Train a t5 model

If you have 10 question and answer pairs for each context, Your input structure should be
source_text->context
target_text->QA pairs separated by any tokens , for example → [SEP] like that