search

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

Easily train a good VC model with voice data <= 10 mins!
MIT License
24.96k stars 3.65k forks source link

Feature Extract move to cuda no feature to do #1891

Open resawlin opened 8 months ago

resawlin commented 8 months ago

runtime\python.exe extract_feature_print.py cuda:0 1 0 0 C:\Mangio-RVC-v23.7.0/logs/cemadrian v2 ['extract_feature_print.py', 'cuda:0', '1', '0', '0', 'C:\Mangio-RVC-v23.7.0/logs/cemadrian', 'v2'] C:\Mangio-RVC-v23.7.0/logs/cemadrian load model(s) from hubert_base.pt move model to cuda no-feature-todo ['extract_f0_print.py', 'C:\Mangio-RVC-v23.7.0/logs/cemadrian', '8', 'rmvpe', '64'] no-f0-todo no-f0-todo no-f0-todo no-f0-todo no-f0-todo no-f0-todo no-f0-todo no-f0-todo ['extract_feature_print.py', 'cuda:0', '1', '0', '0', 'C:\Mangio-RVC-v23.7.0/logs/cemadrian', 'v2'] C:\Mangio-RVC-v23.7.0/logs/cemadrian load model(s) from hubert_base.pt move model to cuda no-feature-todo

thats the output i get when i hit the feature extraction button. preprocessing has no problems but this does.

Rekek9 commented 8 months ago

Me too XDDDD

Alidk1999 commented 6 months ago

I have this problem too

jazzmaestro88 commented 4 months ago

I had the same issue, before realizing that preprocess was not succeeding. I replaced audio.py file in infer/lib and preprocess.py in infer/modules/train with archived copies from an older build (AMD/Intel from a few months ago) and then everything was successful 🙌 Hope this helps!

dimpablo commented 4 months ago

I had the same issue, before realizing that preprocess was not succeeding. I replaced audio.py file in infer/lib and preprocess.py in infer/modules/train with archived copies from an older build (AMD/Intel from a few months ago) and then everything was successful 🙌 Hope this helps!

Maestro, plz send a link to download that correct audio.py you've used 🙏 I've tried previous stable version (link) and it helps a bit, thus I don't see " no-feature-todo" message anymore. But it does not save "*.index" file too.

Vana2793 commented 1 month ago

I'm using the RVC1006Nvidia build. Same problem here. No idea how to fix the fact that it's not preprocessing. Maybe need to try another build since jazzmaestro88 said they used a older replacement for the audio.py and preprocess.py. I will give that a shot and edit comment with findings.

Edit 1: Tried the RVC0813Nvidia. There is no infer folder on that build. And I tried the web ui. Same problem. Going to try another older build. Might even try taking the audio.py and preprocess from an amd build if they're compatible.

Edit 2: So I might have done something stupid and I want people to double check just in case, but the folder I was using as my data set was actually empty, as I forgot to put my sample in there and that's why preprocess wasn't working.

That said, if you're still having issues, you should be able to DL other versions of RVC or take the preprocess/audio.py's from other builds that use RVC (just be sure to back up any files you change). Before I checked my folder, I was going to borrow the preprocess and audio.py's from Jarod's build of ai-voice-cloning-3.0 as both files are in that build. If it's too annoying to DL the whole thing, I'd be happy to put both files on a google drive and share them if it could help anyone. Just let me know.

jazzmaestro88 commented 1 month ago

Here are the two files I mentioned that have been working for me. those 2 files.zip