tokenizer/tokenizer_config.json

{"tokenizer_class": "MarianTokenizer", "source_lang": "en", "target_lang": "kin", "model_max_length": 128}