Debug online endpoints locally in Visual Studio Code

This guide assumes you have the following items installed locally on your PC.

• Docker
• Visual Studio Code
• Azure CLI
• Azure CLI ml extension (v2)

For more information, see the guide on how to prepare your system to deploy online endpoints.

The examples in this article are based on code samples contained in the azureml-examples GitHub repository. To run the commands locally without having to copy/paste YAML and other files, clone the repo and then change directories to azureml-examples/cli:

git clone https://github.com/Azure/azureml-examples --depth 1
cd azureml-examples/cli

If you haven't already set the defaults for the Azure CLI, save your default settings. To avoid passing in the values for your subscription, workspace, and resource group multiple times, use the following commands. Replace the following parameters with values for your specific configuration:

• Replace <subscription> with your Azure subscription ID.
• Replace <workspace> with your Azure Machine Learning workspace name.
• Replace <resource-group> with the Azure resource group that contains your workspace.
• Replace <location> with the Azure region that contains your workspace.

az account set --subscription <subscription>
az configure --defaults workspace=<workspace> group=<resource-group> location=<location>

APPLIES TO: Python SDK azure-ai-ml v2 (current)

This guide assumes you have the following items installed locally on your PC.

• Docker
• Visual Studio Code
• Azure CLI
• Azure CLI ml extension (v2)
• Azure Machine Learning Python SDK (v2)

For more information, see the guide on how to prepare your system to deploy online endpoints.

The examples in this article can be found in the Jupyter notebook called Debug online endpoints locally in Visual Studio Code within the azureml-examples repository. To run the code locally, clone the repo and then change directories to the notebook's parent directory sdk/endpoints/online/managed.

git clone https://github.com/Azure/azureml-examples --depth 1
cd azureml-examples
cd sdk/python/endpoints/online/managed

Open the Jupyter notebook and import the required modules:

from azure.ai.ml import MLClient
from azure.ai.ml.entities import (
    ManagedOnlineEndpoint,
    ManagedOnlineDeployment,
    Model,
    CodeConfiguration,
    Environment,
)
from azure.identity import DefaultAzureCredential

Set up variables for the workspace and endpoint:

subscription_id = "<SUBSCRIPTION_ID>"
resource_group = "<RESOURCE_GROUP>"
workspace_name = "<AML_WORKSPACE_NAME>"

endpoint_name = "<ENDPOINT_NAME>"

Azure Machine Learning local endpoints use Docker and Visual Studio Code development containers (dev containers) to build and configure a local debugging environment. With dev containers, you can take advantage of Visual Studio Code features from inside a Docker container. For more information on dev containers, see Create a development container.

To debug online endpoints locally in Visual Studio Code, use the --vscode-debug flag when creating or updating and Azure Machine Learning online deployment. The following command uses a deployment example from the examples repo:

az ml online-deployment create --file endpoints/online/managed/sample/blue-deployment.yml --local --vscode-debug

A Docker image is built locally. Any environment configuration or model file errors are surfaced at this stage of the process.

Once the image successfully builds, your dev container opens in a Visual Studio Code window.

You'll use a few Visual Studio Code extensions to debug your deployments in the dev container. Azure Machine Learning automatically installs these extensions in your dev container.

• Inference Debug
• Pylance
• Jupyter
• Python

Azure Machine Learning local endpoints use Docker and Visual Studio Code development containers (dev containers) to build and configure a local debugging environment. With dev containers, you can take advantage of Visual Studio Code features from inside a Docker container. For more information on dev containers, see Create a development container.

Get a handle to the workspace:

credential = DefaultAzureCredential()
ml_client = MLClient(
    credential,
    subscription_id=subscription_id,
    resource_group_name=resource_group,
    workspace_name=workspace_name,
)

To debug online endpoints locally in Visual Studio Code, set the vscode-debug and local flags when creating or updating an Azure Machine Learning online deployment. The following code mirrors a deployment example from the examples repo:

deployment = ManagedOnlineDeployment(
    name="blue",
    endpoint_name=endpoint_name,
    model=Model(path="../model-1/model"),
    code_configuration=CodeConfiguration(
        code="../model-1/onlinescoring", scoring_script="score.py"
    ),
    environment=Environment(
        conda_file="../model-1/environment/conda.yml",
        image="mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04:20210727.v1",
    ),
    instance_type="Standard_DS2_v2",
    instance_count=1,
)

deployment = ml_client.online_deployments.begin_create_or_update(
    deployment,
    local=True,
    vscode_debug=True,
)

A Docker image is built locally. Any environment configuration or model file errors are surfaced at this stage of the process.

Once the image successfully builds, your dev container opens in a Visual Studio Code window.

You'll use a few Visual Studio Code extensions to debug your deployments in the dev container. Azure Machine Learning automatically installs these extensions in your dev container.

• Inference Debug
• Pylance
• Jupyter
• Python

Now that your application is running in the debugger, try making a prediction to debug your scoring script.

Use the ml extension invoke command to make a request to your local endpoint.

az ml online-endpoint invoke --name <ENDPOINT-NAME> --request-file <REQUEST-FILE> --local

In this case, <REQUEST-FILE> is a JSON file that contains input data samples for the model to make predictions on similar to the following JSON:

{"data": [
    [1,2,3,4,5,6,7,8,9,10], 
    [10,9,8,7,6,5,4,3,2,1]
]}

az ml online-endpoint show --name <ENDPOINT-NAME> --local

{
 "auth_mode": "aml_token",
 "location": "local",
 "name": "my-new-endpoint",
 "properties": {},
 "provisioning_state": "Succeeded",
 "scoring_uri": "http://localhost:5001/score",
 "tags": {},
 "traffic": {},
 "type": "online"
}

At this point, any breakpoints in your run function are caught. Use the debug actions to step through your code. For more information on debug actions, see the debug actions guide.

Now that your application is running in the debugger, try making a prediction to debug your scoring script.

Use the invoke method on your MLClient object to make a request to your local endpoint.

endpoint = ml_client.online_endpoints.get(name=endpoint_name, local=True)

request_file_path = "../model-1/sample-request.json"

ml_client.online_endpoints.invoke(endpoint_name, request_file_path, local=True)

In this case, <REQUEST-FILE> is a JSON file that contains input data samples for the model to make predictions on similar to the following JSON:

{"data": [
    [1,2,3,4,5,6,7,8,9,10], 
    [10,9,8,7,6,5,4,3,2,1]
]}

print(endpoint)

{
 "auth_mode": "aml_token",
 "location": "local",
 "name": "my-new-endpoint",
 "properties": {},
 "provisioning_state": "Succeeded",
 "scoring_uri": "http://localhost:5001/score",
 "tags": {},
 "traffic": {},
 "type": "online"
}

At this point, any breakpoints in your run function are caught. Use the debug actions to step through your code. For more information on debug actions, see the debug actions guide.

As you debug and troubleshoot your application, there are scenarios where you need to update your scoring script and configurations.

To apply changes to your code:

1. Update your code.
2. Restart your debug session using the Developer: Reload Window command in the command palette. For more information, see the command palette documentation.

For more extensive changes involving updates to your environment and endpoint configuration, use the ml extension update command. Doing so triggers a full image rebuild with your changes.

az ml online-deployment update --file <DEPLOYMENT-YAML-SPECIFICATION-FILE> --local --vscode-debug

Once the updated image is built and your development container launches, use the Visual Studio Code debugger to test and troubleshoot your updated endpoint.

As you debug and troubleshoot your application, there are scenarios where you need to update your scoring script and configurations.

To apply changes to your code:

1. Update your code.
2. Restart your debug session using the Developer: Reload Window command in the command palette. For more information, see the command palette documentation.

For more extensive changes involving updates to your environment and endpoint configuration, use your MLClient's online_deployments.update module/method. Doing so triggers a full image rebuild with your changes.

new_deployment = ManagedOnlineDeployment(
    name="green",
    endpoint_name=endpoint_name,
    model=Model(path="../model-2/model"),
    code_configuration=CodeConfiguration(
        code="../model-2/onlinescoring", scoring_script="score.py"
    ),
    environment=Environment(
        conda_file="../model-2/environment/conda.yml",
        image="mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04:20210727.v1",
    ),
    instance_type="Standard_DS2_v2",
    instance_count=2,
)

ml_client.online_deployments.update(new_deployment, local=True, vscode_debug=True)

Once the updated image is built and your development container launches, use the Visual Studio Code debugger to test and troubleshoot your updated endpoint.