Uberduck

if they wanted it to be private, they could just you know

(but don’t actually this makes uvr 1000 times easier)

<@281505163485642752> <@563902837931638784> making some changes to uberduck-ml-dev that might break backwards compatibility (not for the models themselves but just for running them) - have you pinned the version in your applications?  Is anyone else running uberduck-ml-dev that you know of?

I'm using a fork of it, is it something related to torchmoji?

fork should be fine - <@327116758714417153> you may want to add the version pins to the current version - although I think the change should be really simple (replacing _input with *_input in inferences)

Thanks for the heads up, I'll double check my compatibility when I get home

I will check what happens when I try to use it with the new commit, thanks for the heads up too Sam

Theres a good chance that this will break distributed and fp16 training as well - I'm going to try training a notebook before merging but that will just check the main use case not every potential thing. Not thinking anyone is actually using the distributed or fp16 mechanisms anyway...

but I've got an error, I saw the ncll isn't compatible with Windows, then I changed to gloo, but after that I've received this error:

Do you happen to know what It could be? I tried to figure it out with no luck

https://vxtwitter.com/TedsLittleDream/status/1562836627885416449

fuck kinda voice modulator can make you sound like that and not like a 12 year old thinking he sounds like anonymous

clothing physics will never look that real

makes sense, i also noticed they apparently fed the dall-e 2 images to deepfacelab with no consideration that the images are all different results and not consistent enough at all

bye is down chicks isnt under complimentary