Support for RowBatch processing


#1

Hi community

Is there any support for RowBatch processing of given batch size in xgboost. I can see that there is a function GetRowBatches() which returns the RowBatches. But can we specify somewhere the size of the row batch we require in the code (like if I want to process batches of size 2^15)?

Thanks


#2

Take a look at https://xgboost.readthedocs.io/en/latest/tutorials/external_memory.html, which is the functionality that uses GetRowBatches(). Right now, the page size is fixed to 32 MB:


#3

Thanks for the code.