Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

你好,我想问一下linears.py里面的poolerstartlogits是什么用? #7

Open
heroazhe opened this issue Mar 31, 2020 · 8 comments

Comments

@heroazhe
Copy link

endlogits里面的soft_label为啥要加上num_label?

@lonePatient
Copy link
Owner

lonePatient commented Mar 31, 2020 via email

@heroazhe
Copy link
Author

heroazhe commented Mar 31, 2020 via email

@lonePatient
Copy link
Owner

@heroazhe 很简单的一个想法啊,结束位置依赖于开始位置,很自然的将开始位置信息与文本语义向量结合作为结束预测的输入,那相当于有两种方式,直接hard label即argmax(logits),这种的话train时候hard label是真实的,但是预测的时候是argmax的,不一定正确,这个就像seq2seq里面的“Exposure @bias”(好像那么一回事,呵呵瞎哔哔的),那么就可以使用soft label,即softmax预测的logits,这样train和test是同步的,当然可以折中,抽样,加一个概率判断是使用hard还是soft。

@heroazhe
Copy link
Author

heroazhe commented Apr 1, 2020 via email

@heroazhe
Copy link
Author

heroazhe commented Apr 1, 2020 via email

@lonePatient
Copy link
Owner

@heroazhe 不会吧的啊 ,你看下bert的输出

@heroazhe
Copy link
Author

heroazhe commented Apr 2, 2020 via email

@lonePatient
Copy link
Owner

@heroazhe 还是那个问题啊 ,训练的时候标签是可知的,可用可不用,这里需呀做实验才知道哪种好,我这里可能是应该是为了应对eval使用soft label,将hard label转化为one-hot形式,纬度保持一致,具体的需要你个人做实验蔡得知。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants