from transformers import LlamaTokenizer

tokenizer = LlamaTokenizer.from_pretrained("LLM360/Amber", revision="ckpt_356")