responsible-ai-toolbox
responsible-ai-toolbox copied to clipboard
[Error Analysis] Discrepancy between legend and context menu when recreating the dashboard
Reproducibility steps:
- Created the dashboard for clf = LGBMClassifier(n_estimators=5)
- Recreated the dashboard for clf = LGBMClassifier(n_estimators=4)
The top left legend did not get updated.
I can't reproduce this with the current version of ErrorAnalysisDashboard
. Perhaps when you filed this we didn't create separate objects and reused the same port, @nushib ? If you can still reproduce it, could you point me at the notebook you're using? I used erroranalysis-interpretability-dashboard-census and changed n_estimators
there to get different error rates, but the resulting dashboard has consistent numbers in the top left legend and the tooltip when hovering over the top circle/tree root.
If you also can't reproduce please close the issue. Thanks!