Reproducing WavLM results on speaker verification

microsoft / unilm

Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities

https://aka.ms/GeneralAI

MIT License

20.29k stars 2.56k forks source link

Reproducing WavLM results on speaker verification #1568

Open stevehuang52 opened 6 months ago

stevehuang52 commented 6 months ago

Hello,

I'm trying to reproduce the speaker verification results in the WavLM paper using the ECAPA-TDNN baseline, but cannot get close to the scores in the paper. Could you please provide more details (e.g., data processing/augmentation, optim params, scheduler, epochs, etc) on training the base ECAPA-TDNN model as well as the one with WavLM features so that I can try again?

Thanks, Steve