At the moment you would need to do this yourself. It would be possible to have additional preprocessing to accommodate this though. Feel free to add a feature r

At the moment you would need to do this yourself. It would be possible to have additional preprocessing to accommodate this though. Feel free to add a feature request here https://github.com/marqo-ai/marqo/issues https://github.com/marqo-ai/marqo/issues. The other consideration is that you would want the distribution of the content and queries to match what the selected model was trained on.