This notebook contains a special demo. It highlights a new version of GliNER, which historically mainly tried to focus on the named entitiy recognition usecase, but now also supports structured information parsing as well as classification. Think of it as a very lightweight LLM that can still do zero-shot tasks for NLP, but has the benefit of running locally and on a CPU.
This notebook also highlights the flexibility of marimo. This notebook uses pydantic in a clever way so that we can run this one notebook as an interactive webapp, an API or a command-line utility. Neat!
{ "text": ""}