code

app.py

@@ -29,7 +29,9 @@ else:
 def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict]):
    
     output = pipe({"audio": audio[1], "turns": conversation, "sampling_rate": audio[0]},
-                  max_new_tokens=512)
+                  max_new_tokens=512,
+                  device=0)
+    print("output", output)
 
     conversation.append({"role": "user", "content": output["transcription"]})
     conversation.append({"role": "assistant", "content": output["reply"]})