Block Sparse Matrices for Smaller and Faster Language ModelsBy Hugging Face - Blog / September 10, 2020