tokenizer.json

from transformers import T5Tokenizer

tokenizer = T5Tokenizer.from_pretrained('t5-small')  # or your own trained model
tokenizer.save_pretrained('./t5_urdu_translation')