Rename references to "K" as "backend" for consistency.

keras/backend.py

@@ -1661,9 +1661,8 @@ def zeros_like(x, dtype=None, name=None):
   Example:
 
   ```python
-  from tensorflow.keras import backend as K
-  kvar = K.variable(np.random.random((2,3)))
-  kvar_zeros = K.zeros_like(kvar)
+  kvar = tf.keras.backend.variable(np.random.random((2,3)))
+  kvar_zeros = tf.keras.backend.zeros_like(kvar)
   K.eval(kvar_zeros)
   # array([[ 0.,  0.,  0.], [ 0.,  0.,  0.]], dtype=float32)
   ```

keras/callbacks_v1.py

@@ -20,7 +20,7 @@
 
 import os
 import numpy as np
-from keras import backend as K
+from keras import backend
 from keras import callbacks
 from tensorflow.python.platform import tf_logging as logging
 from tensorflow.python.util.tf_export import keras_export
@@ -160,9 +160,10 @@ def _init_writer(self, model):
       self.writer = tf.summary.create_file_writer(self.log_dir)
       if not model.run_eagerly and self.write_graph:
         with self.writer.as_default():
-          tf.summary.graph(K.get_graph())
+          tf.summary.graph(backend.get_graph())
     elif self.write_graph:
-      self.writer = tf.compat.v1.summary.FileWriter(self.log_dir, K.get_graph())
+      self.writer = tf.compat.v1.summary.FileWriter(
+          self.log_dir, backend.get_graph())
     else:
       self.writer = tf.compat.v1.summary.FileWriter(self.log_dir)
 
@@ -176,27 +177,26 @@ def _make_histogram_ops(self, model):
           tf.compat.v1.summary.histogram(mapped_weight_name, weight)
           if self.write_images:
             w_img = tf.compat.v1.squeeze(weight)
-            shape = K.int_shape(w_img)
+            shape = tuple(w_img.shape)
             if len(shape) == 2:  # dense layer kernel case
               if shape[0] > shape[1]:
                 w_img = tf.compat.v1.transpose(w_img)
-                shape = K.int_shape(w_img)
+                shape = tuple(w_img.shape)
               w_img = tf.reshape(w_img, [1, shape[0], shape[1], 1])
             elif len(shape) == 3:  # convnet case
-              if K.image_data_format() == 'channels_last':
+              if backend.image_data_format() == 'channels_last':
                 # switch to channels_first to display
                 # every kernel as a separate image
                 w_img = tf.compat.v1.transpose(w_img, perm=[2, 0, 1])
-                shape = K.int_shape(w_img)
-              w_img = tf.reshape(w_img,
-                                        [shape[0], shape[1], shape[2], 1])
+                shape = tuple(w_img.shape)
+              w_img = tf.reshape(w_img, [shape[0], shape[1], shape[2], 1])
             elif len(shape) == 1:  # bias case
               w_img = tf.reshape(w_img, [1, shape[0], 1, 1])
             else:
               # not possible to handle 3D convnets etc.
               continue
 
-            shape = K.int_shape(w_img)
+            shape = tuple(w_img.shape)
             assert len(shape) == 4 and shape[-1] in [1, 3, 4]
             tf.compat.v1.summary.image(mapped_weight_name, w_img)
 
@@ -421,7 +421,7 @@ def on_epoch_end(self, epoch, logs=None):
         embeddings_data = self.embeddings_data
         n_samples = embeddings_data[0].shape[0]
         i = 0
-        sess = K.get_session()
+        sess = backend.get_session()
         while i < n_samples:
           step = min(self.batch_size, n_samples - i)
           batch = slice(i, i + step)
@@ -436,8 +436,8 @@ def on_epoch_end(self, epoch, logs=None):
 
           feed_dict.update({self.batch_id: i, self.step: step})
 
-          if not isinstance(K.learning_phase(), int):
-            feed_dict[K.learning_phase()] = False
+          if not isinstance(backend.learning_phase(), int):
+            feed_dict[backend.learning_phase()] = False
 
           sess.run(self.assign_embeddings, feed_dict=feed_dict)
           self.saver.save(sess,

keras/layers/core/dense.py

@@ -16,7 +16,7 @@
 # pylint: disable=g-classes-have-attributes,g-direct-tensorflow-import
 
 from keras import activations
-from keras import backend as K
+from keras import backend
 from keras import constraints
 from keras import initializers
 from keras import regularizers
@@ -128,7 +128,7 @@ def __init__(self,
     self.supports_masking = True
 
   def build(self, input_shape):
-    dtype = tf.as_dtype(self.dtype or K.floatx())
+    dtype = tf.as_dtype(self.dtype or backend.floatx())
     if not (dtype.is_floating or dtype.is_complex):
       raise TypeError('A Dense layer can only be built with a floating-point '
                       f'dtype. Received: dtype={dtype}')

keras/layers/core/dropout.py

@@ -15,7 +15,7 @@
 """Contains the dropout layer."""
 # pylint: disable=g-classes-have-attributes,g-direct-tensorflow-import
 
-from keras import backend as K
+from keras import backend
 from keras.engine.base_layer import Layer
 from keras.utils import control_flow_util
 import tensorflow.compat.v2 as tf
@@ -110,7 +110,7 @@ def _get_noise_shape(self, inputs):
 
   def call(self, inputs, training=None):
     if training is None:
-      training = K.learning_phase()
+      training = backend.learning_phase()
 
     def dropped_inputs():
       return tf.nn.dropout(

keras/layers/core/masking.py

@@ -15,7 +15,6 @@
 """Contains the Masking layer."""
 # pylint: disable=g-classes-have-attributes,g-direct-tensorflow-import
 
-from keras import backend as K
 from keras.engine.base_layer import Layer
 import tensorflow.compat.v2 as tf
 from tensorflow.python.util.tf_export import keras_export
@@ -69,10 +68,10 @@ def __init__(self, mask_value=0., **kwargs):
     self._compute_output_and_mask_jointly = True
 
   def compute_mask(self, inputs, mask=None):
-    return K.any(tf.not_equal(inputs, self.mask_value), axis=-1)
+    return tf.reduce_any(tf.not_equal(inputs, self.mask_value), axis=-1)
 
   def call(self, inputs):
-    boolean_mask = K.any(
+    boolean_mask = tf.reduce_any(
         tf.not_equal(inputs, self.mask_value), axis=-1, keepdims=True)
     outputs = inputs * tf.cast(boolean_mask, inputs.dtype)
     # Compute the mask and outputs simultaneously.

keras/layers/core/repeat_vector.py

@@ -15,7 +15,7 @@
 """Contains the RepeatVector layer."""
 # pylint: disable=g-classes-have-attributes,g-direct-tensorflow-import
 
-from keras import backend as K
+from keras import backend
 from keras.engine.base_layer import Layer
 from keras.engine.input_spec import InputSpec
 import tensorflow.compat.v2 as tf
@@ -57,7 +57,7 @@ def compute_output_shape(self, input_shape):
     return tf.TensorShape([input_shape[0], self.n, input_shape[1]])
 
   def call(self, inputs):
-    return K.repeat(inputs, self.n)
+    return backend.repeat(inputs, self.n)
 
   def get_config(self):
     config = {'n': self.n}

keras/layers/core/tf_op_layer.py

@@ -17,7 +17,7 @@
 import tensorflow.compat.v2 as tf
 # pylint: enable=g-bad-import-order
 
-from keras import backend as K
+from keras import backend
 from keras.engine import keras_tensor
 from keras.engine.base_layer import Layer
 
@@ -48,7 +48,7 @@ def __init__(self, cls_ref, method_name, **kwargs):
         get_canonical_name_for_symbol(
             self.cls_ref, api_name='keras', add_prefix_to_v1_names=True))
     if 'name' not in kwargs:
-      kwargs['name'] = K.unique_object_name(
+      kwargs['name'] = backend.unique_object_name(
           'tf.' + self.cls_symbol + '.' + self.method_name,
           zero_based=True,
           avoid_observed_names=True)
@@ -134,7 +134,7 @@ def __init__(self, attr_name, **kwargs):
     self.attr_name = attr_name
 
     if 'name' not in kwargs:
-      kwargs['name'] = K.unique_object_name(
+      kwargs['name'] = backend.unique_object_name(
           'input.' + self.attr_name, zero_based=True, avoid_observed_names=True)
     kwargs['autocast'] = False
 
@@ -217,7 +217,7 @@ def __init__(self, function, **kwargs):
         name = 'tf.' + self.symbol
       else:
         name = self.function.__name__
-      kwargs['name'] = K.unique_object_name(
+      kwargs['name'] = backend.unique_object_name(
           name, zero_based=True, avoid_observed_names=True)
     kwargs['autocast'] = False
 

keras/saving/saved_model/save.py

@@ -18,7 +18,7 @@
 
 import os
 
-from keras import backend as K
+from keras import backend
 from keras.protobuf import saved_metadata_pb2
 from keras.protobuf import versions_pb2
 from keras.saving import saving_utils
@@ -85,7 +85,7 @@ def save(model, filepath, overwrite, include_optimizer, signatures=None,
   # already-set learning phase placeholder.
   # This is needed for compatibility reasons until learning phase setting
   # is removed from the public apis.
-  with K.deprecated_internal_learning_phase_scope(0):
+  with backend.deprecated_internal_learning_phase_scope(0):
     with utils.keras_option_scope(save_traces):
       saved_nodes, node_paths = save_lib.save_and_return_nodes(
           model, filepath, signatures, options)

keras/saving/saved_model/save_impl.py

@@ -22,7 +22,7 @@
 import threading
 import weakref
 
-from keras import backend as K
+from keras import backend
 from keras.engine import base_layer_utils
 from keras.engine import input_spec
 from keras.mixed_precision import autocast_variable
@@ -355,7 +355,7 @@ def tracing_scope():
     while _thread_local_data.trace_queue:
       fn, args, kwargs, training = _thread_local_data.trace_queue.pop()
       if training is not None:
-        with K.deprecated_internal_learning_phase_scope(training):
+        with backend.deprecated_internal_learning_phase_scope(training):
           fn.get_concrete_function(*args, **kwargs)
       else:
         fn.get_concrete_function(*args, **kwargs)
@@ -694,7 +694,7 @@ def _wrap_activity_regularizer(layer):
       layer._activity_regularizer,
       '{}_activity_regularizer'.format(layer.name),
       input_signature=[
-          tf.TensorSpec(None, layer._compute_dtype or K.floatx())
+          tf.TensorSpec(None, layer._compute_dtype or backend.floatx())
       ])
   # pylint: enable=protected-access
 

keras/saving/saved_model/utils.py

@@ -19,7 +19,7 @@
 import itertools
 import threading
 import types
-from keras import backend as K
+from keras import backend
 from keras.engine import base_layer_utils
 from keras.utils import control_flow_util
 from keras.utils import tf_contextlib
@@ -43,7 +43,7 @@ def use_wrapped_call(layer, call_fn, default_training_value=None,
     call_fn: tf.function that takes layer inputs (and possibly a training arg),
       and returns a tuple of (outputs, list of losses).
     default_training_value: Default value of the training kwarg. If `None`, the
-      default is `K.learning_phase()`.
+      default is `tf.keras.backend.learning_phase()`.
     return_method: Whether to return a method bound to the layer.
 
   Returns:
@@ -138,7 +138,8 @@ def maybe_add_training_arg(
     wrapped_call: Wrapped call function.
     expects_training_arg: Whether to include 'training' argument.
     default_training_value: Default value of the training kwarg to include in
-      the arg spec. If `None`, the default is `K.learning_phase()`.
+      the arg spec. If `None`, the default is
+      `tf.keras.backend.learning_phase()`.
 
   Returns:
     Tuple of (
@@ -152,7 +153,7 @@ def wrap_with_training_arg(*args, **kwargs):
     training_arg_index = get_training_arg_index(original_call)
     training = get_training_arg(training_arg_index, args, kwargs)
     if training is None:
-      training = default_training_value or K.learning_phase()
+      training = default_training_value or backend.learning_phase()
 
     args = list(args)
     kwargs = kwargs.copy()

keras/saving/saving_utils.py

@@ -19,7 +19,7 @@
 
 import copy
 import os
-from keras import backend as K
+from keras import backend
 from keras import losses
 from keras import optimizer_v1
 from keras import optimizers
@@ -149,7 +149,7 @@ def model_metadata(model, include_optimizer=True, require_config=True):
 
   metadata = dict(
       keras_version=str(keras_version),
-      backend=K.backend(),
+      backend=backend.backend(),
       model_config=model_config)
   if model.optimizer and include_optimizer:
     if isinstance(model.optimizer, optimizer_v1.TFOptimizer):

keras/utils/tf_utils.py

@@ -20,7 +20,7 @@
 import copy
 import numpy as np
 from tensorflow.python.framework import ops
-from keras import backend as K
+from keras import backend
 from keras.engine import keras_tensor
 from keras.utils import object_identity
 from keras.utils import tf_contextlib
@@ -438,7 +438,7 @@ def maybe_init_scope(layer):
 def graph_context_for_symbolic_tensors(*args, **kwargs):
   """Returns graph context manager if any of the inputs is a symbolic tensor."""
   if any(is_symbolic_tensor(v) for v in list(args) + list(kwargs.values())):
-    with K.get_graph().as_default():
+    with backend.get_graph().as_default():
       yield
   else:
     yield
@@ -450,7 +450,8 @@ def dataset_is_infinite(dataset):
     return tf.equal(
         tf.data.experimental.cardinality(dataset), tf.data.experimental.INFINITE_CARDINALITY)
   else:
-    dataset_size = K.get_session().run(tf.data.experimental.cardinality(dataset))
+    dataset_size = backend.get_session().run(
+        tf.data.experimental.cardinality(dataset))
     return dataset_size == tf.data.experimental.INFINITE_CARDINALITY