https://uberduck.ai/ logo
Powered by
# machine-learning
  • h
    if it said it's out of memory then that means you need to lower your batch size
  • h
    -oh hm
  • g
    My learning rates are 2e-4 and 1e-5. Is that the problem?
  • h
    no it's
  • g
    I was wanting to try lower rates on this to see if that would make it work. In the past, it would come out only able to say "Grimlock" coherently.
  • h
    i think you might have a bad line in your list.txt
  • h
    because it's complaining about one of the files being a directory
  • g
    I thought I fixed that.
  • g
    @User Should I even be doing this when I don't know Python?
  • g
    I have to do this because nobody else will.
  • h
    i mean you can send the dataset to justinjohn for training
  • h
    or me
  • h
    -actually not me sorry
  • g
    He ignores me half the time.
  • h
    oh
  • g
    I'm the only one who cares enough about Transformers and has made models.
  • g
    Nobody else cares, so nobody else is too motivated to help.
  • g
    I've posted some datasets to FakeYou's dataset channel on their server, and people just ignore them.
  • b
    My New Model of Foldy, I used Noise Remover in Audacity but Still a bit of not Working
  • h
    that's way more background noise than should be there
  • h
    did you run the dataset through the vocal extractor
  • b
    Yeah
  • b
    I made Foldy Model better but She Speaks Gibberish but I need Shorter Lines
  • f
    ben be like
  • f
    idk when will i stop but i wanna continue because this happen with my sl bon bon dataset, wasn't good
  • u
    tbh
  • u
    dont even bother training a hifigan for a dataset so small
  • u
    itll 100% force undertrain
  • u
    use an existing one that might fit
  • h
    0 sec? that looks like your batch size is bigger than your wav count
1...751752753...1068Latest