tf.Variable

TensorFlow 2 version View source on GitHub

See the Variables Guide.

Inherits From: Variable

A variable maintains state in the graph across calls to run(). You add a variable to the graph by constructing an instance of the class Variable.

The Variable() constructor requires an initial value for the variable, which can be a Tensor of any type and shape. The initial value defines the type and shape of the variable. After construction, the type and shape of the variable are fixed. The value can be changed using one of the assign methods.

If you want to change the shape of a variable later you have to use an assign Op with validate_shape=False.

Just like any Tensor, variables created with Variable() can be used as inputs for other Ops in the graph. Additionally, all the operators overloaded for the Tensor class are carried over to variables, so you can also add nodes to the graph by just doing arithmetic on variables.

import tensorflow as tf

# Create a variable.
w = tf.Variable(<initial-value>, name=<optional-name>)

# Use the variable in the graph like any Tensor.
y = tf.matmul(w, ...another variable or tensor...)

# The overloaded operators are available too.
z = tf.sigmoid(w + y)

# Assign a new value to the variable with `assign()` or a related method.
w.assign(w + 1.0)
w.assign_add(1.0)

When you launch the graph, variables have to be explicitly initialized before you can run Ops that use their value. You can initialize a variable by running its initializer op, restoring the variable from a save file, or simply running an assign Op that assigns a value to the variable. In fact, the variable initializer op is just an assign Op that assigns the variable's initial value to the variable itself.

# Launch the graph in a session.
with tf.compat.v1.Session() as sess:
    # Run the variable initializer.
    sess.run(w.initializer)
    # ...you now can run ops that use the value of 'w'...

The most common initialization pattern is to use the convenience function global_variables_initializer() to add an Op to the graph that initializes all the variables. You then run that Op after launching the graph.

# Add an Op to initialize global variables.
init_op = tf.compat.v1.global_variables_initializer()

# Launch the graph in a session.
with tf.compat.v1.Session() as sess:
    # Run the Op that initializes global variables.
    sess.run(init_op)
    # ...you can now run any Op that uses variable values...

If you need to create a variable with an initial value dependent on another variable, use the other variable's initialized_value(). This ensures that variables are initialized in the right order.

All variables are automatically collected in the graph where they are created. By default, the constructor adds the new variable to the graph collection GraphKeys.GLOBAL_VARIABLES. The convenience function global_variables() returns the contents of that collection.

When building a machine learning model it is often convenient to distinguish between variables holding the trainable model parameters and other variables such as a global step variable used to count training steps. To make this easier, the variable constructor supports a trainable=<bool> parameter. If True, the new variable is also added to the graph collection GraphKeys.TRAINABLE_VARIABLES. The convenience function trainable_variables() returns the contents of this collection. The various Optimizer classes use this collection as the default list of variables to optimize.

v = tf.Variable(True)
tf.cond(v, lambda: v.assign(False), my_false_fn)  # Note: this is broken.

Here, adding use_resource=True when constructing the variable will fix any nondeterminism issues:

v = tf.Variable(True, use_resource=True)
tf.cond(v, lambda: v.assign(False), my_false_fn)

To use the replacement for variables which does not have these issues:

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, 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. Defaults to True, unless synchronization is set to ON_READ, in which case it defaults to False.
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.
variable_def VariableDef protocol buffer. If not None, recreates the Variable object with its contents, referencing the variable's nodes in the graph, which must already exist. The graph is not changed. variable_def and the other arguments are mutually exclusive.
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.
expected_shape A TensorShape. If set, initial_value is expected to have this shape.
import_scope Optional string. Name scope to add to the Variable. Only used when initializing from protocol buffer.
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.
use_resource whether to use resource variables.
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.
aggregation Indicates how a distributed variable will be aggregated. Accepted values are constants defined in the class tf.VariableAggregation.
shape (optional) The shape of this variable. If None, the shape of initial_value will be used. When setting this argument to tf.TensorShape(None) (representing an unspecified shape), the variable can be assigned with values of different shapes.

ValueError If both variable_def and initial_value are specified.
ValueError If the initial value is not specified, or does not have a shape and validate_shape is True.
RuntimeError If eager execution is enabled.

aggregation

constraint Returns the constraint function associated with this variable.
device The device of this variable.
dtype The DType of this variable.
graph The Graph of this variable.
initial_value Returns the Tensor used as the initial value for the variable.

Note that this is different from initialized_value() which runs the op that initializes the variable before returning its value. This method returns the tensor that is used by the op that initializes the variable.

initializer The initializer operation for this variable.
name The name of this variable.
op The Operation of this variable.
shape The TensorShape of this variable.
synchronization

trainable

Child Classes

class SaveSliceInfo

Methods

assign

View source

Assigns a new value to the variable.

This is essentially a shortcut for assign(self, value).

Args
value A Tensor. The new value for this variable.
use_locking If True, use locking during the assignment.
name The name of the operation to be created
read_value if True, will return something which evaluates to the new value of the variable; if False will return the assign op.

Returns
A Tensor that will hold the new value of this variable after the assignment has completed.

assign_add

View source

Adds a value to this variable.

This is essentially a shortcut for assign_add(self, delta).

Args
delta A Tensor. The value to add to this variable.
use_locking If True, use locking during the operation.
name The name of the operation to be created
read_value if True, will return something which evaluates to the new value of the variable; if False will return the assign op.

Returns
A Tensor that will hold the new value of this variable after the addition has completed.

assign_sub

View source

Subtracts a value from this variable.

This is essentially a shortcut for assign_sub(self, delta).

Args
delta A Tensor. The value to subtract from this variable.
use_locking If True, use locking during the operation.
name The name of the operation to be created
read_value if True, will return something which evaluates to the new value of the variable; if False will return the assign op.

Returns
A Tensor that will hold the new value of this variable after the subtraction has completed.

batch_scatter_update

View source

Assigns tf.IndexedSlices to this variable batch-wise.

Analogous to batch_gather. This assumes that this variable and the sparse_delta IndexedSlices have a series of leading dimensions that are the same for all of them, and the updates are performed on the last dimension of indices. In other words, the dimensions should be the following:

num_prefix_dims = sparse_delta.indices.ndims - 1 batch_dim = num_prefix_dims + 1 sparse_delta.updates.shape = sparse_delta.indices.shape + var.shape[ batch_dim:]

where

sparse_delta.updates.shape[:num_prefix_dims] == sparse_delta.indices.shape[:num_prefix_dims] == var.shape[:num_prefix_dims]

And the operation performed can be expressed as:

var[i_1, ..., i_n, sparse_delta.indices[i_1, ..., i_n, j]] = sparse_delta.updates[ i_1, ..., i_n, j]

When sparse_delta.indices is a 1D tensor, this operation is equivalent to scatter_update.

To avoid this operation one can looping over the first ndims of the variable and using scatter_update on the subtensors that result of slicing the first dimension. This is a valid option for ndims = 1, but less efficient than this implementation.

Args
sparse_delta tf.IndexedSlices to be assigned to this variable.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered assignment has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

count_up_to

View source

Increments this variable until it reaches limit. (deprecated)

When that Op is run it tries to increment the variable by 1. If incrementing the variable would bring it above limit then the Op raises the exception OutOfRangeError.

If no error is raised, the Op outputs the value of the variable before the increment.

This is essentially a shortcut for count_up_to(self, limit).

Args
limit value at which incrementing the variable raises an error.

Returns
A Tensor that will hold the variable value before the increment. If no other Op modifies this variable, the values produced will all be distinct.

eval

View source

In a session, computes and returns the value of this variable.

This is not a graph construction method, it does not add ops to the graph.

This convenience method requires a session where the graph containing this variable has been launched. If no session is passed, the default session is used. See tf.compat.v1.Session for more information on launching a graph and on sessions.

v = tf.Variable([1, 2])
init = tf.compat.v1.global_variables_initializer()

with tf.compat.v1.Session() as sess:
    sess.run(init)
    # Usage passing the session explicitly.
    print(v.eval(sess))
    # Usage with the default session.  The 'with' block
    # above makes 'sess' the default session.
    print(v.eval())

Args
session The session to use to evaluate this variable. If none, the default session is used.

Returns
A numpy ndarray with a copy of the value of this variable.

experimental_ref

View source

Returns a hashable reference object to this Variable.

The primary usecase for this API is to put variables in a set/dictionary. We can't put variables in a set/dictionary as variable.__hash__() is no longer available starting Tensorflow 2.0.

import tensorflow as tf

x = tf.Variable(5)
y = tf.Variable(10)
z = tf.Variable(10)

# The followings will raise an exception starting 2.0
# TypeError: Variable is unhashable if Variable equality is enabled.
variable_set = {x, y, z}
variable_dict = {x: 'five', y: 'ten'}

Instead, we can use variable.experimental_ref().

variable_set = {x.experimental_ref(),
                y.experimental_ref(),
                z.experimental_ref()}

print(x.experimental_ref() in variable_set)
==> True

variable_dict = {x.experimental_ref(): 'five',
                 y.experimental_ref(): 'ten',
                 z.experimental_ref(): 'ten'}

print(variable_dict[y.experimental_ref()])
==> ten

Also, the reference object provides .deref() function that returns the original Variable.

x = tf.Variable(5)
print(x.experimental_ref().deref())
==> <tf.Variable 'Variable:0' shape=() dtype=int32, numpy=5>

from_proto

View source

Returns a Variable object created from variable_def.

gather_nd

View source

Gather slices from params into a Tensor with shape specified by indices.

See tf.gather_nd for details.

Args
indices A Tensor. Must be one of the following types: int32, int64. Index tensor.
name A name for the operation (optional).

Returns
A Tensor. Has the same type as params.

get_shape

View source

Alias of Variable.shape.

initialized_value

View source

Returns the value of the initialized variable. (deprecated)

You should use this instead of the variable itself to initialize another variable with a value that depends on the value of this variable.

# Initialize 'v' with a random tensor.
v = tf.Variable(tf.random.truncated_normal([10, 40]))
# Use `initialized_value` to guarantee that `v` has been
# initialized before its value is used to initialize `w`.
# The random values are picked only once.
w = tf.Variable(v.initialized_value() * 2.0)

Returns
A Tensor holding the value of this variable after its initializer has run.

load

View source

Load new value into this variable. (deprecated)

Writes new value to variable's memory. Doesn't add ops to the graph.

This convenience method requires a session where the graph containing this variable has been launched. If no session is passed, the default session is used. See tf.compat.v1.Session for more information on launching a graph and on sessions.

v = tf.Variable([1, 2])
init = tf.compat.v1.global_variables_initializer()

with tf.compat.v1.Session() as sess:
    sess.run(init)
    # Usage passing the session explicitly.
    v.load([2, 3], sess)
    print(v.eval(sess)) # prints [2 3]
    # Usage with the default session.  The 'with' block
    # above makes 'sess' the default session.
    v.load([3, 4], sess)
    print(v.eval()) # prints [3 4]

Args
value New variable value
session The session to use to evaluate this variable. If none, the default session is used.

Raises
ValueError Session is not passed and no default session

read_value

View source

Returns the value of this variable, read in the current context.

Can be different from value() if it's on another device, with control dependencies, etc.

Returns
A Tensor containing the value of the variable.

scatter_add

View source

Adds tf.IndexedSlices to this variable.

Args
sparse_delta tf.IndexedSlices to be added to this variable.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered addition has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

scatter_div

View source

Divide this variable by tf.IndexedSlices.

Args
sparse_delta tf.IndexedSlices to divide this variable by.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered division has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

scatter_max

View source

Updates this variable with the max of tf.IndexedSlices and itself.

Args
sparse_delta tf.IndexedSlices to use as an argument of max with this variable.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered maximization has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

scatter_min

View source

Updates this variable with the min of tf.IndexedSlices and itself.

Args
sparse_delta tf.IndexedSlices to use as an argument of min with this variable.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered minimization has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

scatter_mul

View source

Multiply this variable by tf.IndexedSlices.

Args
sparse_delta tf.IndexedSlices to multiply this variable by.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered multiplication has completed.

Raises
TypeError if sparse_delta is not an IndexedSlices.

scatter_nd_add

View source

Applies sparse addition to individual values or slices in a Variable.

The Variable has rank P and indices is a Tensor of rank Q.

indices must be integer tensor, containing indices into self. It must be shape [d_0, ..., d_{Q-2}, K] where 0 < K <= P.

The innermost dimension of indices (with length K) corresponds to indices into elements (if K = P) or slices (if K < P) along the Kth dimension of self.

updates is Tensor of rank Q-1+P-K with shape:

[d_0, ..., d_{Q-2}, self.shape[K], ..., self.shape[P-1]].

For example, say we want to add 4 scattered elements to a rank-1 tensor to 8 elements. In Python, that update would look like this:

    v = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8])
    indices = tf.constant([[4], [3], [1] ,[7]])
    updates = tf.constant([9, 10, 11, 12])
    add = v.scatter_nd_add(indices, updates)
    with tf.compat.v1.Session() as sess:
      print sess.run(add)

The resulting update to v would look like this:

[1, 13, 3, 14, 14, 6, 7, 20]

See tf.scatter_nd for more details about how to make updates to slices.

Args
indices The indices to be used in the operation.
updates The values to be used in the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered addition has completed.

scatter_nd_sub

View source

Applies sparse subtraction to individual values or slices in a Variable.

Assuming the variable has rank P and indices is a Tensor of rank Q.

indices must be integer tensor, containing indices into self. It must be shape [d_0, ..., d_{Q-2}, K] where 0 < K <= P.

The innermost dimension of indices (with length K) corresponds to indices into elements (if K = P) or slices (if K < P) along the Kth dimension of self.

updates is Tensor of rank Q-1+P-K with shape:

[d_0, ..., d_{Q-2}, self.shape[K], ..., self.shape[P-1]].

For example, say we want to add 4 scattered elements to a rank-1 tensor to 8 elements. In Python, that update would look like this:

    v = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8])
    indices = tf.constant([[4], [3], [1] ,[7]])
    updates = tf.constant([9, 10, 11, 12])
    op = v.scatter_nd_sub(indices, updates)
    with tf.compat.v1.Session() as sess:
      print sess.run(op)

The resulting update to v would look like this:

[1, -9, 3, -6, -6, 6, 7, -4]

See tf.scatter_nd for more details about how to make updates to slices.

Args
indices The indices to be used in the operation.
updates The values to be used in the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered subtraction has completed.

scatter_nd_update

View source

Applies sparse assignment to individual values or slices in a Variable.

The Variable has rank P and indices is a Tensor of rank Q.

indices must be integer tensor, containing indices into self. It must be shape [d_0, ..., d_{Q-2}, K] where 0 < K <= P.

The innermost dimension of indices (with length K) corresponds to indices into elements (if K = P) or slices (if K < P) along the Kth dimension of self.

updates is Tensor of rank Q-1+P-K with shape:

[d_0, ..., d_{Q-2}, self.shape[K], ..., self.shape[P-1]].

For example, say we want to add 4 scattered elements to a rank-1 tensor to 8 elements. In Python, that update would look like this:

    v = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8])
    indices = tf.constant([[4], [3], [1] ,[7]])
    updates = tf.constant([9, 10, 11, 12])
    op = v.scatter_nd_assign(indices, updates)
    with tf.compat.v1.Session() as sess:
      print sess.run(op)

The resulting update to v would look like this:

[1, 11, 3, 10, 9, 6, 7, 12]

See tf.scatter_nd for more details about how to make updates to slices.

Args
indices The indices to be used in the operation.
updates The values to be used in the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered assignment has completed.

scatter_sub

View source

Subtracts tf.IndexedSlices from this variable.

Args
sparse_delta tf.IndexedSlices to be subtracted from this variable.
use_locking If True, use locking during the operation.
name the name of the operation.

Returns
A Tensor that will hold the new value of this variable after the scattered subtraction has completed.

Raises