tokenizer_config.json

{
  "do_lower_case": false,
  "model_max_length": 512,
  "max_len": 512,
  "tokenizer_class": "BertTokenizer",
  "pretrained_model_name_or_path": "bert-base-multilingual-cased"
}