diff-sampler

diff-sampler is an open-source toolbox designed for fast sampling of diffusion-based generative models. It enables a fair comparison among existing approaches and helps researchers develp better algorithms. diff-sampler supports a wide range of model implementations, numerical solvers, time schedules, and other advanced features.

This repository also includes the official implementations of our following works:

• [arXiv 2023] A Geometric Perspective on Diffusion Models
• [CVPR 2024] Fast ODE-based Sampling for Diffusion Models in Around 5 Steps
• [ICML 2024] On the Trajectory Regularity of ODE-based Diffusion Sampling
• [NeurIPS 2024] Simple and Fast Distillation of Diffusion Models
• [JSTAT 2025] Geometric Regularity in Deterministic Sampling of Diffusion-based Generative Models

News

• 2025-11-25. The official implementation of Geometric Regularity in Deterministic Sampling of Diffusion-based Generative Models has been released in diff-analyzer-main. We provide extended experimental results and include new features, such as sampling with approximated Gaussian scores, automatic trajectory calibration and calculation of curvature and torsion for projected sampling trajectories.
• 2025-10-17. Our paper Geometric Regularity in Deterministic Sampling of Diffusion-based Generative Models got accepted for publication in J. Stat. Mech. 2025, a prestigious journal in statistical physics.
• 2024-11-13. The official implementation of the paper Simple and Fast Distillation of Diffusion Models was released in sfd-main. We simplify and accelerate the distillation of Diffusion Models. The proposed method accelerates the fine-tuning up to 1,000 times while preserving high-quality few-step image generation.
  
• 2024-10-01. Our new work Simple and Fast Distillation of Diffusion Models got accepted by NeurIPS 2024! We will release the code as soon as possible.
• 2024-09-12. We release diff-analyzer as an easy toolbox for analyzing the behavior of the sampling of diffusion models. We offer the code for reproducing some observations in our works A Geometric Perspective on Diffusion Models and On the Trajectory Regularity of ODE-based Diffusion Sampling.
  
• 2024-06-20. GITS has been integrated into ComfyUI as GITSScheduler! We provide a ComfyUI workflow to compare GITS with the baseline time schedule and the concurrent ICML 2024 paper Align Your Steps: Optimizing Sampling Schedules in Diffusion Models. More details are included in our pull request here.
  
• 2024-06-18. The official implementation of the paper On the Trajectory Regularity of ODE-based Diffusion Sampling was released in gits-main.
  • This new repo is built on the basis of diff-solvers-main, supporting baseline sampling as well as customized time schedule. Our method (GITS) offers a more efficient time schedule for fast sampling of diffusion models.
  • For text-to-image generation, we provide a colab script using Diffusers 🧨 where we also compare our methods with the concurrent work Align Your Steps. Below is a 10-step comparison.
• 2024-06-03. The repo has been made easier to use. The current pre-trained models will be automatically downloaded to ./src/dataset_name. Some errors and typos were fixed. Detailed running scripts are provided in launch.sh, where we also added new scripts for evaluation of CLIP score for Stable Diffusion.
• 2024-05-02. Our work On the Trajectory Regularity of ODE-based Diffusion Sampling got accepted by ICML 2024.
• 2024-04-05. Our work Fast ODE-based Sampling for Diffusion Models in Around 5 Steps was selected as a Highlight paper (Top 2.8%) by CVPR 2024!
• 2024-03-25. The official implementation of the paper Fast ODE-based Sampling for Diffusion Models in Around 5 Steps (amed-solver-main) and a toolbox (diff-solvers-main) for fast sampling of diffusion models is released. We include the reference statistics for FID evaluation here. Hope that this repo can facilitate researchers on fast sampling of diffusion models!
• 2024-02-27. Our work Fast ODE-based Sampling for Diffusion Models in Around 5 Steps got accepted by CVPR 2024.

Supported Fast Samplers for Diffusion Models

Question

If you have any questions regarding the implementations, please contact zhyzhou@zju.edu.cn, defchern@zju.edu.cn.

Citation

If you find this repository useful, please consider citing the following papers (reverse chronological order):

@article{chen2025geometric,
  title={Geometric Regularity in Deterministic Sampling of Diffusion-based Generative Models},
  author={Chen, Defang and Zhou, Zhenyu and Wang, Can and Lyu, Siwei},
  journal={arXiv preprint arXiv:2506.10177},
  year={2025}
}

@article{zhou2024simple,
  title={Simple and Fast Distillation of Diffusion Models},
  author={Zhou, Zhenyu and Chen, Defang and Wang, Can and Chen, Chun and Lyu, Siwei},
  journal={arXiv preprint arXiv:2409.19681},
  year={2024}
}

@article{chen2024trajectory,
  title={On the Trajectory Regularity of ODE-based Diffusion Sampling},
  author={Chen, Defang and Zhou, Zhenyu and Wang, Can and Shen, Chunhua and Lyu, Siwei},
  journal={arXiv preprint arXiv:2405.11326},
  year={2024}
}

@article{zhou2023fast,
  title={Fast ODE-based Sampling for Diffusion Models in Around 5 Steps},
  author={Zhou, Zhenyu and Chen, Defang and Wang, Can and Chen, Chun},
  journal={arXiv preprint arXiv:2312.00094},
  year={2023}
}

@article{chen2023geometric,
  title={A geometric perspective on diffusion models},
  author={Chen, Defang and Zhou, Zhenyu and Mei, Jian-Ping and Shen, Chunhua and Chen, Chun and Wang, Can},
  journal={arXiv preprint arXiv:2305.19947},
  year={2023}
}

Star History