Classes and functions for building TensorFlow graphs.
class tf.Graph
A TensorFlow computation, represented as a dataflow graph.
A Graph
contains a set of
Operation
objects,
which represent units of computation; and
Tensor
objects, which represent
the units of data that flow between operations.
A default Graph
is always registered, and accessible by calling
tf.get_default_graph()
.
To add an operation to the default graph, simply call one of the functions
that defines a new Operation
:
c = tf.constant(4.0)
assert c.graph is tf.get_default_graph()
Another typical usage involves the
Graph.as_default()
context manager, which overrides the current default graph for the
lifetime of the context:
g = tf.Graph()
with g.as_default():
# Define operations and tensors in `g`.
c = tf.constant(30.0)
assert c.graph is g
Important note: This class is not thread-safe for graph construction. All operations should be created from a single thread, or external synchronization must be provided. Unless otherwise specified, all methods are not thread-safe.
tf.Graph.__init__()
Creates a new, empty Graph.
tf.Graph.as_default()
Returns a context manager that makes this Graph
the default graph.
This method should be used if you want to create multiple graphs
in the same process. For convenience, a global default graph is
provided, and all ops will be added to this graph if you do not
create a new graph explicitly. Use this method the with
keyword
to specify that ops created within the scope of a block should be
added to this graph.
The default graph is a property of the current thread. If you
create a new thread, and wish to use the default graph in that
thread, you must explicitly add a with g.as_default():
in that
thread's function.
The following code examples are equivalent:
# 1. Using Graph.as_default():
g = tf.Graph()
with g.as_default():
c = tf.constant(5.0)
assert c.graph is g
# 2. Constructing and making default:
with tf.Graph().as_default() as g:
c = tf.constant(5.0)
assert c.graph is g
A context manager for using this graph as the default graph.
tf.Graph.as_graph_def(from_version=None)
Returns a serialized GraphDef
representation of this graph.
The serialized GraphDef
can be imported into another Graph
(using import_graph_def()
) or used with the
C++ Session API.
This method is thread-safe.
from_version
: Optional. If this is set, returns a GraphDef
containing only the nodes that were added to this graph since
its version
property had the given value.A GraphDef
protocol buffer.
tf.Graph.finalize()
Finalizes this graph, making it read-only.
After calling g.finalize()
, no new operations can be added to
g
. This method is used to ensure that no operations are added
to a graph when it is shared between multiple threads, for example
when using a QueueRunner
.
tf.Graph.finalized
True if this graph has been finalized.
tf.Graph.control_dependencies(control_inputs)
Returns a context manager that specifies control dependencies.
Use with the with
keyword to specify that all operations constructed
within the context should have control dependencies on
control_inputs
. For example:
with g.control_dependencies([a, b, c]):
# `d` and `e` will only run after `a`, `b`, and `c` have executed.
d = ...
e = ...
Multiple calls to control_dependencies()
can be nested, and in
that case a new Operation
will have control dependencies on the union
of control_inputs
from all active contexts.
with g.control_dependencies([a, b]):
# Ops declared here run after `a` and `b`.
with g.control_dependencies([c, d]):
# Ops declared here run after `a`, `b`, `c`, and `d`.
N.B. The control dependencies context applies only to ops that are constructed within the context. Merely using an op or tensor in the context does not add a control dependency. The following example illustrates this point:
# WRONG
def my_func(pred, tensor):
t = tf.matmul(tensor, tensor)
with tf.control_dependencies([pred]):
# The matmul op is created outside the context, so no control
# dependency will be added.
return t
# RIGHT
def my_func(pred, tensor):
with tf.control_dependencies([pred]):
# The matmul op is created in the context, so a control dependency
# will be added.
return tf.matmul(tensor, tensor)
control_inputs
: A list of Operation
or Tensor
objects, which
must be executed or computed before running the operations
defined in the context.A context manager that specifies control dependencies for all operations constructed within the context.
TypeError
: If control_inputs
is not a list of Operation
or
Tensor
objects.
tf.Graph.device(device_name_or_function)
Returns a context manager that specifies the default device to use.
The device_name_or_function
argument may either be a device name
string, a device function, or None:
For example:
with g.device('/gpu:0'):
# All operations constructed in this context will be placed
# on GPU 0.
with g.device(None):
# All operations constructed in this context will have no
# assigned device.
# Defines a function from `Operation` to device string.
def matmul_on_gpu(n):
if n.type == "MatMul":
return "/gpu:0"
else:
return "/cpu:0"
with g.device(matmul_on_gpu):
# All operations of type "MatMul" constructed in this context
# will be placed on GPU 0; all other operations will be placed
# on CPU 0.
device_name_or_function
: The device name or function to use in
the context.A context manager that specifies the default device to use for newly created ops.
tf.Graph.name_scope(name)
Returns a context manager that creates hierarchical names for operations.
A graph maintains a stack of name scopes. A with name_scope(...):
statement pushes a new name onto the stack for the lifetime of the context.
The name
argument will be interpreted as follows:
name
is appended to the prefix of all operations created in the
context. If name
has been used before, it will be made unique by
calling self.unique_name(name)
.with g.name_scope(...) as scope:
statement will be treated as an "absolute" name scope, which
makes it possible to re-enter existing scopes.None
or the empty string will reset the current name scope
to the top-level (empty) name scope.For example:
with tf.Graph().as_default() as g:
c = tf.constant(5.0, name="c")
assert c_1.name == "c"
c_1 = tf.constant(6.0, name="c")
assert c_1.name == "c_1"
# Creates a scope called "nested"
with g.name_scope("nested") as scope:
nested_c = tf.constant(10.0, name="c")
assert nested_c.name == "nested/c"
# Creates a nested scope called "inner".
with g.name_scope("inner"):
nested_inner_c = tf.constant(20.0, name="c")
assert nested_inner_c.name == "nested/inner/c"
# Create a nested scope called "inner_1".
with g.name_scope("inner"):
nested_inner_1_c = tf.constant(30.0, name="c")
assert nested_inner_1_c.name == "nested/inner_1/c"
# Treats `scope` as an absolute name scope, and
# switches to the "nested/" scope.
with g.name_scope(scope):
nested_d = tf.constant(40.0, name="d")
assert nested_d.name == "nested/d"
with g.name_scope(""):
e = tf.constant(50.0, name="e")
assert e.name == "e"
The name of the scope itself can be captured by with g.name_scope(...) as scope:
, which stores the name of the scope
in the variable scope
. This value can be used to name an
operation that represents the overall result of executing the ops
in a scope. For example:
inputs = tf.constant(...)
with g.name_scope('my_layer') as scope:
weights = tf.Variable(..., name="weights")
biases = tf.Variable(..., name="biases")
affine = tf.matmul(inputs, weights) + biases
output = tf.nn.relu(affine, name=scope)
name
: A name for the scope.A context manager that installs name
as a new name scope.
A Graph
instance supports an arbitrary number of "collections"
that are identified by name. For convenience when building a large
graph, collections can store groups of related objects: for
example, the tf.Variable
uses a collection (named
tf.GraphKeys.VARIABLES
) for
all variables that are created during the construction of a graph. The caller
may define additional collections by specifying a new name.
tf.Graph.add_to_collection(name, value)
Stores value
in the collection with the given name
.
name
: The key for the collection. For example, the GraphKeys
class
contains many standard names for collections.value
: The value to add to the collection.
tf.Graph.get_collection(name, scope=None)
Returns a list of values in the collection with the given name
.
key
: The key for the collection. For example, the GraphKeys
class
contains many standard names for collections.scope
: (Optional.) If supplied, the resulting list is filtered to include
only items whose name begins with this string.The list of values in the collection with the given name
, or
an empty list if no value has been added to that collection. The
list contains the values in the order under which they were
collected.
tf.Graph.as_graph_element(obj, allow_tensor=True, allow_operation=True)
Returns the object referred to by obj
, as an Operation
or Tensor
.
This function validates that obj
represents an element of this
graph, and gives an informative error message if it is not.
This function is the canonical way to get/validate an object of one of the allowed types from an external argument reference in the Session API.
This method may be called concurrently from multiple threads.
obj
: A Tensor
, an Operation
, or the name of a tensor or operation.
Can also be any object with an _as_graph_element()
method that returns
a value of one of these types.allow_tensor
: If true, obj
may refer to a Tensor
.allow_operation
: If true, obj
may refer to an Operation
.The Tensor
or Operation
in the Graph corresponding to obj
.
TypeError
: If obj
is not a type we support attempting to convert
to types.ValueError
: If obj
is of an appropriate type but invalid. For
example, an invalid string.KeyError
: If obj
is not an object in the graph.
tf.Graph.get_operation_by_name(name)
Returns the Operation
with the given name
.
This method may be called concurrently from multiple threads.
name
: The name of the Operation
to return.The Operation
with the given name
.
TypeError
: If name
is not a string.KeyError
: If name
does not correspond to an operation in this graph.
tf.Graph.get_tensor_by_name(name)
Returns the Tensor
with the given name
.
This method may be called concurrently from multiple threads.
name
: The name of the Tensor
to return.The Tensor
with the given name
.
TypeError
: If name
is not a string.KeyError
: If name
does not correspond to a tensor in this graph.
tf.Graph.get_operations()
Return the list of operations in the graph.
You can modify the operations in place, but modifications to the list such as inserts/delete have no effect on the list of operations known to the graph.
This method may be called concurrently from multiple threads.
A list of Operations.
tf.Graph.get_default_device()
Returns the default device.
A string.
tf.Graph.seed
tf.Graph.unique_name(name)
Return a unique Operation name for "name".
Note: You rarely need to call unique_name() directly. Most of the time you just need to create "with g.name_scope()" blocks to generate structured names.
unique_name
is used to generate structured names, separated by "/",
to help identify Operations when debugging a Graph. Operation names
are displayed in error messages reported by the TensorFlow runtime,
and in various visualization tools such as TensorBoard.
name
: The name for an Operation
.A string to be passed to create_op()
that will be used
to name the operation being created.
tf.Graph.version
Returns a version number that increases as ops are added to the graph.
tf.Graph.create_op(op_type, inputs, dtypes, input_types=None, name=None, attrs=None, op_def=None, compute_shapes=True)
Creates an Operation
in this graph.
This is a low-level interface for creating an Operation
. Most
programs will not call this method directly, and instead use the
Python op constructors, such as tf.constant()
, which add ops to
the default graph.
op_type
: The Operation
type to create. This corresponds to the
OpDef.name
field for the proto that defines the operation.inputs
: A list of Tensor
objects that will be inputs to the Operation
.dtypes
: A list of DType
objects that will be the types of the tensors
that the operation produces.input_types
: (Optional.) A list of DType
s that will be the types of
the tensors that the operation consumes. By default, uses the base
DType
of each input in inputs
. Operations that expect
reference-typed inputs must specify input_types
explicitly.name
: (Optional.) A string name for the operation. If not specified, a
name is generated based on op_type
.attrs
: (Optional.) A list of AttrValue
protos for the attr
field of
the NodeDef
proto that will represent the operation.op_def
: (Optional.) The OpDef
proto that describes the op_type
that
the operation will have.compute_shapes
: (Optional.) If True, shape inference will be performed
to compute the shapes of the outputs.TypeError
: if any of the inputs is not a Tensor
.An Operation
object.
tf.Graph.gradient_override_map(op_type_map)
EXPERIMENTAL: A context manager for overriding gradient functions.
This context manager can be used to override the gradient function that will be used for ops within the scope of the context.
For example:
@tf.RegisterGradient("CustomSquare")
def _custom_square_grad(op, inputs):
# ...
with tf.Graph().as_default() as g:
c = tf.constant(5.0)
s_1 = tf.square(c) # Uses the default gradient for tf.square.
with g.gradient_override_map({"Square": "CustomSquare"}):
s_2 = tf.square(s_2) # Uses _custom_square_grad to compute the
# gradient of s_2.
op_type_map
: A dictionary mapping op type strings to alternative op
type strings.A context manager that sets the alternative op type to be used for one or more ops created in that context.
TypeError
: If op_type_map
is not a dictionary mapping strings to
strings.
class tf.Operation
Represents a graph node that performs computation on tensors.
An Operation
is a node in a TensorFlow Graph
that takes zero or
more Tensor
objects as input, and produces zero or more Tensor
objects as output. Objects of type Operation
are created by
calling a Python op constructor (such as
tf.matmul()
)
or Graph.create_op()
.
For example c = tf.matmul(a, b)
creates an Operation
of type
"MatMul" that takes tensors a
and b
as input, and produces c
as output.
After the graph has been launched in a session, an Operation
can
be executed by passing it to
Session.run()
.
op.run()
is a shortcut for calling tf.get_default_session().run(op)
.
tf.Operation.name
The full name of this operation.
tf.Operation.type
The type of the op (e.g. "MatMul"
).
tf.Operation.inputs
The list of Tensor
objects representing the data inputs of this op.
tf.Operation.control_inputs
The Operation
objects on which this op has a control dependency.
Before this op is executed, TensorFlow will ensure that the
operations in self.control_inputs
have finished executing. This
mechanism can be used to run ops sequentially for performance
reasons, or to ensure that the side effects of an op are observed
in the correct order.
A list of Operation
objects.
tf.Operation.outputs
The list of Tensor
objects representing the outputs of this op.
tf.Operation.device
The name of the device to which this op has been assigned, if any.
The string name of the device to which this op has been assigned, or None if it has not been assigned to a device.
tf.Operation.graph
The Graph
that contains this operation.
tf.Operation.run(feed_dict=None, session=None)
Runs this operation in a Session
.
Calling this method will execute all preceding operations that produce the inputs needed for this operation.
N.B. Before invoking Operation.run()
, its graph must have been
launched in a session, and either a default session must be
available, or session
must be specified explicitly.
feed_dict
: A dictionary that maps Tensor
objects to feed values.
See Session.run()
for a description of the valid feed values.session
: (Optional.) The Session
to be used to run to this operation. If
none, the default session will be used.
tf.Operation.get_attr(name)
Returns the value of the attr of this op with the given name
.
name
: The name of the attr to fetch.The value of the attr, as a Python object.
ValueError
: If this op does not have an attr with the given name
.
tf.Operation.traceback
Returns the call stack from when this operation was constructed.
tf.Operation.__init__(node_def, g, inputs=None, output_types=None, control_inputs=None, input_types=None, original_op=None, op_def=None)
Creates an Operation
.
NOTE: This constructor validates the name of the Operation (passed as "node_def.name"). Valid Operation names match the following regular expression:
[A-Za-z0-9.][A-Za-z0-9_.-/]*
node_def
: graph_pb2.NodeDef. NodeDef for the Operation.
Used for attributes of graph_pb2.NodeDef, typically "name",
"op", and "device". The "input" attribute is irrelevant here
as it will be computed when generating the model.g
: Graph. The parent graph.inputs
: list of Tensor objects. The inputs to this Operation.output_types
: list of types_pb2.DataType. List of the types of the
Tensors computed by this operation. The length of this list indicates
the number of output endpoints of the Operation.control_inputs
: list of operations or tensors from which to have a
control dependency.input_types
: List of types_pb2.DataType representing the
types of the Tensors accepted by the Operation. By default
uses [x.dtype.base_dtype for x in inputs]. Operations that expect
reference-typed inputs must specify these explicitly.original_op
: Optional. Used to associate the new Operation with an
existing Operation (for example, a replica with the op that was
replicated).op_def
: Optional. The op_def_pb2.OpDef proto that describes the
op type that this Operation represents.TypeError
: if control inputs are not Operations or Tensors,
or if node_def is not a NodeDef,
or if g is not a Graph,
or if inputs are not Tensors,
or if inputs and input_types are incompatible.ValueError
: if the node_def name is not valid.
tf.Operation.node_def
Returns a serialized NodeDef
representation of this operation.
A
NodeDef
protocol buffer.
tf.Operation.op_def
Returns the OpDef
proto that represents the type of this op.
An
OpDef
protocol buffer.
tf.Operation.values()
DEPRECATED: Use outputs.
class tf.Tensor
Represents a value produced by an Operation
.
A Tensor
is a symbolic handle to one of the outputs of an
Operation
. It does not hold the values of that operation's output,
but instead provides a means of computing those values in a
TensorFlow Session
.
This class has two primary purposes:
A Tensor
can be passed as an input to another Operation
.
This builds a dataflow connection between operations, which
enables TensorFlow to execute an entire Graph
that represents a
large, multi-step computation.
Tensor
can be computed by passing it to
Session.run()
.
t.eval()
is a shortcut for calling
tf.get_default_session().run(t)
.In the following example, c
, d
, and e
are symbolic Tensor
objects, whereas result
is a numpy array that stores a concrete
value:
# Build a dataflow graph.
c = tf.constant([[1.0, 2.0], [3.0, 4.0]])
d = tf.constant([[1.0, 1.0], [0.0, 1.0]])
e = tf.matmul(c, d)
# Construct a `Session` to execut the graph.
sess = tf.Session()
# Execute the graph and store the value that `e` represents in `result`.
result = sess.run(e)
tf.Tensor.dtype
The DType
of elements in this tensor.
tf.Tensor.name
The string name of this tensor.
tf.Tensor.value_index
The index of this tensor in the outputs of its Operation
.
tf.Tensor.graph
The Graph
that contains this tensor.
tf.Tensor.op
The Operation
that produces this tensor as an output.
tf.Tensor.consumers()
Returns a list of Operation
s that consume this tensor.
A list of Operation
s.
tf.Tensor.eval(feed_dict=None, session=None)
Evaluates this tensor in a Session
.
Calling this method will execute all preceding operations that produce the inputs needed for the operation that produces this tensor.
N.B. Before invoking Tensor.eval()
, its graph must have been
launched in a session, and either a default session must be
available, or session
must be specified explicitly.
feed_dict
: A dictionary that maps Tensor
objects to feed values.
See Session.run()
for a
description of the valid feed values.session
: (Optional.) The Session
to be used to evaluate this tensor. If
none, the default session will be used.A numpy array corresponding to the value of this tensor.
tf.Tensor.get_shape()
Returns the TensorShape
that represents the shape of this tensor.
The shape is computed using shape inference functions that are
registered for each Operation
type using tf.RegisterShape
.
See TensorShape
for more
details of what a shape represents.
The inferred shape of a tensor is used to provide shape information without having to launch the graph in a session. This can be used for debugging, and providing early error messages. For example:
c = tf.constant([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
print c.get_shape()
==> TensorShape([Dimension(2), Dimension(3)])
d = tf.constant([[1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0]])
print d.get_shape()
==> TensorShape([Dimension(4), Dimension(2)])
# Raises a ValueError, because `c` and `d` do not have compatible
# inner dimensions.
e = tf.matmul(c, d)
f = tf.matmul(c, d, transpose_a=True, transpose_b=True)
print f.get_shape()
==> TensorShape([Dimension(3), Dimension(4)])
In some cases, the inferred shape may have unknown dimensions. If
the caller has additional information about the values of these
dimensions, Tensor.set_shape()
can be used to augment the
inferred shape.
A TensorShape
representing the shape of this tensor.
tf.Tensor.set_shape(shape)
Updates the shape of this tensor.
This method can be called multiple times, and will merge the given
shape
with the current shape of this tensor. It can be used to
provide additional information about the shape of this tensor that
cannot be inferred from the graph alone. For example, this can be used
to provide additional information about the shapes of images:
_, image_data = tf.TFRecordReader(...).read(...)
image = tf.image.decode_png(image_data, channels=3)
# The height and width dimensions of `image` are data dependent, and
# cannot be computed without executing the op.
print image.get_shape()
==> TensorShape([Dimension(None), Dimension(None), Dimension(3)])
# We know that each image in this dataset is 28 x 28 pixels.
image.set_shape([28, 28, 3])
print image.get_shape()
==> TensorShape([Dimension(28), Dimension(28), Dimension(3)])
shape
: A TensorShape
representing the shape of this tensor.ValueError
: If shape
is not compatible with the current shape of
this tensor.tf.Tensor.__init__(op, value_index, dtype)
Creates a new Tensor
.
op
: An Operation
. Operation
that computes this tensor.value_index
: An int
. Index of the operation's endpoint that produces
this tensor.dtype
: A types.DType
. Type of data stored in this tensor.TypeError
: If the op is not an Operation
.
tf.Tensor.device
The name of the device on which this tensor will be produced, or None.
class tf.DType
Represents the type of the elements in a Tensor
.
The following DType
objects are defined:
tf.float32
: 32-bit single-precision floating-point.tf.float64
: 64-bit double-precision floating-point.tf.bfloat16
: 16-bit truncated floating-point.tf.complex64
: 64-bit single-precision complex.
tf.int8
: 8-bit signed integer.tf.uint8
: 8-bit unsigned integer.tf.int32
: 32-bit signed integer.tf.int64
: 64-bit signed integer.
tf.bool
: Boolean.
tf.string
: String.
tf.qint8
: Quantized 8-bit signed integer.tf.quint8
: Quantized 8-bit unsigned integer.tf.qint32
: Quantized 32-bit signed integer.In addition, variants of these types with the _ref
suffix are
defined for reference-typed tensors.
The tf.as_dtype()
function converts numpy types and string type
names to a DType
object.
tf.DType.is_compatible_with(other)
Returns True if the other
DType will be converted to this DType.
The conversion rules are as follows:
DType(T) .is_compatible_with(DType(T)) == True
DType(T) .is_compatible_with(DType(T).as_ref) == True
DType(T).as_ref.is_compatible_with(DType(T)) == False
DType(T).as_ref.is_compatible_with(DType(T).as_ref) == True
other
: A DType
(or object that may be converted to a DType
).True if a Tensor of the other
DType
will be implicitly converted to
this DType
.
tf.DType.name
Returns the string name for this DType
.
tf.DType.base_dtype
Returns a non-reference DType
based on this DType
.
tf.DType.is_ref_dtype
Returns True
if this DType
represents a reference type.
tf.DType.as_ref
Returns a reference DType
based on this DType
.
tf.DType.is_integer
Returns whether this is a (non-quantized) integer type.
tf.DType.is_quantized
Returns whether this is a quantized data type.
tf.DType.as_numpy_dtype
Returns a numpy.dtype
based on this DType
.
tf.DType.as_datatype_enum
Returns a types_pb2.DataType
enum value based on this DType
.
tf.DType.__init__(type_enum)
Creates a new DataType
.
NOTE(mrry): In normal circumstances, you should not need to construct a DataType object directly. Instead, use the types.as_dtype() function.
type_enum
: A types_pb2.DataType
enum value.TypeError
: If type_enum
is not a value types_pb2.DataType
.
tf.DType.max
Returns the maximum representable value in this data type.
TypeError
: if this is a non-numeric, unordered, or quantized type.
tf.DType.min
Returns the minimum representable value in this data type.
TypeError
: if this is a non-numeric, unordered, or quantized type.
tf.as_dtype(type_value)
Converts the given type_value
to a DType
.
type_value
: A value that can be converted to a tf.DType
object. This may currently be a tf.DType
object, a
DataType
enum,
a string type name, or a numpy.dtype
.A DType
corresponding to type_value
.
TypeError
: If type_value
cannot be converted to a DType
.tf.device(dev)
Wrapper for Graph.device()
using the default graph.
See
Graph.name_scope()
for more details.
device_name_or_function
: The device name or function to use in
the context.A context manager that specifies the default device to use for newly created ops.
tf.name_scope(name)
Wrapper for Graph.name_scope()
using the default graph.
See
Graph.name_scope()
for more details.
name
: A name for the scope.A context manager that installs name
as a new name scope in the
default graph.
tf.control_dependencies(control_inputs)
Wrapper for Graph.control_dependencies()
using the default graph.
See Graph.control_dependencies()
for more details.
control_inputs
: A list of Operation
or Tensor
objects, which
must be executed or computed before running the operations
defined in the context.A context manager that specifies control dependencies for all operations constructed within the context.
tf.convert_to_tensor(value, dtype=None, name=None)
Converts the given value
to a Tensor
.
This function converts Python objects of various types to Tensor
objects. It accepts Tensor
objects, numpy arrays, Python lists,
and Python scalars. For example:
import numpy as np
array = np.random.rand((32, 100, 100))
def my_func(arg):
arg = tf.convert_to_tensor(arg, dtype=tf.float32)
return tf.matmul(arg, arg) + arg
# The following calls are equivalent.
value_1 = my_func(tf.constant([[1.0, 2.0], [3.0, 4.0]]))
value_2 = my_func([[1.0, 2.0], [3.0, 4.0]])
value_3 = my_func(np.array([[1.0, 2.0], [3.0, 4.0]], dtype=np.float32))
This function can be useful when composing a new operation in Python
(such as my_func
in the example above). All standard Python op
constructors apply this function to each of their Tensor-valued
inputs, which allows those ops to accept numpy arrays, Python lists,
and scalars in addition to Tensor
objects.
value
: An object whose type has a registered Tensor
conversion function.dtype
: Optional element type for the returned tensor. If missing, the
type is inferred from the type of value
.name
: Optional name to use if a new Tensor
is created.A Tensor
based on value
.
TypeError
: If no conversion function is registered for value
.RuntimeError
: If a registered conversion function returns an invalid value.
tf.get_default_graph()
Returns the default graph for the current thread.
The returned graph will be the innermost graph on which a
Graph.as_default()
context has been entered, or a global default
graph if none has been explicitly created.
N.B. The default graph is a property of the current thread. If you
create a new thread, and wish to use the default graph in that
thread, you must explicitly add a with g.as_default():
in that
thread's function.
The default Graph
being used in the current thread.
tf.import_graph_def(graph_def, input_map=None, return_elements=None, name=None, op_dict=None)
Imports the TensorFlow graph in graph_def
into the Python Graph
.
This function provides a way to import a serialized TensorFlow
GraphDef
protocol buffer, and extract individual objects in the GraphDef
as
Tensor
and Operation
objects. See
Graph.as_graph_def()
for a way to create a
GraphDef
proto.
graph_def
: A GraphDef
proto containing operations to be imported into
the default graph.input_map
: A dictionary mapping input names (as strings) in graph_def
to Tensor
objects. The values of the named input tensors in the
imported graph will be re-mapped to the respective Tensor
values.return_elements
: A list of strings containing operation names in
graph_def
that will be returned as Operation
objects; and/or
tensor names in graph_def
that will be returned as Tensor
objects.name
: (Optional.) A prefix that will be prepended to the names in
graph_def
. Defaults to "import"
.op_dict
: (Optional.) A dictionary mapping op type names to OpDef
protos.
Must contain an OpDef
proto for each op type named in graph_def
.
If omitted, uses the OpDef
protos registered in the global registry.A list of Operation
and/or Tensor
objects from the imported graph,
corresponding to the names in `return_elements'.
TypeError
: If graph_def
is not a GraphDef
proto,
input_map' is not a dictionary mapping strings to
Tensorobjects, or
return_elements` is not a list of strings.ValueError
: If input_map
, or return_elements
contains names that
do not appear in graph_def
, or graph_def
is not well-formed (e.g.
it refers to an unknown tensor).tf.add_to_collection(name, value)
Wrapper for Graph.add_to_collection()
using the default graph.
See Graph.add_to_collection()
for more details.
name
: The key for the collection. For example, the GraphKeys
class
contains many standard names for collections.value
: The value to add to the collection.
tf.get_collection(key, scope=None)
Wrapper for Graph.get_collection()
using the default graph.
See Graph.get_collection()
for more details.
key
: The key for the collection. For example, the GraphKeys
class
contains many standard names for collections.scope
: (Optional.) If supplied, the resulting list is filtered to include
only items whose name begins with this string.The list of values in the collection with the given name
, or
an empty list if no value has been added to that collection. The
list contains the values in the order under which they were
collected.
class tf.GraphKeys
Standard names to use for graph collections.
The standard library uses various well-known names to collect and
retrieve values associated with a graph. For example, the
tf.Optimizer
subclasses default to optimizing the variables
collected under tf.GraphKeys.TRAINABLE_VARIABLES
if none is
specified, but it is also possible to pass an explicit list of
variables.
The following standard keys are defined:
VARIABLES
: the Variable
objects that comprise a model, and
must be saved and restored together. See
tf.all_variables()
for more details.TRAINABLE_VARIABLES
: the subset of Variable
objects that will
be trained by an optimizer. See
tf.trainable_variables()
for more details.SUMMARIES
: the summary Tensor
objects that have been created in the
graph. See
tf.merge_all_summaries()
for more details.QUEUE_RUNNERS
: the QueueRunner
objects that are used to
produce input for a computation. See
tf.start_queue_runners()
for more details.class tf.RegisterGradient
A decorator for registering the gradient function for an op type.
This decorator is only used when defining a new op type. For an op
with m
inputs and n
inputs, the gradient function is a function
that takes the original Operation
and n
Tensor
objects
(representing the gradients with respect to each output of the op),
and returns m
Tensor
objects (representing the partial gradients
with respect to each input of the op).
For example, assuming that operations of type "Sub"
take two
inputs x
and y
, and return a single output x - y
, the
following gradient function would be registered:
@tf.RegisterGradient("Sub")
def _sub_grad(unused_op, grad):
return grad, tf.Neg(grad)
The decorator argument op_type
is the string type of an
operation. This corresponds to the OpDef.name
field for the proto
that defines the operation.
tf.RegisterGradient.__init__(op_type)
Creates a new decorator with op_type
as the Operation type.
op_type
: The string type of an operation. This corresponds to the
OpDef.name
field for the proto that defines the operation.
tf.NoGradient(op_type)
Specifies that ops of type op_type
do not have a defined gradient.
This function is only used when defining a new op type. It may be
used for ops such as tf.size()
that are not differentiable. For
example:
tf.NoGradient("Size")
op_type
: The string type of an operation. This corresponds to the
OpDef.name
field for the proto that defines the operation.TypeError
: If op_type
is not a string.
class tf.RegisterShape
A decorator for registering the shape function for an op type.
This decorator is only used when defining a new op type. A shape
function is a function from an Operation
object to a list of
TensorShape
objects, with one TensorShape
for each output of the
operation.
For example, assuming that operations of type "Sub"
take two
inputs x
and y
, and return a single output x - y
, all with the
same shape, the following shape function would be registered:
@tf.RegisterShape("Sub")
def _sub_shape(op):
return [op.inputs[0].get_shape().merge_with(op.inputs[1].get_shape())]
The decorator argument op_type
is the string type of an
operation. This corresponds to the OpDef.name
field for the proto
that defines the operation.
tf.RegisterShape.__init__(op_type)
Saves the "op_type" as the Operation type.
class tf.TensorShape
Represents the shape of a Tensor
.
A TensorShape
represents a possibly-partial shape specification for a
Tensor
. It may be one of the following:
If a tensor is produced by an operation of type "Foo"
, its shape
may be inferred if there is a registered shape function for
"Foo"
. See tf.RegisterShape()
for details of shape
functions and how to register them. Alternatively, the shape may be set
explicitly using Tensor.set_shape()
.
tf.TensorShape.merge_with(other)
Returns a TensorShape
combining the information in self
and other
.
The dimensions in self
and other
are merged elementwise,
according to the rules defined for Dimension.merge_with()
.
other
: Another TensorShape
.A TensorShape
containing the combined information of self
and
other
.
ValueError
: If self
and other
are not compatible.
tf.TensorShape.concatenate(other)
Returns the concatenation of the dimension in self
and other
.
N.B. If either self
or other
is completely unknown,
concatenation will discard information about the other shape. In
future, we might support concatenation that preserves this
information for use with slicing.
other
: Another TensorShape
.A TensorShape
whose dimensions are the concatenation of the
dimensions in self
and other
.
tf.TensorShape.ndims
Returns the rank of this shape, or None if it is unspecified.
tf.TensorShape.dims
Returns a list of Dimensions, or None if the shape is unspecified.
tf.TensorShape.as_list()
Returns a list of integers or None for each dimension.
tf.TensorShape.is_compatible_with(other)
Returns True iff self
is compatible with other
.
Two possibly-partially-defined shapes are compatible if there exists a fully-defined shape that both shapes can represent. Thus, compatibility allows the shape inference code to reason about partially-defined shapes. For example:
TensorShape(None) is compatible with all shapes.
TensorShape([None, None]) is compatible with all two-dimensional shapes, such as TensorShape([32, 784]), and also TensorShape(None). It is not compatible with, for example, TensorShape([None]) or TensorShape([None, None, None]).
TensorShape([32, None]) is compatible with all two-dimensional shapes with size 32 in the 0th dimension, and also TensorShape([None, None]) and TensorShape(None). It is not compatible with, for example, TensorShape([32]), TensorShape([32, None, 1]) or TensorShape([64, None]).
The compatibility relation is reflexive and symmetric, but not transitive. For example, TensorShape([32, 784]) is compatible with TensorShape(None), and TensorShape(None) is compatible with TensorShape([4, 4]), but TensorShape([32, 784]) is not compatible with TensorShape([4, 4]).
other
: Another TensorShape.True iff self
is compatible with other
.
tf.TensorShape.is_fully_defined()
Returns True iff self
is fully defined in every dimension.
tf.TensorShape.with_rank(rank)
Returns a shape based on self
with the given rank.
This method promotes a completely unknown shape to one with a known rank.
rank
: An integer.A shape that is at least as specific as self
with the given rank.
ValueError
: If self
does not represent a shape with the given rank
.
tf.TensorShape.with_rank_at_least(rank)
Returns a shape based on self
with at least the given rank.
rank
: An integer.A shape that is at least as specific as self
with at least the given
rank.
ValueError
: If self
does not represent a shape with at least the given
rank
.
tf.TensorShape.with_rank_at_most(rank)
Returns a shape based on self
with at most the given rank.
rank
: An integer.A shape that is at least as specific as self
with at most the given
rank.
ValueError
: If self
does not represent a shape with at most the given
rank
.
tf.TensorShape.assert_has_rank(rank)
Raises an exception if self
is not compatible with the given rank
.
rank
: An integer.ValueError
: If self
does not represent a shape with the given rank
.
tf.TensorShape.assert_same_rank(other)
Raises an exception if self
and other
do not have compatible ranks.
other
: Another TensorShape
.ValueError
: If self
and other
do not represent shapes with the
same rank.
tf.TensorShape.assert_is_compatible_with(other)
Raises exception if self
and other
do not represent the same shape.
This method can be used to assert that there exists a shape that both
self
and other
represent.
other
: Another TensorShape.ValueError
: If self
and other
do not represent the same shape.
tf.TensorShape.assert_is_fully_defined()
Raises an exception if self
is not fully defined in every dimension.
ValueError
: If self
does not have a known value for every dimension.tf.TensorShape.__init__(dims)
Creates a new TensorShape with the given dimensions.
dims
: A list of Dimensions, or None if the shape is unspecified.DEPRECATED
: A single integer is treated as a singleton list.
tf.TensorShape.as_dimension_list()
DEPRECATED: use as_list().
tf.TensorShape.num_elements()
Returns the total number of elements, or none for incomplete shapes.
class tf.Dimension
Represents the value of one dimension in a TensorShape.
tf.Dimension.__init__(value)
Creates a new Dimension with the given value.
tf.Dimension.assert_is_compatible_with(other)
Raises an exception if other
is not compatible with this Dimension.
other
: Another Dimension.ValueError
: If self
and other
are not compatible (see
is_compatible_with).
tf.Dimension.is_compatible_with(other)
Returns true if other
is compatible with this Dimension.
Two known Dimensions are compatible if they have the same value. An unknown Dimension is compatible with all other Dimensions.
other
: Another Dimension.True if this Dimension and other
are compatible.
tf.Dimension.merge_with(other)
Returns a Dimension that combines the information in self
and other
.
Dimensions are combined as follows:
Dimension(n) .merge_with(Dimension(n)) == Dimension(n) Dimension(n) .merge_with(Dimension(None)) == Dimension(n) Dimension(None).merge_with(Dimension(n)) == Dimension(n) Dimension(None).merge_with(Dimension(None)) == Dimension(None) Dimension(n) .merge_with(Dimension(m)) raises ValueError for n != m
other
: Another Dimension.A Dimension containing the combined information of self
and
other
.
ValueError
: If self
and other
are not compatible (see
is_compatible_with).
tf.Dimension.value
The value of this dimension, or None if it is unknown.
tf.op_scope(values, name, default_name)
Returns a context manager for use when defining a Python op.
This context manager validates that the given values
are from the
same graph, ensures that that graph is the default graph, and pushes a
name scope.
For example, to define a new Python op called my_op
:
def my_op(a, b, c, name=None):
with tf.op_scope([a, b, c], name, "MyOp") as scope:
a = tf.convert_to_tensor(a, name="a")
b = tf.convert_to_tensor(b, name="b")
c = tf.convert_to_tensor(c, name="c")
# Define some computation that uses `a`, `b`, and `c`.
return foo_op(..., name=scope)
values
: The list of Tensor
arguments that are passed to the op function.name
: The name argument that is passed to the op function.default_name
: The default name to use if the name
argument is None
.A context manager for use in defining a Python op.
tf.get_seed(op_seed)
Returns the local seeds an operation should use given an op-specific seed.
Given operation-specific seed, op_seed
, this helper function returns two
seeds derived from graph-level and op-level seeds. Many random operations
internally use the two seeds to allow user to change the seed globally for a
graph, or for only specific operations.
For details on how the graph-level seed interacts with op seeds, see
set_random_seed
.
op_seed
: integer.A tuple of two integers that should be used for the local seed of this operation.