Thread by @AtharvaIngle7

I am using @weights_biases from past 4-5 months and I am in love with the product. It makes working with Deep Learning projects super easy, trackable and fun. Here are some of my favourite features of wandb 👇

1. It is super easy to integrate Weights and Biases with any framework of your choice

2. You can literally log anything to wandb from any metrics you want to track, to model's weights, data, images, configurations, etc. Basically anything you can think of

3. You can even monitor your model's gradients by using their watch function to inspect any bugs in your model (gradient explosion, etc)

4. You can set name to your experiments and group them accordingly in your dashboard to see the performance selectively.

5. Weights and biases also provide a cool table where you can check all your logged information, apply filters to it, sort according to fields, and also export the table as CSV for further inspection.

6. They also provide data and model versioning (same as you version code in Git)

7. Wandb sweep allows you to automate hyperparameter optimization

8. It is super easy to work in teams on wandb with features like collaborative reports

9. You can even send alerts about your training progress to your email or slack channel (for ex: run is finished, accuracy increased in 2nd epoch, etc)
So you don't have to always watch the model training.

10. This is kind of a Easter egg feature: you can turn on dark mode on your profile by writing 'night' in your profile's description

11. They have some amazing blogs, and are doing great community work on their YouTube channel. I am in love with Weights and Biases.
They have many more amazing features.
Check out their website to know more about them👇

wandb.ai