Machine Learning Engineering with MLflow E-Book

Natu Lauchande

• Herausgeber: Packt Publishing
• Kategorie: Fachliteratur
• Sprache: Englisch

Seitenzahl: 200

Veröffentlichungsjahr: 2021

Beschreibung

MLflow is a platform for the machine learning life cycle that enables structured development and iteration of machine learning models and a seamless transition into scalable production environments.
This book will take you through the different features of MLflow and how you can implement them in your ML project. You will begin by framing an ML problem and then transform your solution with MLflow, adding a workbench environment, training infrastructure, data management, model management, experimentation, and state-of-the-art ML deployment techniques on the cloud and premises. The book also explores techniques to scale up your workflow as well as performance monitoring techniques. As you progress, you’ll discover how to create an operational dashboard to manage machine learning systems. Later, you will learn how you can use MLflow in the AutoML, anomaly detection, and deep learning context with the help of use cases. In addition to this, you will understand how to use machine learning platforms for local development as well as for cloud and managed environments. This book will also show you how to use MLflow in non-Python-based languages such as R and Java, along with covering approaches to extend MLflow with Plugins.
By the end of this machine learning book, you will be able to produce and deploy reliable machine learning algorithms using MLflow in multiple environments.

Leseprobe

Machine Learning Engineering with MLflow

Manage the end-to-end machine learning life cycle with MLflow

Natu Lauchande

BIRMINGHAM—MUMBAI

Machine Learning Engineering with MLflow

Copyright © 2021 Packt Publishing

All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, without the prior written permission of the publisher, except in the case of brief quotations embedded in critical articles or reviews.

Every effort has been made in the preparation of this book to ensure the accuracy of the information presented. However, the information contained in this book is sold without warranty, either express or implied. Neither the author, nor Packt Publishing or its dealers and distributors, will be held liable for any damages caused or alleged to have been caused directly or indirectly by this book.

Packt Publishing has endeavored to provide trademark information about all of the companies and products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot guarantee the accuracy of this information.

Publishing Product Manager: Reshma Raman

Senior Editor: David Sugarman

Content Development Editor: Sean Lobo

Technical Editor: Manikandan Kurup

Copy Editor: Safis Editing

Project Coordinator: Aparna Ravikumar Nair

Proofreader: Safis Editing

Indexer: Pratik Shirodkar

Production Designer: Sinhayna Bais

First published: August 2021

Production reference: 1220721

Published by Packt Publishing Ltd.

Livery Place

35 Livery Street

Birmingham

B3 2PB, UK.

ISBN 978-1-80056-079-6

www.packt.com

Contributors

About the author

Natu Lauchande is a principal data engineer in the fintech space currently tackling problems at the intersection of machine learning, data engineering, and distributed systems. He has worked in diverse industries, including biomedical/pharma research, cloud, fintech, and e-commerce/mobile. Along the way, he had the opportunity to be granted a patent (as co-inventor) in distributed systems, publish in a top academic journal, and contribute to open source software. He has also been very active as a speaker at machine learning/tech conferences and meetups.

About the reviewer

Hitesh Hinduja is an ardent AI enthusiast working as a Senior Manager in AI at Ola Electric, where he leads a team of 20+ people in the areas of machine learning, deep learning, statistics, computer vision, natural language processing, and reinforcement learning. He has filed 14+ patents in India and the US and has numerous research publications under his name. Hitesh has been associated in research roles at India's top B-schools: Indian School of Business, Hyderabad, and the Indian Institute of Management, Ahmedabad. He is also actively involved in training and mentoring and has been invited as a guest speaker by various corporates and associations across the globe.

Table of Contents

Preface

Section 1: Problem Framing and Introductions

Chapter 1: Introducing MLflow

Technical requirements

What is MLflow?

Getting started with MLflow

Developing your first model with MLflow

Exploring MLflow modules

Exploring MLflow projects

Exploring MLflow tracking

Exploring MLflow Models

Exploring MLflow Model Registry

Summary

Further reading

Chapter 2: Your Machine Learning Project

Technical requirements 

Exploring the machine learning process

Framing the machine learning problem

Problem statement

Success and failure definition

Model output

Output usage

Heuristics

Data layer definition

Introducing the stock market prediction problem

Stock movement predictor

Problem statement

Success and failure definition

Model output

Output usage

Heuristics

Data layer definition

Sentiment analysis of market influencers

Problem statement

Success and failure definition

Model output

Output usage

Heuristics

Data layer definition

Developing your machine learning baseline pipeline

Summary

Further reading

Section 2: Model Development and Experimentation

Chapter 3: Your Data Science Workbench

Technical requirements 

Understanding the value of a data science workbench

Creating your own data science workbench

Building our workbench

Using the workbench for stock prediction

Starting up your environment

Updating with your own algorithms

Summary

Further reading

Chapter 4: Experiment Management in MLflow

Technical requirements

Getting started with the experiments module

Defining the experiment

Exploring the dataset

Adding experiments

Steps for setting up a logistic-based classifier

Comparing different models

Tuning your model with hyperparameter optimization

Summary

Further reading

Chapter 5: Managing Models with MLflow

Technical requirements

Understanding models in MLflow

Exploring model flavors in MLflow

Custom models

Managing model signatures and schemas

Introducing Model Registry

Adding your best model to Model Registry

Managing the model development life cycle

Summary

Further reading

Section 3: Machine Learning in Production

Chapter 6: Introducing ML Systems Architecture

Technical requirements

Understanding challenges with ML systems and projects

Surveying state-of-the-art ML platforms

Getting to know Michelangelo

Getting to know Kubeflow

Architecting the PsyStock ML platform

Describing the features of the ML platform

High-level systems architecture

MLflow and other ecosystem tools

Summary

Further reading

Chapter 7: Data and Feature Management

Technical requirements

Structuring your data pipeline project

Acquiring stock data

Checking data quality

Generating a feature set and training data

Running your end-to-end pipeline

Using a feature store

Summary

Further reading

Chapter 8: Training Models with MLflow

Technical requirements

Creating your training project with MLflow

Implementing the training job

Evaluating the model

Deploying the model in the Model Registry

Creating a Docker image for your training job

Summary

Further reading

Chapter 9: Deployment and Inference with MLflow

Technical requirements

Starting up a local model registry

Setting up a batch inference job

Creating an API process for inference

Deploying your models for batch scoring in Kubernetes

Making a cloud deployment with AWS SageMaker

Summary

Further reading

Section 4: Advanced Topics

Chapter 10: Scaling Up Your Machine Learning Workflow

Technical requirements

Developing models with a Databricks Community Edition environment

Integrating MLflow with Apache Spark

Integrating MLflow with NVIDIA RAPIDS (GPU)

Integrating MLflow with the Ray platform

Summary

Further reading

Chapter 11: Performance Monitoring

Technical requirements

Overview of performance monitoring for machine learning models

Monitoring data drift and model performance

Monitoring data drift

Monitoring target drift

Monitoring model drift

Infrastructure monitoring and alerting

Summary

Further reading

Chapter 12: Advanced Topics with MLflow

Technical requirements

Exploring MLflow use cases with AutoML

AutoML pyStock classification use case

AutoML – anomaly detection in fraud

Integrating MLflow with other languages

MLflow Java example

MLflow R example

Understanding MLflow plugins

Summary

Further reading

Other Books You May Enjoy

Section 1: Problem Framing and Introductions

This section will introduce you to a framework for stating machine learning problems in a concise and clear manner using MLflow.

The following chapters are covered in this section:

Chapter 1, Introducing MLflowChapter 2, Your Machine Learning Project