How does T5 handle sampling? #1009
Unanswered
antonio-mastropaolo
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello guys, I was wondering how T5 picks data when we create a multi-task mixture with a default rate of 1.
Is the sampling performed randomly among all the instances featuring the datasets of these different tasks?
In contrast, when we use a balanced sampling, meaning that we take into account the size of the datasets when creating batches, are these instances picked randomly according to their respective dataset sizes?
Beta Was this translation helpful? Give feedback.
All reactions