search

jaywalnut310 / vits

VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech
https://jaywalnut310.github.io/vits-demo/index.html
MIT License
6.48k stars 1.21k forks source link

[Question] How many iterations for the available pretrained model? #116

Open shivammehta25 opened 1 year ago

shivammehta25 commented 1 year ago

Hello,

Thank you for the fantastic work :D On loading the checkpoint I see that value for the iteration key is set to zero. How long and with what batch size was the provided pretrained model trained for?

image

nikich340 commented 1 year ago

"All models in the ablation study were trained up to 300k steps" From paper.

shivammehta25 commented 1 year ago

So the available checkpoint in the GitHub repository is also trained up to 300k steps?

nikich340 commented 1 year ago

So the available checkpoint in the GitHub repository is also trained up to 300k steps?

Must be so. Train dataset is 12.5k, so it's ~1500 iterations over the whole dataset (it may be more correct to reach this value instead of batch steps, which rely on your trainset size and batch size).