search

ashawkey / stable-dreamfusion

Text-to-3D & Image-to-3D & Mesh Exportation with NeRF + Diffusion.
Apache License 2.0
8k stars 710 forks source link

why is first stage autoencoder model scaled differently #14

Closed nikjetchev closed 1 year ago

nikjetchev commented 1 year ago

https://github.com/ashawkey/stable-dreamfusion/blob/a0c430803192fc8fb5e96b1cd94200601bc16441/nerf/sd.py#L139 Why do we have the 0.18215 constant there? This would make the current repository different than the standard way to use StableDiffusion? Can you please explain the motivation for that? Thanks a lot and keep-up the good work!

ashawkey commented 1 year ago

@nikjetchev Hi, I simply followed diffusers' example here.

nikjetchev commented 1 year ago

thanks, i found the reference code, it is all correct.