SqueezeAttention-PyTorch icon indicating copy to clipboard operation
SqueezeAttention-PyTorch copied to clipboard

Published 20 hours ago • verified publisher icon its-mayank

关于论文

Open doudou123456 opened this issue 4 years ago • 2 comments

个人感觉 你的实现 没有什么问题，但是我咋感觉这篇论文写得逻辑有问题呢，就是很多细节忽略了

doudou123456 avatar Sep 23 '20 09:09 doudou123456

@doudou123456 I am really sorry, I do not understand Chinese however I did translate with google translate, and from the conversion, I understand that you have something about the explanation in the paper. I don't know how much I agree with you on these terms as I have performed experiments myself and it seems to work in fact we are using these blocks in one of our papers which we have submitted.

its-mayank avatar Oct 01 '20 19:10 its-mayank

It would definitely work since it adopts a similar structure with seblock. The main key is the multiple loss functions. As for the SAblock, I don't think it is convincing.

zyxu1996 avatar Dec 15 '20 08:12 zyxu1996