I'm surprised first order model isn't talked about more on here. Probably the easiest way to create deepfakes from single images.

Oh my sweet summer child

I dont understand the analogy lol

During training in colab it disconnected .so i try to reconnect it. Bt it shows "connect without GPU".any solution?

The analogy xD

You don't get 1:1 vram but in what I was reading with ppl doing non-DFL deep learning, they were getting like equiv 11 gigs by putting two 8 gig cards together

Yeah, I feel like Windows is the cumshot directly into only one eye and Linux is a 50 man bukakke party in terms of vram

Or rather all of it, unlike in Windows.

You should have a bit more VRAM available from what I heard in Linux.

(native)

Gonna use a spare drive to install it

I'm downloading ubuntu right now. Don't even have another 2070S but hey, I'll at least see I can actually figure out how to run DFL

Are you running Linux natively or in a VM?

Want to try running DFL on linux to see if I could put two 2070 Super's together and have more working vram, which...apparently, depending....works on Linux even though it doesn't on windows.

Any linux users on?

Don't think it's needed though.

Looks like you can adjust batch. internal\DeepFaceLab\models\Model_XSeg : #override def on_initialize_options(self): self.set_batch_size(4)

I just left some xSeg training overnight for the first time, results weren't bad

an xSeg FAQ sounds like a good idea if there isnt one