safetensors split into six sefatensors
The safetensor files are now split into 6, how do I use them on my workflow in Windows?
Thanks
You need to use a proper diffusers-native loader.
I'll ssume you're using Comfy. Comfy community is, for unknown reasons, weirdly stuck on using single giant files and copying them into special folders inside Comfy, which also often means you have to have a whole entire copy of the model if you use it in more than one application, or requires setting up symlinks. It's a nightmare unfortunately.
Diffusers loader nodes do exist, and the best part is all you need to do is type the mode "account_name/model_name" and it just downloads it if it isn't downloaded already, and then you're just done. No downloading from some site, cutting and pasting it into the right folder or guessing where it is supposed to be.
Diffusers uses a common Huggingface cache directory to download files locally, which means all your applications can share model files instead of having tons of copies all over your disks or having to setup a bunch of symlinks like is typical with Comfy/Forge/etc. You can set the cache directory with an environment variable, which generally you'd only do once on one computer and it works forever.
Lines 116-132 in the subject2video.py file of our GitHub project demonstrate how to load multiple safetensor models.