Hugging Face's logo

Datasets:
allenai
/
c4

Tasks:
Text Generation
Fill-Mask
Sub-tasks: language-modeling masked-language-modeling
Languages: Afrikaans Amharic Arabic
Multilinguality: multilingual
Size Categories: n<1K 1K<n<10K 10K<n<100K
Language Creators: found
Annotations Creators: no-annotation
Source Datasets: original
ArXiv:
License:
Dataset card Files Community
10

Does allenai/c4 and the subset C4 in allenai/dolma is the same dataset?

#10
by speiqin - opened
Discussion
speiqin

hi, all. I want to reproducing the scaling-law in C4 dataset, but I found the val-set maybe duplicate with train-set, have anyone can tell me how to cleanup?
And, dolma is high-quality dataset, it also has C4, Is the C4 subset same with allenai/c4?

soldni
 Allen Institute for AI org

it is not; dolma/c4 is a subset of c4 that has been processed like the rest of dolma/cc.

speiqin

@soldni Thanks very much for your kind reply! And can I has any method to only download the dolma/c4 subset? Because dolma is so big.

soldni
 Allen Institute for AI org

After getting access to Dolma, you can download just the C4 subsets from urls listed here.

speiqin

OK. thanks you very much.

speiqin changed discussion status to closed

Sign up or log in to comment