Explore the text-2-text framework as proposed in the T5 paper to see how it performs on non text-2-text tasks by casting them in text-2-text settings.
Write a generic trainer that can be used for any problem which can be formulated as text-2-text. No need to change model, hyperparameters or add a task specific head. Just change the dataset and that's it!!
Don't forget to tag
your comment, otherwise they may not be notified.