Preparing the SageMaker notebook instance for multiple deep learning local experiments
When working with deep learning experiments in Amazon SageMaker, it is important to note that the custom scripts developed and used to train and deploy our models can be tested inside a running deep learning container using local mode. This allows us to fix any issues in the custom scripts right away without having to use dedicated ML training instances. However, working with deep learning containers involves pulling container images, which may cause disk space issues. That said, it is critical that we prepare the SageMaker notebook instance first and configure it to prevent any disk space issues later on.
In this recipe, we will (1) modify the volume size of the notebook instance, (2) create the directories where we will store the notebooks and scripts in this chapter, and (3) configure the Docker service to help us prevent potential disk space issues when we are pulling container images and...
