Question regarding mode deployment

[cjohn001]

Hello @james77777778,
seems like your project is the next missing piece of the puzzle I am looking for :) Seems like you are a specialist in the topic, hence I would like to ask you a question in regards to model deployment. I am currently stuck with Keras2 as I have not found an alternative in Keras3 for tensorflow_model_optimization. I am specifically looking for quantization aware training and pruning preserving quantization aware training. Have you maybe found a solution for this already?

[james77777778]

Hey @cjohn001

Regarding your question, unfortunately, there is no existing public solution for QAT and other advanced QAT with Keras3.
For now, it is still better to stick with Keras2 if you need QAT.

However, QAT is on the roadmap for Keras3. You can check out this PR: keras-team/keras#20641

[cjohn001]

Hello @james77777778 ,
thanks for the directions. I suspected that. Than for the moment I will stay on Keras2 or see if I have more luck by looking into the pytorch/onnx direction. Is your model zoo known to work with Keras2 as well?