https://uberduck.ai/ logo
Powered by
# machine-learning
  • k
    To see the pipeline
  • h
    the default model is already trained on ljspeech
  • h
    so you won't see a difference
  • h
    if you wanna get a hang of it i'd recommend using one of the datasets from #835647732453605376
  • h
    or i can convert a pony dataset for you
  • k
    How do you set default mode
  • h
    nono like
  • h
    the ai doesn't start from scratch
  • h
    it takes a model that was already trained for weeks on ljspeech and tells it "ok you know english cool now be twilight sparkle"
  • k
    oh shit. Now I see . Maybe please get me a ponny dataset
  • h
    mmmactually heck i forgot i'm running training rn so i can't really
  • h
    i was gonna use my training notebook that converts the things, then zip that up and send it to you so you can use it in the normal one
  • h
    but now that i think about it that doesn't super make sense
  • h
    since mine's similar enough that you might as well use it directly
  • h
    idk if any of that made sense
  • k
    Well it did not. Maybe the first part' coz I kno training takes a long time
  • k
    So ya
  • h
    basically i took the notebook that you and many other people are using, and i added Stuff to it
  • h
    for example a pony dataset loader that converts from the 4chan format to ours
  • h
    my plan was to use it to convert a dataset, then download that, and upload it for you to use in the normal notebook
  • h
    but you already know how to load datasets so that wouldn't really teach you anything
  • h
    so instead you could just use my version of the notebook to convert the dataset yourself, and then also use said notebook to train
  • k
    oh alright I see. Let me switch google colab accounts. Just gotten session timed out since I guess my GPU usage for the day has maxed out.
  • h
  • h
    for the dataset link let's say uhhh princess cadance (don't download, but rather paste in the my notebook)
  • h
    small dataset so it'll train quickly
  • k
    Where exactly do I paste the link
  • k
    Thank you
  • h
    step 2, 
    dataset_source
  • k
    Ohh thanks let me run it now
1...425426427...1068Latest