search

RVC-Project / Retrieval-based-Voice-Conversion-WebUI

Easily train a good VC model with voice data <= 10 mins!
MIT License
23.73k stars 3.52k forks source link

Training stuck at 1 epoch #1655

Closed Haxhom closed 5 months ago

Haxhom commented 9 months ago
WindowsTerminal_HJhqPtXU6e

I've been training 4 hours but its still on 0%...

This is the configuration:

explorer_VctcZa7cc6

The audio quality is a bit low but the dataset is around 20-30min long cutten into 9s files. My RAM is 4GB..

How to fix this?

H46K3RL0L commented 9 months ago

Before reading please note I am not an expert, so do not take anything I tell you to do as fact. First of all, you need more ram, I have 16GB, and it works fairly well. What is your GPU? you need a good gpu to use this. I have an RTX 4070 laptop gpu, which has 8GB Vram and it takes about a minute per half hour. This gpu is 14.7 Teraflops (How many trillions of floating point operations per second.) Also, try increasing batch size per gpu. Default is 4, but for me putting it between 9 and 11 gets the highest speed.

Haxhom commented 9 months ago

I have 4GB of ram 💀

github-actions[bot] commented 5 months ago

This issue was closed because it has been inactive for 15 days since being marked as stale.