Machine Learning Experiment Tracking

added 05.06.2020 02:52

Image of article 'Machine Learning Experiment Tracking'

At Toyota Research, the wandb experiment link is used as the official record of every ML model that gets built.

At Latent Space, every team project meeting starts with a review of the latest wandb experiment report and a discussion around how well the current approach is working and what experiments should be tried next.

Looking across experiments is so important that wandb lets you build workspaces where you can select groups of graphs in visualizations like a scatterplot and then immediately view comparisons of the selected runs Aggregate metrics are good, but it is essential to look at specific examples.

In Weights and Biases, you can create reports to track notes alongside model metrics, and to share your findings and progress with your team.

Weights and Biases (wandb) is a simple tool that helps individuals to track their experiments — I talked to several machine learning leaders of different size teams about how they use wandb to track their experiments.