Keras Core: Keras for TensorFlow, Jax, and PyTorch

I think that is pretty cool - literally made me screen &quot;Yes!&quot; when I saw it and I don&#x27;t do this for your everyday framework.<p>I think the beauty of keras was the perfect balance between simplicity&#x2F;abstraction and flexibility. I moved to PyTorch eventually but one thing I always missed was this. And now, to have it leapfrog the current fragmentation and just achieve what seems to be a true multi-backend is pretty awesome.<p>Looking forward to the next steps!

Keras was already that some years ago. It supported tensorflow, theano, mxnet if my memory is right. And then they ditched everything for tensorflow. At the time it was really hard to use keras without calling backend directly for lots of optimisation, unsupported feature on they API etc... This make the use of Keras not agnostic at all. What&#x27;s different now ?

I worked on the project, happy to answer any questions!

As someone who has dealt with countless breaking changes in keras and wasted days of my life attempting to upgrade, no thank you.<p>My pytorch code from years ago still works with no issues, my old keras code would break all the time even in minor releases.

From the announcement:<p><i>&quot;We&#x27;re excited to share with you a new library called Keras Core, a preview version of the future of Keras. In Fall 2023, this library will become Keras 3.0. Keras Core is a full rewrite of the Keras codebase that rebases it on top of a modular backend architecture. It makes it possible to run Keras workflows on top of arbitrary frameworks — starting with TensorFlow, JAX, and PyTorch.&quot;</i><p>Excited about this one. Please let us know if you have any questions.

Keras and PyTorch! I thought I&#x27;d never see the day! Glad to see the two communities bury the hatchet.

If I want to use this brave new keras with torch.compile, what does that look like?

Does this mean the weights output can be backend agnostic?<p>Also, are there any examples using this for the coral TPU?

Can someone ELI5 the relationship between Keras and TensorFlow&#x2F;Jax&#x2F;PyTorch&#x2F;etc? I kinda get the idea the Keras is the &quot;frontend&quot; and TF&#x2F;Jax&#x2F;PyTorch are the &quot;backend&quot; but I&#x27;m looking to solidify my understanding of the relationship. It might help to also comment on the key differences between TF&#x2F;Jax&#x2F;PyTorch&#x2F;etc. Thank you.

Supporting multiple backends (especially Jax) is nice! Makes experimenting&#x2F;migrating between them so much more approachable. Any timeline on when can we expect support for distributed Jax training? The doc currently seems to indicate only TF is supported for distributed training.

IIRC, Keras was added officially added to Tensorflow as part of the version 2.0 release. With Keras reverting to its backend-agnostic state, will it be removed from Tensorflow? Is this a divorce or are TF &amp; Keras just opening up their relationship?

Wouldn&#x27;t a multi-framework wrapper be a subset of any supported framework&#x27;s features common among all frameworks?<p>Additionally would it always be at least a step behind any framework depending on the wrapper&#x27;s release cycle?

I must admit i&#x27;ve never actually used keras but this is interesting to see how they are implementing it with Jax, definitely worth a reminder to dig into one of these days.

demo on JAX: <a href="https:&#x2F;&#x2F;bit.ly&#x2F;keras-on-jax-demo" rel="nofollow noreferrer">https:&#x2F;&#x2F;bit.ly&#x2F;keras-on-jax-demo</a>

Like everything keras, this will promise a lot and only deliver on conops deemed worthy by the keras team.

will keras be backward compatible, or as always and now google&#x2F;tf ecosystem will have 3 gens of frameworks: tf1, tf2 + keras, keras core 3.