問題文
The memo embeddings are 1024 dimensions of single-precision values. Capacity planning asks how the vector size affects storage, and whether switching to half-precision would hurt search quality. Which pair of statements matches the documented guidance?
選択肢
- Vector storage is independent of the page size because vectors are always stored outside the row, so only the query cost changes.
- Storage can be reduced by declaring the column as a sparse vector, which is the supported way to compress high dimension counts.
- A data page holds up to 8060 bytes, so a vector of that size limits how many vectors fit in one page; and reducing precision usually has minimal impact on the quality of similarity comparisons.
- Reducing precision must be avoided because embeddings are highly sensitive to small numerical changes.