Databricks Machine Learning Associate Question 624
Select 3A data science team has trained a machine learning model using Databricks and now wants to deploy it to serve real-time predictions via REST APIs. The team plans to use Databricks Model Serving. Which of the following steps are necessary to deploy and test their model using Databricks Model Serving?
- A
Register the model to the Databricks Model Registry.
- B
Enable Databricks Model Serving for the registered model.
- C
Export the model to an external serving platform before deployment.
- D
Test the deployed model by sending HTTP requests to the serving endpoint.
- E
Manually write a custom API server to serve predictions.
Show answer and explanation
Correct answers: A, B, D
Explanation
To deploy a model using Databricks Model Serving, the model must first be registered in the Model Registry. After registering, Databricks Model Serving must be enabled to generate a REST API endpoint for serving predictions. Finally, the endpoint can be tested by sending HTTP requests. There is no need to export the model to an external platform or manually write an API server, as Databricks provides a fully managed serving solution.
- A. Correct.
This is correct. To deploy a model via Databricks Model Serving, the model must first be registered in the Databricks Model Registry.
- B. Correct.
This is correct. Enabling Databricks Model Serving for the registered model is a necessary step to expose a REST API for real-time predictions.
- C. Incorrect.
This is incorrect. Databricks Model Serving does not require exporting the model to an external platform, as it provides built-in serving capabilities.
- D. Correct.
This is correct. After the model is deployed, it can be tested by sending HTTP requests to the generated REST API endpoint.
- E. Incorrect.
This is incorrect. Databricks Model Serving eliminates the need to manually create a custom API server by providing an out-of-the-box solution for serving models.