Databricks Certified Generative AI Engineer - Associate
Get started today
Ultimate access to all questions.
Question: 34
You are developing an AI-powered knowledge base application for a global research organization. The application will generate detailed technical reports based on user queries. Evaluation metrics include perplexity (response quality), throughput (tokens generated per second), and memory usage. The LLM must deliver highly accurate, contextually relevant information, while minimizing resource consumption. Which of the following LLM configurations would best meet the application's requirements for high accuracy, moderate throughput, and efficient memory usage?
Real Exam
Community
LLeetQuiz
Explanation:
Explanation
A 13-billion parameter model with low perplexity, moderate memory usage, and moderate throughput is the best choice for this use case because it provides a balance of high accuracy (low perplexity), sufficient performance (moderate throughput), and manageable resource consumption (moderate memory usage). This configuration ensures that the LLM can generate contextually relevant and accurate technical reports while operating efficiently in terms of memory and response speed.
Why not the others:
- A. 6-billion parameter model: While it has high throughput and low memory usage, its moderate perplexity indicates lower accuracy, which compromises the quality of technical reports.
- B. 1-billion parameter model: Although it offers very high throughput and low memory usage, its high perplexity suggests poor response quality, making it unsuitable for generating detailed and accurate reports.
- D. 30-billion parameter model: This model provides very low perplexity (high accuracy) but consumes significant memory and has low throughput, making it inefficient for practical use in a resource-constrained environment.
C strikes the right balance between accuracy, performance, and resource efficiency, meeting the requirements of the application effectively.
Comments
Loading comments...