https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • h

    HoodyP

    05/18/2021, 6:25 PM
    Wait can you use a Google Notebook on Kaggle?, cuz I’m not experienced in coding
  • m

    mega b

    05/18/2021, 6:26 PM
    Yes, you can but not any google drive features. but you can download your finished model and upload it locally
  • h

    HoodyP

    05/18/2021, 6:27 PM
    Oh...
  • m

    mega b

    05/18/2021, 6:27 PM
    there is a cool google colab RDP that i found, it has alot of storage and is free to use
  • m

    mega b

    05/18/2021, 6:27 PM
    i use that to transfer large files into my drive
  • h

    HoodyP

    05/18/2021, 6:28 PM
    Oh, cool!
  • m

    mega b

    05/18/2021, 6:28 PM
    one sec
  • m

    mega b

    05/18/2021, 6:28 PM
    https://github.com/PradyumnaKrishna/Colab-Hacks
  • m

    mega b

    05/18/2021, 6:28 PM
    https://colab.research.google.com/github/PradyumnaKrishna/Colab-Hacks/blob/master/Colab%20RDP/Colab%20RDP.ipynb
  • h

    HoodyP

    05/18/2021, 6:31 PM
    I’ll check that out soon, anyways, maybe I could just use another account to continue training my model (since I have more accounts), or will Colab recognise it and limit my resources even more?
  • h

    HoodyP

    05/18/2021, 6:37 PM
    Okay, I’m gonna use another account... eventhough its tedious, I guess I’ll just have to do what I gotta do lol
  • b

    brilliam.de

    05/18/2021, 6:51 PM
    testing \from my new mrbeast model
  • h

    HoodyP

    05/18/2021, 7:07 PM
    Update, I’m using another account to continue training and the GPU I’m using now (P100) is actually have a slightly more fast drastic change in validation loss compared to T4!
  • h

    HoodyP

    05/18/2021, 7:10 PM
    Bc, when the “spike” happened, it went back to it’s original lowest validation point in just under 10 epochs compared to T4, which usually takes at least around 15-20 epochs to get back in track
  • m

    mega b

    05/18/2021, 11:16 PM
    "Artificial noisy data" in style tokens!
  • m

    mega b

    05/18/2021, 11:16 PM
    that is super cool
  • p

    pi

    05/18/2021, 11:17 PM
    explain?
  • p

    pi

    05/18/2021, 11:17 PM
    or link?
  • m

    mega b

    05/18/2021, 11:17 PM
    https://google.github.io/tacotron/publications/global_style_tokens/
  • m

    mega b

    05/18/2021, 11:17 PM
    bottom
  • m

    mega b

    05/18/2021, 11:24 PM
    https://github.com/syang1993/gst-tacotron
  • m

    mega b

    05/18/2021, 11:25 PM
    and
  • m

    mega b

    05/18/2021, 11:25 PM
    https://github.com/jinhan/tacotron2-gst
  • m

    mega b

    05/18/2021, 11:25 PM
    community implementations
  • g

    GamingGuy1133

    05/19/2021, 3:40 AM
    Is gst tacotron tacotron 1?
  • p

    pi

    05/19/2021, 3:58 AM
    no
  • u

    user

    05/19/2021, 4:00 AM
    yall
  • u

    user

    05/19/2021, 4:00 AM
    i did it
  • u

    user

    05/19/2021, 4:00 AM
    model doesn't sound like ass
  • u

    user

    05/19/2021, 4:01 AM
    That's from an interview I did with a family member last year so you won't recognize the voice but this is the first time it hasn't sounded windy and noisy and had lots of pauses
1...979899...1068Latest