Assume you have a large CLIP model that you want to remove from VRAM (without unloading it into low system RAM) before loading your diffusion model to avoid OOM errors: Connect the output from the ...