---
license: mit
library_name: diffusers
tags:
- text-to-image
- stable-diffusion
- diffusion distillation
---

![image/png](https://cdn-uploads.huggingface.co/production/uploads/63943c882b9483beb473ec25/f8ws6nGK2ZkPEiizha2t9.png)

> [**Distribution Backtracking Builds A Faster Convergence Trajectory for One-step Diffusion Distillation**](https://github.com/SYZhang0805/DisBack),            
> *[Shengyuan Zhang](https://github.com/SYZhang0805)<sup>1</sup>, [Ling Yang](https://github.com/YangLing0818)<sup>2</sup>, [Zejian Li*](https://zejianli.github.io/)<sup>1</sup>, An Zhao<sup>1</sup>, Chenye Meng<sup>1</sup>, Changyuan Yang<sup>3</sup>, Guang Yang<sup>3</sup>, Zhiyuan Yang<sup>3</sup>, [Lingyun Sun](https://person.zju.edu.cn/sly)<sup>1</sup>*  
> <sup>1</sup>Zhejiang University <sup>2</sup>Peking University <sup>3</sup>Alibaba Group*
> 
## Contact 

Feel free to contact us if you have any questions about the paper!

Shengyuan Zhang [zhangshengyuan@zju.edu.cn](mailto:zhangshengyuan@zju.edu.cn)

## Usage

For one-step text-to-image generation, DisBack can use the standard diffuser pipeline:

```python
import torch
from diffusers import DiffusionPipeline, UNet2DConditionModel, LCMScheduler
from huggingface_hub import hf_hub_download

base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "SYZhang0805/DisBack"
ckpt_name = "SDXL_DisBack.bin"

unet = UNet2DConditionModel.from_config(base_model_id, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(torch.load(hf_hub_download(repo_name, ckpt_name), map_location="cuda"))

pipe = DiffusionPipeline.from_pretrained(base_model_id, unet=unet, torch_dtype=torch.float16, use_safetensors=True, variant="fp16").to("cuda")
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
prompt="A photo of a dog." 
image=pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, timesteps=[399], height=1024, width=1024).images[0]
image.save('output.png', 'PNG')
```

For more details, please refer to our [github repository](https://github.com/SYZhang0805/DisBack)

## License

DisBack is released under [MIT license](https://choosealicense.com/licenses/mit/)

## Citation
If you find our paper useful or relevant to your research, please kindly cite our papers:
```bib
@article{zhang2024distributionbacktrackingbuildsfaster,
    title={Distribution Backtracking Builds A Faster Convergence Trajectory for One-step Diffusion Distillation}, 
    author={Shengyuan Zhang and Ling Yang and Zejian Li and An Zhao and Chenye Meng and Changyuan Yang and Guang Yang and Zhiyuan Yang and Lingyun Sun},
    journal={arXiv 2408.15991},
    year={2024}
}
```

Paper: https://huggingface.co/papers/2408.15991

## Credits

DisBack is highly built on the following amazing open-source projects:

[DMD2](https://tianweiy.github.io/dmd2/): Improved Distribution Matching Distillation for Fast Image Synthesis

[Diff-Instruct](https://github.com/pkulwj1994/diff_instruct/tree/main): Diff-Instruct: A Universal Approach for Transferring Knowledge From Pre-trained Diffusion Models

[ScoreGAN](https://github.com/White-Link/gpm): Unifying GANs and Score-Based Diffusion as Generative Particle Models

Thanks to the maintainers of these projects for their contribution to this project!