Use automatic-speech-recognition pipeline as baseline, add zero gpu inference. 8ecac59 verified wetdog commited on about 1 month ago