Datasets:

Modalities:
Text
ArXiv:
Libraries:
Datasets
Together org
No description provided.
Together org

Allows to load smaller samples of the RPv2, reflecting the overall distribution by subsampling {1, 10, 100} shards per CC snapshot. This leads to subsets with {10B, 100B, 1T} tokens respectively.

mauriceweber changed pull request status to open
mauriceweber changed pull request status to merged

Sign up or log in to comment