search

Forced-Alignment-and-Vowel-Extraction / fave-asr

Interface for automated transcription and time alignment of conversational interview data
https://forced-alignment-and-vowel-extraction.github.io/fave-asr/
GNU General Public License v3.0
3 stars 0 forks source link

Restore MacOS support #1

Closed chrisbrickhouse closed 5 months ago

chrisbrickhouse commented 11 months ago

In our chain of dependencies openai-whisper depends pyannote-audio depends onnxruntime-gpu>=1.16.0

onnxruntime-gpu does not support MacOS (issue). This dependency appears to have been added to pyannote-audio in the 3.0.1 release (issue).

For the moment we will need to drop plans for MacOS support. The short-term resolution is to pin pyannote-audio==3.0.0 but this is not ideal given the potential conflicts as other dependencies bump versions. The long-term solution would be upstream changes, but we have no control over what upstream does.

For the moment I will chalk this up as code debt given that I have no need to run this on Mac, but it will be a hurdle when we move towards an initial release.

JoFrhwld commented 10 months ago

fyi, it looks like pyannote has an open PR about this: https://github.com/pyannote/pyannote-audio/pull/1540

JoFrhwld commented 10 months ago

And, it's been released! https://github.com/pyannote/pyannote-audio/releases/tag/3.1.0