Merge pull request #39252 from PiyushDatta:datapi_tflow

PiperOrigin-RevId: 338686808
Change-Id: I9faf849a3dce26d452f506c569caefd67a01e838

tensorflow/python/keras/BUILD

@@ -130,6 +130,7 @@ py_library(
     srcs_version = "PY2AND3",
     deps = [
         ":backend",
+        "//tensorflow/python/keras/layers:advanced_activations",
         "//tensorflow/python/keras/utils:engine_utils",
     ],
 )

tensorflow/python/keras/activations.py

@@ -26,6 +26,7 @@ from tensorflow.python.ops import math_ops
 from tensorflow.python.ops import nn
 from tensorflow.python.util import dispatch
 from tensorflow.python.util.tf_export import keras_export
+from tensorflow.python.keras.layers import advanced_activations
 
 # b/123041942
 # In TF 2.x, if the `tf.nn.softmax` is used as an activation function in Keras
@@ -529,9 +530,17 @@ def deserialize(name, custom_objects=None):
       ValueError: `Unknown activation function` if the input string does not
       denote any defined Tensorflow activation function.
   """
+  globs = globals()
+
+  # only replace missing activations
+  advanced_activations_globs = advanced_activations.get_globals()
+  for key, val in advanced_activations_globs.items():
+    if key not in globs:
+      globs[key] = val
+
   return deserialize_keras_object(
       name,
-      module_objects=globals(),
+      module_objects=globs,
       custom_objects=custom_objects,
       printable_module_name='activation function')
 

tensorflow/python/keras/activations_test.py

@@ -65,12 +65,19 @@ class KerasActivationsTest(test.TestCase, parameterized.TestCase):
     activation = advanced_activations.LeakyReLU(alpha=0.1)
     layer = core.Dense(3, activation=activation)
     config = serialization.serialize(layer)
+    # with custom objects
     deserialized_layer = serialization.deserialize(
         config, custom_objects={'LeakyReLU': activation})
     self.assertEqual(deserialized_layer.__class__.__name__,
                      layer.__class__.__name__)
     self.assertEqual(deserialized_layer.activation.__class__.__name__,
                      activation.__class__.__name__)
+    # without custom objects
+    deserialized_layer = serialization.deserialize(config)
+    self.assertEqual(deserialized_layer.__class__.__name__,
+                     layer.__class__.__name__)
+    self.assertEqual(deserialized_layer.activation.__class__.__name__,
+                     activation.__class__.__name__)
 
   def test_softmax(self):
     x = backend.placeholder(ndim=2)

tensorflow/python/keras/layers/advanced_activations.py

@@ -30,6 +30,10 @@ from tensorflow.python.ops import math_ops
 from tensorflow.python.util.tf_export import keras_export
 
 
+def get_globals():
+  return globals()
+
+
 @keras_export('keras.layers.LeakyReLU')
 class LeakyReLU(Layer):
   """Leaky version of a Rectified Linear Unit.