Error calling pipeline.from_pretrained() with AttributeError :'NoneType' object has no attribute 'eval'

[RyoTakahashi]

Hi all,

I tried speaker diarization on my own data with following code.

from pyannote.audio import Audio, Pipeline

pipeline = Pipeline.from_pretrained("pyannote/[email protected]", use_auth_token="<MY TOKEN>")
audio_file = "<MY WAV FILE>"
diarization = pipeline(audio_file)

However, it failed with the following error.

Could not download 'pyannote/segmentation' model.
It might be because the model is private or gated so make
sure to authenticate. Visit https://hf.co/settings/tokens to
create your access token and retry with:

   >>> Model.from_pretrained('pyannote/segmentation',
   ...                       use_auth_token=YOUR_AUTH_TOKEN)

If this still does not work, it might be because the model is gated:
visit https://hf.co/pyannote/segmentation to accept the user conditions.
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
[<ipython-input-5-d90017bc0af2>](https://localhost:8080/#) in <module>
      1 # load whisper model and pyannote speaker diarization model
      2 model = whisper.load_model("medium")
----> 3 pipeline = Pipeline.from_pretrained("pyannote/[email protected]", use_auth_token=hf_token)

2 frames
[/usr/local/lib/python3.8/dist-packages/pyannote/audio/pipelines/utils/getter.py](https://localhost:8080/#) in get_model(model, use_auth_token)
     87         )
     88 
---> 89     model.eval()
     90     return model
     91 

AttributeError: 'NoneType' object has no attribute 'eval'

My access token is valid and I use pyannote.audio-2.1.1.

Do you have any idea to fix it?

[hbredin]

Did you fill the user access form for pyannote/segmentation model?

It looks like you did for pyannote/speaker-diarization but not for pyannote/segmentation (which is used internally by pyannote/speaker-diarization).

[RyoTakahashi]

I tried it after authenticating https://hf.co/pyannote/segmentation and it worked.
I was not aware that checks were needed for pyannote/segmentation as well.

Thank you for your support!