Boundary-Protection W8A8 HiFloat8 Quantization for Large-Scale Text-to-Video Diffusion Transformers 事件

Boundary-Protection W8A8 HiFloat8 Quantization for Large-Scale Text-to-Video Diffusion Transformers arXiv:2606.00957v1 Announce Type: new Abstract: We present a post-training quantization (PTQ) approach for Wan2.1-T2V-14B, a 14-billion-parameter text-to-video diffusion transformer, targeting the W8A8 HiFloat8 (HiF8) format on Ascend 910B NPUs. A central challenge in quantizing video DiT models is the heterogeneous activation distribution across transformer blocks: boundary blocks (the first and