Update app.py

app.py

@@ -49,7 +49,7 @@ def process_audio(microphone, state, task="transcribe"):
     with torch.no_grad():
         vicuna_input = vicuna_tokenizer(prompt, return_tensors="pt").to('cuda')
         vicuna_output = vicuna_model.generate(**vicuna_input, max_new_tokens=192)
-        vicuna_response = vicuna_tokenizer.decode(vicuna_output, skip_special_tokens=True)
+        vicuna_response = vicuna_tokenizer.decode(vicuna_output[0], skip_special_tokens=True) # Access the first sequence [0]
         vicuna_response = vicuna_response.replace(prompt, "").strip()
     updated_state = state + "\n" + vicuna_response