Need help understanding copy_attn_force

pytorch

(Moonlightlane) #1

Hi there,

I am trying to figure out what this option copy_attn_force does? I understand that it is optionally used when you want to use the copy mechanism, in CopyGeneratorCriterion class in copyGenerator.py for calculating the output probability. However, I didn’ understand the comments for NOT using copy_attn_force: Regular prob (no unks and unks that can't be copied). What does this comment mean? Some explanation is greatly appreciated!!


(Guillaume Klein) #2

Hello,

Could you open an issue on GitHub with your question? There are more PyTorch devs there to answer questions about code details.

Thanks.


(Moonlightlane) #3

OK, will do. Thanks!