Refine GPU Resource Allocation for YOLOv11 Inference

**Summary:** This pull request optimizes GPU resource allocation in the SAHI + YOLOv11 demo by removing the duration parameter from the @spaces.GPU decorators and explicitly setting the device to cuda:0 in the load_yolo_model function.

app.py

@@ -41,7 +41,7 @@ sahi.utils.file.download_from_url(
 # Global model variable
 model = None
 
-@spaces.GPU(duration=60)
+@spaces.GPU()
 def load_yolo_model(model_name, confidence_threshold=0.5):
     """
     Loads a YOLOv11 detection model.
@@ -56,12 +56,12 @@ def load_yolo_model(model_name, confidence_threshold=0.5):
     global model
     model_path = model_name
     model = AutoDetectionModel.from_pretrained(
-        model_type="ultralytics", model_path=model_path, device=None, # auto device selection
+        model_type="ultralytics", model_path=model_path, device='cuda:0', 
         confidence_threshold=confidence_threshold, image_size=IMAGE_SIZE
     )
     return model
 
-@spaces.GPU(duration=60)
+@spaces.GPU()
 def sahi_yolo_inference(
     image,
     yolo_model_name,