Installing xgboost and LightGBM on Mac OSX can be tedious and horrifying for someone like me who’s not familiar with compilers.
Prepare GCC
Let’s start with xgboost, which is the trickier one. Run the following command:
brew reinstall gcc --without-multilib
After 30-60 mins you will have the latest version of gcc installed.
Check GCC Availability
First, run the command:
gcc -v
The default C compiler for Mac OSX is clang. Normally it shows something like
Apple LLVM version 7.0.2 (clang-700.1.81)
We need to change the default C compiler from clang to gcc now.
To check the installed gcc version, run the following command:
ls /usr/local/bin/*
Throughout the huge list you will find directories like:
/usr/local/bin/g++-7
/usr/local/bin/gcc-7
Remember it for future use. Now run the following command:
sudo vi ~/.bash_profile
And add the following lines to the sourcing file:
alias gcc='gcc-7'
alias cc='gcc-7'
alias g++='g++-7'
alias c++='c++-7'
Quit editing with :wq, run source ~/.bash_profile followed by gcc -v again, now the default C compiler should be gcc.
gcc version 7.2.0 (Homebrew GCC 7.2.0)
Prepare MPI
Run ls /usr/local/bin/* again and try to find some directories like:
/usr/local/bin/mpicxx
If it exists, skip this part. If not, you should install the HPC library MPI instead.
Download here or check MPI Official website for the latest version and download the .tar.gz file. Run the following commands step by step:
tar zxvf openmpi-1.8.tar
cd openmpi-1.8
./configure --prefix=/usr/local
make all
sudo make install
If you can see the abovementioned directory, then we are fine with MPI.
Prepare xgboost
Clone xgboost from Github repo and try to compile:
git clone --recursive https://github.com/dmlc/xgboost
cd xgboost
cp make/config.mk ./config.mk
make -j4
Now you are on the way to make a 4-core xgboost toolkit.
Debug
If you just had success with compiling, skip this part.
errorunsupported option ‘-fopenmp’
errorunsupported option '-fopenmp' could still happen in rare cases even if you did switch from clang to gcc. You need to edit the Makefile:
vi Makefile
Look for the following lines:
# linux defaults
ifndef CC
export CC = gcc
endif
ifndef CXX
export CXX = g++
Change them to:
# linux defaults
ifndef CC
export CC = /usr/local/bin/gcc-7
endif
ifndef CXX
export CXX = /usr/local/bin/g++-7
The directories should correspond to the gcc version that you have just installed. Do the same with make/config.mk and try to find the following lines:
# choice of compiler, by default use system preference.
# export CC = gcc
# export CXX = g++
# export MPICXX = mpicxx
Change them to (be sure to uncomment them!!!):
# choice of compiler, by default use system preference.
export CC = /usr/local/bin/gcc-7
export CXX = /usr/local/bin/g++-7
export MPICXX = /usr/local/bin/mpicxx
This should fix the compiler bug.
collect2: error: ld returned 1 exit status
I don’t understand why this error pops up but the solution is:
vi Makefile
Search for the following line:
export LDFLAGS= -pthread -lm $(ADD_LDFLAGS) $(DMLC_LDFLAGS) $(PLUGIN_LDFLAGS)
Now add a link before it:
export LDFLAGS= -pthread -lm -mmacosx-version-min=10.9
Quit with :wq, clean up with make clean, then do make -j4. No errors should be raised anymore.
Install Python library
To run xgboost with Python, you need to install the library:
cd python-package
sudo python setup.py install
Try import xgboost in Python notebook and it should be fine.
Install LightGBM
After all what we have done for xgboost it’s effortless now to work on LightGBM. The only thing we need to prepare is cmake.
brew install cmake
The following steps are pretty much generic.
git clone --recursive https://github.com/Microsoft/LightGBM
cd LightGBM
export CXX=g++-7 CC=gcc-7
mkdir build
cd build
cmake ..
make -j4
LightGBM is available through pip for python:
pip install --no-binary :all: lightgbm