LMFResearchSociety

/

CheckpointArchive

like 101

Follow

Literally Me FRFR Research Society 187

Hey I found this repo is this it?
https://huggingface.co/stablediffusionapi/linkedin-diffusion

note: it's in Diffusers format

All the txt2img repos with linkedin in their names

• https://huggingface.co/stablediffusionapi/linkedin-diffusion
• https://huggingface.co/SahilBhoite/linkedin_women
• https://huggingface.co/caraxe5025/linkedin
  • https://huggingface.co/akshaysabale07169/linkedin
  • https://huggingface.co/dcrey7/linkedin

Thanks, yeah, looks like I need to learn how to convert diffusers back into a single checkpoint file, the solutions I've tried only convert each individual file into safetensors, which isn't useful without a way to join text encoder+unet+VAE into a file like it was before being converted to diffusers. Would love to see a huggingface space that could do that.

Thanks, yeah, looks like I need to learn how to convert diffusers back into a single checkpoint file, the solutions I've tried only convert each individual file into safetensors, which isn't useful without a way to join text encoder+unet+VAE into a file like it was before being converted to diffusers. Would love to see a huggingface space that could do that.

That's pretty easy.
https://gist.github.com/jachiam/8a5c0b607e38fcc585168b90c686eb05

Thanks for leading in the right direction! I wonder if a space like this one https://huggingface.co/spaces/Yntec/sd-to-diffusers-backup could be created where the diffusers files are included and the output files open a PR so they could be added to a Model Card in the same way checkpoint files are turned into diffusers versions.
That would be the dream as I don't know what to do with script files and don't see a place in your linked code to place the path of the diffusers files and such, and a space like this would allow everyone to do it (there are many other models that don't have a checkpoint file that can be download anymore but they have their diffusers version) so it may be more worthwhile for me to figure out how to build such a space than just running this script for this model.

It's done! @John6666 created this space https://huggingface.co/spaces/John6666/convert_repo_to_safetensors_sd that can be used to convert these kind of models back to safetensors to rescue them from oblivion! So this model has been saved and it's now live here for everyone to enjoy!:

https://huggingface.co/Yntec/lnkdn

Thanks for all the help.