Conceptfp32 / 单精度浮点数
float32
A numerical precision format that represents a floating-point number in 32 bits (4 bytes). It offers the highest precision but the largest storage size, and is often used as the baseline precision for vector storage.
Timeline (last 90 days)
Oct 18
Under float32 storage, each 1024-dimensional vector occupies 4KB, so an index of a million-scale documents requires at least 4GB of storage.
Unverified50%