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

Clean up the ResourceVariable inheritance hierarchy a bit.

Browse Source 
PiperOrigin-RevId: 253592362
This commit is contained in:
Alexandre Passos 2019-06-17 09:08:06 -07:00 committed by TensorFlower Gardener
parent a5b860a7cc
commit aa93ea6441
13 changed files with 649 additions and 585 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tensorflow/contrib/eager/python/parameter_server.py
View File

@ -63,7 +63,7 @@ def _eager_safe_variable_handle(shape, dtype, shared_name, name, graph_mode):
return handle return handle
class SharedVariable(resource_variable_ops.ResourceVariable): class SharedVariable(resource_variable_ops.BaseResourceVariable):
"""Experimental Variable designed for parameter server training. """Experimental Variable designed for parameter server training.
A SharedVariable has a name and two instances of SharedVariable with the A SharedVariable has a name and two instances of SharedVariable with the
@ -231,7 +231,7 @@ class SharedVariable(resource_variable_ops.ResourceVariable):
self._graph_element = None self._graph_element = None
self._cached_value = None self._cached_value = None
self._handle_deleter = None self._handle_deleter = object()
self._cached_shape_as_list = None self._cached_shape_as_list = None

4
tensorflow/python/distribute/cross_device_ops.py
View File

@ -48,7 +48,7 @@ def check_destinations(destinations):
Boolean which is True if `destinations` is not empty. Boolean which is True if `destinations` is not empty.
""" """
# Calling bool() on a ResourceVariable is not allowed. # Calling bool() on a ResourceVariable is not allowed.
if isinstance(destinations, resource_variable_ops.ResourceVariable): if isinstance(destinations, resource_variable_ops.BaseResourceVariable):
return bool(destinations.device) return bool(destinations.device)
return bool(destinations) return bool(destinations)
@ -56,7 +56,7 @@ def check_destinations(destinations):
def validate_destinations(destinations): def validate_destinations(destinations):
if not isinstance(destinations, if not isinstance(destinations,
(value_lib.DistributedValues, (value_lib.DistributedValues,
resource_variable_ops.ResourceVariable, resource_variable_ops.BaseResourceVariable,
value_lib.AggregatingVariable, value_lib.AggregatingVariable,
six.string_types, six.string_types,
value_lib.TPUMirroredVariable, value_lib.TPUMirroredVariable,

2
tensorflow/python/distribute/parameter_server_strategy.py
View File

@ -479,7 +479,7 @@ class ParameterServerStrategyExtended(distribute_lib.StrategyExtendedV1):
def _update(self, var, fn, args, kwargs, group): def _update(self, var, fn, args, kwargs, group):
if isinstance(var, values.AggregatingVariable): if isinstance(var, values.AggregatingVariable):
var = var.get() var = var.get()
if not isinstance(var, resource_variable_ops.ResourceVariable): if not isinstance(var, resource_variable_ops.BaseResourceVariable):
raise ValueError( raise ValueError(
"You can not update `var` %r. It must be a Variable." % var) "You can not update `var` %r. It must be a Variable." % var)
with ops.colocate_with(var), distribute_lib.UpdateContext(var.device): with ops.colocate_with(var), distribute_lib.UpdateContext(var.device):

4
tensorflow/python/distribute/tpu_strategy.py
View File

@ -519,7 +519,7 @@ class TPUExtended(distribute_lib.StrategyExtendedV1):
def _update(self, var, fn, args, kwargs, group): def _update(self, var, fn, args, kwargs, group):
assert isinstance(var, values.TPUMirroredVariable) or isinstance( assert isinstance(var, values.TPUMirroredVariable) or isinstance(
var, resource_variable_ops.ResourceVariable) var, resource_variable_ops.BaseResourceVariable)
if values._enclosing_tpu_context() is not None: # pylint: disable=protected-access if values._enclosing_tpu_context() is not None: # pylint: disable=protected-access
if group: if group:
return fn(var, *args, **kwargs) return fn(var, *args, **kwargs)
@ -540,7 +540,7 @@ class TPUExtended(distribute_lib.StrategyExtendedV1):
def read_var(self, var): def read_var(self, var):
assert isinstance(var, values.TPUMirroredVariable) or isinstance( assert isinstance(var, values.TPUMirroredVariable) or isinstance(
var, resource_variable_ops.ResourceVariable) var, resource_variable_ops.BaseResourceVariable)
return var.read_value() return var.read_value()
def _local_results(self, val): def _local_results(self, val):

2
tensorflow/python/eager/def_function.py
View File

@ -111,6 +111,8 @@ class UnliftedInitializerVariable(resource_variable_ops.UninitializedVariable):
shape and `validate_shape` is `True`. shape and `validate_shape` is `True`.
RuntimeError: If called outside of a function definition. RuntimeError: If called outside of a function definition.
""" """
with ops.init_scope():
self._in_graph_mode = not context.executing_eagerly()
if not ops.inside_function(): if not ops.inside_function():
# If we've been init_scope()d out of the function definition nothing to do # If we've been init_scope()d out of the function definition nothing to do
# here; we can't really do the capturing or conditional logic. # here; we can't really do the capturing or conditional logic.

4
tensorflow/python/eager/function.py
View File

@ -598,7 +598,7 @@ class ConcreteFunction(object):
return self._call_flat( return self._call_flat(
(t for t in nest.flatten((args, kwargs), expand_composites=True) (t for t in nest.flatten((args, kwargs), expand_composites=True)
if isinstance(t, (ops.Tensor, if isinstance(t, (ops.Tensor,
resource_variable_ops.ResourceVariable))), resource_variable_ops.BaseResourceVariable))),
self.captured_inputs) self.captured_inputs)
def _call_flat(self, args, captured_inputs): def _call_flat(self, args, captured_inputs):
@ -632,7 +632,7 @@ class ConcreteFunction(object):
tensor_inputs = [] tensor_inputs = []
variables_used = set([]) variables_used = set([])
for i, arg in enumerate(args): for i, arg in enumerate(args):
if isinstance(arg, resource_variable_ops.ResourceVariable): if isinstance(arg, resource_variable_ops.BaseResourceVariable):
# We can pass a variable more than once, and in this case we need to # We can pass a variable more than once, and in this case we need to
# pass its handle only once. # pass its handle only once.
if arg.handle in variables_used: if arg.handle in variables_used:

2
tensorflow/python/eager/wrap_function.py
View File

@ -156,7 +156,7 @@ def _lift_unlifted_variables(graph, variable_holder):
def _should_lift_variable(v): def _should_lift_variable(v):
return ((v._in_graph_mode # pylint: disable=protected-access return ((v._in_graph_mode # pylint: disable=protected-access
and v.graph.building_function) and v.graph.building_function)
and isinstance(v, resource_variable_ops.ResourceVariable) and isinstance(v, resource_variable_ops.BaseResourceVariable)
and v.handle not in existing_captures) and v.handle not in existing_captures)
for old_variable in global_collection_variables: for old_variable in global_collection_variables:

4
tensorflow/python/framework/func_graph.py
View File

@ -794,7 +794,7 @@ def func_graph_from_py_func(name,
inputs = [] inputs = []
for arg in (nest.flatten(func_args, expand_composites=True) + for arg in (nest.flatten(func_args, expand_composites=True) +
nest.flatten(func_kwargs, expand_composites=True)): nest.flatten(func_kwargs, expand_composites=True)):
if isinstance(arg, resource_variable_ops.ResourceVariable): if isinstance(arg, resource_variable_ops.BaseResourceVariable):
# Even if an argument variable was not used in the function, we've # Even if an argument variable was not used in the function, we've
# already manually captured the resource Tensor when creating argument # already manually captured the resource Tensor when creating argument
# placeholders. # placeholders.
@ -1003,7 +1003,7 @@ def _get_defun_inputs(args, names, structure, flat_shapes=None):
"_user_specified_name", "_user_specified_name",
attr_value_pb2.AttrValue(s=compat.as_bytes(requested_name))) attr_value_pb2.AttrValue(s=compat.as_bytes(requested_name)))
function_inputs.append(placeholder) function_inputs.append(placeholder)
elif isinstance(arg, resource_variable_ops.ResourceVariable): elif isinstance(arg, resource_variable_ops.BaseResourceVariable):
# Capture arg variables to create placeholders for them. These will be # Capture arg variables to create placeholders for them. These will be
# removed as captures after the function is traced (since otherwise we'd # removed as captures after the function is traced (since otherwise we'd
# just add it back with a new placeholder when the variable was # just add it back with a new placeholder when the variable was

2
tensorflow/python/framework/function.py
View File

@ -790,7 +790,7 @@ class _FuncGraph(ops.Graph):
collections=collections, collections=collections,
use_resource=use_resource) use_resource=use_resource)
self.extra_vars.append(var) self.extra_vars.append(var)
if (isinstance(var, resource_variable_ops.ResourceVariable) and if (isinstance(var, resource_variable_ops.BaseResourceVariable) and
self._capture_resource_var_by_value): self._capture_resource_var_by_value):
# For resource-based variables read the variable outside the function # For resource-based variables read the variable outside the function
# and pass in the value. This ensures that the function is pure and # and pass in the value. This ensures that the function is pure and

1177
tensorflow/python/ops/resource_variable_ops.py
View File

File diff suppressed because it is too large Load Diff

2
tensorflow/python/saved_model/function_deserialization.py
View File

@ -38,7 +38,7 @@ from tensorflow.python.util import tf_inspect
def _is_tensor(t): def _is_tensor(t):
return isinstance(t, (ops.Tensor, resource_variable_ops.ResourceVariable)) return isinstance(t, (ops.Tensor, resource_variable_ops.BaseResourceVariable))
def _call_concrete_function(function, inputs): def _call_concrete_function(function, inputs):

9
tensorflow/python/training/optimizer.py
View File

@ -604,9 +604,9 @@ class Optimizer(
# We colocate all ops created in _apply_dense or _apply_sparse # We colocate all ops created in _apply_dense or _apply_sparse
# on the same device as the variable. # on the same device as the variable.
# TODO(apassos): figure out how to get the variable name here. # TODO(apassos): figure out how to get the variable name here.
if context.executing_eagerly() or isinstance( if (context.executing_eagerly() or
var, isinstance(var, resource_variable_ops.BaseResourceVariable)
resource_variable_ops.ResourceVariable) and not var._in_graph_mode: # pylint: disable=protected-access and not var._in_graph_mode): # pylint: disable=protected-access
scope_name = "" scope_name = ""
else: else:
scope_name = var.op.name scope_name = var.op.name
@ -617,7 +617,8 @@ class Optimizer(
else: else:
with ops.control_dependencies([self._finish(update_ops, "update")]): with ops.control_dependencies([self._finish(update_ops, "update")]):
with ops.colocate_with(global_step): with ops.colocate_with(global_step):
if isinstance(global_step, resource_variable_ops.ResourceVariable): if isinstance(
global_step, resource_variable_ops.BaseResourceVariable):
# TODO(apassos): the implicit read in assign_add is slow; consider # TODO(apassos): the implicit read in assign_add is slow; consider
# making it less so. # making it less so.
apply_updates = resource_variable_ops.assign_add_variable_op( apply_updates = resource_variable_ops.assign_add_variable_op(

8
tensorflow/python/training/saving/saveable_object_util.py
View File

@ -179,7 +179,7 @@ def saveable_objects_for_op(op, name):
# pylint: enable=protected-access # pylint: enable=protected-access
else: else:
# A variable or tensor. # A variable or tensor.
if isinstance(op, resource_variable_ops.ResourceVariable): if isinstance(op, resource_variable_ops.BaseResourceVariable):
# pylint: disable=protected-access # pylint: disable=protected-access
if op._in_graph_mode: if op._in_graph_mode:
variable = op._graph_element variable = op._graph_element
@ -233,7 +233,7 @@ def op_list_to_dict(op_list, convert_variable_to_tensor=True):
# pylint: disable=protected-access # pylint: disable=protected-access
for var in op_list: for var in op_list:
resource_or_ref_variable = ( resource_or_ref_variable = (
isinstance(var, resource_variable_ops.ResourceVariable) or isinstance(var, resource_variable_ops.BaseResourceVariable) or
isinstance(var, variables.RefVariable)) isinstance(var, variables.RefVariable))
if isinstance(var, saveable_object.SaveableObject): if isinstance(var, saveable_object.SaveableObject):
@ -263,7 +263,7 @@ def op_list_to_dict(op_list, convert_variable_to_tensor=True):
# indicating whether they were created in a graph building context. We # indicating whether they were created in a graph building context. We
# also get Tensors when graph building, which do not have this property. # also get Tensors when graph building, which do not have this property.
if not getattr(var, "_in_graph_mode", True): if not getattr(var, "_in_graph_mode", True):
if not isinstance(var, resource_variable_ops.ResourceVariable): if not isinstance(var, resource_variable_ops.BaseResourceVariable):
raise ValueError( raise ValueError(
"Can only save/restore ResourceVariables when eager execution " "Can only save/restore ResourceVariables when eager execution "
"is enabled, type: %s." % type(var)) "is enabled, type: %s." % type(var))
@ -277,7 +277,7 @@ def op_list_to_dict(op_list, convert_variable_to_tensor=True):
(var._shared_name,)) (var._shared_name,))
else: else:
if convert_variable_to_tensor: if convert_variable_to_tensor:
if isinstance(var, resource_variable_ops.ResourceVariable): if isinstance(var, resource_variable_ops.BaseResourceVariable):
var = var._graph_element # pylint: disable=protected-access var = var._graph_element # pylint: disable=protected-access
else: else:
var = ops.internal_convert_to_tensor(var, as_ref=True) var = ops.internal_convert_to_tensor(var, as_ref=True)