ZJULearning/resa

为什么要 x.clone()

jay-z20 opened this issue · 1 comments

jay-z20 commented

resa/models/resa.py

Line 59 in 14b0fea

x = x.clone()

x = x.clone()

Turoad commented

When using previous torch version, it will cause strange error when gradient propagation if we don't add the clone(). You can also remove it now.