Open Models Almanac
Methodsoptimizationtraining parallelism

single mention · optimization

NS-FLOP-balanced static parameter partitioning

A static partitioner assigns whole parameters to data-parallel ranks to equalize estimated NS FLOPs.

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 An -balanced static partitioner.

Filed under optimization › training parallelism