Common errors (and how to fix them)

  1. OSError: [Errno 24] Too many open files
    This is an error that pops up when speakers are probably not properly specified. If there are not enough speakers, and the dataset is large enough, Mac and Linux have issues. Make sure the number of speakers reported when setting up the corpus matches the number of speakers that should be there.

<unk> words in the output

See the files oovs_found.txt and utterance_oovs.txt in the output directory for a list of all out of vocabulary words and a list of utterances with out of vocabulary words in those utterances.