Skip to content
Xiaoyu Xu edited this page Jul 23, 2024 · 107 revisions

OneDiff Community and Feedback

Status

Performance Report in English

felixsanz

https://www.felixsanz.dev/articles/ultimate-guide-to-optimizing-stable-diffusion-xl

(onediff) "improves the visual quality of the result, almost halves inference time and the only penalty is a small wait at compilation time. What a great job!"

Omniinfer

https://medium.com/@admin_13769/stable-diffusion-accelerated-up-to-211-2-202b36ac8090

(onediff) "achieves a relative speedup of 211.2% compared to Xformers on RTX 3090, and a speedup of 205.6% on RTX 4090."

Performance Report in Chinese

Feedbacks

Online Playground

Adoptions of OneDiff Community Edition

OneDiff Community Edition is trusted by the pioneering AIGC Tech organizations/companies to power their production:

30+ more AIGC organizations/companies have deployed the Community Edition and established technical connections with OneDiff team.

Contributors

OneDiff is trying to provide real-time inference for generative vision models, thus enabling high-quality models from research environments to be deployed in production environments.

Thanks to our contributors:

Special Interest Groups(SIGs) and Related Issues

All community issues:

  • SIG1: algorithms and algorithms dev tools
    • Model IDEs
      • HF diffusers (onediffx)
      • ComfyUI
      • SD Web UI
    • Quantization and Enterprise Edition
    • Quality evaluation
    • Optimization algorithms
    • New algorithm
  • SIG2: kernel&runtime, compiler and build&ci
    • kernel and runtime
    • compiler
    • build&ci
  • SIG3: release, deployment, communication
Clone this wiki locally