nlu/base_model/tokenizer_config.json

{
  "do_lower_case": false,
  "vocab_type": "spm"
}