提出高效时间感知量化方法，让扩散Transformer模型更快更省电。
TQ-DiT: Efficient Time-Aware Quantization for Diffusion Transformers
- 分区域+时间分组量化，精准适配扩散模型特征
- W8A8时FID仅增0.29，W6A6性能优于现有方法
- 适合部署在资源受限的实时生成场景
扩散Transformer（DiTs）结合了Transformer架构与扩散模型，但其计算复杂度限制了实时应用和AI系统的可持续性。本文通过模型量化提升计算效率，将权重与激活值用低精度表示。提出多区域量化（MRQ），针对DiT模块中网络值分布不均问题，为子区域分配两个缩放参数；同时提出时间分组量化（TGQ），降低激活值随时间变化带来的量化误差。实验表明，所提算法在W8A8条件下，FID仅增加0.29，性能接近全精度模型；在W6A6下优于其他基线方法，验证了其在低比特量化中的适用性。结果表明该方法具备实现高效实时生成模型的潜力。
原文摘要 · Abstract (English)
Diffusion transformers (DiTs) combine transformer architectures with diffusion models. However, their computational complexity imposes significant limitations on real-time applications and sustainability of AI systems. In this study, we aim to enhance the computational efficiency through model quantization, which represents the weights and activation values with lower precision. Multi-region quantization (MRQ) is introduced to address the asymmetric distribution of network values in DiT blocks by allocating two scaling parameters to sub-regions. Additionally, time-grouping quantization (TGQ) is proposed to reduce quantization error caused by temporal variation in activations. The experimental results show that the proposed algorithm achieves performance comparable to the original full-precision model with only a 0.29 increase in FID at W8A8. Furthermore, it outperforms other baselines at W6A6, thereby confirming its suitability for low-bit quantization. These results highlight the potential of our method to enable efficient real-time generative models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。