issues
search
lyttonhao
/
Neural-Style-MMD
MXNet Code For Demystifying Neural Style Transfer (IJCAI 2017)
83
stars
22
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
gnorm = mx.nd.norm(model_executor.data_grad).asscalar()
#9
fido20160817
opened
1 year ago
0
Have you ever tried the setting of --bn_loss?
#8
wadx2019
opened
3 years ago
0
Why cannot get good results with Gaussian kernel?
#7
wzc118
opened
6 years ago
0
mmd_loss line 43
#6
chanshing
opened
6 years ago
1
python version problem
#5
cicido
opened
6 years ago
0
How do you set the bandwidth in Gaussian Kernel
#4
Aixile
closed
7 years ago
2
Batch Processing
#3
ChoclateRain
closed
7 years ago
1
confusion in Linear Kernel Loss functions
#2
allugrasm
opened
7 years ago
1
The paper might has a mistake
#1
michaelhuang74
opened
7 years ago
1