PrimaCora

0 Followers · 0 Following

• Joined on 2023-10-23

Repositories Projects Packages Public Activity Starred Repositories

PrimaCora commented on issue ecker/ai-voice-cloning#404 2023-10-23 21:23:19 +00:00

Deepspeed - Windows (Yes I know)

WSL will be the easiest way if it is an option. You will need to restart the tool and WSL every hour or two, but aside from that, it will work and a bit faster.

PrimaCora commented on issue ecker/ai-voice-cloning#423 2023-10-23 19:28:20 +00:00

20 min audio, 500 epochs, underwhelming result

Given the images, it looks like it only effectively trained for 100 steps. The default multistep has a steep drop. At epoch 100 you can see it approach (hit) zero. So it isn't learning anything.…