Chapter 4Choosing the Right ML Infrastructure
In this chapter, we will discuss the hardware requirements for training on different types of data and for the various different scenarios of deployment. The hardware requirements depend on the type of machine learning (ML) algorithm you choose, the training method (custom, AutoML, or pretrained), and the data type as well.
Generally, ML models for video and images require a large number of computations for each loop of training, including matrix multiplications, additions, subtractions, and differentials. These operations are performed on very large matrices, which sometimes could be several megabytes for each data point. When you have millions of data points, it becomes a problem for traditional general‐purpose CPUs. So, GPUs (which were originally used in rendering videos or in gaming consoles) were repurposed to train ML models. Later Google created custom ...
Get Official Google Cloud Certified Professional Machine Learning Engineer Study Guide now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
