{
  "clean_up_tokenization_spaces": true,  
  "tokenizer_class": "sentencepiece_ja.SentencePieceJA"
}