1 paper
Liu Hanzuo, Chaofan Lin, Weixuan Sun +4
Semi-structured sparsity provides a practical path to accelerate large language models (LLMs) with native hardware support, but post-training semi-structured pruning often suffers…