search

sarulab-speech / UTMOS22

UT-Sarulab MOS prediction system using SSL models
MIT License
169 stars 12 forks source link

Develop/stacking #10

Closed hyama5 closed 1 year ago

hyama5 commented 2 years ago

9

Takaaki-Saeki commented 2 years ago

Thanks! Can wav2vec large2 and xlsr work with the fairseq version described in the environmental setting?

hyama5 commented 1 year ago

Yes. This was solved by the following modification on Apr. https://github.com/sarulab-speech/fairseq/commit/9028a19131e3d8f4b27f4e13ece4cb0b678ff7ff But I did not modify the run script in that time.