Repository implementing MLJ interface for scikit-learn models (via PythonCall.jl).
If you are using Linux and Julia version <=1.8.3, the libstdcxx version is not compatible with scikit-learn>=1.2. To get around this issue, you have to swap out the libstdcxx version that is loaded in with Julia. There is two methods to do this. The first is to build an environment with Conda.jl and use that as your LD_LIBRARY_PATH.
ROOT_ENV=`julia -e "using Conda; print(Conda.ROOTENV)"`export LD_LIBRARY_PATH=$ROOT_ENV"/lib":$LD_LIBRARY_PATHAnother method is to link your OS's version of libstdcxx instead of the one that comes with Julia. More details can be found here.
In both cases, it is recommended to use a Julia version >=1.8.4, but these fixes can be used as a last resort if you have to stay on an older Julia version.
- ScikitLearn.jl - a Julia implementation of the scikit-learn API