For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python module

max.pipelines.speculative.depth_schedule

How many drafted tokens to verify, as a function of decode batch size.

For now we draft K tokens but only verify a subset of them depending on the schedule.

Type aliases

DepthScheduleEntry	One inclusive batch-size range and the draft depth to use across it.

Functions

build_depth_lookup	Expands a schedule into a dense batch_size -> depth lookup.
normalize_depth_schedule	Validates a schedule and returns it sorted by batch size.