Exam Professional Machine Learning Engineer topic 1 question 175 discussion - ExamTopics

Question:

The question asks about the best approach to deploy a pre-trained scikit-learn model on Google Vertex AI for both online and batch prediction, while minimizing extra code. The model requires input data preprocessing.

Options:

• A: Upload the model using a prebuilt scikit-learn prediction container to the Vertex AI Model Registry. Deploy to Vertex AI Endpoints, and use instanceConfig.instanceType for data transformation in a batch prediction job.
• B: Wrap the model in a custom prediction routine (CPR), build a container image, upload to Vertex AI Model Registry, and deploy to Endpoints for a batch prediction job.
• C: Create a custom container and serving function for the model, upload to the Model Registry, and deploy to Endpoints for a batch prediction job.
• D: Create a custom container, upload to the Model Registry, deploy to Endpoints, and use instanceConfig.instanceType for data transformation in a batch prediction job.

Correct Answer:

The suggested answer is B. Wrapping the model in a custom prediction routine and building a container image is recommended for optimal deployment and efficiency.