FlatCTR is a high-performance toolkit to train Logistic Regression(LR) or Factorization Machines(FM) on large-scale sparse dataset.
git clone --recursive https://github.com/maxuewei2/FlatCTR.git
cd FlatCTR
mkdir build && cd build
cmake ..
make
- Run
./flatctr -h
to display a list of all supported options. - Run
./flatctr
to train on the sample dataset. - For online training, use the
-i
option to load a trained model:./flatctr -i ../output/model.txt
The input data should be in the libsvm format.