Brainoidlabs's picture
Create README.md
3ff1a0d verified
|
Raw
History Blame Contribute Delete
345 Bytes
This Dataset is in ready to use format shuffled, contains 50B+ gpt2 tokens.
Designed for training **Rubin-series models**.
English-only dataset
Contains eval dataset (around 100 million tokens) and 50B training tokens.
Both are well shuffled into sequences of **1024** tokens.
The file is cleaned and contain raw tokens in **unit16** format