Sergey Kurdakov

Results 22 comments of Sergey Kurdakov
trafficstars

What will happen with layers which are in tfa, but not in other keras frameworks and which do not work with Keras 3 (I'm intrested in WeightNormalization layer)

with functional Model while weights could be set (considering fix to serialization in previous comment so that _naked_clone_layer could be built and also removal of as_list in computed shape )...

What will happen with layers which are in tfa, but not in other keras frameworks and which do not work with Keras 3 (I'm intrested in WeightNormalization layer)

finally changing `with tf.control_dependencies(self._initialize_weights(inputs)):` to `with tf.control_dependencies([self._initialize_weights(inputs)]):` I see tensorflow running