A training technique that divides sequences into chunks and processes them in parallel to speed up computation.