T8 – Compression of Deep Learning Models for NLP