try to fix shape

api.py

@@ -215,14 +215,15 @@ async def process_audio(audio_data: bytes, language: str = "auto") -> str:
             resampler = torchaudio.transforms.Resample(sample_rate, 16000)
             input_wav = resampler(torch.from_numpy(input_wav)[None, :])[0, :].numpy()
         
-        print(input_wav.shape)
 
         target_length = 90 * 16000 
         current_length = input_wav.shape[1]
         if current_length < target_length:
             padding_length = target_length - current_length
             padding = np.zeros((1, padding_length), dtype=np.float32)
-            input_wav = np.concatenate((input_wav, padding))
+            print(input_wav.shape)
+            print(padding.shape)
+            input_wav = np.concatenate((input_wav, padding), axis=0)
 
         # Model inference
         text = model.generate(