If you look at here, we pass child data itself not its indices. When doing this, we make a copy of data, and it consumes a lot of time! In PR #197, I fixed partially by making a copy of child data with subsampled features, so it lowers the size of copied data.
If you look at here, we pass child data itself not its indices. When doing this, we make a copy of data, and it consumes a lot of time! In PR #197, I fixed partially by making a copy of child data with subsampled features, so it lowers the size of copied data.