You can do this if you select the `pipe.enable_model_cpu_offload()` option. See this https://huggingface.co/stabilityai/stable-diffusion-xl-base-...