The Definitive Guide toAI Data Centers
Ask the GuideAboutAccount
GuideGlossaryLoRA

LoRA · Low-Rank Adaptation

A fine-tuning method that trains small low-rank adapter matrices instead of all weights, slashing cost and memory.

Current numbers

~0.1%share of parameters trained by a LoRA adapter vs full fine-tune (model-dependent)as of 2021 · register ↗

← All terms