# Copyright 2015 The TensorFlow Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== """A class to store named variables and a scope operator to manage sharing.""" from __future__ import absolute_import from __future__ import division from __future__ import print_function import collections as collections_lib import copy import enum # pylint: disable=g-bad-import-order import functools import sys import threading import traceback import six from six import iteritems from six.moves import xrange # pylint: disable=redefined-builtin from tensorflow.python.eager import context from tensorflow.python.framework import dtypes from tensorflow.python.framework import ops from tensorflow.python.framework import tensor_shape from tensorflow.python.ops import array_ops from tensorflow.python.ops import init_ops from tensorflow.python.ops import resource_variable_ops from tensorflow.python.ops import variables from tensorflow.python.platform import tf_logging as logging from tensorflow.python.util import function_utils from tensorflow.python.util import tf_contextlib from tensorflow.python.util.tf_export import tf_export __all__ = [ "AUTO_REUSE", "VariableScope", "get_variable_scope", "get_variable", "get_local_variable", "variable_scope", "variable_op_scope", "no_regularizer", "VariableSynchronization", "VariableAggregation" ] class _PartitionInfo(object): """Holds partition info used by initializer functions. """ def __init__(self, full_shape, var_offset): """Constructor. Args: full_shape: Tuple or list of `int` indicating the full combined shape of the partitioned variables. var_offset: Tuple or list of `int` specifying offset of this partition with respect to the full variable for each dimension. Raises: TypeError: If `full_shape` or `var_offset` is not a sequence. ValueError: If `full_shape` or `var_offset` differ in length. If `var_offset` exceeds `full_shape` in any dimension. """ if not isinstance(full_shape, collections_lib.Sequence) or isinstance( full_shape, six.string_types): raise TypeError( "`full_shape` must be a sequence (like tuple or list) instead of " + type(full_shape).__name__) if not isinstance(var_offset, collections_lib.Sequence) or isinstance( var_offset, six.string_types): raise TypeError( "`var_offset` must be a sequence (like tuple or list) instead of " + type(var_offset).__name__) if len(var_offset) != len(full_shape): raise ValueError( "Expected equal length, but `var_offset` is of length {} while " "full_shape is of length {}.".format( len(var_offset), len(full_shape))) for i in xrange(len(full_shape)): offset = var_offset[i] shape = full_shape[i] if offset < 0 or offset >= shape: raise ValueError( "Expected 0 <= offset < shape but found offset={}, shape={} for " "var_offset={}, full_shape={}".format(offset, shape, var_offset, full_shape)) self._full_shape = full_shape self._var_offset = var_offset @property def full_shape(self): return self._full_shape @property def var_offset(self): return self._var_offset def single_offset(self, shape): """Returns the offset when the variable is partitioned in at most one dim. Args: shape: Tuple or list of `int` indicating the shape of one specific variable partition. Returns: `int` representing the offset in the dimension along which the variable is partitioned. Returns 0 if the variable is not being partitioned. Raises: ValueError: Depending on self.single_slice_dim(). """ single_slice_dim = self.single_slice_dim(shape) # If this variable is not being partitioned at all, single_slice_dim() could # return None. if single_slice_dim is None: return 0 return self.var_offset[single_slice_dim] def single_slice_dim(self, shape): """Returns the slice dim when the variable is partitioned only in one dim. Args: shape: Tuple or list of `int` indicating the shape of one specific variable partition. Returns: `int` representing the dimension that the variable is partitioned in, or `None` if the variable doesn't seem to be partitioned at all. Raises: TypeError: If `shape` is not a sequence. ValueError: If `shape` is not the same length as `self.full_shape`. If the variable is partitioned in more than one dimension. """ if not isinstance(shape, collections_lib.Sequence) or isinstance( shape, six.string_types): raise TypeError( "`shape` must be a sequence (like tuple or list) instead of " + type(shape).__name__) if len(shape) != len(self.full_shape): raise ValueError( "Expected equal length, but received shape={} of length {} while " "self.full_shape={} is of length {}.".format(shape, len( shape), self.full_shape, len(self.full_shape))) for i in xrange(len(shape)): if self.var_offset[i] + shape[i] > self.full_shape[i]: raise ValueError( "With self.var_offset={}, a partition of shape={} would exceed " "self.full_shape={} in dimension {}.".format( self.var_offset, shape, self.full_shape, i)) slice_dim = None for i in xrange(len(shape)): if shape[i] == self.full_shape[i]: continue if slice_dim is not None: raise ValueError( "Cannot use single_slice_dim() with shape={} and " "self.full_shape={} since slice dim could be either dimension {} " "or {}.".format(shape, self.full_shape, i, slice_dim)) slice_dim = i return slice_dim class _ReuseMode(enum.Enum): """Mode for variable access within a variable scope.""" # Indicates that variables are to be fetched if they already exist or # otherwise created. AUTO_REUSE = 1 # TODO(alive): For TensorFlow 2.0, Deprecate True/False/None API in favor of # enum values. # REUSE_FALSE = 2 # REUSE_TRUE = 3 @tf_export("VariableSynchronization") class VariableSynchronization(enum.Enum): """Indicates when a distributed variable will be synced.""" # Indicates that the synchronization will be determined by the current # `DistributionStrategy` (eg. With `MirroredStrategy` this would be # `ON_WRITE`). AUTO = 0 # Indicates that there will only be one copy of the variable, so there is no # need to sync. NONE = 1 # Indicates that the variable will be aggregated across devices # every time it is updated. ON_WRITE = 2 # Indicates that the variable will be aggregated across devices # when it is read (eg. when checkpointing or when evaluating an op that uses # the variable). ON_READ = 3 @tf_export("VariableAggregation") class VariableAggregation(enum.Enum): """Indicates how a distributed variable will be aggregated.""" NONE = 0 SUM = 1 MEAN = 2 AUTO_REUSE = _ReuseMode.AUTO_REUSE tf_export("AUTO_REUSE").export_constant(__name__, "AUTO_REUSE") AUTO_REUSE.__doc__ = """ When passed in as the value for the `reuse` flag, AUTO_REUSE indicates that get_variable() should create the requested variable if it doesn't exist or, if it does exist, simply return it. """ class _VariableStore(object): """Variable store that carries a number of named Variables. New variable names and new variables can be created; all stored variables are initialized with the initializer passed to __init__. Attributes: vars: a dictionary with string names (same as passed in GetVar) as keys and the corresponding TensorFlow Variables as values. """ def __init__(self): """Create a variable store.""" self._vars = {} # A dictionary of the stored TensorFlow variables. self._partitioned_vars = {} # A dict of the stored PartitionedVariables. self._store_eager_variables = False def get_variable(self, name, shape=None, dtype=dtypes.float32, initializer=None, regularizer=None, reuse=None, trainable=None, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, custom_getter=None, constraint=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): """Gets an existing variable with these parameters or create a new one. If a variable with the given name is already stored, we return the stored variable. Otherwise, we create a new one. Set `reuse` to `True` when you only want to reuse existing Variables. Set `reuse` to `False` when you only want to create new Variables. Set `reuse` to None (the default) or tf.AUTO_REUSE when you want variables to be created if they don't exist or returned if they do. If initializer is `None` (the default), the default initializer passed in the constructor is used. If that one is `None` too, we use a new `glorot_uniform_initializer`. If initializer is a Tensor, we use it as a value and derive the shape from the initializer. If a partitioner is provided, a `PartitionedVariable` is returned. Accessing this object as a `Tensor` returns the shards concatenated along the partition axis. Some useful partitioners are available. See, e.g., `variable_axis_size_partitioner` and `min_max_variable_partitioner`. Args: name: The name of the new or existing variable. shape: Shape of the new or existing variable. dtype: Type of the new or existing variable (defaults to `DT_FLOAT`). initializer: Initializer for the variable. regularizer: A (Tensor -> Tensor or None) function; the result of applying it on a newly created variable will be added to the collection GraphKeys.REGULARIZATION_LOSSES and can be used for regularization. reuse: a Boolean, None, or tf.AUTO_REUSE. Controls reuse or creation of variables. When eager execution is enabled this argument is always forced to be False. trainable: If `True` also add the variable to the graph collection `GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`). `trainable` defaults to `True` unless `synchronization` is set to `ON_READ`. collections: List of graph collections keys to add the `Variable` to. Defaults to `[GraphKeys.GLOBAL_VARIABLES]` (see `tf.Variable`). caching_device: Optional device string or function describing where the Variable should be cached for reading. Defaults to the Variable's device. If not `None`, caches on another device. Typical use is to cache on the device where the Ops using the `Variable` reside, to deduplicate copying through `Switch` and other conditional statements. partitioner: Optional callable that accepts a fully defined `TensorShape` and dtype of the `Variable` to be created, and returns a list of partitions for each axis (currently only one axis can be partitioned). validate_shape: If False, allows the variable to be initialized with a value of unknown shape. If True, the default, the shape of initial_value must be known. use_resource: If False, creates a regular Variable. If True, creates instead an experimental ResourceVariable which has well-defined semantics. Defaults to False (will later change to True). When eager execution is enabled this argument is always forced to be true. custom_getter: Callable that takes as a first argument the true getter, and allows overwriting the internal get_variable method. The signature of `custom_getter` should match that of this method, but the most future-proof version will allow for changes: `def custom_getter(getter, *args, **kwargs)`. Direct access to all `get_variable` parameters is also allowed: `def custom_getter(getter, name, *args, **kwargs)`. A simple identity custom getter that simply creates variables with modified names is: ```python def custom_getter(getter, name, *args, **kwargs): return getter(name + '_suffix', *args, **kwargs) ``` constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. synchronization: Indicates when a distributed a variable will be aggregated. Accepted values are constants defined in the class @{tf.VariableSynchronization}. By default the synchronization is set to `AUTO` and the current `DistributionStrategy` chooses when to synchronize. If `synchronization` is set to `ON_READ`, `trainable` must not be set to `True`. aggregation: Indicates how a distributed variable will be aggregated. Accepted values are constants defined in the class @{tf.VariableAggregation}. Returns: The created or existing `Variable` (or `PartitionedVariable`, if a partitioner was used). Raises: ValueError: when creating a new variable and shape is not declared, when reusing a variable and specifying a conflicting shape, or when violating reuse during variable creation. RuntimeError: when eager execution is enabled and not called from an EagerVariableStore. """ if custom_getter is not None and not callable(custom_getter): raise ValueError( "Passed a custom_getter which is not callable: %s" % custom_getter) with ops.init_scope(): if context.executing_eagerly(): # Variable creation and initialization takes place in `init_scope`s; # as such, if an `init_scope` lifts us into the eager context, then we # need to use `ResourceVariable`s. use_resource = True # Note that it's fine to reuse eager variables whose initialization was # lifted from a function-building graph into the eager context (that's why # the following clause is not wrapped in an `init_scope`); lifted variables # are tracked by the graph's `VariableStore`. if context.executing_eagerly(): if not self._store_eager_variables and reuse: raise RuntimeError( "When eager execution is enabled variable reuse is only supported" " when an EagerVariableStore is active. See the documentation on" " EagerVariableStore for example usage.") if self._store_eager_variables: reuse = AUTO_REUSE # If a *_ref type is passed in an error would be triggered further down the # stack. We prevent this using base_dtype to get a non-ref version of the # type, before doing anything else. When _ref types are removed in favor of # resources, this line can be removed. try: dtype = dtype.base_dtype except AttributeError: # .base_dtype not existing means that we will try and use the raw dtype # which was passed in - this might be a NumPy type which is valid. pass # This is the main logic of get_variable. However, custom_getter # may override this logic. So we save it as a callable and pass # it to custom_getter. # Note: the parameters of _true_getter, and their documentation, match # *exactly* item-for-item with the docstring of this method. def _true_getter( # pylint: disable=missing-docstring name, shape=None, dtype=dtypes.float32, initializer=None, regularizer=None, reuse=None, trainable=None, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, constraint=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): is_scalar = (shape is not None and isinstance(shape, collections_lib.Sequence) and not shape) # Partitioned variable case if partitioner is not None and not is_scalar: if not callable(partitioner): raise ValueError( "Partitioner must be callable, but received: %s" % partitioner) with ops.name_scope(None): return self._get_partitioned_variable(name=name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint) # Special case for partitioned variable to allow reuse without having to # specify partitioner. if (reuse is True and partitioner is None and name in self._partitioned_vars): return self._get_partitioned_variable(name=name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=None, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint) # Single variable case if "%s/part_0" % name in self._vars: raise ValueError( "No partitioner was provided, but a partitioned version of the " "variable was found: %s/part_0. Perhaps a variable of the same " "name was already created with partitioning?" % name) return self._get_single_variable( name=name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint, synchronization=synchronization, aggregation=aggregation) # Set trainable value based on synchronization value. trainable = _get_trainable_value( synchronization=synchronization, trainable=trainable) if custom_getter is not None: # Handle backwards compatibility with getter arguments that were added # to the API after users started writing custom getters. custom_getter_kwargs = { "getter": _true_getter, "name": name, "shape": shape, "dtype": dtype, "initializer": initializer, "regularizer": regularizer, "reuse": reuse, "trainable": trainable, "collections": collections, "caching_device": caching_device, "partitioner": partitioner, "validate_shape": validate_shape, "use_resource": use_resource, "synchronization": synchronization, "aggregation": aggregation, } # `fn_args` can handle functions, `functools.partial`, `lambda`. if "constraint" in function_utils.fn_args(custom_getter): custom_getter_kwargs["constraint"] = constraint return custom_getter(**custom_getter_kwargs) else: return _true_getter( name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint, synchronization=synchronization, aggregation=aggregation) def _get_partitioned_variable(self, name, partitioner, shape=None, dtype=dtypes.float32, initializer=None, regularizer=None, reuse=None, trainable=None, collections=None, caching_device=None, validate_shape=True, use_resource=None, constraint=None): """Gets or creates a sharded variable list with these parameters. The `partitioner` must be a callable that accepts a fully defined `TensorShape` and returns a sequence of integers (the `partitions`). These integers describe how to partition the given sharded `Variable` along the given dimension. That is, `partitions[1] = 3` means split the `Variable` into 3 shards along dimension 1. Currently, sharding along only one axis is supported. If the list of variables with the given name (prefix) is already stored, we return the stored variables. Otherwise, we create a new one. Set `reuse` to `True` when you only want to reuse existing Variables. Set `reuse` to `False` when you only want to create new Variables. Set `reuse` to None (the default) or tf.AUTO_REUSE when you want variables to be created if they don't exist or returned if they do. If initializer is `None` (the default), the default initializer passed in the constructor is used. If that one is `None` too, we use a new `glorot_uniform_initializer`. If initializer is a Tensor, we use it as a value and derive the shape from the initializer. If the initializer is a callable, then it will be called for each shard. Otherwise the initializer should match the shape of the entire sharded Variable, and it will be sliced accordingly for each shard. Some useful partitioners are available. See, e.g., `variable_axis_size_partitioner` and `min_max_variable_partitioner`. Args: name: the name of the new or existing sharded variable. partitioner: Optional callable that accepts a fully defined `TensorShape` and `dtype` of the Variable to be created, and returns a list of partitions for each axis (currently only one axis can be partitioned). shape: shape of the new or existing sharded variable. dtype: type of the new or existing sharded variable (defaults to `DT_FLOAT`). initializer: initializer for the sharded variable. regularizer: a (Tensor -> Tensor or None) function; the result of applying it on a newly created variable will be added to the collection GraphKeys.REGULARIZATION_LOSSES and can be used for regularization. reuse: a Boolean, None, or tf.AUTO_REUSE. Controls reuse or creation of variables. trainable: If `True` also add the variable to the graph collection `GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`). collections: List of graph collections keys to add the Variable to. Defaults to `[GraphKeys.GLOBAL_VARIABLES]` (see `tf.Variable`). caching_device: Optional device string or function describing where the Variable should be cached for reading. Defaults to the Variable's device. If not `None`, caches on another device. Typical use is to cache on the device where the Ops using the Variable reside, to deduplicate copying through `Switch` and other conditional statements. validate_shape: If False, allows the variable to be initialized with a value of unknown shape. If True, the default, the shape of initial_value must be known. use_resource: If False, creates a regular Variable. If True, creates an experimental ResourceVariable which has well-defined semantics. Defaults to False (will later change to True). constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. Returns: A `PartitionedVariable` object. Raises: ValueError: when creating a new variable and shape is not declared, when reusing a variable and specifying a conflicting shape, when violating reuse during variable creation, or if an existing sharded variable exists for the given name but with different sharding. """ if context.executing_eagerly(): raise NotImplementedError("Partitioned variables are not yet supported " "when eager execution is enabled.") initializing_from_value = initializer is not None and isinstance( initializer, ops.Tensor) reuse_without_partition = reuse and not partitioner if name in self._vars: raise ValueError( "A partitioner was provided, but an unpartitioned version of the " "variable was found: %s. Perhaps a variable of the same name was " "already created without partitioning?" % name) shape = tensor_shape.as_shape(shape) if initializing_from_value: shape = shape.merge_with(initializer.get_shape()) if not reuse_without_partition: if not shape.is_fully_defined(): raise ValueError("Shape of a new partitioned variable (%s) must be " "fully defined, but instead was %s." % (name, shape)) if shape.ndims < 1: raise ValueError("A partitioned Variable must have rank at least 1, " "shape: %s" % shape) partitions = partitioner(shape=shape, dtype=dtype) if not isinstance(partitions, collections_lib.Sequence): raise ValueError("Partitioner must return a sequence, but saw: %s" % partitions) if len(partitions) != shape.ndims: raise ValueError( "Partitioner returned a partition list that does not match the " "Variable's rank: %s vs. %s" % (partitions, shape)) if any([p < 1 for p in partitions]): raise ValueError( "Partitioner returned zero partitions for some axes: %s" % partitions) if name in self._partitioned_vars: if reuse is False: raise ValueError( "Partitioned variable with name %s already exists. Did you mean to " "set reuse=True or reuse=tf.AUTO_REUSE in VarScope?" % name) existing_var = self._partitioned_vars[name] if not shape.is_compatible_with(existing_var.get_shape()): raise ValueError( "Trying to reuse partitioned variable %s, but specified shape %s " "and found shape %s." % (name, shape, existing_var.get_shape())) if not dtype.is_compatible_with(existing_var.dtype): raise ValueError( "Trying to reuse partitioned variable %s, but specified dtype %s " "and found dtype %s." % (name, dtype.name, existing_var.dtype.name)) # pylint: disable=protected-access if (not reuse_without_partition and existing_var._get_partitions() != partitions): raise ValueError( "Trying to reuse partitioned variable %s, but specified partitions " "%s and found partitions %s." % (name, partitions, existing_var._get_partitions())) # pylint: enable=protected-access return existing_var if reuse is True: raise ValueError("PartitionedVariable %s does not exist, or was not " "created with tf.get_variable(). Did you mean to set " "reuse=False or reuse=tf.AUTO_REUSE in VarScope?" % name) slice_dim, slice_shape = _compute_slice_dim_and_shape( shape.as_list(), partitions) vs = [] num_slices = partitions[slice_dim] num_slices_with_excess = shape[slice_dim].value % num_slices slice_offset = [0] * shape.ndims if "%s/part_0" % name in self._vars: if "%s/part_%d" % (name, num_slices - 1) not in self._vars: raise ValueError( "Partitioner returned a different partitioning than what was " "already found. Partitioner returned %d shards, and shard " "%s/part_0 was found, but %s/part_%d was not." % (num_slices, name, name, num_slices - 1)) if "%s/part_%d" % (name, num_slices) in self._vars: raise ValueError( "Partitioner returned a different partitioning than what was " "already found. Partitioner returned %d shards, and shard " "%s/part_0 was found, but so was the extra shard %s/part_%d." % (num_slices, name, name, num_slices)) for i in xrange(num_slices): var_shape = slice_shape[:] var_offset = slice_offset[:] partition_info = _PartitionInfo( full_shape=shape.as_list(), var_offset=var_offset) if i < num_slices_with_excess: var_shape[slice_dim] += 1 slice_offset[slice_dim] += var_shape[slice_dim] var_full_name = "%s/part_%d" % (name, i) with ops.name_scope(var_full_name + "/PartitionedInitializer"): # Create the tensor to initialize the variable with default value. if initializer is None: init, initializing_from_value = self._get_default_initializer( name=name, shape=shape, dtype=dtype) if initializing_from_value: init_shape = None else: init_shape = var_shape elif callable(initializer): init = initializer init_shape = var_shape elif isinstance(initializer, ops.Tensor): init = array_ops.slice(initializer, var_offset, var_shape) # Use the dtype of the given tensor. dtype = init.dtype.base_dtype init_shape = None else: init = ops.convert_to_tensor(initializer, dtype=dtype) init = array_ops.slice(init, var_offset, var_shape) init_shape = None with ops.name_scope(None): var = self._get_single_variable( name=var_full_name, shape=init_shape, dtype=dtype, initializer=init, partition_info=partition_info, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint) # pylint: disable=protected-access var._set_save_slice_info(variables.Variable.SaveSliceInfo( name, shape.as_list(), var_offset, var_shape)) vs.append(var) # pylint: enable=protected-access # pylint: disable=protected-access partitioned_var = variables.PartitionedVariable(name=name, shape=shape, dtype=dtype, variable_list=vs, partitions=partitions) # pylint: enable=protected-access self._partitioned_vars[name] = partitioned_var return partitioned_var def _get_single_variable(self, name, shape=None, dtype=dtypes.float32, initializer=None, regularizer=None, partition_info=None, reuse=None, trainable=None, collections=None, caching_device=None, validate_shape=True, use_resource=None, constraint=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): """Get or create a single Variable (e.g. a shard or entire variable). See the documentation of get_variable above (ignore partitioning components) for details. Args: name: see get_variable. shape: see get_variable. dtype: see get_variable. initializer: see get_variable. regularizer: see get_variable. partition_info: _PartitionInfo object. reuse: see get_variable. trainable: see get_variable. collections: see get_variable. caching_device: see get_variable. validate_shape: see get_variable. use_resource: see get_variable. constraint: see get_variable. synchronization: see get_variable. aggregation: see get_variable. Returns: A Variable. See documentation of get_variable above. Raises: ValueError: See documentation of get_variable above. """ # Set to true if initializer is a constant. initializing_from_value = False if initializer is not None and not callable(initializer): initializing_from_value = True if shape is not None and initializing_from_value: raise ValueError("If initializer is a constant, do not specify shape.") dtype = dtypes.as_dtype(dtype) shape = tensor_shape.as_shape(shape) if name in self._vars: # Here we handle the case when returning an existing variable. if reuse is False: tb = self._vars[name].op.traceback[::-1] # Throw away internal tf entries and only take a few lines. tb = [x for x in tb if "tensorflow/python" not in x[0]][:3] raise ValueError("Variable %s already exists, disallowed." " Did you mean to set reuse=True or " "reuse=tf.AUTO_REUSE in VarScope? " "Originally defined at:\n\n%s" % ( name, "".join(traceback.format_list(tb)))) found_var = self._vars[name] if not shape.is_compatible_with(found_var.get_shape()): raise ValueError("Trying to share variable %s, but specified shape %s" " and found shape %s." % (name, shape, found_var.get_shape())) if not dtype.is_compatible_with(found_var.dtype): dtype_str = dtype.name found_type_str = found_var.dtype.name raise ValueError("Trying to share variable %s, but specified dtype %s" " and found dtype %s." % (name, dtype_str, found_type_str)) return found_var # The code below handles only the case of creating a new variable. if reuse is True: raise ValueError("Variable %s does not exist, or was not created with " "tf.get_variable(). Did you mean to set " "reuse=tf.AUTO_REUSE in VarScope?" % name) if not shape.is_fully_defined() and not initializing_from_value: raise ValueError("Shape of a new variable (%s) must be fully defined, " "but instead was %s." % (name, shape)) # Create the tensor to initialize the variable with default value. if initializer is None: initializer, initializing_from_value = self._get_default_initializer( name=name, shape=shape, dtype=dtype) # Enter an init scope when creating the initializer. with ops.init_scope(): if initializing_from_value: init_val = initializer variable_dtype = None else: # Instantiate initializer if provided initializer is a type object. if isinstance(initializer, type(init_ops.Initializer)): initializer = initializer(dtype=dtype) init_val = lambda: initializer( # pylint: disable=g-long-lambda shape.as_list(), dtype=dtype, partition_info=partition_info) variable_dtype = dtype.base_dtype # Create the variable. if use_resource is None: # Set the default value if unspecified. use_resource = False v = variable( initial_value=init_val, name=name, trainable=trainable, collections=collections, caching_device=caching_device, dtype=variable_dtype, validate_shape=validate_shape, constraint=constraint, use_resource=use_resource, synchronization=synchronization, aggregation=aggregation) if context.executing_eagerly() and self._store_eager_variables: if collections: ops.add_to_collections(collections, v) else: ops.add_to_collection(ops.GraphKeys.GLOBAL_VARIABLES, v) if trainable: ops.add_to_collection(ops.GraphKeys.TRAINABLE_VARIABLES, v) if not context.executing_eagerly() or self._store_eager_variables: # In eager mode we do not want to keep default references to Variable # objects as this will prevent their memory from being released. self._vars[name] = v logging.vlog(1, "Created variable %s with shape %s and init %s", v.name, format(shape), initializer) # Run the regularizer if requested and save the resulting loss. if regularizer: with ops.colocate_with(v): with ops.name_scope(name + "/Regularizer/"): loss = regularizer(v) if loss is not None: if context.executing_eagerly(): v_name = "v_%s" % type(v) loss_name = "loss_%s" % type(loss) else: v_name = v.name loss_name = loss.name logging.vlog(1, "Applied regularizer to %s and added the result %s " "to REGULARIZATION_LOSSES.", v_name, loss_name) ops.add_to_collection(ops.GraphKeys.REGULARIZATION_LOSSES, loss) return v # Initialize variable when no initializer provided def _get_default_initializer(self, name, shape=None, dtype=dtypes.float32): """Provide a default initializer and a corresponding value. Args: name: see get_variable. shape: see get_variable. dtype: see get_variable. Returns: initializer and initializing_from_value. See get_variable above. Raises: ValueError: When giving unsupported dtype. """ del shape # If dtype is DT_FLOAT, provide a uniform unit scaling initializer if dtype.is_floating: initializer = init_ops.glorot_uniform_initializer() initializing_from_value = False # If dtype is DT_INT/DT_UINT, provide a default value `zero` # If dtype is DT_BOOL, provide a default value `FALSE` elif (dtype.is_integer or dtype.is_unsigned or dtype.is_bool or dtype == dtypes.string): initializer = init_ops.zeros_initializer() initializing_from_value = False # NOTES:Do we need to support for handling DT_STRING and DT_COMPLEX here? else: raise ValueError("An initializer for variable %s of %s is required" % (name, dtype.base_dtype)) return initializer, initializing_from_value # To stop regularization, use this regularizer @tf_export("no_regularizer") def no_regularizer(_): """Use this function to prevent regularization of variables.""" return None # TODO(alive): support caching devices and partitioned variables in Eager mode. @tf_export("VariableScope") class VariableScope(object): """Variable scope object to carry defaults to provide to `get_variable`. Many of the arguments we need for `get_variable` in a variable store are most easily handled with a context. This object is used for the defaults. Attributes: name: name of the current scope, used as prefix in get_variable. initializer: default initializer passed to get_variable. regularizer: default regularizer passed to get_variable. reuse: Boolean, None, or tf.AUTO_REUSE, setting the reuse in get_variable. When eager execution is enabled this argument is always forced to be False. caching_device: string, callable, or None: the caching device passed to get_variable. partitioner: callable or `None`: the partitioner passed to `get_variable`. custom_getter: default custom getter passed to get_variable. name_scope: The name passed to `tf.name_scope`. dtype: default type passed to get_variable (defaults to DT_FLOAT). use_resource: if False, create a normal Variable; if True create an experimental ResourceVariable with well-defined semantics. Defaults to False (will later change to True). When eager execution is enabled this argument is always forced to be True. constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. """ def __init__(self, reuse, name="", initializer=None, regularizer=None, caching_device=None, partitioner=None, custom_getter=None, name_scope="", dtype=dtypes.float32, use_resource=None, constraint=None): """Creates a new VariableScope with the given properties.""" self._name = name self._initializer = initializer self._regularizer = regularizer self._reuse = reuse self._caching_device = caching_device self._partitioner = partitioner self._custom_getter = custom_getter self._name_scope = name_scope self._dtype = dtype self._use_resource = use_resource self._constraint = constraint if context.executing_eagerly(): if self._caching_device is not None: raise NotImplementedError("Caching devices is not yet supported " "when eager execution is enabled.") if self._partitioner is not None: raise NotImplementedError("Partitioned variables are not yet supported " "when eager execution is enabled.") self._reuse = AUTO_REUSE self._use_resource = True @property def name(self): return self._name @property def original_name_scope(self): return self._name_scope @property def reuse(self): return self._reuse @property def initializer(self): return self._initializer @property def dtype(self): return self._dtype @property def use_resource(self): return self._use_resource @property def regularizer(self): return self._regularizer @property def caching_device(self): return self._caching_device @property def partitioner(self): return self._partitioner @property def custom_getter(self): return self._custom_getter @property def constraint(self): return self._constraint def reuse_variables(self): """Reuse variables in this scope.""" self._reuse = True def set_initializer(self, initializer): """Set initializer for this scope.""" self._initializer = initializer def set_dtype(self, dtype): """Set data type for this scope.""" self._dtype = dtype def set_use_resource(self, use_resource): """Sets whether to use ResourceVariables for this scope.""" if context.executing_eagerly() and not use_resource: raise ValueError("When eager execution is enabled, " "use_resource cannot be set to false.") self._use_resource = use_resource def set_regularizer(self, regularizer): """Set regularizer for this scope.""" self._regularizer = regularizer def set_caching_device(self, caching_device): """Set caching_device for this scope.""" if context.executing_eagerly(): raise NotImplementedError("Caching devices are not yet supported " "when eager execution is enabled.") self._caching_device = caching_device def set_partitioner(self, partitioner): """Set partitioner for this scope.""" if partitioner and context.executing_eagerly(): raise NotImplementedError("Partitioned variables are not yet supported " "when eager execution is enabled.") self._partitioner = partitioner def set_custom_getter(self, custom_getter): """Set custom getter for this scope.""" self._custom_getter = custom_getter def get_collection(self, name): """Get this scope's variables.""" scope = self._name + "/" if self._name else "" return ops.get_collection(name, scope) def trainable_variables(self): """Get this scope's trainable variables.""" return self.get_collection(ops.GraphKeys.TRAINABLE_VARIABLES) def global_variables(self): """Get this scope's global variables.""" return self.get_collection(ops.GraphKeys.GLOBAL_VARIABLES) def local_variables(self): """Get this scope's local variables.""" return self.get_collection(ops.GraphKeys.LOCAL_VARIABLES) def get_variable(self, var_store, name, shape=None, dtype=None, initializer=None, regularizer=None, reuse=None, trainable=None, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, custom_getter=None, constraint=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): """Gets an existing variable with this name or create a new one.""" if regularizer is None: regularizer = self._regularizer if caching_device is None: caching_device = self._caching_device if partitioner is None: partitioner = self._partitioner if custom_getter is None: custom_getter = self._custom_getter if context.executing_eagerly(): reuse = False use_resource = True else: if reuse is None: reuse = self._reuse if use_resource is None: use_resource = self._use_resource full_name = self.name + "/" + name if self.name else name # Variable names only depend on variable_scope (full_name here), # not name_scope, so we reset it below for the time of variable creation. with ops.name_scope(None): # Check that `initializer` dtype and `dtype` are consistent before # replacing them with defaults. if (dtype is not None and initializer is not None and not callable(initializer)): init_dtype = ops.convert_to_tensor(initializer).dtype.base_dtype if init_dtype != dtype: raise ValueError("Initializer type '%s' and explicit dtype '%s' " "don't match." % (init_dtype, dtype)) if initializer is None: initializer = self._initializer if constraint is None: constraint = self._constraint if dtype is None: dtype = self._dtype return var_store.get_variable( full_name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=reuse, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, custom_getter=custom_getter, constraint=constraint, synchronization=synchronization, aggregation=aggregation) def _get_partitioned_variable(self, var_store, name, shape=None, dtype=None, initializer=None, regularizer=None, trainable=None, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, constraint=None): """Gets an existing variable with this name or create a new one.""" if context.executing_eagerly(): raise NotImplementedError("Partitioned variables are not yet supported " "when eager execution is enabled.") if initializer is None: initializer = self._initializer if regularizer is None: regularizer = self._regularizer if constraint is None: constraint = self._constraint if caching_device is None: caching_device = self._caching_device if partitioner is None: partitioner = self._partitioner if dtype is None: dtype = self._dtype if use_resource is None: use_resource = self._use_resource if self._custom_getter is not None: raise ValueError( "Private access to _get_partitioned_variable is not allowed when " "a custom getter is set. Current custom getter: %s. " "It is likely that you're using create_partitioned_variables. " "If so, consider instead using get_variable with a non-empty " "partitioner parameter instead." % self._custom_getter) if partitioner is None: raise ValueError("No partitioner was specified") # This allows the variable scope name to be used as the variable name if # this function is invoked with an empty name arg, for backward # compatibility with create_partitioned_variables(). full_name_list = [] if self.name: full_name_list.append(self.name) if name: full_name_list.append(name) full_name = "/".join(full_name_list) # Variable names only depend on variable_scope (full_name here), # not name_scope, so we reset it below for the time of variable creation. with ops.name_scope(None): # pylint: disable=protected-access return var_store._get_partitioned_variable( full_name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, reuse=self.reuse, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint) # pylint: enable=protected-access _VARSTORE_KEY = ("__variable_store",) _VARSCOPESTORE_KEY = ("__varscope",) class _VariableScopeStore(threading.local): """A thread local store for the current variable scope and scope counts.""" def __init__(self): super(_VariableScopeStore, self).__init__() self.current_scope = VariableScope(False) self.variable_scopes_count = {} def open_variable_scope(self, scope_name): if scope_name in self.variable_scopes_count: self.variable_scopes_count[scope_name] += 1 else: self.variable_scopes_count[scope_name] = 1 def close_variable_subscopes(self, scope_name): for k in list(self.variable_scopes_count.keys()): if scope_name is None or k.startswith(scope_name + "/"): self.variable_scopes_count[k] = 0 def variable_scope_count(self, scope_name): return self.variable_scopes_count.get(scope_name, 0) def get_variable_scope_store(): """Returns the variable scope store for current thread.""" scope_store = ops.get_collection(_VARSCOPESTORE_KEY) if not scope_store: scope_store = _VariableScopeStore() ops.add_to_collection(_VARSCOPESTORE_KEY, scope_store) else: scope_store = scope_store[0] return scope_store @tf_export("get_variable_scope") def get_variable_scope(): """Returns the current variable scope.""" return get_variable_scope_store().current_scope def _get_default_variable_store(): store = ops.get_collection(_VARSTORE_KEY) if store: return store[0] store = _VariableStore() ops.add_to_collection(_VARSTORE_KEY, store) return store @tf_contextlib.contextmanager def with_variable_store(store): store_collection = ops.get_collection_ref(_VARSTORE_KEY) old = list(store_collection) store_collection[:] = [store] try: yield finally: store_collection[:] = old class EagerVariableStore(object): """Wrapper allowing functional layers to be used with eager execution. When eager execution is enabled Variables get deleted when they go out of scope, and are not stored in global collections by default. A lot of code (mostly the functional layers in tf.layers) assumes that variables are kept in a global list. EagerVariableStore can be used in conjunction with this code to make it eager-friendly. For example, to create a dense layer, use: ``` container = tfe.EagerVariableStore() for input in dataset_iterator: with container.as_default(): x = tf.layers.dense(input, name="l1") print(container.variables) # Should print the variables used in the layer. ``` """ def __init__(self, store=None): if store is not None: if not store._store_eager_variables: # pylint: disable=protected-access raise ValueError("Cannot construct EagerVariableStore from a " "VariableStore object that does not hold eager " "variables.") self._store = store else: self._store = _VariableStore() self._store._store_eager_variables = True # pylint: disable=protected-access def as_default(self): return with_variable_store(self._store) def variables(self): return sorted(self._store._vars.values(), key=lambda x: x.name) # pylint: disable=protected-access def trainable_variables(self): # pylint: disable=protected-access return sorted([x for x in self._store._vars.values() if x.trainable], key=lambda x: x.name) # pylint: enable=protected-access def non_trainable_variables(self): # pylint: disable=protected-access return sorted([x for x in self._store._vars.values() if not x.trainable], key=lambda x: x.name) # pylint: enable=protected-access def copy(self): """Copy this variable store and all of its contents. Variables contained in this store will be copied over to the new variable store, meaning that they can be modified without affecting the variables in this store. Returns: A new EagerVariableStore instance containing copied variables. """ # pylint: disable=protected-access new_store = EagerVariableStore() for key, var in iteritems(self._store._vars): # Strip device out of variable name. try: index = var.name.index(":") except ValueError: stripped_var_name = var.name else: stripped_var_name = var.name[:index] # Create new variable with same value, name, and "trainable" flag. new_var = resource_variable_ops.ResourceVariable( var.read_value(), name=stripped_var_name, trainable=var.trainable) new_store._store._vars[key] = new_var return new_store # pylint: enable=protected-access # The argument list for get_variable must match arguments to get_local_variable. # So, if you are updating the arguments, also update arguments to # get_local_variable below. @tf_export("get_variable") def get_variable(name, shape=None, dtype=None, initializer=None, regularizer=None, trainable=None, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, custom_getter=None, constraint=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): return get_variable_scope().get_variable( _get_default_variable_store(), name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, custom_getter=custom_getter, constraint=constraint, synchronization=synchronization, aggregation=aggregation) get_variable_or_local_docstring = ( """%s %sThis function prefixes the name with the current variable scope and performs reuse checks. See the @{$variables$Variable Scope How To} for an extensive description of how reusing works. Here is a basic example: ```python def foo(): with tf.variable_scope("foo", reuse=tf.AUTO_REUSE): v = tf.get_variable("v", [1]) return v v1 = foo() # Creates v. v2 = foo() # Gets the same, existing v. assert v1 == v2 ``` If initializer is `None` (the default), the default initializer passed in the variable scope will be used. If that one is `None` too, a `glorot_uniform_initializer` will be used. The initializer can also be a Tensor, in which case the variable is initialized to this value and shape. Similarly, if the regularizer is `None` (the default), the default regularizer passed in the variable scope will be used (if that is `None` too, then by default no regularization is performed). If a partitioner is provided, a `PartitionedVariable` is returned. Accessing this object as a `Tensor` returns the shards concatenated along the partition axis. Some useful partitioners are available. See, e.g., `variable_axis_size_partitioner` and `min_max_variable_partitioner`. Args: name: The name of the new or existing variable. shape: Shape of the new or existing variable. dtype: Type of the new or existing variable (defaults to `DT_FLOAT`). initializer: Initializer for the variable if one is created. Can either be an initializer object or a Tensor. If it's a Tensor, its shape must be known unless validate_shape is False. regularizer: A (Tensor -> Tensor or None) function; the result of applying it on a newly created variable will be added to the collection @{tf.GraphKeys.REGULARIZATION_LOSSES} and can be used for regularization. %scollections: List of graph collections keys to add the Variable to. Defaults to `[%s]` (see `tf.Variable`). caching_device: Optional device string or function describing where the Variable should be cached for reading. Defaults to the Variable's device. If not `None`, caches on another device. Typical use is to cache on the device where the Ops using the Variable reside, to deduplicate copying through `Switch` and other conditional statements. partitioner: Optional callable that accepts a fully defined `TensorShape` and `dtype` of the Variable to be created, and returns a list of partitions for each axis (currently only one axis can be partitioned). validate_shape: If False, allows the variable to be initialized with a value of unknown shape. If True, the default, the shape of initial_value must be known. For this to be used the initializer must be a Tensor and not an initializer object. use_resource: If False, creates a regular Variable. If true, creates an experimental ResourceVariable instead with well-defined semantics. Defaults to False (will later change to True). When eager execution is enabled this argument is always forced to be True. custom_getter: Callable that takes as a first argument the true getter, and allows overwriting the internal get_variable method. The signature of `custom_getter` should match that of this method, but the most future-proof version will allow for changes: `def custom_getter(getter, *args, **kwargs)`. Direct access to all `get_variable` parameters is also allowed: `def custom_getter(getter, name, *args, **kwargs)`. A simple identity custom getter that simply creates variables with modified names is: ```python def custom_getter(getter, name, *args, **kwargs): return getter(name + '_suffix', *args, **kwargs) ``` Returns: The created or existing `Variable` (or `PartitionedVariable`, if a partitioner was used). Raises: ValueError: when creating a new variable and shape is not declared, when violating reuse during variable creation, or when `initializer` dtype and `dtype` don't match. Reuse is set inside `variable_scope`. """) get_variable.__doc__ = get_variable_or_local_docstring % ( "Gets an existing variable with these parameters or create a new one.", "", "trainable: If `True` also add the variable to the graph collection\n" " `GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).\n ", "GraphKeys.GLOBAL_VARIABLES") # The argument list for get_local_variable must match arguments to get_variable. # So, if you are updating the arguments, also update arguments to get_variable. @tf_export("get_local_variable") def get_local_variable( # pylint: disable=missing-docstring name, shape=None, dtype=None, initializer=None, regularizer=None, trainable=False, # pylint: disable=unused-argument collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE, custom_getter=None, constraint=None): if collections: collections += [ops.GraphKeys.LOCAL_VARIABLES] else: collections = [ops.GraphKeys.LOCAL_VARIABLES] return get_variable( name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, trainable=False, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, synchronization=synchronization, aggregation=aggregation, custom_getter=custom_getter, constraint=constraint) get_local_variable.__doc__ = get_variable_or_local_docstring % ( "Gets an existing *local* variable or creates a new one.", "Behavior is the same as in `get_variable`, except that variables are\n" "added to the `LOCAL_VARIABLES` collection and `trainable` is set to\n" "`False`.\n", "", "GraphKeys.LOCAL_VARIABLES") def _get_partitioned_variable(name, shape=None, dtype=None, initializer=None, regularizer=None, trainable=True, collections=None, caching_device=None, partitioner=None, validate_shape=True, use_resource=None, constraint=None): """Gets or creates a sharded variable list with these parameters. The `partitioner` must be a callable that accepts a fully defined `TensorShape` and returns a sequence of integers (the `partitions`). These integers describe how to partition the given sharded `Variable` along the given dimension. That is, `partitions[1] = 3` means split the `Variable` into 3 shards along dimension 1. Currently, sharding along only one axis is supported. If the list of variables with the given name (prefix) is already stored, we return the stored variables. Otherwise, we create a new one. If initializer is `None` (the default), the default initializer passed in the constructor is used. If that one is `None` too, we use a new `glorot_uniform_initializer`. If initializer is a Tensor, we use it as a value and derive the shape from the initializer. If the initializer is a callable, then it will be called for each shard. Otherwise the initializer should match the shape of the entire sharded Variable, and it will be sliced accordingly for each shard. Some useful partitioners are available. See, e.g., `variable_axis_size_partitioner` and `min_max_variable_partitioner`. Args: name: The name of the new or existing variable. shape: Shape of the new or existing variable. dtype: Type of the new or existing variable (defaults to `DT_FLOAT`). initializer: Initializer for the variable if one is created. regularizer: A (Tensor -> Tensor or None) function; the result of applying it on a newly created variable will be added to the collection GraphKeys.REGULARIZATION_LOSSES and can be used for regularization. trainable: If `True` also add the variable to the graph collection `GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`). collections: List of graph collections keys to add the Variable to. Defaults to `[GraphKeys.GLOBAL_VARIABLES]` (see `tf.Variable`). caching_device: Optional device string or function describing where the Variable should be cached for reading. Defaults to the Variable's device. If not `None`, caches on another device. Typical use is to cache on the device where the Ops using the Variable reside, to deduplicate copying through `Switch` and other conditional statements. partitioner: Optional callable that accepts a fully defined `TensorShape` and `dtype` of the Variable to be created, and returns a list of partitions for each axis (currently only one axis can be partitioned). validate_shape: If False, allows the variable to be initialized with a value of unknown shape. If True, the default, the shape of initial_value must be known. use_resource: If False, creates a regular Variable. If True, creates an experimental ResourceVariable instead which has well-defined semantics. Defaults to False (will later change to True). constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. Returns: A tuple `(shards, partitions)` where `shards` is the list of `Variable` shards and `partitions` is the output of the partitioner on the input shape. Raises: ValueError: when creating a new variable and shape is not declared, or when violating reuse during variable creation. Reuse is set inside `variable_scope`. """ # pylint: disable=protected-access scope = get_variable_scope() if scope.custom_getter is not None: raise ValueError( "Private access to _get_partitioned_variable is not allowed when " "a custom getter is set. Current custom getter: %s. " "It is likely that you're using create_partitioned_variables. " "If so, consider instead using get_variable with a non-empty " "partitioner parameter instead." % scope.custom_getter) return scope._get_partitioned_variable( _get_default_variable_store(), name, shape=shape, dtype=dtype, initializer=initializer, regularizer=regularizer, trainable=trainable, collections=collections, caching_device=caching_device, partitioner=partitioner, validate_shape=validate_shape, use_resource=use_resource, constraint=constraint) # pylint: enable=protected-access # Named like a function for compatibility with the previous # @tf_contextlib.contextmanager definition. class _pure_variable_scope(object): # pylint: disable=invalid-name """A context for the variable_scope, see `variable_scope` for docs.""" def __init__(self, name_or_scope, reuse=None, initializer=None, regularizer=None, caching_device=None, partitioner=None, custom_getter=None, old_name_scope=None, dtype=dtypes.float32, use_resource=None, constraint=None): """Creates a context for the variable_scope, see `variable_scope` for docs. Note: this does not create a name scope. Args: name_or_scope: `string` or `VariableScope`: the scope to open. reuse: `True` or None, or tf.AUTO_REUSE; if `None`, we inherit the parent scope's reuse flag. initializer: default initializer for variables within this scope. regularizer: default regularizer for variables within this scope. caching_device: default caching device for variables within this scope. partitioner: default partitioner for variables within this scope. custom_getter: default custom getter for variables within this scope. old_name_scope: the original name scope when re-entering a variable scope. dtype: type of the variables within this scope (defaults to `DT_FLOAT`). use_resource: If False, variables in this scope will be regular Variables. If True, experimental ResourceVariables will be creates instead, with well-defined semantics. Defaults to False (will later change to True). constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. """ self._name_or_scope = name_or_scope self._reuse = reuse self._initializer = initializer self._regularizer = regularizer self._caching_device = caching_device self._partitioner = partitioner self._custom_getter = custom_getter self._old_name_scope = old_name_scope self._dtype = dtype self._use_resource = use_resource self._constraint = constraint self._var_store = _get_default_variable_store() self._var_scope_store = get_variable_scope_store() if isinstance(self._name_or_scope, VariableScope): self._new_name = self._name_or_scope.name name_scope = self._name_or_scope._name_scope # pylint: disable=protected-access # Handler for the case when we jump to a shared scope. We create a new # VariableScope (self._var_scope_object) that contains a copy of the # provided shared scope, possibly with changed reuse and initializer, if # the user requested this. variable_scope_object = VariableScope( self._name_or_scope.reuse if not self._reuse else self._reuse, name=self._new_name, initializer=self._name_or_scope.initializer, regularizer=self._name_or_scope.regularizer, caching_device=self._name_or_scope.caching_device, partitioner=self._name_or_scope.partitioner, dtype=self._name_or_scope.dtype, custom_getter=self._name_or_scope.custom_getter, name_scope=name_scope, use_resource=self._name_or_scope.use_resource, constraint=self._constraint) if self._initializer is not None: variable_scope_object.set_initializer(self._initializer) if self._regularizer is not None: variable_scope_object.set_regularizer(self._regularizer) if self._caching_device is not None: variable_scope_object.set_caching_device(self._caching_device) if self._partitioner is not None: variable_scope_object.set_partitioner(self._partitioner) if self._custom_getter is not None: variable_scope_object.set_custom_getter( _maybe_wrap_custom_getter( self._custom_getter, self._name_or_scope.custom_getter)) if self._dtype is not None: variable_scope_object.set_dtype(self._dtype) if self._use_resource is not None: variable_scope_object.set_use_resource(self._use_resource) self._cached_variable_scope_object = variable_scope_object def __enter__(self): """Begins the scope block. Returns: A VariableScope. Raises: ValueError: when trying to reuse within a create scope, or create within a reuse scope, or if reuse is not `None` or `True`. TypeError: when the types of some arguments are not appropriate. """ self._old = self._var_scope_store.current_scope if isinstance(self._name_or_scope, VariableScope): self._var_scope_store.open_variable_scope(self._new_name) self._old_subscopes = copy.copy( self._var_scope_store.variable_scopes_count) variable_scope_object = self._cached_variable_scope_object else: # Handler for the case when we just prolong current variable scope. # VariableScope with name extended by the provided one, and inherited # reuse and initializer (except if the user provided values to set). self._new_name = ( self._old.name + "/" + self._name_or_scope if self._old.name else self._name_or_scope) self._reuse = (self._reuse or self._old.reuse) # Re-using is inherited by sub-scopes. if self._old_name_scope is None: name_scope = self._name_or_scope else: name_scope = self._old_name_scope variable_scope_object = VariableScope( self._reuse, name=self._new_name, initializer=self._old.initializer, regularizer=self._old.regularizer, caching_device=self._old.caching_device, partitioner=self._old.partitioner, dtype=self._old.dtype, use_resource=self._old.use_resource, custom_getter=self._old.custom_getter, name_scope=name_scope, constraint=self._constraint) if self._initializer is not None: variable_scope_object.set_initializer(self._initializer) if self._regularizer is not None: variable_scope_object.set_regularizer(self._regularizer) if self._caching_device is not None: variable_scope_object.set_caching_device(self._caching_device) if self._partitioner is not None: variable_scope_object.set_partitioner(self._partitioner) if self._custom_getter is not None: variable_scope_object.set_custom_getter( _maybe_wrap_custom_getter(self._custom_getter, self._old.custom_getter)) if self._dtype is not None: variable_scope_object.set_dtype(self._dtype) if self._use_resource is not None: variable_scope_object.set_use_resource(self._use_resource) self._var_scope_store.open_variable_scope(self._new_name) self._var_scope_store.current_scope = variable_scope_object return variable_scope_object def __exit__(self, type_arg, value_arg, traceback_arg): # If jumping out from a non-prolonged scope, restore counts. if isinstance(self._name_or_scope, VariableScope): self._var_scope_store.variable_scopes_count = self._old_subscopes else: self._var_scope_store.close_variable_subscopes(self._new_name) self._var_scope_store.current_scope = self._old def _maybe_wrap_custom_getter(custom_getter, old_getter): """Wrap a call to a custom_getter to use the old_getter internally.""" if old_getter is None: return custom_getter # The new custom_getter should call the old one def wrapped_custom_getter(getter, *args, **kwargs): # Call: # custom_getter( # lambda: old_getter(true_getter, ...), *args, **kwargs) # which means custom_getter will call old_getter, which # will call the true_getter, perform any intermediate # processing, and return the results to the current # getter, which will also perform additional processing. return custom_getter( functools.partial(old_getter, getter), *args, **kwargs) return wrapped_custom_getter def _get_unique_variable_scope(prefix): """Get a name with the given prefix unique in the current variable scope.""" var_scope_store = get_variable_scope_store() current_scope = get_variable_scope() name = current_scope.name + "/" + prefix if current_scope.name else prefix if var_scope_store.variable_scope_count(name) == 0: return prefix idx = 1 while var_scope_store.variable_scope_count(name + ("_%d" % idx)) > 0: idx += 1 return prefix + ("_%d" % idx) # Named like a function for backwards compatibility with the # @tf_contextlib.contextmanager version, which was switched to a class to avoid # some object creation overhead. @tf_export("variable_scope") # pylint: disable=invalid-name class variable_scope(object): """A context manager for defining ops that creates variables (layers). This context manager validates that the (optional) `values` are from the same graph, ensures that graph is the default graph, and pushes a name scope and a variable scope. If `name_or_scope` is not None, it is used as is. If `name_or_scope` is None, then `default_name` is used. In that case, if the same name has been previously used in the same scope, it will be made unique by appending `_N` to it. Variable scope allows you to create new variables and to share already created ones while providing checks to not create or share by accident. For details, see the @{$variables$Variable Scope How To}, here we present only a few basic examples. Simple example of how to create a new variable: ```python with tf.variable_scope("foo"): with tf.variable_scope("bar"): v = tf.get_variable("v", [1]) assert v.name == "foo/bar/v:0" ``` Simple example of how to reenter a premade variable scope safely: ```python with tf.variable_scope("foo") as vs: pass # Re-enter the variable scope. with tf.variable_scope(vs, auxiliary_name_scope=False) as vs1: # Restore the original name_scope. with tf.name_scope(vs1.original_name_scope): v = tf.get_variable("v", [1]) assert v.name == "foo/v:0" c = tf.constant([1], name="c") assert c.name == "foo/c:0" ``` Basic example of sharing a variable AUTO_REUSE: ```python def foo(): with tf.variable_scope("foo", reuse=tf.AUTO_REUSE): v = tf.get_variable("v", [1]) return v v1 = foo() # Creates v. v2 = foo() # Gets the same, existing v. assert v1 == v2 ``` Basic example of sharing a variable with reuse=True: ```python with tf.variable_scope("foo"): v = tf.get_variable("v", [1]) with tf.variable_scope("foo", reuse=True): v1 = tf.get_variable("v", [1]) assert v1 == v ``` Sharing a variable by capturing a scope and setting reuse: ```python with tf.variable_scope("foo") as scope: v = tf.get_variable("v", [1]) scope.reuse_variables() v1 = tf.get_variable("v", [1]) assert v1 == v ``` To prevent accidental sharing of variables, we raise an exception when getting an existing variable in a non-reusing scope. ```python with tf.variable_scope("foo"): v = tf.get_variable("v", [1]) v1 = tf.get_variable("v", [1]) # Raises ValueError("... v already exists ..."). ``` Similarly, we raise an exception when trying to get a variable that does not exist in reuse mode. ```python with tf.variable_scope("foo", reuse=True): v = tf.get_variable("v", [1]) # Raises ValueError("... v does not exists ..."). ``` Note that the `reuse` flag is inherited: if we open a reusing scope, then all its sub-scopes become reusing as well. A note about name scoping: Setting `reuse` does not impact the naming of other ops such as mult. See related discussion on [github#6189](https://github.com/tensorflow/tensorflow/issues/6189) Note that up to and including version 1.0, it was allowed (though explicitly discouraged) to pass False to the reuse argument, yielding undocumented behaviour slightly different from None. Starting at 1.1.0 passing None and False as reuse has exactly the same effect. A note about using variable scopes in multi-threaded environment: Variable scopes are thread local, so one thread will not see another thread's current scope. Also, when using `default_name`, unique scopes names are also generated only on a per thread basis. If the same name was used within a different thread, that doesn't prevent a new thread from creating the same scope. However, the underlying variable store is shared across threads (within the same graph). As such, if another thread tries to create a new variable with the same name as a variable created by a previous thread, it will fail unless reuse is True. Further, each thread starts with an empty variable scope. So if you wish to preserve name prefixes from a scope from the main thread, you should capture the main thread's scope and re-enter it in each thread. For e.g. ``` main_thread_scope = variable_scope.get_variable_scope() # Thread's target function: def thread_target_fn(captured_scope): with variable_scope.variable_scope(captured_scope): # .... regular code for this thread thread = threading.Thread(target=thread_target_fn, args=(main_thread_scope,)) ``` """ def __init__(self, name_or_scope, default_name=None, values=None, initializer=None, regularizer=None, caching_device=None, partitioner=None, custom_getter=None, reuse=None, dtype=None, use_resource=None, constraint=None, auxiliary_name_scope=True): """Initialize the context manager. Args: name_or_scope: `string` or `VariableScope`: the scope to open. default_name: The default name to use if the `name_or_scope` argument is `None`, this name will be uniquified. If name_or_scope is provided it won't be used and therefore it is not required and can be None. values: The list of `Tensor` arguments that are passed to the op function. initializer: default initializer for variables within this scope. regularizer: default regularizer for variables within this scope. caching_device: default caching device for variables within this scope. partitioner: default partitioner for variables within this scope. custom_getter: default custom getter for variables within this scope. reuse: `True`, None, or tf.AUTO_REUSE; if `True`, we go into reuse mode for this scope as well as all sub-scopes; if tf.AUTO_REUSE, we create variables if they do not exist, and return them otherwise; if None, we inherit the parent scope's reuse flag. When eager execution is enabled, new variables are always created unless an EagerVariableStore or template is currently active. dtype: type of variables created in this scope (defaults to the type in the passed scope, or inherited from parent scope). use_resource: If False, all variables will be regular Variables. If True, experimental ResourceVariables with well-defined semantics will be used instead. Defaults to False (will later change to True). When eager execution is enabled this argument is always forced to be True. constraint: An optional projection function to be applied to the variable after being updated by an `Optimizer` (e.g. used to implement norm constraints or value constraints for layer weights). The function must take as input the unprojected Tensor representing the value of the variable and return the Tensor for the projected value (which must have the same shape). Constraints are not safe to use when doing asynchronous distributed training. auxiliary_name_scope: If `True`, we create an auxiliary name scope with the scope. If `False`, we don't create it. Note that the argument is not inherited, and it only takes effect for once when creating. You should only use it for re-entering a premade variable scope. Returns: A scope that can be captured and reused. Raises: ValueError: when trying to reuse within a create scope, or create within a reuse scope. TypeError: when the types of some arguments are not appropriate. """ self._name_or_scope = name_or_scope self._default_name = default_name self._values = values self._initializer = initializer self._regularizer = regularizer self._caching_device = caching_device self._partitioner = partitioner self._custom_getter = custom_getter self._reuse = reuse self._dtype = dtype self._use_resource = use_resource self._constraint = constraint if self._default_name is None and self._name_or_scope is None: raise TypeError("If default_name is None then name_or_scope is required") if self._reuse is False: # We don't allow non-inheriting scopes, False = None here. self._reuse = None if not (self._reuse is True or self._reuse is None or self._reuse is AUTO_REUSE): raise ValueError("The reuse parameter must be True or False or None.") if self._values is None: self._values = [] self._in_graph_mode = not context.executing_eagerly() if self._in_graph_mode: self._graph = ops._get_graph_from_inputs(self._values) # pylint: disable=protected-access self._cached_pure_variable_scope = None self._current_name_scope = None if not isinstance(auxiliary_name_scope, bool): raise TypeError("The auxiliary_name_scope must be `True` or `False`, " "while get {}".format(auxiliary_name_scope)) self._auxiliary_name_scope = auxiliary_name_scope def __enter__(self): # If the default graph is building a function, then we should not replace it # with the cached graph. if ops.get_default_graph().building_function: self._building_function = True else: self._building_function = False if self._in_graph_mode and not self._building_function: self._graph_context_manager = self._graph.as_default() self._graph_context_manager.__enter__() if self._cached_pure_variable_scope is not None: # Fast path for re-entering variable_scopes. We've held on to the pure # variable scope from a previous successful __enter__, so we avoid some # overhead by re-using that object. if self._current_name_scope is not None: self._current_name_scope.__enter__() return self._cached_pure_variable_scope.__enter__() try: return self._enter_scope_uncached() except: if self._graph_context_manager is not None: self._graph_context_manager.__exit__(*sys.exc_info()) raise def _enter_scope_uncached(self): """Enters the context manager when there is no cached scope yet. Returns: The entered variable scope. Raises: TypeError: A wrong type is passed as `scope` at __init__(). ValueError: `reuse` is incorrectly set at __init__(). """ if self._auxiliary_name_scope: # Create a new name scope later current_name_scope = None else: # Reenter the current name scope name_scope = ops.get_name_scope() if name_scope: # Hack to reenter name_scope += "/" current_name_scope = ops.name_scope(name_scope) else: # Root scope current_name_scope = ops.name_scope(name_scope) # IMPORTANT: Only assign to self._cached_pure_variable_scope and # self._current_name_scope after successful __enter__() calls. if self._name_or_scope is not None: if not isinstance(self._name_or_scope, (VariableScope,) + six.string_types): raise TypeError("VariableScope: name_or_scope must be a string or " "VariableScope.") if isinstance(self._name_or_scope, six.string_types): name_scope = self._name_or_scope else: name_scope = self._name_or_scope.name.split("/")[-1] if name_scope or current_name_scope: current_name_scope = current_name_scope or ops.name_scope(name_scope) try: current_name_scope_name = current_name_scope.__enter__() except: current_name_scope.__exit__(*sys.exc_info()) raise self._current_name_scope = current_name_scope if isinstance(self._name_or_scope, six.string_types): old_name_scope = current_name_scope_name else: old_name_scope = self._name_or_scope.original_name_scope pure_variable_scope = _pure_variable_scope( self._name_or_scope, reuse=self._reuse, initializer=self._initializer, regularizer=self._regularizer, caching_device=self._caching_device, partitioner=self._partitioner, custom_getter=self._custom_getter, old_name_scope=old_name_scope, dtype=self._dtype, use_resource=self._use_resource, constraint=self._constraint) try: entered_pure_variable_scope = pure_variable_scope.__enter__() except: pure_variable_scope.__exit__(*sys.exc_info()) raise self._cached_pure_variable_scope = pure_variable_scope return entered_pure_variable_scope else: self._current_name_scope = None # This can only happen if someone is entering the root variable scope. pure_variable_scope = _pure_variable_scope( self._name_or_scope, reuse=self._reuse, initializer=self._initializer, regularizer=self._regularizer, caching_device=self._caching_device, partitioner=self._partitioner, custom_getter=self._custom_getter, dtype=self._dtype, use_resource=self._use_resource, constraint=self._constraint) try: entered_pure_variable_scope = pure_variable_scope.__enter__() except: pure_variable_scope.__exit__(*sys.exc_info()) raise self._cached_pure_variable_scope = pure_variable_scope return entered_pure_variable_scope else: # Here name_or_scope is None. Using default name, but made unique. if self._reuse: raise ValueError("reuse=True cannot be used without a name_or_scope") current_name_scope = current_name_scope or ops.name_scope( self._default_name) try: current_name_scope_name = current_name_scope.__enter__() except: current_name_scope.__exit__(*sys.exc_info()) raise self._current_name_scope = current_name_scope unique_default_name = _get_unique_variable_scope(self._default_name) pure_variable_scope = _pure_variable_scope( unique_default_name, initializer=self._initializer, regularizer=self._regularizer, caching_device=self._caching_device, partitioner=self._partitioner, custom_getter=self._custom_getter, old_name_scope=current_name_scope_name, dtype=self._dtype, use_resource=self._use_resource, constraint=self._constraint) try: entered_pure_variable_scope = pure_variable_scope.__enter__() except: pure_variable_scope.__exit__(*sys.exc_info()) raise self._cached_pure_variable_scope = pure_variable_scope return entered_pure_variable_scope def __exit__(self, type_arg, value_arg, traceback_arg): self._cached_pure_variable_scope.__exit__( type_arg, value_arg, traceback_arg) if self._current_name_scope: self._current_name_scope.__exit__(type_arg, value_arg, traceback_arg) if self._in_graph_mode and not self._building_function: self._graph_context_manager.__exit__(type_arg, value_arg, traceback_arg) # pylint: disable=g-doc-return-or-yield @tf_export("variable_op_scope") @tf_contextlib.contextmanager def variable_op_scope(values, name_or_scope, default_name=None, initializer=None, regularizer=None, caching_device=None, partitioner=None, custom_getter=None, reuse=None, dtype=None, use_resource=None, constraint=None): """Deprecated: context manager for defining an op that creates variables.""" logging.warn("tf.variable_op_scope(values, name, default_name) is deprecated," " use tf.variable_scope(name, default_name, values)") with variable_scope(name_or_scope, default_name=default_name, values=values, initializer=initializer, regularizer=regularizer, caching_device=caching_device, partitioner=partitioner, custom_getter=custom_getter, reuse=reuse, dtype=dtype, use_resource=use_resource, constraint=constraint) as scope: yield scope def _compute_slice_dim_and_shape(full_shape, slicing): """Computes which dimension is being sliced and the typical slice shape.""" slice_shape = [0] * len(full_shape) slice_dim = None for dim, num_slices in enumerate(slicing): dim_size = full_shape[dim] if num_slices <= 0 or dim_size < num_slices: raise ValueError("Cannot create %d slices for size %d. shape: %s, " "slicing: %s" % (num_slices, full_shape[dim], full_shape, slicing)) if num_slices == 1: # Not slicing in this dimension. slice_shape[dim] = dim_size elif slice_dim is not None: # We only support slicing along one of the dimensions. raise ValueError("Can only slice a variable along one dimension: " "shape: %s, slicing: %s" % (full_shape, slicing)) else: # Note: We will add any extras onto the last slice, later. slice_dim = dim slice_shape[dim] = dim_size // num_slices # Degenerate case: If "slicing" was all ones, pretend we are slicing along # the first dimension. if slice_dim is None: slice_dim = 0 return slice_dim, slice_shape def _get_trainable_value(synchronization, trainable): """Computes the trainable value based on the given arguments.""" if synchronization == VariableSynchronization.ON_READ: if trainable: raise ValueError( "Synchronization value can be set to " "VariableSynchronization.ON_READ only for non-trainable variables. " "You have specified trainable=True and " "synchronization=VariableSynchronization.ON_READ.") else: # Set trainable to be false when variable is to be synced on read. trainable = False elif trainable is None: trainable = True return trainable def default_variable_creator(next_creator=None, **kwargs): """Default variable creator.""" assert next_creator is None initial_value = kwargs.get("initial_value", None) trainable = kwargs.get("trainable", None) collections = kwargs.get("collections", None) validate_shape = kwargs.get("validate_shape", True) caching_device = kwargs.get("caching_device", None) name = kwargs.get("name", None) dtype = kwargs.get("dtype", None) constraint = kwargs.get("constraint", None) use_resource = kwargs.get("use_resource", None) # Set trainable value based on synchronization value. synchronization = kwargs.get("synchronization", VariableSynchronization.AUTO) trainable = _get_trainable_value( synchronization=synchronization, trainable=trainable) if use_resource is None: use_resource = get_variable_scope().use_resource if use_resource or (use_resource is None and context.executing_eagerly()): return resource_variable_ops.ResourceVariable( initial_value=initial_value, trainable=trainable, collections=collections, validate_shape=validate_shape, caching_device=caching_device, name=name, dtype=dtype, constraint=constraint) elif not use_resource and context.executing_eagerly(): raise RuntimeError( "VariableScope should use resource variable when eager execution is" " enabled, but use_resource is False." ) else: return variables.Variable( initial_value=initial_value, trainable=trainable, collections=collections, validate_shape=validate_shape, caching_device=caching_device, name=name, dtype=dtype, constraint=constraint) def _make_getter(captured_getter, captured_previous): """Gets around capturing loop variables in python being broken.""" return lambda **kwargs: captured_getter(captured_previous, **kwargs) def variable(initial_value=None, trainable=None, collections=None, validate_shape=True, caching_device=None, name=None, dtype=None, constraint=None, use_resource=None, synchronization=VariableSynchronization.AUTO, aggregation=VariableAggregation.NONE): previous_getter = lambda **kwargs: default_variable_creator(None, **kwargs) for getter in ops.get_default_graph()._variable_creator_stack: # pylint: disable=protected-access previous_getter = _make_getter(getter, previous_getter) # Reset `aggregation` that is explicitly set as `None` to the enum None value. if aggregation is None: aggregation = VariableAggregation.NONE return previous_getter( initial_value=initial_value, trainable=trainable, collections=collections, validate_shape=validate_shape, caching_device=caching_device, name=name, dtype=dtype, constraint=constraint, use_resource=use_resource, synchronization=synchronization, aggregation=aggregation) @tf_contextlib.contextmanager def variable_creator_scope(variable_creator): """Scope which defines a variable creation function to be used by variable(). variable_creator is expected to be a function with the following signature: ``` def variable_creator(next_creator, **kwargs) ``` The creator is supposed to eventually call the next_creator to create a variable if it does want to create a variable and not call Variable or ResourceVariable directly. This helps make creators composable. A creator may choose to create multiple variables, return already existing variables, or simply register that a variable was created and defer to the next creators in line. Creators can also modify the keyword arguments seen by the next creators. Custom getters in the variable scope will eventually resolve down to these custom creators when they do create variables. The valid keyword arguments in kwds are: initial_value: A `Tensor`, or Python object convertible to a `Tensor`, which is the initial value for the Variable. The initial value must have a shape specified unless `validate_shape` is set to False. Can also be a callable with no argument that returns the initial value when called. In that case, `dtype` must be specified. (Note that initializer functions from init_ops.py must first be bound to a shape before being used here.) trainable: If `True`, the default, also adds the variable to the graph collection `GraphKeys.TRAINABLE_VARIABLES`. This collection is used as the default list of variables to use by the `Optimizer` classes. `trainable` defaults to `True` unless `synchronization` is set to `ON_READ`. collections: List of graph collections keys. The new variable is added to these collections. Defaults to `[GraphKeys.GLOBAL_VARIABLES]`. validate_shape: If `False`, allows the variable to be initialized with a value of unknown shape. If `True`, the default, the shape of `initial_value` must be known. caching_device: Optional device string describing where the Variable should be cached for reading. Defaults to the Variable's device. If not `None`, caches on another device. Typical use is to cache on the device where the Ops using the Variable reside, to deduplicate copying through `Switch` and other conditional statements. name: Optional name for the variable. Defaults to `'Variable'` and gets uniquified automatically. dtype: If set, initial_value will be converted to the given type. If `None`, either the datatype will be kept (if `initial_value` is a Tensor), or `convert_to_tensor` will decide. constraint: A constraint function to be applied to the variable after updates by some algorithms. use_resource: if True, a ResourceVariable is always created. synchronization: Indicates when a distributed a variable will be aggregated. Accepted values are constants defined in the class @{tf.VariableSynchronization}. By default the synchronization is set to `AUTO` and the current `DistributionStrategy` chooses when to synchronize. If `synchronization` is set to `ON_READ`, `trainable` must not be set to `True`. aggregation: Indicates how a distributed variable will be aggregated. Accepted values are constants defined in the class @{tf.VariableAggregation}. This set may grow over time, so it's important the signature of creators is as mentioned above. Args: variable_creator: the passed creator Yields: A scope in which the creator is active """ with ops.get_default_graph()._variable_creator_scope(variable_creator): # pylint: disable=protected-access yield