Introducing Amazon SageMaker training
In Chapter 5, you learned about SageMaker Studio and how to process data and train a model using SageMaker Data Wrangler and SageMaker Studio notebooks, respectively. Studio notebooks are great for experimentation on smaller datasets and testing your training scripts locally before running them on the full dataset. While SageMaker Studio notebooks provide a choice of GPU-powered accelerated computing, it is sometimes more cost effective to run the training as a job outside the notebook. It is also an architectural best practice to decouple the development environment from the training environment so they can scale independently from each other. This is where SageMaker training comes in. Let us now understand the basics of SageMaker training.
Understanding the SageMaker training architecture
SageMaker provides options to scale your training job in a managed environment decoupled from your development environment (such as SageMaker Studio notebooks...
