Update app.py

app.py

@@ -70,7 +70,8 @@ def run_inference_localization(
     messages_for_template: List[dict[str, Any]], 
     pil_image_for_processing: Image.Image
 ) -> str:
-    # torch.cuda.set_device("cuda")
+    model.to("cuda")
+    torch.cuda.set_device(0)
     """
     Runs inference using the Holo1 model.
     - messages_for_template: The prompt structure, potentially including the PIL image object