lhotse had problems preparing musan data and cannot produce supervisions data

[wangkaisine]

lhotse prepare musan download/musan data/manifests
WARNING:root:There are 15 recordings that do not have any corresponding supervisions in the SupervisionSet.

In data/manifests floder, there is only supervisions_music.json but no supervisions_noise.json，supervisions_speech.json.

[pkufool]

lhotse prepare musan download/musan data/manifests
WARNING:root:There are 15 recordings that do not have any corresponding supervisions in the SupervisionSet.

It's ok, just ignore this warning.

In data/manifests floder, there is only supervisions_music.json but no supervisions_noise.json，supervisions_speech.json.

There is no supervisions for noise & speech in musan dataset. Don't warry, it doesn't matter.

[csukuangfj]

In data/manifests floder, there is only supervisions_music.json but no supervisions_noise.json，supervisions_speech.json.

I think that is the expected behavior as the code in lhotse does not produce supervisions for noise and speech.
See the code
https://github.com/lhotse-speech/lhotse/blob/master/lhotse/recipes/musan.py#L70-L78

    if 'music' in parts:
        manifests['music'] = prepare_music(corpus_dir, use_vocals=use_vocals)
        validate_recordings_and_supervisions(**manifests['music'])
    if 'speech' in parts:
        manifests['speech'] = {'recordings': scan_recordings(corpus_dir / 'speech')}
        validate(manifests['speech']['recordings'])
    if 'noise' in parts:
        manifests['noise'] = {'recordings': scan_recordings(corpus_dir / 'noise')}
        validate(manifests['noise']['recordings'])

---

Some directories in noise and speech don't have ANNOTATIONS. Maybe @pzelasko has more to say
why not to generate supervisions_*.json for speech and noise.

(py38) fangjun:/ceph-fj/open-source/icefall4/egs/librispeech/ASR/download/musan$ find . -name ANNOTATIONS
./noise/free-sound/ANNOTATIONS
./music/jamendo/ANNOTATIONS
./music/fma-western-art/ANNOTATIONS
./music/fma/ANNOTATIONS
./music/rfm/ANNOTATIONS
./music/hd-classical/ANNOTATIONS
./speech/librivox/ANNOTATIONS
(py38) fangjun:/ceph-fj/open-source/icefall4/egs/librispeech/ASR/download/musan$ ls noise/
README  free-sound  sound-bible
(py38) fangjun:/ceph-fj/open-source/icefall4/egs/librispeech/ASR/download/musan$ ls speech/
README  librivox  us-gov
(py38) fangjun:/ceph-fj/open-source/icefall4/egs/librispeech/ASR/download/musan$ ls music/
README  fma  fma-western-art  hd-classical  jamendo  rfm

[pzelasko]

That’s right, IIRC there was no supervision data for noise and speech.