Open Models Almanac
Methodsinference & servingKV cache management

single mention · inference & serving

Hierarchical KV caching across HBM and pinned host memory

Keeps cache state in HBM during GPU use and in a pinned host-memory pool during tool use.

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 hierarchical extension of the cache.

Filed under inference & serving › KV cache management