Open Models Almanac
Methodsinference & servingdecoding strategy

single mention · inference & serving

DSpark draft head

A released draft head intended for compatible speculative-decoding serving implementations.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See decoding strategy for the methods filed beside it.

Also called DSpark speculative-decoding draft head, draft head.

Filed under inference & serving › decoding strategy