AttributeError: 'str' object has no attribute 'dim' on Offline model Segmentation

[MustafaCQN]

I wanted to run offline model because of online model is slow on big voice files (22 minutes of talk took 28 minutes to diarize). Then I found the segmentation api has an offline model and I downloaded that model for usage.

Whenever I am trying to run below script using offline model that installed here (pytorch_model.bin)

from pyannote.audio import Model
import os

PYANNOTE_API_KEY = os.environ.get('PYANNOTE_API_KEY')
offline_model = Model.from_pretrained('pytorch_model.bin', use_auth_token=PYANNOTE_API_KEY)

offline_model.to("cuda")

offline_segmentation = offline_model('loudness-fixed.wav')

I get this error

AttributeError                            Traceback (most recent call last)
Cell In[23], line 1
----> 1 offline_segmentation = offline_model('loudness-fixed.wav')

File [c:\Users\Musta\Desktop\stt-bluerise\SpeechToText\.venv\lib\site-packages\torch\nn\modules\module.py:1194](file:///C:/Users/Musta/Desktop/stt-bluerise/SpeechToText/.venv/lib/site-packages/torch/nn/modules/module.py:1194), in Module._call_impl(self, *input, **kwargs)
   1190 # If we don't have any hooks, we want to skip the rest of the logic in
   1191 # this function, and just call forward.
   1192 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
   1193         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1194     return forward_call(*input, **kwargs)
   1195 # Do not call functions when jit is used
   1196 full_backward_hooks, non_full_backward_hooks = [], []

File [c:\Users\Musta\Desktop\stt-bluerise\SpeechToText\.venv\lib\site-packages\pyannote\audio\models\segmentation\PyanNet.py:164](file:///C:/Users/Musta/Desktop/stt-bluerise/SpeechToText/.venv/lib/site-packages/pyannote/audio/models/segmentation/PyanNet.py:164), in PyanNet.forward(self, waveforms)
    152 def forward(self, waveforms: torch.Tensor) -> torch.Tensor:
    153     """Pass forward
    154 
    155     Parameters
   (...)
    161     scores : (batch, frame, classes)
    162     """
--> 164     outputs = self.sincnet(waveforms)
    166     if self.hparams.lstm["monolithic"]:
    167         outputs, _ = self.lstm(
...
--> 148     if input.dim() not in (2, 3):
    149         raise ValueError('expected 2D or 3D input (got {}D input)'
    150                          .format(input.dim()))

AttributeError: 'str' object has no attribute 'dim'

I have checked for similar errors but I couldn't find any. Can anybody help me on this?

[hbredin]

Models expect tensors as input.
If you do want to apply a model on the whole file, the Inference class can be used to wrap the model.

from pyannote.audio import Inference
inference = Inference(offline_model)
output = inference('loudness-fixed.wav')

Please read the documentation for more details and options.