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 function

normalize_depth_schedule

normalize_depth_schedule()

max.pipelines.speculative.depth_schedule.normalize_depth_schedule(entries)

source

Validates a schedule and returns it sorted by batch size.

Parameters:

entries (Sequence[Sequence[int]]) – Triples of (batch_start, batch_end, depth).

Returns:

The entries as tuples, sorted ascending by batch_start.

Raises:

ValueError – If entries is empty, an entry is not a triple, a range is not positive and ascending, a depth is negative, ranges overlap, or the first range does not start at batch size 1. Starting at 1 is required so every runtime batch size resolves to a depth.

Return type:

list[tuple[int, int, int]]