tokenizer_config.json

{"additional_special_tokens": ["<e>", "</e>"], "model_max_length": 512}