Update app.py

app.py

@@ -203,7 +203,11 @@ def generate_music(genre_txt, lyrics_txt, max_new_tokens=5, run_n_segments=2, us
     decoded_instrumentals = codec_model.decode(torch.as_tensor(instrumentals.astype(np.int16), dtype=torch.long).unsqueeze(0).permute(1, 0, 2).to(device)).cpu().squeeze(0)
 
     mixed_audio = (decoded_vocals + decoded_instrumentals) / 2
-    return (16000, mixed_audio.detach().numpy())
+    mixed_audio_np = mixed_audio.detach().numpy()  # Convert to NumPy array
+    mixed_audio_int16 = (mixed_audio_np * 32767).astype(np.int16)  # Convert to int16
+    
+    # Return the sample rate and the converted audio data
+    return (16000, mixed_audio_int16)
 
 @spaces.GPU(duration=120)
 def infer(genre_txt_content, lyrics_txt_content, num_segments=2, max_new_tokens=10):