Datasets:

---

EleutherAI

/

the_pile_deduplicated

like 104

Follow

EleutherAI 1.08k

Hi there,

I was wondering how the deduplication has been done exactly? Namely, was it done on only the train split of the original Pile (leaving the test/validation split untouched) or was it done before splitting it on the entire set of train/test/validation?

Thanks a lot!