Jon Durbin

airoboros. Customizable implementation of the self-instruct paper.

bagel. A bagel, with everything.

qlora. QLoRA: Efficient Finetuning of Quantized LLMs

dre2. DRE2 - A fast, lightweight regex engine in c.

llama-recipes. Examples and recipes for Llama 2 model

trl. Train transformer language models with reinforcement learning.