Creating a Compute Instance in Azure ML Using CLI v2

This guide walks you through the process of creating a compute instance in Azure Machine Learning (Azure ML) using the Azure CLI v2.

---

Prerequisites

Ensure the following before proceeding:

1. Azure CLI Installed: Download and install the Azure CLI from Azure CLI Installation Guide.

2. Azure ML Extension Installed: Add the Azure ML extension by running:

   az extension add -n ml

3. Log in to Azure: Authenticate your Azure CLI session:

   az login

4. Access to Azure ML Workspace: You need an existing Azure ML workspace to create a compute instance.

---

Step 1: Verify Your Azure ML Workspace

List all available Azure ML workspaces in your resource group:

az ml workspace list -g <your-resource-group>

Replace <your-resource-group> with your actual resource group name. Note the workspaceName for the next steps.

---

Step 2: Create a Compute Instance

Use the following command to create a compute instance:

az ml compute create --name compute-cli01 `
                     --type ComputeInstance `
                     --size Standard_DS3_v2 `
                     --workspace-name ws_demo_pipeline03 `
                     --resource-group rg_demo03

Parameters Explained

• --name: The name of the compute instance (e.g., ml-compute-demo).
• --type: Always set to ComputeInstance.
• --size: Specify the virtual machine size. Common options include Standard_DS11_v2 or Standard_DS3_v2.
• --workspace-name: The name of your Azure ML workspace.
• --resource-group: The resource group containing your Azure ML workspace.

Example

az ml compute create --name my-compute-instance `
                     --type ComputeInstance `
                     --size Standard_DS11_v2 `
                     --workspace-name my-ml-workspace `
                     --resource-group my-resource-group

The compute instance will now be provisioned. This may take a few minutes.

---

Step 3: Check the Status of the Compute Instance

After creating the compute instance, verify its status using:

az ml compute show --name <compute-instance-name> `
                   --workspace-name <workspace-name> `
                   --resource-group <resource-group-name>

This command will display details about the compute instance, including its current status (e.g., Creating, Running, Stopped).

---

Step 4: List All Compute Instances

To view all compute resources in your Azure ML workspace:

az ml compute list --workspace-name <workspace-name> `
                   --resource-group <resource-group-name>

---

Step 5: Stop the Compute Instance

Stopping a compute instance when not in use can help save costs. Use the following command:

az ml compute stop --name <compute-instance-name> `
                   --workspace-name <workspace-name> `
                   --resource-group <resource-group-name>

---

Step 6: Delete the Compute Instance

If you no longer need the compute instance, delete it to free up resources:

az ml compute delete --name <compute-instance-name> `
                     --workspace-name <workspace-name> `
                     --resource-group <resource-group-name> --yes

The --yes flag confirms the deletion without further prompts.

---

Additional Tips

• Always stop compute instances when they are not in use to save costs.
• Use parameterized scripts to automate the creation and management of compute instances.
• Regularly monitor your compute resources to optimize resource usage.

---

By following this guide, you have successfully learned how to create and manage compute instances in Azure ML using the CLI v2. If you have any questions, feel free to explore the Azure CLI Documentation for more details!