Datasets:

tiiuae

/

falcon-refinedweb

like 809

Follow

Technology Innovation Institute 919

Hi there!
Would love to know how this open released refinedweb portion is sampled from your full corpus? Is this uniformly sampled from your full refinedweb, in which case we can use this tiiuae/falcon-refinedweb instead of other common-crawl pretrain data?

Or if this is sampled with some weighted distribution, could you elaborate that as well?

Thank you and looking forward to your reply!