Resource-Efficient Pruning for Transformer via Low-Rank Importance Estimation
基于低秩重要性估计的Transformer资源高效剪枝
机构 * School of Computer Science and Technology, Guangdong University of Technology(广东工业大学计算机科学与技术学院) ; College of Information and Artificial Intelligence, Yangzhou University(扬州大学信息与人工智能学院)
专题命中 效率与部署 :language model(abstract);分类 cs.AI、cs.LG
AI总结 针对Transformer模型部署的高资源消耗问题,提出基于LoRA低秩梯度的REP-LIE剪枝方法,结合稳定性分数迭代剪枝,在LLaMA-7B等模型上实现高效剪枝并保持竞争力性能。
Comments 15 pages, 9 figures
Journal ref IEEE Transactions on Emerging Topics in Computational Intelligence, 2026