search

MiukkaZh / MGT

Learning Domain-Invariant Transformation for Speaker Verification.
8 stars 0 forks source link

epoch #3

Open wangsheng3 opened 1 year ago

wangsheng3 commented 1 year ago

How many epochs did you train to converge, and could you provide a model that you have trained well?

MiukkaZh commented 1 year ago

Please carefully review the config.yaml, where you will find some reference hyperparameters. A MGT model with gmlp architecture, which is trained on CN-Celeb, has been uploaded on CN-Celeb for reference.