Closed liuzhaoo closed 3 years ago
I successfully calculated the d-vector and used it for uisrnn, but the loss during training was nan. I checked and found that some d-vectors have a value of 0. Is this normal?
This is related to the other repo.
See this https://github.com/CorentinJ/Real-Time-Voice-Cloning/pull/678
I successfully calculated the d-vector and used it for uisrnn, but the loss during training was nan. I checked and found that some d-vectors have a value of 0. Is this normal?