search

hayeong0 / DDDM-VC

Official Pytorch Implementation for "DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion" (AAAI 2024)
https://hayeong0.github.io/DDDM-VC-demo/
160 stars 18 forks source link

Model is not converging even after 50+ epochs! #11

Closed Ashigarg123 closed 2 months ago

hayeong0 commented 3 months ago

I have trained a model for over 200 epochs with a batch size of 64, using the LibriTTS-100 and 360 datasets. From my experience with training diffusion-based VC models, I have observed that sufficient training is necessary, and the LR, depending on the batch size, also affects the convergence of the training.

Ashigarg123 commented 2 months ago

Thank you, it performs well. I was looking at the loss printed for each epoch. However, results after performing inference with a saved checkpoint looks good.