single mention · inference & serving
Fine-grained prefix hashing within cache pages
Hashes prefixes in fine-grained token blocks while using larger physical pages as the allocation unit.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See KV cache management for the methods filed beside it.
Also called Prefix hashing runs on fine hash blocks.
Filed under inference & serving › KV cache management