# machine-learning
- m
mega b
04/02/2022, 7:53 PMwhat did you freeze? - h
hecko
04/02/2022, 7:54 PMearlier everything but speaker_embedding.weight - h
hecko
04/02/2022, 7:54 PMnow trying to freeze everything but the pictured ones - m
mega b
04/02/2022, 7:54 PMhuh - m
mega b
04/02/2022, 7:54 PMstrange - m
mega b
04/02/2022, 7:54 PMlemme know how it turns out - h
hecko
04/02/2022, 7:58 PMthis feels weird - h
hecko
04/02/2022, 7:59 PMbut without it it'll yell about it not being iterable - h
hecko
04/02/2022, 7:59 PMi mean it's also wrong but - m
mega b
04/02/2022, 8:00 PMdoes thing[1].requires_grad(True) work? - h
hecko
04/02/2022, 8:01 PMoh is that how it's done - m
mega b
04/02/2022, 8:01 PMuh i think = true does the same thing too - m
mega b
04/02/2022, 8:01 PMidk - h
hecko
04/02/2022, 8:02 PMwell it did a different thing - h
hecko
04/02/2022, 8:02 PMvs previous (i had removed the thing[0] by then) - h
hecko
04/02/2022, 8:02 PMi don't know what i'm doing - m
mega b
04/02/2022, 8:02 PMhmm - m
mega b
04/02/2022, 8:03 PMme neither - h
hecko
04/02/2022, 8:05 PMooh think i'm getting somewhere - h
hecko
04/02/2022, 8:06 PMno error this time - h
hecko
04/02/2022, 8:06 PMbut will the loss drop - h
hecko
04/02/2022, 8:08 PMnot sure but it will do this [cw loud] - m
mega b
04/02/2022, 8:09 PMuhh - m
mega b
04/02/2022, 8:09 PMyeah - h
hecko
04/02/2022, 8:10 PMif nothing else i've found a new way to summon demons - h
hecko
04/02/2022, 8:11 PMmaybe removing speaker embedding was a mistake yeah let's train both - h
hecko
04/02/2022, 8:21 PMwell loss is going down but - m
mega b
04/02/2022, 8:21 PMhmmm - h
hecko
04/02/2022, 8:21 PMclearly that means i'm messing with too many parameters on a one-wav dataset but that's not gonna stop me - h
hecko
04/02/2022, 8:21 PM,,ok it might stop me