Fetching the paper…

QuartzNet: Deep Automatic Speech Recognition with 1D Time-Channel Separable Convolutions · Around