Support of Keras 2.7.0 in X-CUBE-AI.7.1.0

[F F]

Hi,

I'm trying to import a .h5 model generated by an online AI platform in X-CUBE-AI.7.1.0, receiving the error:

E010(InvalidModelError): Model saved with Keras 2.7.0 but <= 2.6.0 is supported

Will there be any support on this version? 

Is there any workaround to e.g. convert a .h5 to an older version?

[fauvarque.daniel · Accepted solution]

This is a known issue of X-Cube-AI 7.1.0, the supported version of Keras is 2.6.0 while the supported version of Tensorflow Lite is 2.7.0.

X-Cube-AI 7.2.0 (end of 2022Q2) will support the latest version of Tensorflow (2.8 or 2.9) and the corresponding version of Keras

the workaround is to transform your keras model into a tflite model (see https://www.tensorflow.org/lite/convert) and import it as tflite in X-Cube-AI.

Regards

Daniel

[F F]

Hi Daniel,

Thank you for your answer.

Do you know if the X-CUBE-AI 7.2.0. release is still on track and can be expected soon?

I'm struggling with the tflite import (model output not as expected). Therefore I'm really looking forward to the release.

Regards

Frank

[fauvarque.daniel]

Yes it is still on track for end of next week and it will be based on tensorflow 2.9

Regards

Daniel

[F F]

Hi Daniel,

That's great!

Question, is it maybe possible to get my hands already on the beta release of X-CUBE-AI 7.2.0?

Regards,

Frank

[F F]

Hi Daniel,

As my project is really depending on this release, can you please tell if the X-CUBE-AI 7.2.0 package is still expected to release this week?

If not is there any way to get my hands on the beta release already to validate my model.

Thanks in advance!