https://uberduck.ai/ logo
Powered by
# machine-learning
  • s
    sure thingy
  • s
    i'll upload my adorabat dataset to drive in a little bit
  • i
    if you can leave things training at work/overnight then its better for you to train it, since i cant do those
  • s
    no wait i already did
  • s
    hmm
  • s
  • s
    also wouldn't letting it run for a long time be a bad idea?
  • i
    nah from my experiments it stops when its been trained enough
  • s
    aight
  • p
    no it doesnt
  • p
    it stops after the set amount of epochs have been reached
  • p
    which is 10000 as default
  • p
    so don't do that
  • i
    bruh i think my attempts stopped waayy before that, what happened
  • i
    also you know how long you should train it for pi?
  • p
    how do you mean that?
  • p
    you want some reading tips on overfitting?
  • i
    for how many iterations should you train a hifi-gan model
  • p
    as always: dont overdo it 🙂
  • p
    ?
  • i
    that too probably, my knowledge on overfitting needs to be bigger
  • p
    I assume the wikipedia article for Bias–variance tradeoff would be good intro
  • s
    zzzzz
  • s
    take 4
  • s
    fourth times the charm
  • s
    ?
  • s
    v1 was trained for too long v2 i forgot what went wrong v3 was glitchy as fuck v4 should HOPEFULLY be better somewhat
  • s
    i merged the short lines into longer ones and gave him a few new lines of dialogue
  • s
    hopefully we get our king going well this time
  • s
    it's forming something?
1...222223224...1068Latest