LoRA

Also called: low-rank adaptation, low rank adaptation, LoRAs, QLoRA, adapter, adapters

Low-rank adaptation. A cheap way to fine-tune a model by training a small add-on set of weights instead of changing the whole model. The add-on file is small and can be swapped in and out.

For example: A company trains a LoRA on its brand voice in a few hours on one GPU and loads it on top of an open-weights model.