This repository has been archived by the owner on Sep 18, 2024. It is now read-only.
Udpated concat axis to match image_data_format in keras #2946
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What would you like to be added:
The concatenation axis in src/sdk/pynni/nni/nas/tensorflow/mutator.py is fixed to 0 which does not reflect image_data_format setting in keras.json config.
Why is this needed:
This will cause the model to concatenate multiple inputs in the skip connections in the wrong dimensions. Users might get Incompatible shapes error if he uses multiple filter sizes.
Without this feature, how does current nni work：
nni only allows concatenating in the first dimension. This will fail if image_data_format is channel_last.
Components that may involve changes:
src/sdk/pynni/nni/nas/tensorflow/mutator.py