lleaves 0.2.0

LLVM-based compiler for LightGBM models

lleaves 🍃

A LLVM-based compiler for LightGBM decision trees.

lleaves converts trained LightGBM models to optimized machine code, speeding-up inference by ≥10x.

Example

lgbm_model = lightgbm.Booster(model_file="NYC_taxi/model.txt")
%timeit lgbm_model.predict(df)
# 12.77s

llvm_model = lleaves.Model(model_file="NYC_taxi/model.txt")
llvm_model.compile()
%timeit llvm_model.predict(df)
# 0.90s 

Why lleaves?

• Speed: Both low-latency single-row prediction and high-throughput batch-prediction.
• Drop-in replacement: The interface of lleaves.Model is a subset of LightGBM.Booster.
• Dependencies: llvmlite and numpy. LLVM comes statically linked.

Some LightGBM features are not yet implemented: multiclass prediction, linear models.

Installation

conda install -c conda-forge lleaves or pip install lleaves (Linux and MacOS only).

Benchmarks

Ran on a dedicated Intel i7-4770 Haswell, 4 cores. Stated runtime is the minimum over 20.000 runs.

Dataset: NYC-taxi

mostly numerical features.

batchsize	1	10	100
LightGBM	52.31μs	84.46μs	441.15μs
ONNX	11.00μs	36.74μs	190.87μs
Treelite	28.03μs	40.81μs	94.14μs
lleaves	9.61μs	14.06μs	31.88μs

Dataset: MTPL2

mix of categorical and numerical features.

batchsize	10,000	100,000	1,000,000
LightGBM	95.14ms	992.472ms	7034.65ms
ONNX	38.83ms	381.40ms	2849.42ms
Treelite	38.15ms	414.15ms	2854.10ms
lleaves	5.90ms	56.96ms	388.88ms

Development

conda env create
conda activate lleaves
pip install -e .
pre-commit install
pytest
# (optional) benchmark dependencies
conda install -c conda-forge treelite onnxruntime onnxmltools