Book Image

Practical Deep Learning at Scale with MLflow

By : Yong Liu
5 (1)
Book Image

Practical Deep Learning at Scale with MLflow

5 (1)
By: Yong Liu

Overview of this book

The book starts with an overview of the deep learning (DL) life cycle and the emerging Machine Learning Ops (MLOps) field, providing a clear picture of the four pillars of deep learning: data, model, code, and explainability and the role of MLflow in these areas. From there onward, it guides you step by step in understanding the concept of MLflow experiments and usage patterns, using MLflow as a unified framework to track DL data, code and pipelines, models, parameters, and metrics at scale. You’ll also tackle running DL pipelines in a distributed execution environment with reproducibility and provenance tracking, and tuning DL models through hyperparameter optimization (HPO) with Ray Tune, Optuna, and HyperBand. As you progress, you’ll learn how to build a multi-step DL inference pipeline with preprocessing and postprocessing steps, deploy a DL inference pipeline for production using Ray Serve and AWS SageMaker, and finally create a DL explanation as a service (EaaS) using the popular Shapley Additive Explanations (SHAP) toolbox. By the end of this book, you’ll have built the foundation and gained the hands-on experience you need to develop a DL pipeline solution from initial offline experimentation to final deployment and production, all within a reproducible and open source framework.

Related Content you might be interested in

Current Title:

Small book image
Practical Deep Learning at Scale with MLflow
Yong Liu
Jul, 2022 | 288 pages
Small book image
Machine Learning Engineering with MLflow.
Natu Lauchande
Aug, 2021 | 248 pages
Small book image
Practical Machine Learning on Databricks
Debu Sinha
Nov, 2023 | 244 pages
Small book image
Distributed Data Systems with Azure Databricks
Alan Bernardo Palacio
May, 2021 | 414 pages
Table of Contents (17 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
Share Your Thoughts
1
Section 1 - Deep Learning Challenges and MLflow Prime
Section 1 - Deep Learning Challenges and MLflow Prime
Free Chapter
2
Chapter 1: Deep Learning Life Cycle and MLOps Challenges
Chapter 1: Deep Learning Life Cycle and MLOps Challenges
Technical requirements
Understanding the DL life cycle and MLOps challenges
Understanding DL data challenges
Understanding DL model challenges
Understanding DL code challenges
Understanding DL explainability challenges
Summary
Further reading
3
Chapter 2: Getting Started with MLflow for Deep Learning
Chapter 2: Getting Started with MLflow for Deep Learning
Technical requirements
Setting up MLflow
Implementing our first DL experiment with MLflow autologging
Exploring MLflow's components and usage patterns
Summary
Further reading
4
Section 2 –
Tracking a Deep Learning Pipeline at Scale
Section 2 –
Tracking a Deep Learning Pipeline at Scale
5
Chapter 3: Tracking Models, Parameters, and Metrics
Chapter 3: Tracking Models, Parameters, and Metrics
Technical requirements
Setting up a full-fledged local MLflow tracking server
Tracking model provenance
Tracking model metrics
Tracking model parameters
Summary
Further reading
6
Chapter 4: Tracking Code and Data Versioning
Chapter 4: Tracking Code and Data Versioning
Technical requirements
Tracking notebook and pipeline versioning
Tracking locally, privately built Python libraries
Tracking data versioning in Delta Lake
Summary
Further reading
7
Section 3 –
Running Deep Learning Pipelines at Scale
Section 3 –
Running Deep Learning Pipelines at Scale
8
Chapter 5: Running DL Pipelines in Different Environments
Chapter 5: Running DL Pipelines in Different Environments
Technical requirements
An overview of different execution scenarios and environments
Running locally with local code
Running remote code in GitHub locally
Running local code remotely in the cloud
Running remotely in the cloud with remote code in GitHub
Summary
Further reading
9
Chapter 6: Running Hyperparameter Tuning at Scale
Chapter 6: Running Hyperparameter Tuning at Scale
Technical requirements
Understanding automatic HPO for DL pipelines
Creating HPO-ready DL models with Ray Tune and MLflow
Running the first Ray Tune HPO experiment with MLflow
Running HPO with Ray Tune using Optuna and HyperBand
Summary
Further reading
10
Section 4 –
Deploying a Deep Learning Pipeline at Scale
Section 4 –
Deploying a Deep Learning Pipeline at Scale
11
Chapter 7: Multi-Step Deep Learning Inference Pipeline
Chapter 7: Multi-Step Deep Learning Inference Pipeline
Technical requirements
Understanding patterns of DL inference pipelines
Implementing a custom MLflow Python model
Implementing preprocessing and postprocessing steps in a DL inference pipeline
Implementing an inference pipeline as a new entry point in the main MLproject
Summary
Further reading
12
Chapter 8: Deploying a DL Inference Pipeline at Scale
Chapter 8: Deploying a DL Inference Pipeline at Scale
Technical requirements
Understanding different deployment tools and host environments
Deploying locally for batch and web service inference
Deploying using Ray Serve and MLflow deployment plugins
Deploying to AWS SageMaker – a complete end-to-end guide
Summary
Further reading
13
Section 5 – Deep Learning Model Explainability at Scale
Section 5 – Deep Learning Model Explainability at Scale
14
Chapter 9: Fundamentals of Deep Learning Explainability
Chapter 9: Fundamentals of Deep Learning Explainability
Technical requirements
Understanding the categories and audience of explainability
Exploring the SHAP Explainability toolbox
Exploring the Transformers Interpret toolbox
Summary
Further reading
15
Chapter 10: Implementing DL Explainability with MLflow
Chapter 10: Implementing DL Explainability with MLflow
Technical requirements
Understanding current MLflow explainability integration
Implementing a SHAP explanation using the MLflow artifact logging API
Implementing a SHAP explainer using the MLflow pyfunc API
Summary
Further reading
Why subscribe?
16
Other Books You May Enjoy
Other Books You May Enjoy
Packt is searching for authors like you
Share Your Thoughts

Chapter 1: Deep Learning Life Cycle and MLOps Challenges

The past few years have seen great success in Deep Learning (DL) for solving practical business, industrial, and scientific problems, particularly for tasks such as Natural Language Processing (NLP), image, video, speech recognition, and conversational understanding. While research in these areas has made giant leaps, bringing these DL models from offline experimentation to production and continuously improving the models to deliver sustainable values is still a challenge. For example, a recent article by VentureBeat (https://venturebeat.com/2019/07/19/why-do-87-of-data-science-projects-never-make-it-into-production/) found that 87% of data science projects never make it to production. While there might be business reasons for such a low production rate, a major contributing factor is the difficulty caused by the lack of experiment management and a mature model production and feedback platform.

This chapter will help us to understand the challenges and bridge these gaps by learning the concepts, steps, and components that are commonly used in the full life cycle of DL model development. Additionally, we will learn about the challenges of an emerging field known as Machine Learning Operations (MLOps), which aims to standardize and automate ML life cycle development, deployment, and operation. Having a solid understanding of these challenges will motivate us to learn the skills presented in the rest of this book using MLflow, an open source, ML full life cycle platform. The business values of adopting MLOps' best practices are numerous; they include faster time-to-market of model-derived product features, lower operating costs, agile A/B testing, and strategic decision making to ultimately improve customer experience. By the end of this chapter, we will have learned about the critical role that MLflow plays in the four pillars of MLOps (that is, data, model, code, and explainability), implemented our first working DL model, and grasped a clear picture of the challenges with data, models, code, and explainability in DL.

In this chapter, we're going to cover the following main topics:

  • Understanding the DL life cycle and MLOps challenges
  • Understanding DL data challenges
  • Understanding DL model challenges
  • Understanding DL code challenges
  • Understanding DL explainability challenges
Previous Section
End of Section 1
Next Section