Describe the bug
If the cudf has a column of type datetime[ms] and if I use cudf.to_pandas then I get the column but with an incorrect time resolution. In this case it seems it always returns datetime[ns]
Steps/Code to reproduce bug
Load a csv file that contains a date64 column
Print the cudf and you will see the column type is datetime64[ms] (Note here the ms: milliseconds)
Use cudf.to_pandas and check the resulting dataframe, it will have the datetime column but with ns (nanoseconds)
Describe the bug If the cudf has a column of type datetime[ms] and if I use cudf.to_pandas then I get the column but with an incorrect time resolution. In this case it seems it always returns datetime[ns]
Steps/Code to reproduce bug
Here the script that reproduce the issue:
Expected behavior to_pandas should return a dataframe with correct time resolution if the column is a datetime.
Environment overview (please complete the following information)
Environment details Please run and paste the output of the
cudf/print_env.sh
script here, to gather any other relevant environment detailsClick here to see environment details
Additional context This issue also affects dask-cudf when the dask work want to process a cudf with type datetime.