I want to use vocab file in PairedDataloader, but the the save_vocab function of SentencePieceTokenizer only save the model file.
The model file can't be load by Dataloader because of decoding error.
In sentencepiece_tokenizer.py, I saw you delete the vocab file.