PermLLM: Learnable Channel Permutation for N:M Sparse Large Language Models

Farzan Farnia (The Chinese University of Hong Kong) · Bei Yu · Tsung-Yi Ho (Department of Computer Science and Engineering, The Chinese University of Hong Kong) · Lancheng Zou (The Chinese University of Hong Kong) · Shuo Yin (Department of Computer Science and Engineering, The Chinese University of Hong Kong) · Zehua Pei (Department of Computer Science and Engineering, The Chinese University of Hong Kong)

block-wise channel permutationchannel permutationcomputational complexitydifferentiable soft permutationdiscrete permutation matricesend-to-end optimizationlearnable channel permutationllama seriesmodel performancen:m sparsityone-shot pruningpruning frameworkpruning-induced errorssinkhorn normalizationweight matrices

Channel permutation is a powerful technique for enhancing the accuracy of N:M sparse models by reordering the channels of weight matrices to prioritize the retention of important weights. However, traditional channel permutation methods rely on handcrafted quality metrics, which often fail to accurately capture the true impact of pruning on model performance. To address this limitation, we propose PermLLM, a novel post-training pruning framework that introduces learnable channel permutation (LCP) for N:M sparsity. LCP leverages Sinkhorn normalization to transform discrete permutation matrices into differentiable soft permutation matrices, enabling end-to-end optimization. Additionally, PermLLM incorporates an efficient block-wise channel permutation strategy, which significantly reduces the number of learnable parameters and computational complexity. PermLLM seamlessly integrates with existing one-shot pruning methods to adaptively optimize channel permutations, effectively mitigating pruning-induced errors. Extensive experiments on the LLaMA series, Qwen, and OPT models demonstrate that PermLLM achieves superior performance in optimizing N:M sparse models.