Book Image

Learn Amazon SageMaker

By : Julien Simon
Book Image

Learn Amazon SageMaker

By: Julien Simon

Overview of this book

Amazon SageMaker enables you to quickly build, train, and deploy machine learning (ML) models at scale, without managing any infrastructure. It helps you focus on the ML problem at hand and deploy high-quality models by removing the heavy lifting typically involved in each step of the ML process. This book is a comprehensive guide for data scientists and ML developers who want to learn the ins and outs of Amazon SageMaker. You’ll understand how to use various modules of SageMaker as a single toolset to solve the challenges faced in ML. As you progress, you’ll cover features such as AutoML, built-in algorithms and frameworks, and the option for writing your own code and algorithms to build ML models. Later, the book will show you how to integrate Amazon SageMaker with popular deep learning libraries such as TensorFlow and PyTorch to increase the capabilities of existing models. You’ll also learn to get the models to production faster with minimum effort and at a lower cost. Finally, you’ll explore how to use Amazon SageMaker Debugger to analyze, detect, and highlight problems to understand the current model state and improve model accuracy. By the end of this Amazon book, you’ll be able to use Amazon SageMaker on the full spectrum of ML workflows, from experimentation, training, and monitoring to scaling, deployment, and automation.
Table of Contents (19 chapters)
Section 1: Introduction to Amazon SageMaker
Section 2: Building and Training Models
Section 3: Diving Deeper on Training
Section 4: Managing Models in Production

Streaming datasets with pipe mode

The default setting of estimators is to copy the dataset to training instances, which is known as File Mode. Instead, pipe mode streams it directly from S3. The name of the feature comes from its use of Unix named pipes (also known as FIFOs): at the beginning of each epoch, one pipe is created per input channel.

Pipe mode removes the need to copy any data to training instances. Obviously, training jobs start quicker. They generally run faster too, as pipe mode is highly optimized. Another benefit is that you won't have to provision any storage for the dataset on training instances.

Cutting down on training time and storage means that you'll save money. The larger the dataset, the more you'll save. You can find benchmarks at

In practice, you can start experimenting with pipe mode for datasets in the hundreds of...