gpu request tryout

app.py

@@ -230,6 +230,7 @@ def run_inference(word, image, mapping):
     )
 
     # Start generation in a separate thread with torch.no_grad()
+    @spaces.GPU
     def generate_with_no_grad():
         with torch.no_grad():
             model.generate(**generation_kwargs)
@@ -331,10 +332,5 @@ with gr.Blocks() as demo:
         [run_button, loading_indicator, word_input, submit_button, encoded_word_display],
     )
 
-
-@spaces.GPU
-def main():
-    demo.launch()
-
 if __name__ == "__main__":
-    main()
+    demo.launch()