How to use narrowband?

[johnchienbronci]

Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/mixed.py", line 285, in getattr) = self._assert_one_data_cut_with_attr_and_return_it_with_track_index(name)
File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/mixed.py", line 378, in
_assert_one_data_cut_with_attr_and_return_it_with_track_index assert len(non_padding_cuts_with_custom_attr) == 1, ( AssertionError: This MixedCut has 0 non-padding cuts with a custom attribute 'narrowband'. We currently don't support mixing custom attributes. Consider dropping the attribute on all but one of DataCuts. Problematic cut:

    if config.perturb_speed:
        cuts = CutSet.mux(
            cuts,
            cuts.perturb_speed(0.9),
            cuts.perturb_speed(1.1),
        )
    if config.perturb_narrow:
        cuts = CutSet.mux(
            cuts,
            cuts.narrowband(codec="lpc10", restore_orig_sr=True),
        )

[pzelasko]

I couldn't reproduce your issue with the following snippet:

from lhotse import *

cuts = CutSet.from_file("libri-train-5.jsonl.gz")

cuts = CutSet.mux(
    cuts,
    cuts.perturb_speed(0.9),
    cuts.perturb_speed(1.1),
)

cuts = CutSet.mux(
    cuts,
    cuts.narrowband(codec="mulaw", restore_orig_sr=True),
)

for c in cuts:
    c.load_audio()

Can you update lhotse to the latest version and try again, or share more details?

[johnchienbronci]

lhotse version use 1.29.0

set cuts = CutSet.mux( cuts, cuts.narrowband(codec="mulaw", restore_orig_sr=True)) encountered no issues, but an error occurs when training starts.

Maybe nemo training issue ?

the detail msg:

Error executing job with overrides: []
Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/mixed.py", line 300, in __getattr__
    ) = self._assert_one_data_cut_with_attr_and_return_it_with_track_index(name)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/mixed.py", line 401, in _assert_one_data_cut_with_attr_and_return_it_with_track_index
    assert len(non_padding_cuts_with_custom_attr) == 1, (
AssertionError: This MixedCut has 0 non-padding cuts with a custom attribute 'narrowband'. We currently don't support mixing custom attributes. Consider dropping the attribute on all but one of DataCuts. Problematic cut:
  MixedCut(id='4a84eb03-8d1f-d9b7-4d2b-9deb1beb3711_sp0.9_vp0.3', tracks=[MixTrack(cut=MonoCut(id='1759edc3-72ae-2244-8b01-63c1cd9d2b7d_sp0.9_vp0.3', start=0.0, duration=30.77775, channel=0, supervisions=[SupervisionSegment(id='/corpus/20250110_0392-0002.wav-000000-002770_sp0.9_vp0.3', recording_id='/corpus/202501/20250110/20250110_0392-0002.wav_sp0.9_vp0.3', start=0.0, duration=30.77775, channel=0, text='xxxxx', language=None, speaker=None, gender=None, custom={'tokens': array([ 996,    0,  996,    0,  996,    0,  996,    0,  996,    0,  996, 0,  996,    0,  996,    0,  765, 1009, 1008,  541,  689,  996, 0])}, alignment=None)], features=None, recording=Recording(id='/corpus/20250110_0392-0002.wav_sp0.9_vp0.3', sources=[AudioSource(type ='file', channels=[0], source='/corpus/20250110_0392-0002.wav')], sampling_rate=16000, num_samples=492444, duration=30.77775, channel_ids=[0], transforms=[Speed(factor=0.9), Volume(factor=0.3)]), custom={'text': 'xxxxx'}), type='MonoCut', offset=0.0, snr=None), MixTrack(cut =MonoCut(id='e005b860-51ef-1922-fe43-c49e149818d1_sp0.9_vp0.3', start=0.0, duration=3.4539375, channel=0, supervisions=[SupervisionSegment(id='noise-free-sound-0630_sp0.9_vp0.3', recording_id='noise-free-sound-0630_sp0.9_vp0.3', start=0.0, duration=3.4539375, channel=0, text='', language=None, speaker=None, gender=None, custom=None, alignment=None)], features=None, recording=Re cording(id='noise-free-sound-0630_sp0.9_vp0.3', sources=[AudioSource(type='file', channels=[0], source='/corpus/noise/free-sound/noise-free-sound-0630.wav')], sampling_rate=16000, num_samples=55263, duration=3.4539375, channel_ids=[0], transforms=[Speed(factor=0.9), Volume(factor=0.3)]), custom={'text': ''}), type='MonoCut', offset=0.0, snr=12.423998112834207), MixTrack(cut=MonoCut(id='7d41e602-eece-328b-ff7b-118e820865d6_sp0.9_vp0.3', start=35.5460625, duration=27.26825, channel=0, supervisions=[SupervisionSegment(id='music-fma-wa-0059_sp0.9_vp0.3', recording_id='music-fma-wa-0059_sp0.9_vp0.3', start=-35.5460625, duration=169.7088125, channel=0, text='', language=None, speaker=None, gender=None, custom=None, alignment=None)], features=None, recording=Recording(id='music-fma-wa-0059_sp0.9_vp0.3', sources=[AudioSource(type='file', channels=[0], source='/corpus/music/fma-western-art/music-fma-wa-0059.wav')], sampling_rate=16000, num_samples=2715341, duration=169.7088125, channel_ids=[0], transforms=[Speed(factor=0.9), Volume(factor=0.3)]), custom={'text': ''}), type='MonoCut', offset=3.4539375, snr=12.423998112834207)], transforms=None)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/trainer/call.py", line 43, in _call_and_handle_interrupt
    return trainer.strategy.launcher.launch(trainer_fn, *args, trainer=trainer, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/strategies/launchers/subprocess_script.py", line 105, in launch
    return function(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/trainer/trainer.py", line 579, in _fit_impl
    self._run(model, ckpt_path=ckpt_path)
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/trainer/trainer.py", line 986, in _run
    results = self._run_stage()
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/trainer/trainer.py", line 1030, in _run_stage
    self.fit_loop.run()
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/loops/fit_loop.py", line 197, in run
    self.setup_data()
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/loops/fit_loop.py", line 263, in setup_data
    iter(self._data_fetcher)  # creates the iterator inside the fetcher
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/loops/fetchers.py", line 104, in __iter__
    super().__iter__()
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/loops/fetchers.py", line 51, in __iter__
    self.iterator = iter(self.combined_loader)
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/utilities/combined_loader.py", line 351, in __iter__
    iter(iterator)
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/utilities/combined_loader.py", line 92, in __iter__
    super().__iter__()
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/utilities/combined_loader.py", line 43, in __iter__
    self.iterators = [iter(iterable) for iterable in self.iterables]
  File "/usr/local/lib/python3.10/dist-packages/pytorch_lightning/utilities/combined_loader.py", line 43, in <listcomp>
    self.iterators = [iter(iterable) for iterable in self.iterables]
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 484, in __iter__
    return self._get_iterator()
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 415, in _get_iterator
    return _MultiProcessingDataLoaderIter(self)
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 1191, in __init__
    self._reset(loader, first_iter=True)
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 1228, in _reset
    self._try_put_index()
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 1471, in _try_put_index
    index = self._next_index()
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 691, in _next_index
    return next(self._sampler_iter)  # may raise StopIteration
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/base.py", line 299, in __next__
    batch = self._next_batch()
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 299, in _next_batch
    batch = next(self.cuts_iter)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 588, in __iter__
    self._collect_cuts_in_buckets(self.buffer_size)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 768, in _collect_cuts_in_buckets
    cuts = next(self.cuts_iter)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic.py", line 377, in __iter__
    for item in self.iterator:
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 2710, in __iter__
    yield from self.cuts
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 2710, in __iter__
    yield from self.cuts
  File "/usr/local/lib/python3.10/dist-packages/lhotse/lazy.py", line 443, in __iter__
    item = next(selected)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 2710, in __iter__
    yield from self.cuts
  File "/usr/local/lib/python3.10/dist-packages/lhotse/lazy.py", line 695, in __iter__
    ans = self.fn(item)
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 1609, in <lambda>
    lambda cut: cut.narrowband(
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/mixed.py", line 303, in __getattr__
    raise AttributeError(
AttributeError: No such attribute: 'narrowband' (note: custom attributes are not supported when a MixedCut consists of more than one MonoCut with that attribute).

[pzelasko]

Which NeMo version is this and at which exact spot did you insert the mux with narrowband call? It'll help me nail down what's going on.

[johnchienbronci]

NeMo: v2.0.0

https://github.com/NVIDIA/NeMo/blob/v2.0.0/nemo/collections/common/data/lhotse/dataloader.py
line 244:
add code:

cuts = CutSet.mux(
    cuts,
    cuts.narrowband(codec="lpc10", restore_orig_sr=True),
)