dmlc / treelite

Universal model exchange and serialization format for decision tree forests
https://treelite.readthedocs.io/en/latest/
Apache License 2.0
723 stars 98 forks source link

[sklearn] Use more stable interface to get base_scores #526

Closed hcho3 closed 8 months ago

hcho3 commented 8 months ago

_loss.get_init_raw_predictions is removed in the recent refactoring of HistGradientBoosting (https://github.com/scikit-learn/scikit-learn/pull/26278). Replace it with _raw_predict_init, which would probably stick around for a while.

codecov[bot] commented 8 months ago

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (309263b) 83.28% compared to head (256035f) 69.72%. Report is 1 commits behind head on v4.

:exclamation: Current head 256035f differs from pull request most recent head 03551cd. Consider uploading reports for the commit 03551cd to get more accurate results

Additional details and impacted files ```diff @@ Coverage Diff @@ ## v4 #526 +/- ## =========================================== - Coverage 83.28% 69.72% -13.56% =========================================== Files 67 66 -1 Lines 5551 4965 -586 Branches 426 0 -426 =========================================== - Hits 4623 3462 -1161 - Misses 928 1503 +575 ``` | [Files](https://app.codecov.io/gh/dmlc/treelite/pull/526?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=dmlc) | Coverage Δ | | |---|---|---| | [python/treelite/sklearn/importer.py](https://app.codecov.io/gh/dmlc/treelite/pull/526?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=dmlc#diff-cHl0aG9uL3RyZWVsaXRlL3NrbGVhcm4vaW1wb3J0ZXIucHk=) | `91.94% <ø> (ø)` | | ... and [51 files with indirect coverage changes](https://app.codecov.io/gh/dmlc/treelite/pull/526/indirect-changes?src=pr&el=tree-more&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=dmlc)

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.