Integrating Persephone into a production ML pipeline

[babat00nday]

I am trying to integrate models built with Persephone into my ML pipeline. I started by developing models using the corpus.Corpus(...) and experiment.train_ready(...) functions. Is there an API available for me to apply the preprocessing steps in corpus.Corpus(...) to a single wav file and then use something similar to experiment.transcribe(...) to get the predicted model transcription for that same wav file?

cc: @oadams

[oadams]

Hi, did you face issues with this approach?

[babat00nday]

Yes I did. I was able to add a flag to support a custom behavior of loading a corpus with no train/test/valid data only untranscribed data for inference.

[shuttle1987]

@toonday were you able to make progress? It would be good to know what you tried and what issues you encountered.