Olica: Efficient Structured Pruning of Large Language Models without Retraining

Jiujun He, Huazhen Lin

Most existing structured pruning methods for Large Language Models (LLMs) require substantial computational and data resources for retraining to reestablish the corrupted correlati…