Cloud Native Machine Learning
Deploying PyTorch models on AWS
Carl Osipov
  • MEAP began May 2020
  • Publication in Spring 2021 (estimated)
  • ISBN 9781617297762
  • 250 pages (estimated)
  • printed in black & white

It's clear the author has street cred and has done quality work in the trenches.

Todd Cook
Deploying a machine learning model into a fully realized production system usually requires painstaking work by an operations team creating and managing custom servers. Cloud Native Machine Learning helps you bridge that gap by using the pre-built services provided by cloud platforms like Azure and AWS to assemble your ML system’s infrastructure. Following a real-world use case for calculating taxi fares, you’ll learn how to get a serverless ML pipeline up and running using AWS services. Clear and detailed tutorials show you how to develop reliable, flexible, and scalable machine learning systems without time-consuming management tasks or the costly overheads of physical hardware.

About the Technology

Your new machine learning model is ready to put into production, and suddenly all your time is taken up by setting up your server infrastructure. Serverless machine learning offers a productivity-boosting alternative. It eliminates the time-consuming operations tasks from your machine learning lifecycle, letting out-of-the-box cloud services take over launching, running, and managing your ML systems. With the serverless capabilities of major cloud vendors handling your infrastructure, you’re free to focus on tuning and improving your models.

About the book

Cloud Native Machine Learning is a guide to bringing your experimental machine learning code to production using serverless capabilities from major cloud providers. You’ll start with best practices for your datasets, learning to bring VACUUM data-quality principles to your projects, and ensure that your datasets can be reproducibly sampled. Next, you’ll learn to implement machine learning models with PyTorch, discovering how to scale up your models in the cloud and how to use PyTorch Lightning for distributed ML training. Finally, you’ll tune and engineer your serverless machine learning pipeline for scalability, elasticity, and ease of monitoring with the built-in notification tools of your cloud platform. When you’re done, you’ll have the tools to easily bridge the gap between ML models and a fully functioning production system.
Table of Contents detailed table of contents

Part 1: Mastering the dataset

1 Introduction to serverless machine learning

1.1 What is a machine learning platform?

1.2 Challenges when designing a machine learning platform

1.3 Public clouds for machine learning platforms

1.4 What is serverless machine learning?

1.5 Why serverless machine learning?

1.5.1 Serverless vs. IaaS and PaaS

1.5.2 Serverless machine learning lifecycle

1.6 Who is this book for?

1.6.1 Audience

1.6.2 What you can get out of this book

1.7 How does this book teach?

1.8 When is this book not for you?

1.9 Conclusions

1.10 Summary

2 Importing the dataset

2.1 Introducing the Washington, DC taxi rides dataset

2.1.1 What is the business use case?

2.1.2 What are the business rules?

2.1.3 What is the schema for the business service?

2.1.4 What are the options for implementing the business service?

2.1.5 What data assets are available for the business service?

2.1.6 Downloading and unzipping the dataset

2.2 Starting with object storage for the dataset

2.2.1 Understanding object storage vs. file systems

2.2.2 Authenticating with Amazon Web Services

2.2.3 Creating a serverless object storage bucket

2.3 Discovering the schema for the dataset

2.3.1 Introducing AWS Glue

2.3.2 Authorizing the crawler to access your objects

2.3.3 Using a crawler to discover the data schema

2.4 Migrating to columnar storage for more efficient analytics

2.4.1 Introducing column-oriented data formats for analytics

2.4.2 Migrating to a column-oriented data format

2.5 Summary

3 Exploring and preparing the dataset

3.1 Getting started with interactive querying

3.1.1 Choosing the right use case for interactive querying

3.1.2 Introducing AWS Athena

3.1.3 Preparing a sample dataset

3.1.4 Interactive querying using a sample dataset

3.1.5 Querying the DC taxi dataset

3.2 Getting started with data quality

3.2.1 From "garbage-in garbage-out" to data quality

3.2.2 Before starting with data quality

3.2.3 Normative principles for data quality

3.3 Applying VACUUM to DC taxi data

3.3.1 Enforcing the schema to ensure valid values

3.3.2 Cleaning up invalid fare amounts

3.3.3 Improving the accuracy

3.4 Implementing VACUUM in a PySpark job

3.5 Summary

4 More exploratory data analysis and data preparation

4.1 Interactive notebooks with AWS Sagemaker

4.1.1 Interactive Python programming with Jupyter notebooks

4.1.2 Using Sagemaker for data exploration with Jupyter notebooks

4.2 Getting started with data sampling

4.2.1 Exploring the summary statistics of the cleaned up dataset

4.2.2 Choosing the right sample size for the test dataset

4.2.3 Exploring the statistics of alternative sample sizes

4.2.4 Calculating p-values for a specific sample size

4.3 Summary

Part 2: PyTorch for Serverless Machine Learning

5 Introducing PyTorch: Tensor Basics

5.1 Getting started with tensors

5.2 Getting started with PyTorch tensor creation operations

5.3 Creating PyTorch tensors of pseudo-random and interval values

5.4 PyTorch tensor operations and broadcasting

5.5 PyTorch tensors vs. native Python lists

5.6 Summary

6 Core PyTorch: Autograd, Optimizers, and Utilities

7 Serverless machine learning at scale

Part 3: Serverless machine learning pipeline

8 Feature engineering

9 Hyperparameter tuning

10 Serverless machine learning pipeline


Appendix A: Machine learning foundations

Appendix B: Public cloud computing basics

Appendix C: Docker essentials

What's inside

  • Extracting, transforming, and loading datasets
  • Querying datasets with SQL
  • Understanding automatic differentiation in PyTorch
  • Deploying trained models and pipelines as a service endpoint
  • Monitoring and managing your pipeline’s life cycle
  • Measuring performance improvements

About the reader

For data professionals with intermediate Python skills and basic familiarity with machine learning. No cloud experience required.

About the author

Carl Osipov has spent over 15 years working on big data processing and machine learning in multi-core, distributed systems, such as service-oriented architecture and cloud computing platforms. While at IBM, Carl helped IBM Software Group to shape its strategy around the use of Docker and other container-based technologies for serverless computing using IBM Cloud and Amazon Web Services. At Google, Carl learned from the world’s foremost experts in machine learning and also helped manage the company’s efforts to democratize artificial intelligence. You can learn more about Carl from his blog Clouds With Carl.

placing your order...

Don't refresh or navigate away from the page.
Manning Early Access Program (MEAP) Read chapters as they are written, get the finished eBook as soon as it’s ready, and receive the pBook long before it's in bookstores.
print book $24.99 $49.99 pBook + eBook + liveBook
Additional shipping charges may apply
Cloud Native Machine Learning (print book) added to cart
continue shopping
go to cart

eBook $19.99 $39.99 3 formats + liveBook
Cloud Native Machine Learning (eBook) added to cart
continue shopping
go to cart

Prices displayed in rupees will be charged in USD when you check out.
customers also reading

This book 1-hop 2-hops 3-hops

FREE domestic shipping on three or more pBooks