Inference speed mechanism or Model size compression ?

[Xiuyu-Li]

Hi, at the moment, only simulated quantization is implemented in this repository for FID evaluation. We will be releasing a significantly updated arXiv paper that includes comprehensive data on model sizes.

[continue-revolution]

Do you plan to build an end-to-end quantization for stable diffusion? If yes, I want to know the approximate time.

[Xiuyu-Li]

@continue-revolution Hi Chengsong,

End-to-end quantization is on our roadmap, and we're aiming to release it in the coming months. Please stay tuned for updates, and thanks for your interest in the project! We also welcome contributions from the open-source community to help expedite this process.

[continue-revolution]

@Xiuyu-Li Glad to contribute in some way. Have sent an email to your berkeley address.

[Jack47]

can't wait to see "a significantly updated arXiv paper that includes comprehensive data on model sizes". cheers~

[sravanthOppo27]

When can we expect the code for Calibration ?

[lucasjinreal]

Any date we expected the speedup with quantization?