experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Remove automatic control dep wrapping from layers in v2.

Browse Source 
PiperOrigin-RevId: 316929712
Change-Id: Ic1a7d125776eeb0c7654e321dd6f2351c8656a16
This commit is contained in:
Pavithra Vijay 2020-06-17 11:32:39 -07:00 committed by TensorFlower Gardener
parent 9eed8b3cf5
commit 653131dd38
1 changed files with 1 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
tensorflow/python/keras/engine/base_layer.py
View File

@ -40,7 +40,6 @@ from tensorflow.python.eager import context
from tensorflow.python.eager import execute
from tensorflow.python.eager import function
from tensorflow.python.eager import monitoring
from tensorflow.python.framework import auto_control_deps
from tensorflow.python.framework import constant_op
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import errors
@ -1105,16 +1104,6 @@ class Layer(module.Module, version_utils.LayerVersionSelector):
try:
with ops.enable_auto_cast_variables(self._compute_dtype_object):
# Add auto_control_deps in V2 when they are not already added by
# a `tf.function`.
if (ops.executing_eagerly_outside_functions() and
not base_layer_utils.is_in_eager_or_tf_function()):
with auto_control_deps.AutomaticControlDependencies() as acd:
outputs = call_fn(cast_inputs, *args, **kwargs)
# Wrap Tensors in `outputs` in `tf.identity` to avoid
# circular dependencies.
outputs = base_layer_utils.mark_as_return(outputs, acd)
else:
outputs = call_fn(cast_inputs, *args, **kwargs)
except errors.OperatorNotAllowedInGraphError as e: