This notebook can be used to generate section level embeddings using huggingface's sentence-transformer. This task to move this code into a python package. The goal is to enable faster iteration when working with embeddings, including
- using different granularity of wikitext
- using different models to generate embeddings
- scaling computation of embeddings in wmf infrastructure
- sharing code between projects making use of embeddings
This is one already well-defined task of a larger effort to create tooling and infrastructure to work with embeddings, which is in planning.