diff --git a/lpmc/music_captioning/datasets/msd.py b/lpmc/music_captioning/datasets/msd.py
index cfc03ed..887d5ae 100644
--- a/lpmc/music_captioning/datasets/msd.py
+++ b/lpmc/music_captioning/datasets/msd.py
@@ -46,7 +46,6 @@ def load_audio(self, audio_path, file_type):
             audio = pad
         random_idx = random.randint(0, audio.shape[-1]-self.n_samples)
         audio_tensor = torch.from_numpy(np.array(audio[random_idx:random_idx+self.n_samples]).astype('float32'))
-        audio_tensor = torch.randn(16000*10)
         return audio_tensor
 
     def __getitem__(self, index):