Is there any unnecessary or debugging files in the XGBoost model to remove and make the model size smaller?

#1

Is there any unnecessary or debugging files in the XGBoost model to remove and make the model size smaller? Thanks!

#2

No, the model file contains only the information necessary to represent the constituent decision trees.

#3

Thank you very much for the prompt response, Philip.

Any suggestions for compressing the XGBoost models? FYI, my input data (training and test) takes integer values and therefore the splitting thresholds within the trees must all be something like *.5. Any easy way to quantize those thresholds to compress the models further?

#4

No, not at this moment.

You may have some success with saving the model using the JSON format and then compressing it using gzip or other compression software.