nunchaku-ai/nunchaku-sana

<p align="center" style="border-radius: 10px"> <img src="https://huggingface.co/datasets/nunchaku-tech/cdn/resolve/main/nunchaku/assets/nunchaku_v2.png" width="30%" alt="Nunchaku Logo"/> </p>

<div align="center"> <a href=https://discord.gg/Wk6PnwX9Sm target="blank"><img src=https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fdiscord.com%2Fapi%2Finvites%2FWk6PnwX9Sm%3Fwithcounts%3Dtrue&query=%24.approximatemembercount&logo=discord&logoColor=white&label=Discord&color=green&suffix=%20total height=22px></a> <a href=https://huggingface.co/datasets/nunchaku-tech/cdn/resolve/main/nunchaku/assets/wechat.jpg target="_blank"><img src=https://img.shields.io/badge/WeChat-07C160?logo=wechat&logoColor=white height=22px></a> </div>

Model Card for nunchaku-sana

This repository contains Nunchaku-quantized versions of SANA-1.6B, designed to generate high-quality images from text prompts. It is optimized for efficient inference while maintaining minimal loss in performance.

Model Details

Model Description

• —Developed by: Nunchaku Team
• —Model type: text-to-image
• —License: NVIDIA License
• —Quantized from model: Sana_1600M_1024px

Model Files

• —`svdq-int4_r32-sana1.6b.safetensors`: SVDQuant quantized INT4 SANA-1.6B model. For users with non-Blackwell GPUs (pre-50-series).

Model Sources

• —Inference Engine: nunchaku
• —Quantization Library: deepcompressor
• —Paper: SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models
• —Demo: demo.nunchaku.tech

Usage

See sana1.6b.py.

Performance

Citation

@inproceedings{
  li2024svdquant,
  title={SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models},
  author={Li*, Muyang and Lin*, Yujun and Zhang*, Zhekai and Cai, Tianle and Li, Xiuyu and Guo, Junxian and Xie, Enze and Meng, Chenlin and Zhu, Jun-Yan and Han, Song},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025}
}
@article{
  xie2024sana,
  title={Sana: Efficient high-resolution image synthesis with linear diffusion transformers},
  author={Xie, Enze and Chen, Junsong and Chen, Junyu and Cai, Han and Tang, Haotian and Lin, Yujun and Zhang, Zhekai and Li, Muyang and Zhu, Ligeng and Lu, Yao and others},
  journal={arXiv preprint arXiv:2410.10629},
  year={2024}
}