I read that google publish some BERT per-trained text models.
Are there any per-trained BERT model (for voice) ?