Demonstrate the Ability to Train and Inference CERTIFIED-AI-EXPERT Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A team is training a model and has split the data into training, validation, and test sets. After training, the model performs well on the training and validation sets but poorly on the test set. Which of the following is the most likely cause?

   • AThe validation set was used for training, causing data leakage.
   • BThe model was not trained long enough.
   • CThe model has overfit to the training data and does not generalize to unseen data.
   • DThe test set is too small, so the model cannot learn from it.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Which step is essential when deploying a trained model to serve inference requests in production?

   • AConverting the model to a format that only runs on a specific GPU
   • BDeleting the training data to free storage
   • CSetting up an API endpoint to accept input and return predictions
   • DRetraining the model on every incoming request
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   An e-commerce company has trained a recommendation model and wants to deploy it for real-time inference. The model was trained in Python with a specific library version. The production environment runs a different OS and has limited GPU resources. Which of the following is the most important consideration when deploying this model?

   • ADeploy the model on the same GPU cluster used for training to ensure identical performance.
   • BRetrain the model on the production OS to ensure compatibility with the new environment.
   • CConvert the model to a format that does not require any dependencies, such as a standalone executable.
   • DEnsure the inference environment has the same library versions and dependencies as the training environment to avoid prediction mismatches.
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A model serving team has deployed a model using a containerized service. They notice that the first inference request after a period of inactivity takes significantly longer than subsequent requests. What is the most likely cause and solution?

   • AThe container may have scaled to zero or been cold-started; use a keep-alive mechanism or a minimum number of instances.
   • BThe container is using a GPU that needs to warm up; pre-warm the GPU.
   • CThe model is being retrained on the first request; schedule retraining during off-peak hours.
   • DThe network is slow; upgrade the network bandwidth.
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A company has a model that must be updated monthly with new data. They currently retrain the model from scratch each month, which is time-consuming and costly. Which of the following approaches would best reduce the cost and time of model updates?

   • AContinue using the old model and only update it every six months.
   • BUse a larger GPU cluster to speed up the full retraining process.
   • CRetrain the model from scratch but on a smaller subset of data.
   • DUse incremental learning or fine-tuning on the existing model with the new data.
   Reveal Solution Discussion Select an answer first