SpecAugment w/torchaudio

Hi all,

I just published a Pytorch implementation of Google’s new data augmentation technique in SpecAugment with torch audio. Thought it might be of interest to some people working on audio in the forum.

To implement one of the transforms we ported a Tensorflow function sparse_image_warp to Pytorch. FWIW this was my deep dive into Pytorch and I found the experience really enjoyable with good docs and APIs that were easy to match with their TF counterparts :slight_smile: