Could I jion the program to pull the request??

[Yazooliu]

你好! Good Job!

I reproduce your training model and generate the best model to inference on my special business sample as I DO not have more business data, but found some issue. During inference, the model will show the overcorrection: check the correct word to wrong word like this： { "paragraph": "在本合同中，除上下文另有规定外，下列用语应当具有如下含义：", "error_fragments": [ { "error_init_id": 26, "error_end_id": 27, "src_fragment": "含", -> the correct word in paragraph "tgt_fragment": "涵" -> model output wrong word } ] } So, there will be one white_name_list config and code to fix this overcorrection issue. I'd like to pull this code in tools/inference.py. So are you approve and do you hava any good idea ？ Let's keep talk, thx

BestRegards Yazhou

[gitabtion]

您好，非常感谢！白名单是一个在产品上非常实用的功能，欢迎您提交pr，谢谢您为本仓库的贡献！

[Yazooliu]

this PR is merged and issue is closed

[Yazooliu]

this PR is merged and issue is closed