/conifer

Fast inference of Boosted Decision Trees in FPGAs

Primary LanguagePythonApache License 2.0Apache-2.0

conifer

Conifer translates trained Boosted Decision Trees to FPGA firmware for extreme low latency inference. Check the examples directory for examples to get started with.

Currently models from sklearn, xgboost, and TMVA are supported. FPGA firmware can be produced in Xilinx Vivado HLS or VHDL.

See our paper: https://arxiv.org/abs/2002.02534

Conifer originated as a development for https://hls-fpga-machine-learning.github.io/hls4ml/, and borrows heavily from the code and ideas developed for it.

Installation

git clone https://github.com/thesps/conifer.git
cd conifer
pip install .