benchmark
benchmark copied to clipboard
Use full stable_diffusion model instead of just text_encoder and unet separately