Reproduction: Sparser, Faster, Lighter Transformer Language Models