Datasets:

allenai

/

dolma

like 732

Thanks for releasing your data!

I have a clarifying question about the different versions.
The dataset card mentioned that a difference between 1.5 and 1.6 is the removal of metadata. Is that the only difference?

Relatedly, does that mean v1_6-sample remains a subsample of what was used to train OLMo (v1_5)?
And is that a random sample?