This is the llama3 tokenizer with an added <think> and </think> tokens. Created using this notebook https://colab.research.google.com/drive/15OyZuxGTung7ZwtrYcRw0ItN6d2FFH_H?usp=sharing