Open Models Almanac
Methodsoptimizationlearning-rate schedule

single mention · optimization

Warmup-Stable-Merge

A learning-rate training recipe named Warmup-Stable-Merge; the evidence does not specify its phases or mechanism.

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 Warmup-Stable-Merge learning-rate schedule, Warmup-Stable-Merge (WSM) recipe.

Filed under optimization › learning-rate schedule