Log In
Or create an account ->
Imperial Library
Home
About
News
Upload
Forum
Help
Login/SignUp
Index
Title Page
Copyright and Credits
What's New in TensorFlow 2.0
Contributors
About the authors
About the reviewers
Packt is searching for authors like you
About Packt
Why subscribe?
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
Reviews
Section 1: TensorFlow 2.0 - Architecture and API Changes
Getting Started with TensorFlow 2.0
Technical requirements
What's new?
Changes from TF 1.x
TF 2.0 installation and setup
Installing and using pip
Using Docker
GPU installation
Installing using Docker
Installing using pip
Using TF 2.0
Rich extensions
Ragged Tensors
What are Ragged Tensors, really?
Constructing a Ragged Tensor
Basic operations on Ragged Tensors
New and important packages
Summary
Keras Default Integration and Eager Execution
Technical requirements
New abstractions in TF 2.0
Diving deep into the Keras API
What is Keras?
Building models
The Keras layers API
Simple model building using the Sequential API
Advanced model building using the functional API
Training models
Saving and loading models
Loading and saving architecture and weights separately
Loading and saving architectures
Loading and saving weights
Saving and loading entire models
Using Keras
Using the SavedModel API
Other features
The keras.applications module
The keras.datasets module
An end-to-end Sequential example
Estimators
Evaluating TensorFlow graphs
Lazy loading versus eager execution
Summary
Section 2: TensorFlow 2.0 - Data and Model Training Pipelines
Designing and Constructing Input Data Pipelines
Technical requirements
Designing and constructing the data pipeline
Raw data
Splitting data into train, validation, and test data
Creating TFRecords
TensorFlow protocol messages – tf.Example
tf.data dataset object creation
Creating dataset objects
Creating datasets using TFRecords
Creating datasets using in-memory objects and tensors
Creating datasets using other formats directly without using TFRecords
Transforming datasets
The map function
The flat_map function
The zip function
The concatenate function
The interleave function
The take(count) function
The filter(predicate) function
Shuffling and repeating the use of tf.data.Dataset
Batching
Prefetching
Validating your data pipeline output before feeding it to the model
Feeding the created dataset to the model
Examples of complete end-to-end data pipelines
Creating tfrecords using pickle files
Best practices and the performance optimization of a data pipeline in TF 2.0 
Built-in datasets in TF 2.0
Summary
Further reading
Model Training and Use of TensorBoard
Technical requirements
Comparing Keras and tf.keras
Comparing estimator and tf.keras
A quick review of machine learning taxonomy and TF support
Creating models using tf.keras 2.0
Sequential APIs
Functional APIs
Model subclassing APIs
Model compilation and training
The compile() API
The fit() API
Saving and restoring a model
Saving checkpoints as the training progresses
Manually saving and restoring weights
Saving and restoring an entire model
Custom training logic
Distributed training
TensorBoard
Hooking up TensorBoard with callbacks and invocation
Visualization of scalar, metrics, tensors, and image data
Graph dashboard
Hyperparameter tuning
What-If Tool
Profiling tool
Summary
Questions
Further reading
Section 3: TensorFlow 2.0 - Model Inference and Deployment and AIY
Model Inference Pipelines - Multi-platform Deployments
Technical requirements
Machine learning workflow – the inference phase
Understanding a model from an inference perspective
Model artifact – the SavedModel format
Understanding the core dataflow model
The tf.function API
The tf.autograph function
Exporting your own SavedModel model
Using the tf.function API
Analyzing SavedModel artifacts
The SavedModel command-line interface
Inference on backend servers
TensorFlow Serving
Setting up TensorFlow Serving
Setting up and running an inference server
When TensorFlow.js meets Node.js
Inference in the browser
Inference on mobile and IoT devices
Summary
AIY Projects and TensorFlow Lite
Introduction to TFLite
Getting started with TFLite
Running TFLite on mobile devices
TFLite on Android
TFLite on iOS
Running TFLite on low-power machines
Running TFLite on an Edge TPU processor
Running TF on the NVIDIA Jetson Nano
Comparing TFLite and TF
AIY
The Voice Kit
The Vision Kit
Summary
Section 4: TensorFlow 2.0 - Migration, Summary
Migrating From TensorFlow 1.x to 2.0
Major changes in TF 2.0
Recommended techniques to employ for idiomatic TF 2.0
Making code TF 2.0-native
Converting TF 1.x models
Upgrading training loops
Other things to note when converting
Frequently asked questions
The future of TF 2.0
More resources to look at
Summary
Other Books You May Enjoy
Leave a review - let other readers know what you think
← Prev
Back
Next →
← Prev
Back
Next →