A variant of the all-MiniLM-L6-v2 embedding model that exposes attention weights for interpretability.
No lifecycle dates recorded.