one problem with your train.json and dev.json

[wangzazao]

File "run_contract_qa.py", line 365, in convert_examples_to_features
input_ids = tokenizer.convert_tokens_to_ids(tokens)
File "/home/py/Contract_Elements_Extraction/bert/tokenization.py", line 128, in convert_tokens_to_ids
return convert_by_vocab(self.vocab, tokens)
File "/home/py/Contract_Elements_Extraction/bert/tokenization.py", line 89, in convert_by_vocab
output.append(vocab[item])
KeyError: 'ど'

----------the japnese char 'ど' seems not within the vocab.txt file ，it is quite strange. I verified this problem by ：
cd multilingual_L-12_H-768_A-12
grep ど vocab.txt
and got no result.
buddy,can you help me to run further?I am sorry the ど appears within the json files so frequently

[hoangtuanvu]

@wangzazao Well. I think there are 2 ways to fix it:

1. Add this char to vocab.txt
2. Treat this char as UNK char if this char is not in the answer.

[wangzazao]

hi,handsome:
does this 'ど' exists within your vocab.txt of the multilingual model?