For build, train, test and deploy machine learning model, we need infrastructure in Azure machine learning which is called as Azure machine learning resources.
There are basically three resources required as below:
- Azure machine learning workspace
- Azure Compute resources
- Azure datastore
Azure machine learning workspace:
The top-level resource of the Azure machine learning is called as “Workspace” where you can train and track models, run experiments, jobs, components as well as deploy model as endpoint.
The workspace is used to organize the machine learning work.
The workspace can store references of the compute resources and datastore. It can also store logs, models, outputs, metrics, security setting etc.
What are tasks performed within a workspace?
Below are the some of the tasks which are performed in the workspace.
- Jobs
- Experiments
- Data assets
- Models
- Endpoints
Azure Compute resource:
Compute is one of the important resources for training, experimenting, and deploying model.
Basically, there are four types of compute resources available.
- Compute instances
- Compute clusters
- Inference clusters
- Attached compute
Azure datastore:
When new workspace is created. At the same time, Storage account has been created. This storage account is used to store machine learning artifacts such as jupeter notebooks, jobs, logs, experiments, etc.
