castorini / howl

Wake word detection modeling toolkit for Firefox Voice, supporting open datasets like Speech Commands and Common Voice.
Mozilla Public License 2.0
201 stars 30 forks source link

Optimize the process of computing dataset statistics #96

Closed ljj7975 closed 3 years ago

ljj7975 commented 3 years ago

when there are ~ 1500 samples, compute_dataset_statistics were taking about 8 mins.

Using multiprocessing & passing sample metadata instead of audio data to worker, the process only takes about 1.5 mins