Designing and Implementing a Data Science Solution on Azure | DP-100 Exam Question Answer

Using Databricks Cluster as Training Compute | DP-100 Exam Solution

Prev Question Next Question

Question

You have an existing, powerful Databricks cluster in your local environment and, instead of provisioning an Azure ML compute, you decide to use it in your ML experiments as a training compute.

# Create the compute databricks_cluster = [select the passing code segment here](ws, compute_name, compute_config) databricks_cluster.wait_for_completion(show_output=True)
Which code segment should you choose to complete the code?

Answers

Explanations

Click on the arrows to vote for the correct answer

A. B. C. D.

Answer: C.

Option A is incorrect because the create() method of the ComputeTarget class is used for adding Azure-managed computes to the workspace.

It cannot be used for external computes.

Option B is incorrect because provisioning_configuration is a parameter of the create() method.

Option C is CORRECT becausecompute targets defined outside the Azure space (typically a Databricks cluster) can be added to the workspace by the attach() method.

Option D is incorrect because use is not a valid method name for compute targets.

Reference:

In this scenario, the objective is to use an existing Databricks cluster as a training compute for machine learning experiments instead of provisioning an Azure ML compute. To achieve this, we need to create a ComputeTarget object that represents the Databricks cluster and then use this object in our ML experiments.

The correct method to create a ComputeTarget object that represents the existing Databricks cluster is ComputeTarget.attach(). This method is used to attach an existing compute target to a workspace. We can use this method to attach a Databricks cluster as a compute target.

Therefore, the correct code segment to complete the code is:

makefile
databricks_cluster = ComputeTarget.attach(workspace=ws, name=compute_name)

where "ws" is the workspace object, "compute_name" is the name of the Databricks cluster, and "databricks_cluster" is the ComputeTarget object that represents the Databricks cluster.

After creating the ComputeTarget object, we can use it as a training compute for our ML experiments. The wait_for_completion() method is used to wait for the completion of the creation of the compute target.

Therefore, the complete code would be:

python
from azureml.core.workspace import Workspace from azureml.core.compute import ComputeTarget  # Get the workspace ws = Workspace.from_config()  # Define the name and configuration of the existing Databricks cluster compute_name = "existing-databricks-cluster" compute_config = None # or any configuration specific to the Databricks cluster  # Attach the Databricks cluster as a compute target databricks_cluster = ComputeTarget.attach(workspace=ws, name=compute_name)  # Wait for the completion of the creation of the compute target databricks_cluster.wait_for_completion(show_output=True)

This code segment will allow us to use the existing Databricks cluster as a training compute for our ML experiments without the need to provision an Azure ML compute.

Prev Question Next Question