Intro to Weights and Biases with fastai


I am Thomas, I work at Weights and Biases. We wrote with @jeremy the hyperparameter search for fine-tuning the image models that he showed on the walkthrus. Sweeps are great to explore different training recipes and configurations, so…

I think it’s due time for an introduction to experiment tracking with W&B :sweat_smile:

I am planning on doing an intro to Weights and Biases for fastai users. The idea is to give students a simple walkthru on how to use W&B with fastai. If you want to start reading about W&B and fastai you can start with:

I am planning on doing this intro on Friday, June 24, at 10am CET (18h Brisbane Time). I may do a second one for our US folks, later on.

It will be a chilled-out zoom call with time for questions and demos. My idea is not to overwhelm people with W&B features, just give some value straightaway for fastai users.

Link to the meeting: Launch Meeting - Zoom

The password is fastai creator’s last name.


Had a great time discussing W&B with everyone. For the folks in America (not only USA) do we book another session on a more friendly time schedule?


Yes, please. Thanks.

Yes, please :slight_smile:

1 Like

Would love to join the Western Hemisphere version. But if you made a recording, that’d be great too. :slight_smile:


That would be great! Will there be a recording of the first session? (I understand it may not be possible.)

Sadly, I misclicked the record button and only registered 5 mins. But don’t worry, let’s redo this:

On 2022-06-30T15:30:00Z (8:30 am Los Angeles time)

The link will be the same: Launch Meeting - Zoom

1 Like

Quickly reminder that we are live in 20 min

1 Like

It seems I’ve missed both opportunities :frowning: , any chance for the recording?

I think it was recorded IIRC, but not sure when/if it’ll be published.

1 Like

The link is here:

the password: 6v0YJQ#n


Thanks for sharing the video @tcapelle . It was very informative. I was trying to go through Jeremy’s paddy series using wandb. In the third notebook, Jeremy has tried various models with different transforms.

models = {
    'convnext_large_in22k': {
        (Resize(res), (320,224)),
    }, 'vit_large_patch16_224': {
        (Resize(480, method='squish'), 224),
        (Resize(res), 224),
    }, 'swinv2_large_window12_192_22k': {
        (Resize(480, method='squish'), 192),
        (Resize(res), 192),
    }, 'swin_large_patch4_window7_224': {
        (Resize(res), 224),

Is it possible to do this using a single sweep config file in wandb? If I use grid search then it will try all models with all possible Resize options.

You are right!
You will have to construct this manually, I would not use sweep for something like this.

You could just pass them as list of lists:

    values: [[1,2,3], [4,5,6], [7,8,9]]

but in your code would be extremely ugly:

def train():
  do stuff:
    config.params[0], config.params[1], ...

Thanks for the quick reply. sklearn has this functionality. May be this can be a useful feature for future versions to have :slight_smile: