Input issues with tfjs conversion

[rohanmuplara]

Hi,
I am not sure how the inputs are translated with the tfjs converter with named inputs.
https://colab.research.google.com/drive/1F1xO8950vJrw2JfVUDiid5pXAdOoo0lI?usp=sharing. I run this code multiple times and the order that the inputs appear in the conversion file change everytime(code is the same just reset runtime.) Additionally, the names change everytime, ie sometimes times input_0 and 1, sometimes input_5, input_6(no input 1 through 4) and the order that these appear in the json everytime changes. I was wondering if anything can be done as it means everytime I receonvert my model I have to change my tfjs code. I also have many inputs of same dimensions so is there is any way to figure out the mapping happens.
I also have attached a screen recording explaining my issues: https://share.descript.com/view/5xYQx07Q7T4
Thanks,
Rohan

[pyu10055]

@rohanmuplara I took a look at your colab, looks like you can fix the input layer names by setting the name param for tf.keras.Layers API.
I have seen the signature of model created to stay the same after that change.

[rohanmuplara]

Yes thanks for the tip and this works great. I don't think this as much of a tfjs issue more like regular tensorflow issue and this could be just my opinion but it would be a little more idiomatic/convenient for it to use the dictionary provided in model than to use passing in the names to the input. Right now it never uses the names provided in dictionary to model and will only change signature if you change input layer names.

[google-ml-butler]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you.

[google-ml-butler]

Closing as stale. Please @mention us if this needs more attention.