Class DraftModelSpeculation (1.81.0)

DraftModelSpeculation(mapping=None, *, ignore_unknown_fields=False, **kwargs)

Draft model speculation works by using the smaller model to generate candidate tokens for speculative decoding.

Attribute

Name Description
draft_model str
Required. The resource name of the draft model.

Methods

DraftModelSpeculation

DraftModelSpeculation(mapping=None, *, ignore_unknown_fields=False, **kwargs)

Draft model speculation works by using the smaller model to generate candidate tokens for speculative decoding.