Distilled model for Instructed image editing

fast-codi / CoDi

[CVPR24] CoDi: Conditional Diffusion Distillation for Higher-Fidelity and Faster Image Generation

https://fast-codi.github.io/

87 stars 2 forks source link

Distilled model for Instructed image editing #5

Open llCurious opened 2 months ago

llCurious commented 2 months ago

hi @MKFMIKU . Nice work!

I want to have a try on the 1-step text-conditioned image editing. Would you mind share the trained model checkpoints?

MKFMIKU commented 2 months ago

Hi @llCurious , Thanks for your interests. Unfortunately, this model is trained with internal data and we cannot share it publicly. However, it should be easy to replicate this by modifying our code on the instruct pix2pix data.

llCurious commented 2 months ago

Thx for your quick response!

Regarding the practice

replicate this by modifying our code on the instruct pix2pix data

do you mean runing the Training CoDi on HuggingFace Data script, with the dataset on HF (timbrooks/instructpix2pix-clip-filtered)?

Do i need to modify some code? If so, do you mind point the required code modification logic?