DAGNetwork showing no properties.

2 次查看(过去 30 天)
Hello everyone,
I have a model and I load in on two matlabs one 2019b and one 2018b on two systems. For loading the model on Matlab 2019b, it shows the layers and everything but for Matlab 2018b it shows "DAGNetwork with no properties." and there is no info for layers and connection.
Can anyone tell me what is the issue?!
Thanks
  2 个评论
Asvin Kumar
Asvin Kumar 2020-5-18
Can you please provide a way to reproduce this issue?
If your reproduction steps don't already include them, please provide details on how you're importing/loading and preferably what you're importing/loading too.
faran
faran 2020-5-18
Hi Asvin,
I have the model as ".mat" file and load it like this:
Matlab 2019a:
>> NetworkModel = load(modelpath, '-mat');
>> NetworkModel.net
ans =
DAGNetwork with properties:
Layers: [52×1 nnet.cnn.layer.Layer]
Connections: [55×2 table]
MAtlab 2018b:
>> NetworkModel = load(modelpath, '-mat');
>> NetworkModel.net
ans =
DAGNetwork with no properties.
on 2018b I get this warning too:
Warning: While loading an object of class 'nnet.cnn.layer.ImageInputLayer':
Reference to non-existent field 'IsAverageImageMeanPerChannel'.
Warning: While loading an object of class 'nnet.cnn.layer.TransposedConvolution2DLayer':
Reference to non-existent field 'Cropping'.
Warning: While loading an object of class 'nnet.cnn.layer.TransposedConvolution2DLayer':
Reference to non-existent field 'Cropping'.
Warning: While loading an object of class 'nnet.cnn.layer.TransposedConvolution2DLayer':
Reference to non-existent field 'Cropping'.
Warning: While loading an object of class 'nnet.cnn.layer.TransposedConvolution2DLayer':
Reference to non-existent field 'Cropping'.
Warning: While loading an object of class 'DAGNetwork':
Dot indexing is not supported for variables of this type.

请先登录,再进行评论。

采纳的回答

Asvin Kumar
Asvin Kumar 2020-5-19
Here are the explanations for the three different types of warnings you’re seeing.
Reference to non-existent field 'IsAverageImageMeanPerChannel'.
---> ‘IsAverageImageMeanPerChannel’ was introduced in R2019a [Link to a similar Question]
Reference to non-existent field 'Cropping'.
---> ‘Cropping’ was replaced by Croppingsize in R2019a [Link to Release Notes]
Dot indexing is not supported for variables of this type.
---> As for the dot indexing not being supported, this is a known behaviour. There was a change in the representation of Network Layers from R2018b to R2019a. Models created in R2018b should work fine in R2019a but the other way around will have issues.

更多回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Custom Training Loops 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by