Research into sequence predictions using neural networks holds the key to major advances in speech recognition, machine translation, and language modeling. However, building and training such models is computationally taxing and very expensive. For instance, recent advances in language modeling