I got error when I used your model file "convnext_tf.py" and use trained with callback for model save. Is there model prior save model weight file or error from my side? Need your help and suggestions. Thank you.

[Asif Iqbal Fahim]

[Sayak Paul]

I need to see the full code to know why this error is coming.

[Asif Iqbal Fahim]

Code Link
Callback Link

Thanks for your response please check. Thanks.

[Sayak Paul]

Please provide a Colab Notebook link that would allow me to reproduce the issue.

[Asif Iqbal Fahim]

Colab Notebook Link
[Error happen after one epoch end when try to save model weight file]
In the output, you will find output exists. You can run all to reproduce the same error. Please look at the following cell for error:

[Sayak Paul]

It's probably best to open this issue in the Keras repository. There's a similar issue here: https://stackoverflow.com/questions/70739244/valueerror-unable-to-create-dataset-name-already-exists-while-saving-tensorfl.

[Asif Iqbal Fahim]

Thanks but I tried with other models but didn't get this error.

[Sayak Paul]

I see. Since the model definition itself is a bit contrived it's currently not possible for me to dig deeper. The easiest fix I can suggest is serializing the model as a SavedModel resource rather h5 file.

[Asif Iqbal Fahim]

Thanks sayak for your response will check this hope to solve it anyhow. Thanks for your contribution.

[uranus Yu]

The error is caused by missing get_config() and from_config() methods in Class StochasticDepth.
See Tensorflow Guide.