Update app.py

app.py

@@ -8,7 +8,7 @@ import re
 
 import torch
 from transformers import Qwen2VLForConditionalGeneration, GenerationConfig, AutoProcessor
-
+import spaces
 
 
 def extract_answer_content(text: str) -> str:
@@ -150,8 +150,12 @@ def scale_box(box, scale):
     sw, sh = scale
     return [int(box[0]*sw), int(box[1]*sh), int(box[2]*sw), int(box[3]*sh)]
 
-
+@spaces.GPU
 def generate_sgg(image):
+    device='cuda' if torch.cuda.is_available() else "cpu"
+    if next(model.parameters()).device != torch.device(device):
+        model = model.to(device)
+    
     iw, ih = image.size
     scale_factors = (iw / 1000.0, ih / 1000.0)