发表机构
King's College London (KCL); University of Surrey（伦敦国王学院; 萨里大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究针对基于扩散的文本到音频生成模型AudioLDM计算成本高的问题，应用模型剪枝提高其计算效率，通过分析参数冗余和评估剪枝策略，经实验验证可大幅减少参数和运算量，且能在一定程度上恢复对特定声音事件的生成能力。
AI 中文摘要
基于扩散的文本到音频生成模型，如AudioLDM，能实现高感知质量和强语义一致性，但由于U-Net去噪主干的大量计算成本，其实际部署受到阻碍。在这项工作中，我们应用模型剪枝来提高基于U-Net的文本条件音频潜在扩散模型AudioLDM的计算效率。我们分析了U-Net卷积块中的参数冗余，并评估了一种滤波器剪枝策略。剪枝以基于范数的标准为指导，随后进行轻量级微调以恢复性能损失。实验结果表明，与未剪枝的基线网络相比，U-Net的参数减少了83%，乘加运算减少了39%，同时保持并在某些情况下提高了生成质量。我们发现剪枝会影响AudioLDM生成某些声音事件的能力，如枪声、警笛声、爆炸声等安全关键声音，以及钻孔机、缝纫机等机械声音，还有喷雾声、滴答声等其他声音，这些大多通过对剪枝模型的轻量级微调得以恢复。
英文摘要
Diffusion-based text-to-audio generative models such as AudioLDM achieve high perceptual quality and strong semantic consistency; however, their practical deployment is hindered by the substantial computational cost of the U-Net denoising backbone. In this work, we apply model pruning to improve the computational efficiency of AudioLDM, a U-Net-based text-conditioned audio latent diffusion model. We analyse parameter redundancy across U-Net convolutional blocks and evaluate a filter-pruning strategy. Pruning is guided by norm-based criteria and followed by lightweight finetuning to recover performance losses. Experimental results demonstrate that up to 83% of the parameters and 39% of the multiply-accumulate operations of U-Net have been reduced while maintaining, and in some cases improving, generation quality compared to the baseline unpruned network. We find that pruning affects AudioLDM's ability to generate certain sound events including safety-critical sounds such as gunshots, sirens, and explosions, as well as mechanical sounds such as drills and sewing machines, and other sounds such as sprays and tick-tocks, which are mostly recovered by lightweight finetuning of the pruned model.
CommentsSubmitted to DCASE 2026 Workshop
