GitLab

PiperOrigin-RevId: 322844988

PiperOrigin-RevId: 322828673

PiperOrigin-RevId: 322801804

PiperOrigin-RevId: 322713374

PiperOrigin-RevId: 322707701

PiperOrigin-RevId: 322657835

PiperOrigin-RevId: 322594580

Explicit set the dtype in embedding layer. Otherwise, using the layer in TF1.x will create an int32 tensor.

PiperOrigin-RevId: 322473236

(1) call() consume kwargs and implement _build_from_signature (layers added inside init_scope)
(2) make build/call_attention as public.

PiperOrigin-RevId: 322454619

PiperOrigin-RevId: 322428923

PiperOrigin-RevId: 322415899

PiperOrigin-RevId: 322387911

* update readme according to template

* initial upgrades

* demo_inference updated, need to check tf.flags

* issues with tf.image

* manual checks needed for tf.contrib

* more tf2 updates, need to fix tf.contrib and tf.flags

PiperOrigin-RevId: 322244056

PiperOrigin-RevId: 322234532

PiperOrigin-RevId: 322234001

Make hack in official/vision/detection models that enters the backend keras graph stop happening once we enable the Functional API KerasTensors refactoring:

As a workaround hack for the tf op layer conversion being fragile, the detection models have to explicitly enter the Keras backend graph.

When we enable the KerasTensors refactoring of the Functional API internals, the op layer conversion will be much more reliable and this hack will not be necessary. In addition, the hack actually causes the models to break when we enable the refactoring (because it causes tensors to leak out of a graph).

So, this CL changes the existing hack to stop applying once we've enabled the KerasTensors refactoring.

PiperOrigin-RevId: 322229802

PiperOrigin-RevId: 322216928

PiperOrigin-RevId: 322214979

PiperOrigin-RevId: 322197751

Modify the unit test of encoder scaffold to demonstrate that a subclass embedding network can be used.

PiperOrigin-RevId: 322083775