[QST] CPU memory spike during cudf dataframe conversion

Hi all, I have a dataframe that is ~19K Rows, ~11.4 MB (Profiled using ```df.info(memory_usage = "deep")```). We are currently running into CPU out of memory issues and so profiling our memory using this sample dataset. As you can see in the screenshot attached, there is a jump in mem usage, from 840MiB -> 4148MiB, during the type conversion of ```df```. Image below shows the dataframe memory usage after conversion.

My question is: Why is there a jump in the memory usage when converting a dataframe from pandas to cudf? Furthermore, this memory is not released after, and so increases from this point in following processing steps.