So, i wanted to upload a image generation checkpoint, but i don’t know how to do it correctly so i can use it with the diffusers package.
Do I need to set up that folder structure containing the tokenizer, text_encoders, scheduler,… folders? Cuz when i tried just uploading the .safetensors file like it’s done when we’re uploading loras, it didn’t work on colab…
The official method is to convert to the Diffusers format, but there is no problem with uploading the un-converted file as it is.
In that case, use from_single_file to load it from Diffusers. Note that if the model is not available in its entirety, as in the case of FLUX, the procedure is slightly different.