search

pyannote / pyannote-audio

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding
http://pyannote.github.io
MIT License
6.03k stars 758 forks source link

Issue with segmentation model #1040

Closed petersnias closed 1 year ago

petersnias commented 2 years ago

Hi. I noticed a change to the segmentation model: https://huggingface.co/pyannote/segmentation

Ever since then I'm getting the following error:

Traceback (most recent call last): File "test.py", line 3, in model = Model.from_pretrained("pyannote/segmentation") File "/home/npeters/.local/lib/python3.8/site-packages/pyannote/audio/core/model.py", line 804, in from_pretrained loaded_checkpoint = pl_load(path_for_pl, map_location=map_location) File "/opt/anaconda/lib/python3.8/site-packages/pytorch_lightning/utilities/cloud_io.py", line 38, in load return torch.load(f, map_location=map_location) File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 607, in load return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args) File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 882, in _load result = unpickler.load() File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 875, in find_class return super().find_class(mod_name, name) ModuleNotFoundError: No module named 'torch.torch_version'

Any suggestions?

acidcoma commented 2 years ago

same problem

hbredin commented 2 years ago

You may try to load the previous version using pyannote/segmentation@Interspeech2021 instead of pyannote/segmentation.

petersnias commented 2 years ago

Hi @hbredin

Based on the information on pyannote.audio github, v2.0 produces improved results over v1.1.

I noted that pyannote-audio/speaker_diarization for v2 uses

pyannote/segmentation@2022.07

Rolling pyannote/segmentation (or pyannote/segmentation@2022.07) back to pyannote/segmentation@Interspeech2021 works, but is the new segmentation model responsible (or partly responsible) for the improvement in v2.0? If so, how can we use it?

I tried and get still:

model = Model.from_pretrained("pyannote/segmentation@2022.07") File "/home/npeters/.local/lib/python3.8/site-packages/pyannote/audio/core/model.py", line 804, in from_pretrained loaded_checkpoint = pl_load(path_for_pl, map_location=map_location) File "/opt/anaconda/lib/python3.8/site-packages/pytorch_lightning/utilities/cloud_io.py", line 38, in load return torch.load(f, map_location=map_location) File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 607, in load return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args) File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 882, in _load result = unpickler.load() File "/opt/anaconda/lib/python3.8/site-packages/torch/serialization.py", line 875, in find_class return super().find_class(mod_name, name) ModuleNotFoundError: No module named 'torch.torch_version'

Thank you

hbredin commented 2 years ago

Would you mind sharing a link to a Google Colab containing a minimum reproducible example that I can quickly run to reproduce the error?

stale[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.