Update app.py

app.py

@@ -15,6 +15,7 @@ else:
     torch_dtype = torch.float32
 
 pipe = pipeline(task="automatic-speech-recognition", model=model_repo_id, device=device)
+pipe.model.generation_config.forced_decoder_ids = pipe.processor.get_decoder_prompt_ids(language="zh", task="transcribe")
 
 # @spaces.GPU #[uncomment to use ZeroGPU]
 def infer(
@@ -25,7 +26,7 @@ def infer(
         wav = wav.mean(axis=1)
     wav = wav.astype(np.float32)
     wav /= np.max(np.abs(wav))
-    return pipe(wav, generate_kwargs={'language': 'chinese'})['text']
+    return pipe(wav)['text']
 
 css = """
 #col-container {