torch: TypeError: len() of a 0-d tensor

[asquare]

I'm consistently getting the error below, even for small audio files. Seems like this is due to a change in pytorch as mentioned at [1] and [2].

Thanks!

[1] pytorch/vision#2344
[2] https://fixexception.com/torch/len-of-a-0-d-tensor/

The environment is -

OS: Ubuntu 20.04.5 LTS
Torch: 1.13.0
Python: 3.10.0

    input = prepare_model_input(audio, device=device)
  File ".../.cache/torch/hub/snakers4_silero-models_master/src/silero/utils.py", line 42, in prepare_model_input
    max_seqlength = max(max([len(_) for _ in batch]), 12800)
  File ".../.cache/torch/hub/snakers4_silero-models_master/src/silero/utils.py", line 42, in <listcomp>
    max_seqlength = max(max([len(_) for _ in batch]), 12800)
  File ".../lib/python3.10/site-packages/torch/_tensor.py", line 925, in __iter__
    return iter(self.unbind(0))
TypeError: len() of a 0-d tensor

[asquare]

Sorry, this was user error... I was calling prepare_model_input with the output of read_audio instead of read_batch.