from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("abeja/gpt-neox-japanese-2.7b")