Open Models Almanac
Methodsoptimizationtraining runtime

single mention · optimization

Centralized block-runtime measurement collection

Gathers per-block runtime measurements on GPU or process 0 to obtain detailed statistics.

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

Also called gather the runtimes on GPU/process 0.

Filed under optimization › training runtime