←back to thread

Francois Chollet is leaving Google

(developers.googleblog.com)
377 points xnx | 1 comments | 13 Nov 24 22:28 UTC | HN request time: 0.271s | source
Show context
tadeegan ◴[13 Nov 24 23:18 UTC] No.42131302[source]▶
I guess they realized muilti-backend keras is futile? I never liked the tf.keras apis and the docs always promosed multi backend but then I guess they were never able to deliver that without breaking keras 3 changes. And even now.... "Keras 3 includes a brand new distribution API, the keras.distribution namespace, currently implemented for the JAX backend (coming soon to the TensorFlow and PyTorch backends)". I don't believe it. They are too different to reconcile under 1 api. And even if you could, I dont really see the benefit. Torch and Flax have similar goals to Keras and are imo better.
replies(3): >>42131553 #>>42131787 #>>42134352 #
1. fchollet ◴[14 Nov 24 09:06 UTC] No.42134352[source]▶
Actually, `keras.distribution` is straightforward to implement in TF DTensor and with the experimental PyTorch SPMD API. We haven't done it yet first because these APIs are experimental (only JAX is mature) and second because all the demand for large-model distribution at Google was towards the JAX backend.