Key-value head count mathematical notation

Key-value head count is a recurring research-paper notation family. Counts independently stored key and value heads per layer, which may be smaller than the number of query heads in grouped-query attention.

Key-value head count: Counts independently stored key and value heads per layer, which may be smaller than the number of query heads in grouped-query attention. Example: Each active sequence stores one key vector per KV head and token.