Recent Projects

SpecAugment (In Pytorch) 2019 | Collaboration with qcai2002 at FastAI 2019 A Pytorch implementation of Google Brain’s SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition. As far as I know, this was the first and so far only implementation of this paper in Pytorch. It required porting Tensorflow functionality to Pytorch, including writing many difficult functions from scratch since Pytorch did not support many necessary operations. Blog post