GitLab

Adds an implementation of Emformer, a memory-efficient transformer architecture 
introduced in https://ieeexplore.ieee.org/document/9414560 that targets low-latency 
streaming speech recognition applications.