Auto-Tuning Hyperparameters with Optuna and PyTorch

Поделиться
HTML-код
  • Опубликовано: 19 ноя 2024

Комментарии • 28

  • @Stay.Strong.Keep.Moving
    @Stay.Strong.Keep.Moving 3 месяца назад +2

    One of the best explanations I have been able to find. Thank you for your time and effort!

  • @Jack-dx7qb
    @Jack-dx7qb 2 года назад +10

    This seems to be an incredible tool for ML practitioners. I can't wait to start using it!!!

  • @stephennfernandes
    @stephennfernandes 3 года назад +10

    Amazing video , very clearly explained optuna !

    • @PyTorch
      @PyTorch  3 года назад +3

      Glad it was helpful!

  • @jonathansum9084
    @jonathansum9084 4 года назад +2

    Yeah! We are no longer need to do fine-tuning.
    I used an Adma with 1e-3 with the same setting but I was beaten by an SGD 1e-4 with the same setting.
    That notebook was the Pytorch official fast R-CNN fine-tuning notebook.

  • @boira_dani
    @boira_dani 2 года назад +1

    This tool seems to be incredible, i will be sure to include it in my next project, thanks :)

  • @SwapnilBhatta07
    @SwapnilBhatta07 Год назад +1

    This is great, a lot easier to use than hyperopt.

  • @hajvklkaj6462
    @hajvklkaj6462 3 года назад +3

    Where to find gaussian sampler?

  • @TrueDebendra
    @TrueDebendra 2 месяца назад

    Great explanation

  • @ItsRowen
    @ItsRowen 3 года назад +1

    thank you, very nice presentation

  • @newbiejailer8675
    @newbiejailer8675 4 года назад +3

    want to know how to implement the progress bar at 12:45, it looks pretty cool

    • @offchan
      @offchan 3 года назад +3

      use library called tqdm and then set progress bar mode to ascii

  • @devashishkprasad
    @devashishkprasad 3 года назад

    Great tool ! Thank you for sharing

  • @msughakighir2795
    @msughakighir2795 2 месяца назад +1

    Appreciated

  • @DSee-e1s
    @DSee-e1s 3 месяца назад

    Where can i find quality code examples?

  • @eugeniures7593
    @eugeniures7593 3 года назад

    Excellent video!

  • @angelsancheese
    @angelsancheese 4 года назад

    Thank you for the video!

  • @Mohammadmohammad-jp2gx
    @Mohammadmohammad-jp2gx 4 года назад

    a great useful framework thank u

  • @rodrigocardoso7846
    @rodrigocardoso7846 4 года назад

    Awesome video and for thx to show this tool!

  • @kr.sheelvardhanbanty9136
    @kr.sheelvardhanbanty9136 6 месяцев назад

    The provided link to the source code is not available.

  • @sambitmukherjee1713
    @sambitmukherjee1713 Год назад

    Brilliant.

  • @LFP2024-w7m
    @LFP2024-w7m 3 года назад

    Amazing!

  • @luiztauffer8513
    @luiztauffer8513 3 года назад

    would be awesome to have this integrated with MLFlow!

  • @zhaobryan4441
    @zhaobryan4441 Год назад

    could you share the slides?

  • @sphereron
    @sphereron 4 года назад +1

    Windows pip build seems to be broken, I had to install with conda

    • @UnbenutzerKanalname
      @UnbenutzerKanalname 4 года назад

      that is very common unfortunately, do you have C++ compilers installed on Win?

  • @neighboroldwang
    @neighboroldwang 3 года назад

    14:37 Never though learning rate is so important

  • @ezrachua1317
    @ezrachua1317 Год назад

    11:10 twice as fast WITH tuning.