Execute ONNX models with QNN Execution Provider