lhotse-speech / lhotse

Tools for handling speech data in machine learning projects.
https://lhotse.readthedocs.io/en/latest/
Apache License 2.0
956 stars 219 forks source link

Features not including original recording_id when computing #1411

Closed njellinas closed 3 weeks ago

njellinas commented 4 weeks ago

I run the command: compute_and_store_features on a CutSet and the resulting cuts contain features that do not include the corresponding recording_id and thus I cannot decompose them and save them separately because they do not contain recording information.

pzelasko commented 3 weeks ago

Since you know that cut == recording in your case, you can fix that manually before decomposition with

for cut in cuts:
  cut.features.recording_id = cut.recording.id
njellinas commented 3 weeks ago

Ok, I thought that something like this would be automatic but it's OK.