#Should I continue training this model?

1 messages · Page 1 of 1 (latest)

novel ocean
#

I can't post images in #📑│making-models so this is my next best bet.

This is at 500 epochs right now. My dataset is about >30 minutes long using the original pretrained set at 32 hop and 48k. My mel and kl have been still going down, but I'm scared of overtraining. What should I do to advance?

I'm also using RVC disconnect is that helps.

topaz hearth
#

I would keep training to see what it does. The graph should look a little smoother than that, rather than steps, but it doesn't look like its dying I think

It's helpful if you can share all the graphs on the tensorboard in your screenshot