Databricks Machine Learning Associate Question 622
Select 3You are a data scientist working on deploying a machine learning model in Databricks. The model has been trained and saved as a MLflow Model, and your team wants to deploy it as a REST API to allow real-time predictions. Which of the following steps are essential to achieve this within Databricks?
- A
Register the model in the MLflow Model Registry.
- B
Deploy the model using MLflow’s built-in model serving capability.
- C
Configure a Databricks Job to periodically load the model for batch predictions.
- D
Create a Databricks cluster with GPU support for real-time serving.
- E
Enable Databricks Model Serving and test the REST endpoint with sample data.
Show answer and explanation
Correct answers: A, B, E
Explanation
To deploy a machine learning model as a REST API in Databricks, the model must first be registered in the MLflow Model Registry to manage versions and provide a central location for deployment. MLflow’s built-in serving functionality enables seamless deployment of models as REST endpoints within Databricks. Additionally, enabling Databricks Model Serving activates the REST endpoint for testing and real-time inference. Batch inference or GPU support is not necessary for this specific use case unless additional requirements are explicitly mentioned.
- A. Correct.
Registering the model in the MLflow Model Registry is essential, as it allows the model to be versioned, tracked, and accessed for deployment.
- B. Correct.
MLflow’s built-in serving functionality is a key feature that enables deployment of models as REST endpoints within Databricks.
- C. Incorrect.
Configuring a Databricks Job for batch predictions is not relevant for real-time deployment, as this is meant for scheduled batch inference rather than REST API serving.
- D. Incorrect.
Creating a Databricks cluster with GPU support is not necessary unless the model requires GPU acceleration, which is not explicitly stated in the scenario.
- E. Correct.
Enabling Databricks Model Serving ensures the REST API endpoint is active and accessible for real-time predictions and is critical for testing the deployment.