Version 2.2.2#

  • Sampler class
  • Improve plotting categorical plots

    • Now when plotting categorical with Dataset.plot will display the raw numbers by each bar.

  • Fixed the caching for Loader

    • Error w/ previous versions for caching with Loader when behav == ‘all’

    • Now fixed, and more effienct.

  • More efficient scoring / bug fix

    • Updating scoring logic in evaluate() to stop re-calling predict / predict proba when there was multiple scores.

    • New implementation uses to logic from sklearn multimetric wrapper to handle this better.

    • Can give a big speed up in cases where early pipeline transformation / predictions take a long time and multiple metrics are calculated.

  • Add data files function updated

    • The add data files method is now robust to reference index that are integer

    • The new default subject to file is now ‘auto’.

  • Shading to progress bar

    • The progress bar in BPtEvaluator now is shaded, no real reason, just prettier.

  • Some internal refactoring

    • Internal renaming of modules to more pythonic.

  • Add repr and iter to Compare

    • Just changing how it looks, no functional changes