Adapter

A small set of extra weights inserted into a frozen model and trained alone, so one base model can be specialised many ways without copying it. LoRA is the form almost everyone uses now.

A definition is not an understanding. This is taught properly in Prompt, retrieve or fine-tune, part of Fine-tuning.

Learn this properly