saving memory

app.py

@@ -42,7 +42,11 @@ controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=t
 
 
 pipe = FluxControlNetInpaintPipeline.from_pretrained(base_model, controlnet=controlnet, torch_dtype=torch.bfloat16).to(device)
-
+pipe.to("cuda")
+torch.backends.cuda.matmul.allow_tf32 = True
+pipe.vae.enable_tiling()
+pipe.vae.enable_slicing()
+pipe.enable_model_cpu_offload() # for saving memory
 
 control_mode_ids = {
     "scribble_hed": 0,
@@ -490,4 +494,6 @@ with gr.Blocks() as demo:
             output_json_component
         ]
     )
-demo.queue().launch()
+
+demo.queue(api_open=False)
+demo.launch()