update docs and code on https://github.com/onnx/keras-onnx no more supported about sparseml HOT 4 CLOSED

Hi @pacospace, thank you for raising this to our attention. We'll work on the migration of the ONNX export flows for Keras since that is now deprecated along with the docs. I'll update here once that migration is complete. In the interim, our current Keras export flows are tested with the older versions, so if you're willing to downgrade those should continue to work for you.

Thanks,
Mark

Hi @markurtz, happy to contribute to it. I actually switched to Pytorch where everything works perfectly and managed to deploy a pruned Neural Magic model with deepsparse!

from sparseml.

The cause seems to be the package itself that you suggest to install, that works only with certain version of tf and onnx. Docs should be updated to use https://github.com/onnx/tensorflow-onnx.

from https://github.com/onnx/keras-onnx

We stopped active development of keras2onnx and keras2onnx is now frozen to tf-2.3 and onnx-1.10.
To convert your Keras models you can head over to tf2onnx which can convert Tensorflow, Keras, Tflite and Tensorflow.js models. All keras2onnx unit tests have been added to the tf2onnx ci pipeline to make sure there are no avoidable regressions.

from sparseml.

Hi @pacospace, thank you for raising this to our attention. We'll work on the migration of the ONNX export flows for Keras since that is now deprecated along with the docs. I'll update here once that migration is complete. In the interim, our current Keras export flows are tested with the older versions, so if you're willing to downgrade those should continue to work for you.

Thanks,
Mark

from sparseml.

Closing this issue now. @pacospace great to hear - the keras export migration ended up not landing as keras fell off the roadmap, but happy to take a look at any contributions if you still have them handy.

from sparseml.