2
$\begingroup$

I am working to create a Keras custom layer. this layer needs to use a function that doesn't exist in Keras TF backend, although it exist in TF.

What would be the best way to solve it?

  1. Is there a way to call TF function directly from Keras and not from backend?
  2. Should I work to modify backend to include a new unsupported function?

Any example or pointer how to implement the best solution.

Thank you

$\endgroup$
1
1
$\begingroup$

There is no possible way to directly call tensorflow functions from keras. Keras is a stand-alone project. Although, tensorflow >= 1.12 has keras built in, so you could play with the source of tensorflow and figure it out.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.