It makes sense to use a PartitionedDataSet instance but also maybe it should be easier doing a specific dataset that does this