Have problem (compatibility issue) loading the audios

#1
by rimoKR - opened

After loading the dataset, when I try outputting dataset['train'][0]['audio']
I got this error,

ImportError: To support decoding audio data, please install 'torchcodec'.

After installing torchcodec and downgrading the torch version, I am now getting this error:

ImportError: cannot import name 'register_fake' from 'torch.library' (/usr/local/lib/python3.11/dist-packages/torch/library.py)

Can anyone provide the exact versions of torch and torchcodec needed to view the audios.

Sign up or log in to comment