π
These models are ported from Google repo BERT checkpoint. So you should refer to the Google BERT repository for training code and tokenizer info.
Β· Sign up or log in to comment