Update app.py

app.py

@@ -248,14 +248,16 @@ def generate_music(
     def convert_to_audio(codec_result, rescale):
         with torch.no_grad():
             decoded_waveform = codec_model.decode(
-                torch.as_tensor(codec_result.astype(np.int16), dtype=torch.long).unsqueeze(0).permute(1, 0, 2).to(
-                    device))
+                # Corrected line: Convert numpy array to PyTorch tensor with appropriate type
+                torch.as_tensor(codec_result, dtype=torch.int16).unsqueeze(0).permute(1, 0, 2).to(device)
+            )
         decoded_waveform = decoded_waveform.cpu().squeeze(0)
         
         limit = 0.99
         max_val = decoded_waveform.abs().max()
         scaled_waveform = decoded_waveform * min(limit / max_val, 1) if rescale else decoded_waveform.clamp(-limit, limit)
-        return (16000, (scaled_waveform * 32767).astype(np.int16))
+        # Corrected line: Convert to numpy array before casting to int16
+        return (16000, (scaled_waveform * 32767).detach().cpu().numpy().astype(np.int16))
 
     vocal_audio = convert_to_audio(vocals, rescale)
     instrumental_audio = convert_to_audio(instrumentals, rescale)