Open GuoQuanhao opened 3 years ago
collections.OrderedDict
is not a tensor, this is why torchsummary
is not able to compute its size
One of your network outputs (including the hidden layer outputs) is a dict (IntermediateLayerGetter
). Currently, torchsummary
does not support this case. You could use my modification (#165) to handle this problem. There are also some other solutions in the pull requests.
Edit
I find that your code also has dict
input (FPN
). Currently the input shape is recorded but not used in torchsummary
. I have fixed the problem in an ugly way.
Just switch to torchinfo
much better!
platform: win10 version: 1.6.0
net.py
retinaface.py
so, how can I get the summary