Efficiency

Lightening of models by design
How to lighten NN-based NLP models by design?