Hello,
I am running XGBoost 1.6.0.1 in R on the GPU.
During a grid search hyperparameter tuning I am reaching a memory limit:
R cannot allocate vector of size 2.7 Mb
I reach this limit after several loops in the grid search.
I am wondering if there is a solution to overcome this limit. If I understand correctly there are two solutions for the Phyton version with ussing xgboost.DeviceQuantileDMatrix or external memory version.
As far as if I can see these are not available for the R version.
In a post from January 2017 at R-bloggers I did find this External memory method in alinea 3:
Although I understood this is for the processor version and not for the GPU version.
Are my conclusions correct or is it possible in some way to overcome the memory limit in XGBoost version for R with GPU?
Thanks a lot!