Book Image

Production-Ready Applied Deep Learning

By : Tomasz Palczewski, Jaejun (Brandon) Lee, Lenin Mookiah
Book Image

Production-Ready Applied Deep Learning

By: Tomasz Palczewski, Jaejun (Brandon) Lee, Lenin Mookiah

Overview of this book

Machine learning engineers, deep learning specialists, and data engineers encounter various problems when moving deep learning models to a production environment. The main objective of this book is to close the gap between theory and applications by providing a thorough explanation of how to transform various models for deployment and efficiently distribute them with a full understanding of the alternatives. First, you will learn how to construct complex deep learning models in PyTorch and TensorFlow. Next, you will acquire the knowledge you need to transform your models from one framework to the other and learn how to tailor them for specific requirements that deployment environments introduce. The book also provides concrete implementations and associated methodologies that will help you apply the knowledge you gain right away. You will get hands-on experience with commonly used deep learning frameworks and popular cloud services designed for data analytics at scale. Additionally, you will get to grips with the authors’ collective knowledge of deploying hundreds of AI-based services at a large scale. By the end of this book, you will have understood how to convert a model developed for proof of concept into a production-ready application optimized for a particular production setting.

Related Content you might be interested in

Current Title:

Small book image
Production-Ready Applied Deep Learning
Tomasz Palczewski | Jaejun (Brandon) Lee | Lenin Mookiah
Aug, 2022 | 322 pages
No titles found
Table of Contents (19 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
Part 1 – Building a Minimum Viable Product
Part 1 – Building a Minimum Viable Product
Free Chapter
2
Chapter 1: Effective Planning of Deep Learning-Driven Projects
Chapter 1: Effective Planning of Deep Learning-Driven Projects
Technical requirements
What is DL?
Understanding the role of DL in our daily lives
Overview of DL projects
Planning a DL project
Summary
Further reading
3
Chapter 2: Data Preparation for Deep Learning Projects
Chapter 2: Data Preparation for Deep Learning Projects
Technical requirements
Setting up notebook environments
Data collection, data cleaning, and data preprocessing
Extracting features from data
Performing data visualization
Introduction to Docker
Summary
4
Chapter 3: Developing a Powerful Deep Learning Model
Chapter 3: Developing a Powerful Deep Learning Model
Technical requirements
Going through the basic theory of DL
Components of DL frameworks
Implementing and training a model in PyTorch
Implementing and training a model in TF
Decomposing a complex, state-of-the-art model implementation
Summary
5
Chapter 4: Experiment Tracking, Model Management, and Dataset Versioning
Chapter 4: Experiment Tracking, Model Management, and Dataset Versioning
Technical requirements
Overview of DL project tracking
DL project tracking with Weights & Biases
DL project tracking with MLflow and DVC
Dataset versioning – beyond Weights & Biases, MLflow, and DVC
Summary
6
Part 2 – Building a Fully Featured Product
Part 2 – Building a Fully Featured Product
7
Chapter 5: Data Preparation in the Cloud
Chapter 5: Data Preparation in the Cloud
Technical requirements
Data processing in the cloud
Introduction to Apache Spark
Setting up a single-node EC2 instance for ETL
Setting up an EMR cluster for ETL
Creating a Glue job for ETL
Utilizing SageMaker for ETL
Comparing the ETL solutions in AWS
Summary
8
Chapter 6: Efficient Model Training
Chapter 6: Efficient Model Training
Technical requirements
Training a model on a single machine
Training a model on a cluster
Training a model using SageMaker
Training a model using Horovod
Training a model using Ray
Training a model using Kubeflow
Summary
9
Chapter 7: Revealing the Secret of Deep Learning Models
Chapter 7: Revealing the Secret of Deep Learning Models
Technical requirements
Obtaining the best performing model using hyperparameter tuning
Understanding the behavior of the model with Explainable AI
Summary
10
Part 3 – Deployment and Maintenance
Part 3 – Deployment and Maintenance
11
Chapter 8: Simplifying Deep Learning Model Deployment
Chapter 8: Simplifying Deep Learning Model Deployment
Technical requirements
Introduction to ONNX
Conversion between TensorFlow and ONNX
Conversion between PyTorch and ONNX
Summary
12
Chapter 9: Scaling a Deep Learning Pipeline
Chapter 9: Scaling a Deep Learning Pipeline
Technical requirements
Inferencing using Elastic Kubernetes Service
Inferencing using SageMaker
Summary
13
Chapter 10: Improving Inference Efficiency
Chapter 10: Improving Inference Efficiency
Technical requirements
Network quantization – reducing the number of bits used for model parameters
Weight sharing – reducing the number of distinct weight values
Network pruning – eliminating unnecessary connections within the network
Knowledge distillation – obtaining a smaller network by mimicking the prediction
Network Architecture Search – finding the most efficient network architecture
Summary
14
Chapter 11: Deep Learning on Mobile Devices
Chapter 11: Deep Learning on Mobile Devices
Preparing DL models for mobile devices
Creating iOS apps with a DL model
Creating Android apps with a DL model
Summary
15
Chapter 12: Monitoring Deep Learning Endpoints in Production
Chapter 12: Monitoring Deep Learning Endpoints in Production
Technical requirements
Introduction to DL endpoint monitoring in production
Monitoring using CloudWatch
Monitoring a SageMaker endpoint using CloudWatch
Monitoring an EKS endpoint using CloudWatch
Summary
16
Chapter 13: Reviewing the Completed Deep Learning Project
Chapter 13: Reviewing the Completed Deep Learning Project
Reviewing a DL project
Gathering the reusable knowledge, concepts, and artifacts for future projects
Summary
17
Index
Index
Why subscribe?
18
Other Books You May Enjoy
Other Books You May Enjoy
Packt is searching for authors like you
Share Your Thoughts

Training a model using SageMaker

As mentioned in the Utilizing SageMaker for ETL section of Chapter 5, Data Preparation in the Cloud, the motivation of SageMaker is to help engineers and researchers focus on developing high-quality DL pipelines without worrying about infrastructure management. SageMaker manages data storage and computational resources for you, allowing you to utilize a distributed system for model training with minimal effort. In addition, SageMaker supports streaming data to your models for inferencing, hyperparameter tuning, and tracking experiments and artifacts.

SageMaker Studio is the place where you define the logic for your model. The SageMaker Studio notebooks allow you to quickly explore the available data and set up model training logic. When model training takes too long, scaling up to use multiple computational resources and finding the best set of hyperparameters can be efficiently achieved by making a few modifications to the infrastructure’...

Previous Section
End of Section 5
Next Section