You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
My comment was for all of us datasets team, as a reminder: when making a PR, but also when reviewing some other's PR, we should not forget to update the corresponding docstring and doc pages. It is something we can improve if we help each other in reminding about it... 🤗
Our installation documentation page says that one can use Datasets for mp3 only with
torchaudio<0.12
.torchaudio>0.12
is actually supported too but requires a specific version of ffmpeg which is not easily installed on all linux versions but there is a custom ubuntu repo for it, we have insctructions in the code: https://github.com/huggingface/datasets/blob/main/src/datasets/features/audio.py#L327So we should update the doc page. But first investigate this issue.
The text was updated successfully, but these errors were encountered: