Open Models Almanac
Methodsoptimizationlearning-rate schedule

single mention · optimization

Scaling-law fit for learning-rate and batch-size selection

A scaling-law fit is used to guide learning-rate and batch-size choices, although the evidence does not specify its procedure.

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

Also called new scaling-law fit.

Filed under optimization › learning-rate schedule