At Comet.ml, we strive to help data scientists and machine learning engineers speed up the development and productionisation of their machine learning models.

A key part of the machine learning workflow is testing and evaluating model iterations. Visualizations are useful to compare various metrics and to track how these metrics change during training 📈📊

Tracking a keras model — plus I can export this chart as a JPEG to put into my presentation!

Especially since deep learning models run significantly longer than classical machine learning models (e.g. scikit learn), real-time visualizations can save time and compute (read: $$) by identifying errors and performance differences earlier. Instead of waiting 12+ hours to see how your model metrics and sharing screenshots, use real-time visualizations in Comet.ml to track performance and collaborate with your team.

Our original chart implementation overlaid all of the metrics on one chart. Based on user feedback and usability tests, we’ve introduced exciting new features for the Comet.ml chart builder!


With the new chart builder, you can:

  • Add charts for any of your logged metrics
  • Resize charts
  • Compare multiple metrics on a single chart (great for viewing training loss v. test loss)
  • Create, save, and update Saved Views (we start you off with a default view 👍🏼) — plus your entire Comet.ml team will be able to access these Saved Views!
  • Apply your Saved Views to other experiments (so you don’t have to build up the same view across experiments)
  • Export the charts as JPEG files (to put in your presentations)
A view into the charts builder — see a full example with this Comet.ml public project

“Stunning” — Nimrod L. our CTO 😉


We hope you find our new chart builder stunning like Nimrod — try out this multiple chart view with your own Comet.ml account!

Have feedback? We’d love to hear it on our Slack community or our Github repo!

Posted by:Cecelia Shao

Product Lead @ Comet.ml Comet is doing for Machine Learning what GitHub did for software. We allow data science teams to automatically track their datasets, code changes, experimentation history and production models creating efficiency, transparency, and reproducibility. Learn more at www.comet.ml

Leave a Reply

Your email address will not be published. Required fields are marked *