Suppress llama.cpp __del__ errors during pre-load

coderai

@@ -3668,18 +3668,25 @@ def main():
                             model_path = download_model(model_to_use, cache_dir)
                             model_to_use = model_path
                     
-                    # Load with llama.cpp (Vulkan)
+                    # Load with llama.cpp (Vulkan) - suppress __del__ errors
+                    audio_model = None
+                    try:
                         audio_model = Llama(
                             model_path=model_to_use,
                             n_gpu_layers=-1,  # All layers to GPU
                             n_ctx=2048,
                             verbose=False
                         )
+                    except BaseException:
+                        audio_model = None
                     
+                    if audio_model is not None:
                         # Store in multi_model_manager
                         model_key = f"audio:{args.audio_model}"
                         multi_model_manager.add_model(model_key, audio_model)
                         print(f"Audio model loaded successfully (GGUF/Vulkan)")
+                    else:
+                        raise Exception("llama.cpp failed to load model")
                     
                 except Exception as e:
                     print(f"Warning: Could not pre-load audio model: {e}")