nit

app.py

@@ -29,11 +29,15 @@ from qwenimage.models.qwen_fa3_processor import QwenDoubleStreamAttnProcessorFA3
 dtype = torch.bfloat16
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-exp = Qwen_FA3_AoT_int8()
-# exp = Qwen_FA3_AoT_fp8()
-exp.load()
-exp.optimize()
-pipe = exp.pipe
+@spaces.GPU
+def load_pipe():
+    exp = Qwen_FA3_AoT_int8()
+    # exp = Qwen_FA3_AoT_fp8()
+    exp.load()
+    exp.optimize()
+    return exp.pipe
+
+pipe = load_pipe()
 
 
 MAX_SEED = np.iinfo(np.int32).max