Partial installation for inference - is it possible?

#1

Hi,
I have a use case where I only need to do inference with an xgboost model, and I have limited space to install the Xgboost itself.

Is there a way to reduce the size of Xgboost installation by skipping some modules /dependencies?

Thank you

#2

You may opt out of the GPU support in XGBoost by installing XGBoost from source:

python -m pip install --no-binary xgboost xgboost

This will consume at most 10 MB of disk space.

#3

Super helpful. Will try.