Skip to content

deployment.md

Latest commit

 

History

History
206 lines (156 loc) · 5.4 KB

deployment.md

File metadata and controls

206 lines (156 loc) · 5.4 KB
Raw

Deploying Azure Infrastructure for Model Distillation Workshop

This guide provides instructions for deploying the Azure infrastructure required for the Model Distillation Workshop using Bicep templates.

Infrastructure Overview

The deployment includes:

  • Azure AI Foundry (Hub & Project)
  • Azure AI Foundry for DeepSeek V3 model
  • Azure ML Compute resources for model training and inference
  • Storage accounts and Key Vaults for secure data management
  • Necessary role assignments for service access

Deployment Options

Option 1: Using Azure Developer CLI (azd) - Recommended

The Azure Developer CLI (azd) is the simplest way to deploy your infrastructure as it handles parameter substitution and resource naming automatically based on your azure.yaml file.

Prerequisites:

Deployment Steps:

  1. Navigate to the root directory of your project (where azure.yaml is located)

    PowerShell:

    cd path\to\Build25-LAB329

    Bash:

    cd path/to/Build25-LAB329

  2. Initialize your environment (first time only)

    PowerShell/Bash:

    azd init

  3. Deploy the infrastructure

    PowerShell/Bash:

    azd provision

During the azd provision process, you will be prompted to:

  • Select a subscription
  • Provide a unique environment name
  • Choose a location for your resources
  • (Optionally) confirm the deployment

Option 2: Using Azure CLI

If you prefer to use Azure CLI directly:

  1. Login to Azure

    PowerShell/Bash:

    az login

  2. Set your subscription

    PowerShell/Bash:

    az account set --subscription "YOUR_SUBSCRIPTION_ID"

  3. Create a resource group (if it doesn't exist)

    PowerShell:

    $resourceGroupName = "rg-build25-lab329"
$location = "eastus"  # Choose an appropriate region
az group create --name $resourceGroupName --location $location

    Bash:

    resourceGroupName="rg-build25-lab329"
location="eastus"  # Choose an appropriate region
az group create --name $resourceGroupName --location $location

  4. Deploy the Bicep template

    PowerShell:

    az deployment sub create `
  --name "build25-lab329-deployment" `
  --location $location `
  --template-file "infra\main.bicep" `
  --parameters environmentName="build25lab329" `
              location=$location `
              deepSeekV31Location="eastus" `
              principalId=$(az ad signed-in-user show --query id -o tsv)

    Bash:

    az deployment sub create \
  --name "build25-lab329-deployment" \
  --location $location \
  --template-file "infra/main.bicep" \
  --parameters environmentName="build25lab329" \
              location=$location \
              deepSeekV31Location="eastus" \
              principalId=$(az ad signed-in-user show --query id -o tsv)

Important Parameters

The Bicep deployment requires several parameters:

Parameter Description Example Value
environmentName A name for your environment (used in resource naming) "build25lab329"
location Primary location for resources "eastus", "westus2"
deepSeekV31Location Location for the DeepSeek model (must support the model) "eastus"
principalId Object ID of the user/service principal deploying resources auto-generated
computeVmSize VM size for Azure ML compute "STANDARD_NC40ADS_H100_V5"

After Deployment

After successful deployment, the outputs include:

  • Azure ML Workspace information
  • Teacher model endpoint and key
  • Azure AI Project connection string

These outputs can be used to create a local.env file for the workshop notebooks:

TEACHER_MODEL_NAME=DeepSeek-V3
TEACHER_MODEL_ENDPOINT=<from-deployment-output>
TEACHER_MODEL_KEY=<from-deployment-output>
AZUREML_SUBSCRIPTION_ID=<from-deployment-output>
AZUREML_RESOURCE_GROUP=<from-deployment-output>
AZUREML_WS_NAME=<from-deployment-output>

Accessing Deployment Outputs

With Azure Developer CLI:

PowerShell:

# View all outputs
azd env get-values

# Get a specific output
azd env get-values | findstr TEACHER_MODEL_ENDPOINT

Bash:

# View all outputs
azd env get-values

# Get a specific output
azd env get-values | grep TEACHER_MODEL_ENDPOINT

With Azure CLI:

PowerShell/Bash:

# View all outputs
az deployment sub show --name "build25-lab329-deployment" --query "properties.outputs" -o json

Troubleshooting

If deployment fails, check the following:

  • Ensure you have sufficient permissions to create resources in your subscription
  • Verify the selected regions support all required services
  • Check that you have sufficient quota for the GPU compute resources
  • Examine the deployment logs for specific error messages

Resource Cleanup

When you're finished with the workshop, you can remove all resources to prevent further charges:

PowerShell/Bash:

# Using Azure Developer CLI
azd down

PowerShell:

# Using Azure CLI
az group delete --name $resourceGroupName --yes --no-wait

Bash:

# Using Azure CLI
az group delete --name $resourceGroupName --yes --no-wait