Citrine v3.17.0 is released!

In this release of Citrine Python, we've introduced capabilities to further streamline our users' AI-driven explorations. Users can now directly access Feature Effect values used in to create the Feature Impact plots in our web UI for each model output, enriching their understanding of the model's predictions. Additionally, previously UI-exclusive Candidate attributes such as pinned and hidden attributes, candidate name, and comments, are now viewable in a read-only format within Citrine Python, offering a more options for custom automated filtering and visualization. We've also included some updates to enhance our own development and experimentation with platform assets, ensuring we continue to deliver best-in-class materials AI capabilities.

What's New

• Support for accessing Feature Effect values from valid Graph Predictors. The call predictor.feature_effects() will return a FeatureEffects object with feature effect values for each output of the chosen predictor. #982
• Expose Candidate attributes previously only seen in the UI, including pinned and hidden attributes, candidate name, and any candidate Comments. Attributes are read-only in the context of Citrine python. #987, #988

Improvements

• Expose the experimental AttributeAccumulationPredictor node for internal development and testing. #984

Fixes

• Improve unit test efficiency. #986

Full Changelog: v3.11.6...v3.17.0