InfrastructureParameter-Efficient Fine-Tuning
A family of techniques that adapts a pretrained language or vision model to a downstream task by training only a small fraction of its parameters, dramatically reducing compute, memory, and storage requirements compared to full fine-tuning.