COPAL: Continual Pruning in Large Language Generative Models