Features : - add more datasets ( VCTK + LibriSpeech, TEDLIUM release 2). Total # of sentence is now 240,612. ( Previously 36,395 ) - split dataset into train, valid and test to check validation/test loss. - add simple speed variation augmentation - support multiple GPU training - support Docker image
Refactored : - adapted to tensorflow 1.0.0