Reduce GPU duration to 30s (saves quota)

app.py

@@ -63,7 +63,7 @@ async def homepage():
         return f.read()
 
 
-@spaces.GPU
+@spaces.GPU(duration=30)
 def run_inference(prompt, width, height, seed):
     """Run the diffusion pipeline on GPU."""
     generator = torch.Generator("cuda").manual_seed(seed)