Comparison
Full fine-tuningvsParameter-efficient fine-tuning (PEFT)
Full fine-tuning
you update every parameter, need serious GPU memory to do it, and end up with a full-size copy of the model per variant.
Updating every parameter of the model. Maximum effect, maximum cost: for large models it needs serious GPU memory and produces a full-size copy per variant. Rare outside labs and large teams.
Full entry →Parameter-efficient fine-tuning (PEFT)
you train a small fraction of the parameters and the memory bill drops by orders of magnitude.
The family of methods that adapt a model while training only a small fraction of parameters, cutting memory and cost by orders of magnitude. LoRA is the member you will actually meet.
Full entry →