issues
search
gjylt
/
DoubleAttentionNet
PyTorch implementation of Double Attention Net
26
stars
3
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
two bugs
#5
dongzhuoyao
opened
4 years ago
1
Application to temporal data
#4
JohnMBrandt
opened
4 years ago
1
Wrong softmax for attention maps?
#3
nguyenvo09
closed
4 years ago
0
Why do you need to divide the batch size by self.K?
#2
nguyenvo09
closed
4 years ago
0
What does c_m and c_n mean? Does the dimension change when the tensor come cross the layer?
#1
ZealACMer
closed
5 years ago
4