Efficient Contextualized Representation: Language Model Pruning for Sequence Labeling - View it on GitHub