Dealing with T5 for Multiple Choice Task

I wanted to learn how others deal typically with a training text-to-text model for performing multiple choice task such as SWAG. Any pointers will be appreciated.

If anyone is interested: There’s a SWAG notebook in community section notebook that can be used for reference. The input can be context: sentence. 1: option_1, 2: option_2 where labels can be option_numbers.