Open shkarupa-alex opened 1 year ago
@shkarupa-alex, Thank you for reporting. We are investigating the mentioned issue and will provide the resolution and cause for the issue.
Thank you!
Issue persist with TF 2.15.0
@shkarupa-alex, I tried to execute the mentioned code with the Keras3.0 version and was able to get the same output for the weight names. Kindly find the gist of it here.
['kernel', 'bias', 'gamma', 'beta', 'moving_mean', 'moving_variance']
['kernel', 'bias', 'gamma', 'beta', 'moving_mean', 'moving_variance']
['kernel', 'bias', 'gamma', 'beta', 'moving_mean', 'moving_variance']
['kernel', 'bias', 'gamma', 'beta', 'moving_mean', 'moving_variance']
Thank you!
This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you.
Issue is still here
System information.
Describe the problem.
Models with custom layers miss their weight names after restoring from config
Describe the current behavior.
See example below. When replacing BatchNormalization with it's successor (the only thing changes - core vs custom class) after restoring from config we got wrong weight names.
Describe the expected behavior.
Weight names of custom layers should be preserved just like built-in ones.
Standalone code to reproduce the issue.
Source code / logs.
Code posted above will print: