AttributeError: 'Model' object has no attribute '_get_distribution_strategy'

[milansoliya4210]

I am trying to train model in Google Colab using tensorflow 2.x bundle package for Colab and Getting Error as mention Below:

AttributeError:'Model' object has no attribute _get_distribution_strategy'

[hgaiser]

tensorflow/tensorflow#34870

you can also remove the tensorboard callback for now.

[shadow12138]

tensorflow/tensorflow#34870

you can also remove the tensorboard callback for now.

Will the training weights be saved if the callback is removed?

[hgaiser]

Yes they will be. Only thing you lose is the usage of Tensorboard.

ps. you can also use --tensorboard-dir "" on the command line to disable Tensorboard callback.

[SalahAdDin]

It is a problem with keras, right? I mean, you cannot solve it in this implementation, right?

[shadow12138]

It is a problem with keras, right? I mean, you cannot solve it in this implementation, right?

i solved it by --tensorboard-dir ""

[SaltyGlutinousRiceBall]

Yes they will be. Only thing you lose is the usage of Tensorboard.

ps. you can also use --tensorboard-dir "" on the command line to disable Tensorboard callback.

'--tensorboard-dir ""' is not an internal or external command, nor is it a runnable program or batch files.@hgaiser

[hgaiser]

It is a problem with keras, right? I mean, you cannot solve it in this implementation, right?

It's a problem with keras and tensorflow. Tensorflow made some changes that didn't work when using keras, only works when using tensorflow.keras. tensorflow/tensorflow#34870 got merged so this should be automatically fixed with a new tensorflow release.

[stale]

This issue has been automatically marked as stale due to the lack of recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[daryl-user]

Did you try import from tensorflow.keras instead of directly from keras?

[ZhangYuef]

Did you try import from tensorflow.keras instead of directly from keras?

It's the same for me. Both tensorflow.keras and keras raise the error.

[hgaiser]

Tensorflow just released version 2.2, this should fix this issue.

[DarthAmk97]

The issue persists :(

[SalahAdDin]

@DarthAmk97 Use the monkeypatch solution ,it works for me.

[TimNagle-McNaughton]

@SalahAdDin can you provide a link to or any documentation of that solution?

[SalahAdDin]

It is a problem with keras, right? I mean, you cannot solve it in this implementation, right?

i solved it by --tensorboard-dir ""

@TimNagle-McNaughton This one.

[TimNagle-McNaughton]

oh. Sadly that doesn't actually fix the functionality, just removes the functionality and the error. Thanks

[SalahAdDin]

@TimNagle-McNaughton it is related to Tensorflow library more than this package, unfortunately.