下一条:Bridging pre-trained models to continual learning: A hypernetwork based framework with parameter-efficient fine-tuning techniques