Creating a Compute
Understand the concept of Azure compute resources and their types including compute instances and clusters. Learn how to create, configure, and manage compute clusters for machine learning tasks using Azure CLI and the Azure Machine Learning portal, including setting tiers, sizes, and quotas.
We'll cover the following...
What is a compute?
Do we need VMs for training and deploying ML and deep learning models all the time? Not really. That’s where the compute comes into the picture. We can create a compute and use it for ML tasks later. It’s a fully managed workstation maintained on Azure. We can run our training script or host our service deployment using compute target. This location can be our local machine or a cloud-based compute resource.
Azure Machine Learning offers four different types of compute options:
-
Compute instance: This is a single-node compute server. It’s helpful for loading a Jupyter Notebook or SSH to compute.
-
Compute clusters: This is for having more than one node in the compute servers. It offers parallel processing abilities.
-
Inference clusters: These help create a new Azure Kubernetes Service (AKS) cluster or attach an existing one to our workspace and then deploy the model as a REST ...