trying to fix flash-attn issue when using cuda

app.py

@@ -6,7 +6,6 @@
 import gradio as gr
 import cv2
 import tempfile
-from ultralytics import YOLO
 import spaces, subprocess
 
 # Install flash attention, skipping CUDA build if necessary
@@ -16,6 +15,8 @@ subprocess.run(
     shell=True,
 )
 
+from ultralytics import YOLO
+
 @spaces.GPU
 def yolov12_inference(image, video, model_id, image_size, conf_threshold):
     model = YOLO(model_id)