针对文生视频模型优化低比特量化，提升精度与稳定性。
Tail-Aware HiFloat4: W4A4 Post-Training Quantization for Wan2.2
- 采用感知尾部激活的校准策略构建通道掩码
- 在保持原有算子不变的前提下实现W4A4量化
- 适合部署在支持HiFloat4算子的边缘设备上
本文介绍Tail-Aware HiFloat4，我们提交至低比特文生视频生成量化挑战的方法。该方法将公开的ViDiT-Q后训练量化流程适配至Wan2.2模型，采用HiFloat4数值格式。对Wan2.2中变压器模块的主要线性层进行W4A4 HiFloat4伪量化，保留数值敏感的边界模块为高精度，并引入激活尾部感知的百分位校准模块用于通道掩码构建。结合紧凑的PTQ状态恢复机制，该设计在不改变运行时HiFloat4算术和采样流水线的前提下，有效降低罕见校准异常值的影响。
原文摘要 · Abstract (English)
This report describes Tail-Aware HiFloat4, our submission to the low-bit text-to-video generation quantization challenge. Our method adapts the public ViDiT-Q post-training quantization pipeline to Wan2.2 under the HiFloat4 numerical format. We quantize the main linear layers in both Wan2.2 transformer modules with W4A4 HiFloat4 fake quantization, keep numerically sensitive boundary modules in high precision, and introduce an activation-tail-aware percentile calibration module for channel-mask construction. Together with compact PTQ-state restoration, this design reduces the influence of rare calibration outliers while keeping the runtime HiFloat4 arithmetic and sampling pipeline unchanged.
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。