Created by: rmalav15

Hi,

According to my best understanding, the demo_server provided with does not implement any of the improvement discussed in section[7] (Deployment) of the DeepSpeech2 paper, right?

I wanted to know, are the discussed deployment improvements i.e Batch Dispatch, half-precision matrix-matrix multiply kernel, Improved beam-search, row-convolution unidirectional model provided with this Repo? If yes, can you please share the details?

Thanks, Ram