⚡ Supercharge your Training with PyTorch Lightning + Weights & Biases

Поділитися
Вставка
  • Опубліковано 26 лис 2024

КОМЕНТАРІ • 15

  • @WeightsBiases
    @WeightsBiases  4 роки тому

    Thanks to everyone who showed up to watch the premiere!

  • @dddd8947
    @dddd8947 2 роки тому +1

    Great video!
    The integration explanation is good as well as the Lightning explenation

  • @caylasharp3755
    @caylasharp3755 4 роки тому +2

    I get those goosebumps every time, yeah

  • @geeky_explorer9105
    @geeky_explorer9105 3 роки тому

    I was searching for such kind 9f things since 2 weeks, and Stanford slides brough me here, m too small to judge this great great video, just God bless you..., you guys made my day

  • @giannisdimaridis2920
    @giannisdimaridis2920 3 роки тому +2

    This is an amazingly informative video. You made my day, thank you!

  • @user-or7ji5hv8y
    @user-or7ji5hv8y 3 роки тому +1

    These type of lessons are awesome

  • @vulnerablegrowth3774
    @vulnerablegrowth3774 3 роки тому

    I used the custom callback, but it didn't show up in wandb? I only have a few of the charts from the training run. Is there something I should be looking for if it's not on the project page?

  • @ayushthakur736
    @ayushthakur736 4 роки тому

    Great introduction Charles. 🎇

  • @johannahoppe5544
    @johannahoppe5544 2 роки тому

    Hi, in what step are you usually logging your prediction images ? In the train_step or validation_step ?

    • @WeightsBiases
      @WeightsBiases  2 роки тому

      It's best to log after the validation_step. LogPredictionsCallback in this colab shows how to do this: wandb.me/lightning
      It logs them in on_validation_batch_end in the callback.

  • @ogrp5777
    @ogrp5777 3 роки тому

    Hi
    Is it possible to aggregate several graphs in Weight and Biases? Let’ say that I have the results of a model that was trained across 10 folds, so finally I will have the results for each fold and I want to show a plot with the average results

    • @WeightsBiases
      @WeightsBiases  3 роки тому

      We have a few different ways to set grouping! Check out the options here: docs.wandb.ai/guides/track/advanced/grouping

    • @ogrp5777
      @ogrp5777 3 роки тому

      @@WeightsBiases Tx!

  • @Glovali
    @Glovali 4 роки тому

    Is it like DVC?

    • @WeightsBiases
      @WeightsBiases  4 роки тому

      Yep! But if you're doing experiment tracking with W&B, Artifacts allows you to associate models/datasets with training run and evaluation data. Plus you can view all that in a single web UI!