Tail-Aware HiFloat4: W4A4 Post-Training Quantization for Wan2.2
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917533986062336 |
|---|---|
| author | Feng, Zhanfeng Guo, Shuai Di, Xin Peng, Long Cao, Yang Zha, Zhengjun |
| author_facet | Feng, Zhanfeng Guo, Shuai Di, Xin Peng, Long Cao, Yang Zha, Zhengjun |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_26628 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Tail-Aware HiFloat4: W4A4 Post-Training Quantization for Wan2.2 Feng, Zhanfeng Guo, Shuai Di, Xin Peng, Long Cao, Yang Zha, Zhengjun Artificial Intelligence 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. |
| title | Tail-Aware HiFloat4: W4A4 Post-Training Quantization for Wan2.2 |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.26628 |