On the Frictionless Data call today, I bumped a topic that’s been at the top of my mind for a few months: support for the Apertus LLM in ODE. The gist of it is this: we need tangible methods and good tools to address the data literacy gap, and having a safe environment to do that is an important goal of the Open Data Editor, which has native support for Data Packages.
There is already an existing feature to download an LLM and use it to infer the schema or generate other parts of the dataset. In the Pull Request I’ve proposed changing from the Llama 3.2 to the Apertus 1.0 model. Apertus is the project of the team I work with at the Swiss AI Initiative to create a fully open model, with a transparent and reproducible data pipeline. You can read more about it here: https://apertus-ai.org
Since opening the PR, we’ve released smaller versions of Apertus, which fit even better the use case of being able to use ODE on a decent (but not necessarily very powerful) laptop. See our latest blog post here: APERTVS.ai
I would like to in general get your feedback about having a collaboration around this, as I think there are several areas where the Apertus project would also benefit from better support for Frictionless Data, and use cases for ODE could be a great starting point.
Thanks for your feedback!