AI-102 Question 470
Select 4Your company is deploying a generative AI model for summarizing customer support tickets. The model's response time is slow, and it's consuming more compute resources than expected. What actions can you take to optimize the generative AI solution in Azure?
- A
Enable model quantization to reduce the size of the model and improve inference speed.
- B
Use Azure AI Content Safety to filter inappropriate outputs generated by the model.
- C
Implement Azure Cache for Redis to store and retrieve frequently used model predictions.
- D
Switch to a smaller pre-trained model that meets the accuracy requirements.
- E
Leverage batch processing to handle multiple input requests simultaneously.
Show answer and explanation
Correct answers: A, C, D, E
Explanation
Optimizing generative AI solutions involves a combination of techniques such as reducing model size (quantization or smaller pre-trained models), optimizing infrastructure (caching and batch processing), and adjusting the deployment configuration to improve performance and reduce resource usage. These strategies ensure the model operates efficiently while meeting business requirements.
- A. Correct.
Enabling model quantization reduces the precision of the model weights, which can optimize performance without significantly affecting accuracy, making it a valid optimization strategy.
- B. Incorrect.
Azure AI Content Safety is used for filtering inappropriate outputs but does not contribute to optimizing the model's performance or resource consumption.
- C. Correct.
Using Azure Cache for Redis to store frequently used predictions can reduce the computation load by avoiding redundant calls to the model for the same inputs.
- D. Correct.
Switching to a smaller pre-trained model is a common optimization technique, as it reduces resource consumption while maintaining acceptable accuracy.
- E. Correct.
Batch processing allows the system to handle multiple input requests in a single forward pass, improving throughput and reducing resource usage.