Update app.py

app.py

@@ -117,10 +117,10 @@ pipe.to(device)
 
 # Enable CPU offload for the model.
 # turn on if you don't have multiple GPUs or enough GPU memory(such as H100) and it will cost more time in inference, it may also reduce the quality
-# pipe.enable_model_cpu_offload()
-# pipe.enable_sequential_cpu_offload()
-pipe.vae.enable_slicing()
-pipe.vae.enable_tiling()
+pipe.enable_model_cpu_offload()
+pipe.enable_sequential_cpu_offload()
+# pipe.vae.enable_slicing()
+# pipe.vae.enable_tiling()
 
 os.makedirs("./output", exist_ok=True)
 os.makedirs("./gradio_tmp", exist_ok=True)