[QST] How'd this work with multi-threading pandas ?

**What is your question?**
I have a multi-thread pandas process to divide large dataset by date ranges then combine them, trying cudf now for speed .

It does something like this , args contains the start/end_dt for each segment, run to process segments
while Pool() as pool :
result = pool.starmap(run, args)

However , running in cudf it threw this error just now.
__pickle.PicklingError: Can't pickle : it's not the same object as pandas._libs.tslibs.timedeltas._timedelta_unpickle_