Databricks Machine Learning Associate Question 634
Select 3You have trained a custom machine learning model locally and want to deploy it as a REST API endpoint in Databricks to serve predictions. Which of the following steps are necessary to achieve this?
- A
Register the trained model to the Databricks Model Registry.
- B
Create a Databricks MLflow model serving endpoint and associate it with the registered model.
- C
Write a custom scoring script using the Databricks REST API to handle prediction requests.
- D
Enable model serving for the registered model version in the Databricks Model Registry.
- E
Package the model as a Docker container and upload it to the Databricks workspace.
Show answer and explanation
Correct answers: A, B, D
Explanation
Deploying a custom model as an endpoint in Databricks typically involves registering the model in the Model Registry, enabling model serving, and creating a serving endpoint. These steps ensure the model can be accessed via REST API without requiring manual infrastructure management. Custom scoring scripts or additional packaging are not necessary as Databricks provides integrated model serving capabilities.
- A. Correct.
Correct. Registering the trained model in the Databricks Model Registry is required to manage model versions and integrate it with Databricks model serving.
- B. Correct.
Correct. Creating an MLflow model serving endpoint is necessary to expose the model as an API endpoint for inference.
- C. Incorrect.
Incorrect. Databricks model serving automatically handles prediction requests without requiring a custom scoring script.
- D. Correct.
Correct. Enabling model serving for the registered model version is required to activate the REST API endpoint for the model.
- E. Incorrect.
Incorrect. Packaging into a Docker container and uploading it is not necessary for deploying a model with Databricks model serving.