Paul Dubs

At the moment `tf.keras` import is not supported. Only regular Keras. This issues tracks the progress: https://github.com/eclipse/deeplearning4j/issues/8348.

Please share the non tf.keras version of your model file, and we'll take a look.

We need the actual model file you are trying to import to see why it doesn't import like it should. The stacktrace you've provided doesn't make a lot of sense,...

I see, your first layer is a `Flatten` layer, which be definition doesn't have any input size. That isn't a supported model configuration for import. You should be able to...

> Zen supports AVX2 but it requires two clock cycles to complete each AVX2 instruction compared to Intel's one.[70][71] This difference was corrected in Zen 2. The 3500U is based...

That particular quote is from https://en.wikipedia.org/wiki/Zen_(first_generation_microarchitecture)#Performance And yes, there are cases where due to the way it is implemented, it is just compatible enough to execute those binaries, but it...

I can confirm that on my Ryzen 9 5900X (Zen 3) it also doesn't seems to make a difference. @pza94 Thank you for highlighting the issue. We'll try to figure...

There is a bigger example repository restructuring coming in a few days, in order to not loose this PR in the aftermath, I'm keeping it open to add it in...

Please use the latest release, which is 1.0.0-beta7. But in any case, what you are seeing is a regular out of memory error. You either have to allow your program...

What you are seeing is just the effect of no fixed random seed. The weights get initialized randomly on each run of the example, you get different gradients on each...