BERT-NER-Pytorch icon indicating copy to clipboard operation
BERT-NER-Pytorch copied to clipboard

Published 20 hours ago • verified publisher icon lonePatient

为什么“total_loss = (start_loss + end_loss) / 2”

Open possible1402 opened this issue 4 years ago • 0 comments

https://github.com/lonePatient/BERT-NER-Pytorch/blob/13199aeea321aa3bebe4cf2bb4bf37c69937e8ff/models/bert_for_ner.py#L124

直接对两个loss加和除2的意义是什么呢，有没有论文可以参考呢？ 我看ACL2020收录的论文"A Unified MRC Framework for Named Entity Recognition"里面为了计算start 和end的匹配，单独设计了一个span match的目标函数。这里除以2是想表达寻找距离最小的start和end的意思吗？

possible1402 avatar May 09 '20 03:05 possible1402