Issue of `bert_model` arg in `run_classify.py`

[llidev]

Hi,

I am trying to understand the bert_model arg in run_classify.py. In the file, I can see

tokenizer = BertTokenizer.from_pretrained(args.bert_model)

where bert_model is expected to be the vocab text file of the model

However, I also see

model = BertForSequenceClassification.from_pretrained(args.bert_model, len(label_list))

where bert_model is expected to be a archive file containing the model checkpoint and config.

Please help to advice the correct use of bert_model if I have my pretrained model converted locally already.

Thanks!

[thomwolf]

Hi, please read this section of the readme.