Databricks Certified Generative AI Engineer - Associate
Get started today
Ultimate access to all questions.
Question: 7
You are designing a recommendation engine for an e-commerce platform. The source documents consist of product descriptions, customer reviews, and seller guidelines, ranging from 50 to 1000 words. Customer queries are typically short (1-2 sentences) and focus on finding specific products or features. You want to optimize the system for fast, accurate responses to queries while minimizing unnecessary memory usage. Which context length for the embedding model would be most appropriate for your use case?
Real Exam
Community
LLeetQuiz
Explanation:
Explanation
D. 512 tokens - This context length strikes a balance between being large enough to capture meaningful product descriptions, customer reviews, and features, while also being optimized for fast, accurate responses. With customer queries being short (1-2 sentences), 512 tokens provide sufficient context for embedding product descriptions and reviews without using excessive memory.
Why not the others:
-
A. 2048 tokens would be overkill for short queries and relatively small source documents, leading to unnecessary memory usage without significant gains in accuracy.
-
B. 128 tokens would be too short to capture sufficient context from product descriptions and customer reviews, potentially affecting the quality of the recommendations.
-
C. 256 tokens might be borderline for some longer product descriptions and reviews, potentially missing important contextual information that could improve recommendation quality.
Comments
Loading comments...