Hi all,
I want to use xgboost in production, however dll size is quite large. Is there a lighter version (for example inference version) that I can use?
Or any other solution that exists as a smaller size package?
Thanks!
Hi all,
I want to use xgboost in production, however dll size is quite large. Is there a lighter version (for example inference version) that I can use?
Or any other solution that exists as a smaller size package?
Thanks!
The DLL is large because it contains dependencies necessary to enable GPU functionality.
You may save space by install XGBoost using Conda:
conda install -c conda-forge py-xgboost-cpu
which won’t support any GPU algorithm.
Thank you @hcho3
Is there a possibility to perform a similar installation with pip rather than conda?
You can ask pip to use the source package:
pip install -v --no-binary xgboost xgboost
This will invoke the C++ compiler on the host system to build XGBoost from the source and produce a lean library.
Note: this may not work well on Windows