Open Models Almanac
Methodstraining objectivemulti-token prediction objective

single mention · training objective

MTP warm-up

A dedicated 100-billion-token MTP warm-up stage used before subsequent training.

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

Also called dedicated 100-billion-token MTP warm-up.

Filed under training objective › multi-token prediction objective