We are pleased to announce that today we are releasing a new SigOpt experience! 

Not to worry, you’ll be able to continue to use SigOpt Python clients < 8.0.0, SigOpt Java clients <= 6.3.0, and SigOpt R Client 0.0.1 until Q3 2022 without experiencing any breaking changes. 

What’s changing?

With this new version of SigOpt, users will be able to seamlessly leverage SigOpt’s Experiment Management with SigOpt’s Optimization Engine. SigOpt Experiments will now automatically produce Runs and consume Metrics without the intermediate concepts of Suggestions and Observations. You can queue Runs to probe specific parts of your parameter space, analyze your tuning and training Runs across an Experiment and Project, and easily use SigOpt’s core offerings with minimal changes to your modeling. In the future, we hope to build upon this new experience to allow you to leverage your past experimentation to inform your future work, and create integral integrations with other products you use in your modeling workflow. 

Why are we making this change?

From usability testing and interviews, we saw that the concepts in our product didn’t map easily into existing user flows, and that it was difficult to understand how to integrate our 3 products into a singular flow. With this new SigOpt experience, you will be able to seamlessly use tracked Runs, HPO Experiments, and job Orchestration in a singular workflow. We also found that flexibility in integrating SigOpt was really important, and as a result, you’ll still be able to leverage each core product line individually and pick and choose which features to integrate.  As a whole, every SigOpt Experiment will enable detailed model tracking, allowing you to manage and recreate your model tuning with greater detail.

How will this impact users?

• You’ll now have access to all features through 1 Python Client
• You’ll now be able to create Experiments and Runs with fewer lines of code
• You’ll have access to tracking, tuning, and orchestration through 1 SigOpt CLI
• Your Experiment’s History will contain Runs
• Your Experiment’s Analysis plots will contain Runs and link out to tracked information
• You’ll be able to easily track information for a tuning Run and watch the progress on the UI
• You’ll be able to easily find, filter, and analyze Runs associated with an Experiment
• You’ll be able to continue to use SigOpt Python clients <= 8.0.0, SigOpt Java clients <= 6.3.0, and SigOpt R Client 0.0.1 until Q3 2022 without experiencing any breaking changes
• After Q3 2022, any Experiments you created with SigOpt Python clients <= 8.0.0, SigOpt Java clients <= 6.3.0, and SigOpt R Client 0.0.1 will be migrated over to the new SigOpt experience.

How can you start using the new SigOpt?

To get started, check out our docs!

All you have to do is:

• pip install SigOpt 
• Instrument your code with a few lines of code
• And, you’re done!

Try out the new features in the notebook below:

We’d love to know what you think of the new experience- please contact [email protected]  with any feedback you have.

Happy Modeling!

Meghana