from transformers import AutoModelForCausalLM, AutoTokenizer

BASE_MODEL = "google/t5-v1_1-xxl"
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
# The tokenizer will be saved in /tmp/tokenizer/tokenizer.json
tokenizer.save_pretrained("/tmp/tokenizer/")