why its larger then dev file which is 12b
#3
by Ai11Ali - opened
look like have included the text encoder in main file too .
contains the text encoders/CLIP built in.
whats best way to strip them out dont want to waste disk space i've got the gguf's for the encoders.
whats best way to strip them out dont want to waste disk space i've got the gguf's for the encoders.
To get the unet out of the file is very simple:
Just put sd3.5_large_fp8_scaled.safetensors into "checkpoints" folder and load like in the image, left side node. When you run that, the unet without text encoder will be saved into "output/diffusion_models" folder.