Use existing pretrained model for retraining using SpanBERT

[aman31kmr]

Hi,
I have a pretrained BERT model for domain specific corpus, and I am trying to create SpanBERT language model for my corpus to predict ngrams masked terms. It is not allowing me to use my model.bin file and error is :
train.py: error: argument --arch/-a: invalid choice: 'model/model.bin' (choose from 'bert_pair', 'bert_pair_large', 'cased_bert_pair', 'cased_bert_pair_large', 'bert_hf', 'bert_hf_large', 'cased_bert_hf', 'cased_bert_hf_large')