search

yl4579 / StarGANv2-VC

StarGANv2-VC: A Diverse, Unsupervised, Non-parallel Framework for Natural-Sounding Voice Conversion
MIT License
479 stars 107 forks source link

F0 model training on 160 mels #55

Closed ashfaquekhalid closed 2 years ago

ashfaquekhalid commented 2 years ago
Screenshot 2022-06-27 at 6 04 17 PM

I am training the F0 model on 160 mels with the pre-trained f0 model that you provided as a base model. training data is Librispeech-clean-100. is the model overfitting only in 10-20 epochs?

ashfaquekhalid commented 2 years ago

fixed

superhg commented 2 years ago

could you share your solution to this overfiting issue? I met same issue @ashfaquekhalid

MMMMichaelzhang commented 2 years ago

@ashfaquekhalid I met same issue

joan126 commented 2 years ago

@yl4579 could you help to see this issue?

yl4579 commented 2 years ago

@joan126 @MMMMichaelzhang @superhg see https://github.com/yl4579/PitchExtractor/issues/4