lleaves v1.2.3

Release version 1.2.3

lleaves 1.2.2

Safe softmax for multiclass in #83
Thanks @starkwj and Claude!

lleaves 1.2.1

What's Changed

• add single precision (float32) mode by @zjzjwang in #70

New Contributors

• @zjzjwang made their first contribution in #70

Full Changelog: 1.2.0...1.2.1

lleaves 1.1.0

What's Changed

• Implement random forest by @siboehm in #59
• Add ruff by @siboehm in #56

Full Changelog: 1.0.0...1.1.0

lleaves 1.0.0

FIRST MAJOR VERSION RELEASE. There are actually no big changes in this version, but lleaves has been so stable over the past 2 years that not calling it v1 was just me chickening out 🐓

Minor changes:

• C interface: the results array passed via pointer is now overwritten with the results instead of being accumulated into.

lleaves 0.2.7

What's Changed

• Avoid undefined behaviour / poison by checking for NaNs before llvm::fptosi by @siboehm in #23. This broke categorical predictions when NaNs occurred, but only on ARM arch.

Full Changelog: 0.2.6...0.2.7

lleaves 0.2.6

Minor new feature: Allow specification of the root function's name in the compiled binary. This enables linking against multiple lleaves-compiled trees. Thanks @fuyw!

What's Changed

• Chore: Bump pre-commit and Github actions + py3.10 on CI by @siboehm in #22
• add function_name to compiler by @fuyw in #21

New Contributors

• @fuyw made their first contribution in #21

Full Changelog: 0.2.5...0.2.6

lleaves 0.2.5

• Fix a bug when predicting from pandas Dataframes for models with pandas_categorical:null. Thanks @chenglin!

lleaves 0.2.4

Bugfix release: Removes an accidentally introduced dependency on pandas.

lleaves 0.2.3

• Fixed an overflow issue occurring during predictions on very large datasets (>50Mio rows, >20 features)