nvm found it!

ITS WORKINGGGG

im going to be here a while hahahaha

so while this is doing its thing, the video said i want to train it until the number is 0.15 right?

that hasnt changed?

yeah though recommendations vary

quoting dawn: > sub-0.14 for ridiculously large datasets (like 30 minutes or more), sub-0.10 for 150+ wav datasets, sub-0.09 for datasets between 100-30, sub-0.07 for 30 below

alright, so something around 0.1 would be good i think

got about 14 minutes worth over 255 files

getting there!!!!!

basically any time i feel like turning it off now it should be good, woohoo~!

gonna leave it running more anyway, i dont need to turn it off or anything

What does the graph look like?

can i train my own HiFi-GAN for Uberduck Pipeline? i found notebook but its for tacotrone

I think it was mentioned earlier in the channel by [@]hecko that Hifigan doesn't work with the pipeline iirc

however i will attempt to make it work with the pipeline

just got back from going out sorry gjkklhfghlgk

is that okay?

Is it always like this? It’s supposed to look diagonal

it had one iteration where it was all diagonal

but i was out of the house then

okay actually theres been more nice diagonal ones

Yeah, it should be good when its done

a lot of diagonals followed by long purple lines though

read somewhere thats just from shorter samples?

Yeah

okay so thats fine then yeah

i'd say at least 95% of graphs are good solid diagonals ignoring the long end bits

theres some breaks in the steps but not many

oooh we hit 0.10!!!!