Fix device mismatch: load models on same device as GPU computation

app.py

@@ -35,8 +35,8 @@ def load_models_and_components(model_name: str):
     """Load and cache models, tokenizer, vocoder, and feature extractor."""
     global _models_cache, _tokenizer_cache, _vocoder_cache, _feature_extractor_cache
 
-    # Set device (CPU for Spaces)
-    device = torch.device("cpu")
+    # Set device (GPU if available for Spaces GPU acceleration)
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
     if model_name not in _models_cache:
         print(f"Loading {model_name} model...")