Jun, 2024
ViDiT-Q: 图像和视频生成的扩散变压器的高效准确量化
ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation
Tianchen Zhao, Tongcheng Fang, Enshu Liu, Wan Rui, Widyadewi Soedarmadji...
TL;DRDiffusion transformers have challenges in quantization, but the proposed ViDiT-Q method achieves lossless W8A8 quantization and ViDiT-Q-MP achieves W4A8 with negligible visual quality degradation, resulting in memory optimization and latency speedup.