RedPajama replicates LLaMA dataset to build open source, state-of-the-art LLMs

RedPajama, which creates fully open-source large language models, has released a 1.2 trillion token dataset following the LLaMA recipe.

ai art business computer science conversational ai databricks dataset github hugging face language language models large language models llama llms meta ml and deep learning nlp open source open source software recipe redpajama science state