Open Models Almanac
Methodsoptimizationtraining parallelism

single mention · optimization

GPU planning kernel for redundant expert migration

A GPU planning kernel computes near-optimal redundant-expert assignments while respecting the stated expert-to-rank upper bound.

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

Also called online planning of redundant experts.

Filed under optimization › training parallelism