zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-400175517
The main problem stops hybridizing is the Gram Matrix calculation, which
reads the shape.
https://github.com/zhanghang1989/M
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-400034208
The gluon interface enables applying to image with any size. This model is
not hybridizable due to lack of some operations.
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-391796506
The pretrained model is ready now. It will be merged soon. Thanks again for
your patient and interest in this work!
---
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-391597085
Thanks for following up! Sorry for the delay. I have fixed the training in
PR https://github.com/apache/incubator-mxnet/pull/
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-384435679
I will look into it and get back to you.
This is an automat
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-371252011
@piiswrong Need some help about API changes.
In the style transfer example, the `set_data()` function was okay for
back-p
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-371252011
@piiswrong Need some help about API changes.
In the style transfer example, the `set_data()` function was okay for
back-p
zhanghang1989 commented on issue #9989: Cannot train example gluon style
transfer
URL:
https://github.com/apache/incubator-mxnet/issues/9989#issuecomment-371242715
Hi @samhodge , you shouldn't move L82 out of autograd.record scope, because
the gradient of siamese network won't back-propag