ezwiefel / azure-machine-learning-bicep

A set of Bicep templates for Azure Machine Learning.

Bicep files for Azure Machine Learning Advanced Deployment

Bicep is a Domain Specific Language (DSL) for deploying Azure resources declaratively.

This repo aims to leverage Bicep for creating AML - the benefit of Bicep is that it can be output to ARM templates - which can then be leveraged on other areas such as Azure Blueprints, etc.

To test this deployment this in your Azure tenant, you can create a new resource group and then run a group deployment on main.bicep

> az group create -n rg-test -l southcentralus

Location        Name
--------------  -------
southcentralus  rg-test

> az deployment group create \
    --name aml-deployment \
    --mode Incremental \
    --resource-group rg-test \
    --template-file main.bicep \
    --parameters baseResourceName=amlworkspace

Name            ResourceGroup    State      Timestamp                         Mode
--------------  ---------------  ---------  --------------------------------  -----------
aml-deployment  rg-test          Succeeded  {timestamp}  Incremental

To Do for this repo:

• Create Architecture Drawing
• Create template for Compute Instance
• Create template for Compute Resource
• Create template for Datastore / Dataset
• Document the variables
• Document prereqs
  • Quota for Private Link
• Test Extensively

azure-machine-learning-bicep's Issues

Add template for compute instance

For testing purposes - add a JumpBox VM with Azure Bastion

This will allow for testing of workspace functionality without having to configure VPN, etc

Add template for Datastore

Private Endpoint for file storage is required

Workspace storage requires a private endpoint for file on top of blob. The current one creates a private endpoint for blob only.

https://docs.microsoft.com/en-us/azure/machine-learning/how-to-secure-workspace-vnet#secure-azure-storage-accounts-with-private-endpoints

Add template for Dataset

Create (or link) architecture drawing