Thank you for the detailed open-source work! Would it be possible to add the inference code for visual understanding?