Objective 4.1 Deploy a Model AIP-210 Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which of the following is typically included when setting up a runtime environment for a machine learning model?

   • AThe model's training logs
   • BThe original training dataset
   • CThe data annotation guidelines
   • DPython packages and libraries required by the model
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which of the following is a common method for validating a deployed model's performance?

   • ARetraining the model with a larger dataset
   • BIncreasing the number of CPU cores
   • CDeleting the model artifact from the server
   • DSending test requests with known expected outputs and comparing results
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   In a production inference pipeline, what is typically the first step after receiving an input request?

   • ARetrain the model on the new input
   • BStore the input in a data warehouse
   • CPreprocess the input data to match the model's expected format
   • DSend the input to a human reviewer
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   What is the purpose of serializing a trained machine learning model into a format like pickle or ONNX?

   • ATo visualize the model's decision boundaries
   • BTo save the model to disk so it can be loaded later for inference
   • CTo improve the model's accuracy on new data
   • DTo generate training data for the next iteration
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   A team has deployed a new model version that improves accuracy on the test set. However, in production, they observe that the model's predictions are systematically biased for a certain demographic group. What is the most appropriate next step?

   • ARoll back to the previous model version immediately
   • BInvestigate the bias by analyzing the model's predictions across demographic groups and compare with the previous version
   • CIgnore the bias because the model has higher overall accuracy
   • DRetrain the model with more data and redeploy immediately
   Reveal Solution Discussion Select an answer first