Skip to main content
llm.info

RedPajama

Open Source
Datasets

Open reproduction of LLaMA training data

About

RedPajama is Together AI's open dataset reproducing the LLaMA training mix, later expanded to RedPajama-V2 with over 30 trillion filtered tokens across five languages - a large, permissive corpus for open model pretraining.

Details

Category
Datasets
License
Apache-2.0 (code)