With Sequel, you just call the use_cursor method on the dataset before loading records, and it use a cursor and fetch 1000 records at a time until the whole dataset is processed.
You can also provide an option to use_cursor to specify the number of rows to get per cursor fetch.