Hi, I want to apply the GeDi model on a model with BertTokenizer. However, a dimension problem will occur when combining the probability distributions of two models. In addition, the tokens tokenized by two Tokenizers are different, so I can't match the two vocabularies directly. Is there any approach to applying the GeDi model? Thanks!
Hi, I want to apply the GeDi model on a model with BertTokenizer. However, a dimension problem will occur when combining the probability distributions of two models. In addition, the tokens tokenized by two Tokenizers are different, so I can't match the two vocabularies directly. Is there any approach to applying the GeDi model? Thanks!