy models in a robust and stable environment by following recommended practices made available in TensorFlow Enterprise. This book also teaches you how to manage your services better and enhance the performance and reliability of your artificial intelligence (AI) applications. You’ll discover how to use various enterprise-ready services to accelerate your ML and AI workflows on Google Cloud Platform (GCP). Finally, you’ll scale your ML models and handle heavy workloads across CPUs, GPUs, and Cloud TPUs. By the end of this TensorFlow book, you’ll have learned the patterns needed for TensorFlow Enterprise model development, data pipelines, training, and deployment.

Related Content you might be interested in

Current Title:

Small book image
Learn TensorFlow Enterprise
KC Tung
Nov, 2020 | 314 pages
No titles found
Table of Contents (15 chapters)
Preface
Preface
Who this book is for
What this book covers
To get the most out of this book
Download the example code files
Download the color images
Conventions used
Get in touch
1
Section 1 – TensorFlow Enterprise Services and Features
Section 1 – TensorFlow Enterprise Services and Features
Free Chapter
2
Chapter 1: Overview of TensorFlow Enterprise
Chapter 1: Overview of TensorFlow Enterprise
Understanding TensorFlow Enterprise
Configuring cloud environments for TensorFlow Enterprise
Creating a data warehouse
Using TensorFlow Enterprise in AI Platform
Accessing the data sources
Summary
3
Chapter 2: Running TensorFlow Enterprise in Google AI Platform
Chapter 2: Running TensorFlow Enterprise in Google AI Platform
Setting up a notebook environment
Easy parameterized data extraction from BigQuery
Summary
4
Section 2 – Data Preprocessing and Modeling
Section 2 – Data Preprocessing and Modeling
5
Chapter 3: Data Preparation and Manipulation Techniques
Chapter 3: Data Preparation and Manipulation Techniques
Converting tabular data to a TensorFlow dataset
Converting distributed CSV files to a TensorFlow dataset
Handling image data for input pipelines
Decoding TFRecord and reconstructing the image
Handling image data at scale
Summary
6
Chapter 4: Reusable Models and Scalable Data Pipelines
Chapter 4: Reusable Models and Scalable Data Pipelines
Using TensorFlow Hub
Applying models from TensorFlow Hub
Leveraging the TensorFlow Keras API
Working with TensorFlow Estimators
Summary
7
Section 3 – Scaling and Tuning ML Works
Section 3 – Scaling and Tuning ML Works
8
Chapter 5: Training at Scale
Chapter 5: Training at Scale
Using the Cloud TPU through AI Platform
Using the Google Cloud GPU through AI Platform
Summary
9
Chapter 6: Hyperparameter Tuning
Chapter 6: Hyperparameter Tuning
Technical requirements
Delineating hyperparameter types
Understanding the syntax and use of Keras Tuner
Delineating hyperparameter search algorithms
Submitting tuning jobs in a local environment
Submitting tuning jobs in Google's AI Platform
Summary
10
Section 4 – Model Optimization and Deployment
Section 4 – Model Optimization and Deployment
11
Chapter 7: Model Optimization
Chapter 7: Model Optimization
Technical requirements
Understanding the quantization concept
Preparing a full original model for scoring
Converting a full model to a reduced float16 model
Converting a full model to a reduced hybrid quantization model
Converting a full model to an integer quantization model
Summary
12
Chapter 8: Best Practices for Model Training and Performance
Chapter 8: Best Practices for Model Training and Performance
Input handling for loading data
Regularization
Summary
13
Chapter 9: Serving a TensorFlow Model
Chapter 9: Serving a TensorFlow Model
Technical requirements
Running Local Serving
Understanding TensorFlow Serving with Docker
Downloading TensorFlow Serving Docker images
Summary
14
Other Books You May Enjoy
Other Books You May Enjoy
Leave a review - let other readers know what you think

Using the Google Cloud GPU through AI Platform

Having worked through the previous section for utilizing Cloud TPU with AI Platform, we are ready to do the same with the GPU. As it turns out, the formats of training script and invocation commands are very similar. With the exception of a few more parameters and slight differences in the distributed strategy definition, everything else remains the same.

There are several distributed strategies (https://www.tensorflow.org/guide/distributed_training#types_of_strategies) currently available. For a TensorFlow Enterprise distribution in Google AI Platform, MirroredStrategy and TPUStrategy are the only two that are fully supported. All the others are experimental. Therefore, in this section's example, we will use MirroredStrategy. This strategy creates copies of all the variables in the model on each GPU. As these variables are updated at each gradient decent step, the value updates are copied to each GPU synchronously. By default...

Previous Section
End of Section 3
Next Section