Update app.py

app.py

@@ -8,6 +8,7 @@ import json
 import base64
 import soundfile as sf
 import torch
+import spaces
 
 from supertonic import TTS
 from vllm import LLM, SamplingParams
@@ -38,6 +39,7 @@ def get_base64_image(image):
     img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
     return f"data:image/jpeg;base64,{img_str}"
 
+@spaces.GPU
 def extract_vocabulary(pdf_text, images, translit_lang, translit_format, target_lang):
     """Use vLLM to extract vocabulary from text and images."""
     global llm, sampling_params
@@ -116,6 +118,7 @@ def numpy_to_base64_audio(wav, sample_rate):
     audio_base64 = base64.b64encode(buffer.read()).decode('utf-8')
     return f"data:audio/wav;base64,{audio_base64}"
 
+@spaces.GPU
 def process_pdf(pdf_file, translit_lang, translit_format, target_lang):
     global tts, voice_style
     
@@ -453,7 +456,8 @@ def create_demo():
         )
     return demo
 
-if __name__ == "__main__":
+@spaces.GPU
+def main():
     print("Loading Qwen3.5-2B model via vLLM...")
     llm = LLM(
         model="Qwen/Qwen3.5-2B",
@@ -485,3 +489,7 @@ if __name__ == "__main__":
         
     demo = create_demo()
     demo.launch(server_name="0.0.0.0", server_port=7861)
+    
+
+if __name__ == "__main__":
+    main()