面向扩散 Transformer 的可插拔剪枝与连续层蒸馏
Pluggable Pruning with Contiguous Layer Distillation for Diffusion Transformers
对做扩散 Transformer 生成的人，这套「线性探针定位冗余层区间 + 深度/宽度联合剪枝 + 一次性教师-学生交替蒸馏」的流程可直接迁移到自己的 DiT/MMDiT 骨干上，无需为每个剪枝比例重训即可在 50% 参数下保持生成质量。
原文依据 we propose a plug-and-play teacher-student alternating distillation scheme tailored to integrate depth-wise and width-wise pruning within a single training phase
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论扩散 Transformer（DiT）在图像生成上表现优异，但其庞大的参数量带来高昂的计算开销，阻碍了在资源受限环境中的部署。为此，我们提出可插拔剪枝与连续层蒸馏（PPCL），一个专为 DiT 架构设计的灵活结构化剪枝框架。首先，我们通过线性探针机制结合相似度指标的一阶差分趋势分析，识别出冗余的层区间。随后，我们提出一种即插即用的教师-学生交替蒸馏方案，旨在在单一训练阶段内整合深度方向与宽度方向的剪枝。该蒸馏框架支持在不同剪枝比例之间灵活进行知识迁移，无需针对每种配置重新训练。在多个多模态扩散 Transformer 架构模型上的大量实验表明，PPCL 相比完整模型实现了 50% 的参数量削减，关键客观指标下降不足 3%。值得注意的是，我们的方法在更高压缩比下仍保持高质量图像生成能力，使其非常适用于资源受限环境。
英文原文摘要 1284 字符
Diffusion Transformers (DiTs) have shown exceptional performance in image generation, yet their large parameter counts incur high computational costs, impeding deployment in resource-constrained settings. To address this, we propose Pluggable Pruning with Contiguous Layer Distillation (PPCL), a flexible structured pruning framework specifically designed for DiT architectures. First, we identify redundant layer intervals through a linear probing mechanism combined with the first-order differential trend analysis of similarity metrics. Subsequently, we propose a plug-and-play teacher-student alternating distillation scheme tailored to integrate depth-wise and width-wise pruning within a single training phase. This distillation framework enables flexible knowledge transfer across diverse pruning ratios, eliminating the need for per-configuration retraining. Extensive experiments on multiple Multi-Modal Diffusion Transformer architecture models demonstrate that PPCL achieves a 50% reduction in parameter count compared to the full model, with less than 3% degradation in key objective metrics. Notably, our method maintains high-quality image generation capabilities while achieving higher compression ratios, rendering it well-suited for resource-constrained environments.