- sk scorer for binary classification with custom threshold - enable/disable tracker via trava directly - logger handler now tracks logs as artifacts
v.0.2.5 - further refactoring of TrainOnAllTestOnOneFitPredictor - accidental black formatting + official black integration
0.2.4
- fixed plotter when running in jupyter environment
0.2.3
- much more efficient evaluation - now BasicSplitLogic works correctly with validation split - TrainOnAllTestOnOneFitPredictor now supports custom FitPredictorSteps - other minor changes
0.2.2
- mypy & flake8 compliance - pandas results handler - fix for plots that don't support overlapping
0.2.1
- fit_params & predict_params now also go to the tracker. ( currently only numbers, strings and bools are supported )
0.2.0
- better experience with plotter, now PlotItem contains all the information needed to be able to plot metrics - metrics results are now cached between results handlers