final version

api.py

@@ -203,26 +203,20 @@ async def process_audio(audio_data: bytes, language: str = "auto") -> str:
         audio_buffer = BytesIO(audio_data)
         waveform, sample_rate = torchaudio.load(audio_buffer)
         
-        print(1, waveform.shape)
-
         # Convert to mono channel
         if waveform.shape[0] > 1:
             waveform = waveform.mean(dim=0)
         else:
             waveform = np.squeeze(waveform)
-        print(2, waveform.shape)
         
         # Convert to numpy array and normalize
         input_wav = waveform.numpy().astype(np.float32)
-        print(3, input_wav.shape)
         
         # Resample to 16kHz if needed
         if sample_rate != 16000:
             resampler = torchaudio.transforms.Resample(sample_rate, 16000)
             input_wav = resampler(torch.from_numpy(input_wav)[None, :])[0, :].numpy()
         
-        print(4, input_wav.shape)
-
         # Model inference
         text = model.generate(
             input=input_wav,