Task 3.1: Select Deployment Infrastructure Based on Existing Architecture and Requirements MLA-C01 Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   A team already uses Apache Airflow for their data engineering workflows and wants to integrate their ML training and deployment steps into the same orchestration. Which approach is most appropriate?

   • AReplace Airflow with SageMaker Pipelines
   • BRun Airflow on SageMaker
   • CUse AWS Step Functions instead
   • DUse Airflow to trigger SageMaker training and deployment jobs
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A company has a model that is deployed on a SageMaker real-time endpoint. The model is written in R, and the team has a custom R script that loads the model and serves predictions. They want to deploy this on SageMaker with minimal changes. Which container strategy should they use?

   • AUse the SageMaker scikit-learn container and convert the R model to Python
   • BUse the SageMaker Python container and call R from Python
   • CBuild a custom container that implements the SageMaker inference protocol
   • DUse the SageMaker R container
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   A team is setting up a test environment for a machine learning model that will be used for unit testing and small-scale validation. Which compute provisioning approach is most cost-effective for this purpose?

   • AProvision a production-sized endpoint
   • BUse a multi-GPU instance
   • CUse a small CPU instance
   • DUse a large GPU cluster
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   A company needs to generate predictions for a large dataset that is already stored in Amazon S3. The predictions are not needed in real-time. Which deployment strategy is most appropriate?

   • AAsynchronous deployment
   • BBatch deployment
   • CServerless deployment
   • DReal-time deployment
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A team has a PyTorch model with a custom inference script that includes pre- and post-processing logic not available in the default SageMaker inference toolkit. They want to deploy it on SageMaker with minimal changes to their code. Which container strategy should they use?

   • AUse the SageMaker TensorFlow container and convert the model to TensorFlow
   • BBuild a custom container from scratch that implements the SageMaker inference protocol
   • CUse the SageMaker scikit-learn container with the PyTorch model
   • DUse the SageMaker PyTorch inference container and provide a custom inference script
   Reveal Solution Discussion Select an answer first