3RNN/Lib/site-packages/tensorflow/python/framework/tensor.py
2024-05-26 19:49:15 +02:00

1464 lines
51 KiB
Python

# Copyright 2023 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.
# ==============================================================================
"""Tensor and TensorSpec classes."""
from typing import Optional, Type
import numpy as np
from tensorflow.core.framework import attr_value_pb2
from tensorflow.core.function import trace_type
from tensorflow.core.protobuf import struct_pb2
from tensorflow.python import tf2
from tensorflow.python.eager import context
from tensorflow.python.eager import monitoring
from tensorflow.python.eager import record
from tensorflow.python.framework import common_shapes
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import errors
from tensorflow.python.framework import op_callbacks
from tensorflow.python.framework import stack
from tensorflow.python.framework import tensor_conversion_registry
from tensorflow.python.framework import tensor_shape
from tensorflow.python.framework import tensor_util
from tensorflow.python.framework import type_spec
from tensorflow.python.framework import type_spec_registry
from tensorflow.python.ops import handle_data_util
from tensorflow.python.platform import tf_logging as logging
from tensorflow.python.saved_model import nested_structure_coder
from tensorflow.python.types import core as core_tf_types
from tensorflow.python.types import internal
from tensorflow.python.util import compat
from tensorflow.python.util import deprecation
from tensorflow.python.util import object_identity
from tensorflow.python.util.tf_export import tf_export
_tensor_equality_api_usage_gauge = monitoring.BoolGauge(
"/tensorflow/api/enable_tensor_equality",
"Whether ops.enable_tensor_equality() is called.")
def _override_helper(clazz_object, operator, func):
"""Overrides (string) operator on Tensors to call func.
Args:
clazz_object: the class to override for; either Tensor or SparseTensor.
operator: the string name of the operator to override.
func: the function that replaces the overridden operator.
Raises:
ValueError: If operator is not allowed to be overwritten.
"""
if operator not in Tensor.OVERLOADABLE_OPERATORS:
raise ValueError(f"Overriding {operator} is disallowed. "
f"Allowed operators are {Tensor.OVERLOADABLE_OPERATORS}.")
setattr(clazz_object, operator, func)
def _eval_using_default_session(tensors, feed_dict, graph, session=None):
"""Uses the default session to evaluate one or more tensors.
Args:
tensors: A single Tensor, or a list of Tensor objects.
feed_dict: A dictionary that maps Tensor objects (or tensor names) to lists,
numpy ndarrays, TensorProtos, or strings.
graph: The graph in which the tensors are defined.
session: (Optional) A different session to use to evaluate "tensors".
Returns:
Either a single numpy ndarray if "tensors" is a single tensor; or a list
of numpy ndarrays that each correspond to the respective element in
"tensors".
Raises:
ValueError: If no default session is available; the default session
does not have "graph" as its graph; or if "session" is specified,
and it does not have "graph" as its graph.
"""
if session is None:
session = stack.get_default_session()
if session is None:
raise ValueError("Cannot evaluate tensor using `eval()`: No default "
"session is registered. Use `with "
"sess.as_default()` or pass an explicit session to "
"`eval(session=sess)`")
if session.graph is not graph:
raise ValueError("Cannot use the default session to evaluate tensor: "
"the tensor's graph is different from the session's "
"graph. Pass an explicit session to "
"`eval(session=sess)`.")
else:
if session.graph is not graph:
raise ValueError("Cannot use the given session to evaluate tensor: "
"the tensor's graph is different from the session's "
"graph.")
return session.run(tensors, feed_dict)
def _add_error_prefix(msg, *, name=None):
return msg if name is None else f"{name}: {msg}"
class _TensorIterator(object):
"""Iterates over the leading dim of a Tensor. Performs no error checks."""
__slots__ = ["_tensor", "_index", "_limit"]
def __init__(self, tensor, dim0):
self._tensor = tensor
self._index = 0
self._limit = dim0
def __iter__(self):
return self
def __next__(self):
if self._index == self._limit:
raise StopIteration
result = self._tensor[self._index]
self._index += 1
return result
next = __next__ # python2.x compatibility.
@tf_export("Tensor", "experimental.numpy.ndarray", v1=["Tensor"])
class Tensor(internal.NativeObject, core_tf_types.Symbol):
"""A `tf.Tensor` represents a multidimensional array of elements.
All elements are of a single known data type.
When writing a TensorFlow program, the main object that is
manipulated and passed around is the `tf.Tensor`.
A `tf.Tensor` has the following properties:
* a single data type (float32, int32, or string, for example)
* a shape
TensorFlow supports eager execution and graph execution. In eager
execution, operations are evaluated immediately. In graph
execution, a computational graph is constructed for later
evaluation.
TensorFlow defaults to eager execution. In the example below, the
matrix multiplication results are calculated immediately.
>>> # Compute some values using a Tensor
>>> 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)
>>> print(e)
tf.Tensor(
[[1. 3.]
[3. 7.]], shape=(2, 2), dtype=float32)
Note that during eager execution, you may discover your `Tensors` are actually
of type `EagerTensor`. This is an internal detail, but it does give you
access to a useful function, `numpy`:
>>> type(e)
<class '...ops.EagerTensor'>
>>> print(e.numpy())
[[1. 3.]
[3. 7.]]
In TensorFlow, `tf.function`s are a common way to define graph execution.
A Tensor's shape (that is, the rank of the Tensor and the size of
each dimension) may not always be fully known. In `tf.function`
definitions, the shape may only be partially known.
Most operations produce tensors of fully-known shapes if the shapes of their
inputs are also fully known, but in some cases it's only possible to find the
shape of a tensor at execution time.
A number of specialized tensors are available: see `tf.Variable`,
`tf.constant`, `tf.placeholder`, `tf.sparse.SparseTensor`, and
`tf.RaggedTensor`.
Caution: when constructing a tensor from a numpy array or pandas dataframe
the underlying buffer may be re-used:
```python
a = np.array([1, 2, 3])
b = tf.constant(a)
a[0] = 4
print(b) # tf.Tensor([4 2 3], shape=(3,), dtype=int64)
```
Note: this is an implementation detail that is subject to change and users
should not rely on this behaviour.
For more on Tensors, see the [guide](https://tensorflow.org/guide/tensor).
"""
# List of Python operators that we allow to override.
OVERLOADABLE_OPERATORS = {
# Binary.
"__add__",
"__radd__",
"__sub__",
"__rsub__",
"__mul__",
"__rmul__",
"__div__",
"__rdiv__",
"__truediv__",
"__rtruediv__",
"__floordiv__",
"__rfloordiv__",
"__mod__",
"__rmod__",
"__lt__",
"__le__",
"__gt__",
"__ge__",
"__ne__",
"__eq__",
"__and__",
"__rand__",
"__or__",
"__ror__",
"__xor__",
"__rxor__",
"__getitem__",
"__pow__",
"__rpow__",
# Unary.
"__invert__",
"__neg__",
"__abs__",
"__matmul__",
"__rmatmul__"
}
# Whether to allow hashing or numpy-style equality
_USE_EQUALITY = tf2.enabled()
def __getattr__(self, name):
if name in {"T", "astype", "ravel", "transpose", "reshape", "clip", "size",
"tolist", "data"}:
# TODO(wangpeng): Export the enable_numpy_behavior knob
raise AttributeError(
f"{type(self).__name__} object has no attribute '{name}'. " + """
If you are looking for numpy-related methods, please run the following:
tf.experimental.numpy.experimental_enable_numpy_behavior()
""")
self.__getattribute__(name)
@property
def dtype(self):
"""The `DType` of elements in this tensor."""
return self._dtype
@property
def name(self):
return self._name
@property
def shape(self) -> tensor_shape.TensorShape:
"""Returns a `tf.TensorShape` that represents the shape of this tensor.
>>> t = tf.constant([1,2,3,4,5])
>>> t.shape
TensorShape([5])
`tf.Tensor.shape` is equivalent to `tf.Tensor.get_shape()`.
In a `tf.function` or when building a model using
`tf.keras.Input`, they return the build-time shape of the
tensor, which may be partially unknown.
A `tf.TensorShape` is not a tensor. Use `tf.shape(t)` to get a tensor
containing the shape, calculated at runtime.
See `tf.Tensor.get_shape()`, and `tf.TensorShape` for details and examples.
"""
if self._shape_val is None:
dims, unknown_shape = self._shape
if unknown_shape:
self._shape_val = tensor_shape.unknown_shape()
else:
self._shape_val = tensor_shape.TensorShape(dims)
return self._shape_val
@property
def ndim(self):
return self.shape.rank
def _disallow(self, task):
raise errors.OperatorNotAllowedInGraphError(
f"{task} is not allowed."
" You can attempt the following resolutions to the problem:"
" If you are running in Graph mode, use Eager execution mode"
" or decorate this function with @tf.function."
" If you are using AutoGraph, you can try decorating this function"
" with @tf.function. If that does not work, then you may be using"
" an unsupported feature or your source code may not be visible"
" to AutoGraph. See"
" https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/autograph/g3doc/reference/limitations.md#access-to-source-code"
" for more information.")
def _disallow_bool_casting(self):
self._disallow("Using a symbolic `tf.Tensor` as a Python `bool`")
def _disallow_iteration(self):
self._disallow("Iterating over a symbolic `tf.Tensor`")
def __iter__(self):
if not context.executing_eagerly():
self._disallow_iteration()
first_dim = self._get_first_dim()
return _TensorIterator(self, first_dim)
def _get_first_dim(self):
shape = self._shape_tuple()
if shape is None:
raise TypeError("Cannot iterate over a tensor with unknown shape.")
if not shape:
raise TypeError("Cannot iterate over a scalar tensor.")
if shape[0] is None:
raise TypeError(
"Cannot iterate over a tensor with unknown first dimension.")
return shape[0]
def _shape_as_list(self):
if self.shape.ndims is not None:
return [dim.value for dim in self.shape.dims]
else:
return None
def _shape_tuple(self):
shape = self._shape_as_list()
if shape is None:
return None
return tuple(shape)
def _record_tape(self, capture):
"""Connect this graph tensor with capture for gradients calculation."""
record.record_operation(
"captured_value",
[self], [capture],
backward_function=lambda x: [x],
forward_function=lambda x: [x])
def get_shape(self) -> tensor_shape.TensorShape:
"""Returns a `tf.TensorShape` that represents the shape of this tensor.
In eager execution the shape is always fully-known.
>>> a = tf.constant([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
>>> print(a.shape)
(2, 3)
`tf.Tensor.get_shape()` is equivalent to `tf.Tensor.shape`.
When executing in a `tf.function` or building a model using
`tf.keras.Input`, `Tensor.shape` may return a partial shape (including
`None` for unknown dimensions). See `tf.TensorShape` for more details.
>>> inputs = tf.keras.Input(shape = [10])
>>> # Unknown batch size
>>> print(inputs.shape)
(None, 10)
The shape is computed using shape inference functions that are
registered for each `tf.Operation`.
The returned `tf.TensorShape` is determined at *build* time, without
executing the underlying kernel. It is not a `tf.Tensor`. If you need a
shape *tensor*, either convert the `tf.TensorShape` to a `tf.constant`, or
use the `tf.shape(tensor)` function, which returns the tensor's shape at
*execution* time.
This is useful for debugging and providing early errors. For
example, when tracing a `tf.function`, no ops are being executed, shapes
may be unknown (See the [Concrete Functions
Guide](https://www.tensorflow.org/guide/concrete_function) for details).
>>> @tf.function
... def my_matmul(a, b):
... result = a@b
... # the `print` executes during tracing.
... print("Result shape: ", result.shape)
... return result
The shape inference functions propagate shapes to the extent possible:
>>> f = my_matmul.get_concrete_function(
... tf.TensorSpec([None,3]),
... tf.TensorSpec([3,5]))
Result shape: (None, 5)
Tracing may fail if a shape missmatch can be detected:
>>> cf = my_matmul.get_concrete_function(
... tf.TensorSpec([None,3]),
... tf.TensorSpec([4,5]))
Traceback (most recent call last):
...
ValueError: Dimensions must be equal, but are 3 and 4 for 'matmul' (op:
'MatMul') with input shapes: [?,3], [4,5].
In some cases, the inferred shape may have unknown dimensions. If
the caller has additional information about the values of these
dimensions, `tf.ensure_shape` or `Tensor.set_shape()` can be used to augment
the inferred shape.
>>> @tf.function
... def my_fun(a):
... a = tf.ensure_shape(a, [5, 5])
... # the `print` executes during tracing.
... print("Result shape: ", a.shape)
... return a
>>> cf = my_fun.get_concrete_function(
... tf.TensorSpec([None, None]))
Result shape: (5, 5)
Returns:
A `tf.TensorShape` representing the shape of this tensor.
"""
return self.shape
def set_shape(self, shape):
"""Updates the shape of this tensor.
Note: It is recommended to use `tf.ensure_shape` instead of
`Tensor.set_shape`, because `tf.ensure_shape` provides better checking for
programming errors and can create guarantees for compiler
optimization.
With eager execution this operates as a shape assertion.
Here the shapes match:
>>> t = tf.constant([[1,2,3]])
>>> t.set_shape([1, 3])
Passing a `None` in the new shape allows any value for that axis:
>>> t.set_shape([1,None])
An error is raised if an incompatible shape is passed.
>>> t.set_shape([1,5])
Traceback (most recent call last):
...
ValueError: Tensor's shape (1, 3) is not compatible with supplied
shape [1, 5]
When executing in a `tf.function`, or building a model using
`tf.keras.Input`, `Tensor.set_shape` will *merge* the given `shape` with
the current shape of this tensor, and set the tensor's shape to the
merged value (see `tf.TensorShape.merge_with` for details):
>>> t = tf.keras.Input(shape=[None, None, 3])
>>> print(t.shape)
(None, None, None, 3)
Dimensions set to `None` are not updated:
>>> t.set_shape([None, 224, 224, None])
>>> print(t.shape)
(None, 224, 224, 3)
The main use case for this is to provide additional shape information
that cannot be inferred from the graph alone.
For example if you know all the images in a dataset have shape [28,28,3] you
can set it with `tf.set_shape`:
>>> @tf.function
... def load_image(filename):
... raw = tf.io.read_file(filename)
... image = tf.image.decode_png(raw, channels=3)
... # the `print` executes during tracing.
... print("Initial shape: ", image.shape)
... image.set_shape([28, 28, 3])
... print("Final shape: ", image.shape)
... return image
Trace the function, see the [Concrete Functions
Guide](https://www.tensorflow.org/guide/concrete_function) for details.
>>> cf = load_image.get_concrete_function(
... tf.TensorSpec([], dtype=tf.string))
Initial shape: (None, None, 3)
Final shape: (28, 28, 3)
Similarly the `tf.io.parse_tensor` function could return a tensor with
any shape, even the `tf.rank` is unknown. If you know that all your
serialized tensors will be 2d, set it with `set_shape`:
>>> @tf.function
... def my_parse(string_tensor):
... result = tf.io.parse_tensor(string_tensor, out_type=tf.float32)
... # the `print` executes during tracing.
... print("Initial shape: ", result.shape)
... result.set_shape([None, None])
... print("Final shape: ", result.shape)
... return result
Trace the function
>>> concrete_parse = my_parse.get_concrete_function(
... tf.TensorSpec([], dtype=tf.string))
Initial shape: <unknown>
Final shape: (None, None)
Make sure it works:
>>> t = tf.ones([5,3], dtype=tf.float32)
>>> serialized = tf.io.serialize_tensor(t)
>>> print(serialized.dtype)
<dtype: 'string'>
>>> print(serialized.shape)
()
>>> t2 = concrete_parse(serialized)
>>> print(t2.shape)
(5, 3)
Caution: `set_shape` ensures that the applied shape is compatible with
the existing shape, but it does not check at runtime. Setting
incorrect shapes can result in inconsistencies between the
statically-known graph and the runtime value of tensors. For runtime
validation of the shape, use `tf.ensure_shape` instead. It also modifies
the `shape` of the tensor.
>>> # Serialize a rank-3 tensor
>>> t = tf.ones([5,5,5], dtype=tf.float32)
>>> serialized = tf.io.serialize_tensor(t)
>>> # The function still runs, even though it `set_shape([None,None])`
>>> t2 = concrete_parse(serialized)
>>> print(t2.shape)
(5, 5, 5)
Args:
shape: A `TensorShape` representing the shape of this tensor, a
`TensorShapeProto`, a list, a tuple, or None.
Raises:
ValueError: If `shape` is not compatible with the current shape of
this tensor.
"""
# Reset cached shape.
self._shape_val = None
# We want set_shape to be reflected in the C API graph for when we run it.
if not isinstance(shape, tensor_shape.TensorShape):
shape = tensor_shape.TensorShape(shape)
dim_list = []
if shape.dims is None:
unknown_shape = True
else:
unknown_shape = False
for dim in shape.dims:
if dim.value is None:
dim_list.append(-1)
else:
dim_list.append(dim.value)
self._set_shape(dim_list, unknown_shape)
def _as_node_def_input(self):
"""Return a value to use for the NodeDef "input" attribute.
The returned string can be used in a NodeDef "input" attribute
to indicate that the NodeDef uses this Tensor as input.
Raises:
ValueError: if this Tensor's Operation does not have a name.
Returns:
a string.
"""
assert self._op.name
if self.value_index == 0:
return self._op.name
else:
return "%s:%d" % (self._op.name, self.value_index)
def __str__(self):
return "Tensor(\"%s\"%s%s%s)" % (
self.name,
(", shape=%s" %
self.get_shape()) if self.get_shape().ndims is not None else "",
(", dtype=%s" % self._dtype.name) if self._dtype else "",
(", device=%s" % self.device) if self.device else "")
def __repr__(self):
return "<tf.Tensor '%s' shape=%s dtype=%s>" % (self.name, self.get_shape(),
self._dtype.name)
def __hash__(self):
g = getattr(self, "graph", None)
if (Tensor._USE_EQUALITY and (g is None or g.building_function)):
raise TypeError("Tensor is unhashable. "
"Instead, use tensor.ref() as the key.")
else:
return id(self)
# NOTE(mrry): This enables the Tensor's overloaded "right" binary
# operators to run when the left operand is an ndarray, because it
# accords the Tensor class higher priority than an ndarray, or a
# numpy matrix.
# TODO(mrry): Convert this to using numpy's __numpy_ufunc__
# mechanism, which allows more control over how Tensors interact
# with ndarrays.
__array_priority__ = 100
def __array__(self, dtype=None):
del dtype
raise NotImplementedError(
f"Cannot convert a symbolic tf.Tensor ({self.name}) to a numpy array."
f" This error may indicate that you're trying to pass a Tensor to"
f" a NumPy call, which is not supported.")
def __len__(self):
raise TypeError(f"len is not well defined for a symbolic Tensor "
f"({self.name}). Please call `x.shape` rather than "
f"`len(x)` for shape information.")
# TODO(mdan): This convoluted machinery is hard to maintain. Clean up.
@staticmethod
def _override_operator(operator, func):
_override_helper(Tensor, operator, func)
def __bool__(self): # pylint: disable=invalid-bool-returned
"""Dummy method to prevent a tensor from being used as a Python `bool`.
This overload raises a `TypeError` when the user inadvertently
treats a `Tensor` as a boolean (most commonly in an `if` or `while`
statement), in code that was not converted by AutoGraph. For example:
```python
if tf.constant(True): # Will raise.
# ...
if tf.constant(5) < tf.constant(7): # Will raise.
# ...
```
Raises:
`TypeError`.
"""
self._disallow_bool_casting()
def __nonzero__(self):
"""Dummy method to prevent a tensor from being used as a Python `bool`.
This is the Python 2.x counterpart to `__bool__()` above.
Raises:
`TypeError`.
"""
self._disallow_bool_casting()
def eval(self, feed_dict=None, session=None):
"""Evaluates this tensor in a `Session`.
Note: If you are not using `compat.v1` libraries, you should not need this,
(or `feed_dict` or `Session`). In eager execution (or within `tf.function`)
you do not need to call `eval`.
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.
Args:
feed_dict: A dictionary that maps `Tensor` objects to feed values. See
`tf.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.
Returns:
A numpy array corresponding to the value of this tensor.
"""
return _eval_using_default_session(self, feed_dict, self.graph, session)
@deprecation.deprecated(None, "Use ref() instead.")
def experimental_ref(self):
return self.ref()
def ref(self):
# tf.Variable also has the same ref() API. If you update the
# documentation here, please update tf.Variable.ref() as well.
"""Returns a hashable reference object to this Tensor.
The primary use case for this API is to put tensors in a set/dictionary.
We can't put tensors in a set/dictionary as `tensor.__hash__()` is no longer
available starting Tensorflow 2.0.
The following will raise an exception starting 2.0
>>> x = tf.constant(5)
>>> y = tf.constant(10)
>>> z = tf.constant(10)
>>> tensor_set = {x, y, z}
Traceback (most recent call last):
...
TypeError: Tensor is unhashable. Instead, use tensor.ref() as the key.
>>> tensor_dict = {x: 'five', y: 'ten'}
Traceback (most recent call last):
...
TypeError: Tensor is unhashable. Instead, use tensor.ref() as the key.
Instead, we can use `tensor.ref()`.
>>> tensor_set = {x.ref(), y.ref(), z.ref()}
>>> x.ref() in tensor_set
True
>>> tensor_dict = {x.ref(): 'five', y.ref(): 'ten', z.ref(): 'ten'}
>>> tensor_dict[y.ref()]
'ten'
Also, the reference object provides `.deref()` function that returns the
original Tensor.
>>> x = tf.constant(5)
>>> x.ref().deref()
<tf.Tensor: shape=(), dtype=int32, numpy=5>
"""
return object_identity.Reference(self)
def __tf_tracing_type__(self, signature_context):
if self.dtype == dtypes.resource or self.dtype == dtypes.variant:
shape_inference_handle_data = handle_data_util.get_handle_data(self)
handle_data = (
dtypes.HandleData(shape_inference_handle_data)
if shape_inference_handle_data
else None
)
dtype = dtypes.DType(self.dtype._type_enum, handle_data)
else:
dtype = self.dtype
spec = TensorSpec(self.shape, dtype)
return spec
def __tf_tensor__(
self, dtype: Optional[dtypes.DType] = None, name: Optional[str] = None
) -> "Tensor":
if dtype is not None and not dtype.is_compatible_with(self.dtype):
raise ValueError(
_add_error_prefix(
f"Tensor conversion requested dtype {dtype.name} "
f"for Tensor with dtype {self.dtype.name}: {self!r}",
name=name))
return self
@tf_export(v1=["enable_tensor_equality"])
def enable_tensor_equality():
"""Compare Tensors with element-wise comparison and thus be unhashable.
Comparing tensors with element-wise allows comparisons such as
tf.Variable(1.0) == 1.0. Element-wise equality implies that tensors are
unhashable. Thus tensors can no longer be directly used in sets or as a key in
a dictionary.
"""
logging.vlog(1, "Enabling tensor equality")
_tensor_equality_api_usage_gauge.get_cell().set(True)
Tensor._USE_EQUALITY = True # pylint: disable=protected-access
@tf_export(v1=["disable_tensor_equality"])
def disable_tensor_equality():
"""Compare Tensors by their id and be hashable.
This is a legacy behaviour of TensorFlow and is highly discouraged.
"""
logging.vlog(1, "Disabling tensor equality")
_tensor_equality_api_usage_gauge.get_cell().set(False)
Tensor._USE_EQUALITY = False # pylint: disable=protected-access
# TODO(b/249802365): Sanitize all TensorSpec names.
def sanitize_spec_name(name: str) -> str:
"""Sanitizes Spec names. Matches Graph Node and Python naming conventions.
Without sanitization, names that are not legal Python parameter names can be
set which makes it challenging to represent callables supporting the named
calling capability.
Args:
name: The name to sanitize.
Returns:
A string that meets Python parameter conventions.
"""
if not name:
return "unknown"
# Lower case and replace non-alphanumeric chars with '_'
swapped = "".join([c if c.isalnum() else "_" for c in name.lower()])
if swapped[0].isalpha():
return swapped
else:
return "tensor_" + swapped
def get_op_name(tensor_name):
"""Extract the Op name from a Tensor name.
The Op name is everything before a colon, if present,
not including any ^ prefix denoting a control dependency.
Args:
tensor_name: the full name of a Tensor in the graph.
Returns:
The name of the Op of which the given Tensor is an output.
Raises:
ValueError: if tensor_name is None or empty.
"""
if not tensor_name:
raise ValueError(
f"Tensor name cannot be empty or None. Received: {tensor_name}.")
# Control dependency inputs start with ^.
if tensor_name.startswith("^"):
tensor_name = tensor_name[1:]
if ":" in tensor_name:
op_name, _ = tensor_name.split(":")
return op_name
return tensor_name
class DenseSpec(type_spec.TypeSpec):
"""Describes a dense object with shape, dtype, and name."""
__slots__ = ["_shape", "_dtype", "_name"]
_component_specs = property(lambda self: self)
def __init__(self, shape, dtype=dtypes.float32, name=None):
"""Creates a TensorSpec.
Args:
shape: Value convertible to `tf.TensorShape`. The shape of the tensor.
dtype: Value convertible to `tf.DType`. The type of the tensor values.
name: Optional name for the Tensor.
Raises:
TypeError: If shape is not convertible to a `tf.TensorShape`, or dtype is
not convertible to a `tf.DType`.
"""
self._shape = tensor_shape.TensorShape(shape)
self._dtype = dtypes.as_dtype(dtype)
self._name = name
@property
def shape(self):
"""Returns the `TensorShape` that represents the shape of the tensor."""
return self._shape
@property
def dtype(self):
"""Returns the `dtype` of elements in the tensor."""
return self._dtype
@property
def name(self):
"""Returns the (optionally provided) name of the described tensor."""
return self._name
def is_compatible_with(self, spec_or_value):
return (isinstance(spec_or_value, (DenseSpec, self.value_type)) and
self._dtype.is_compatible_with(spec_or_value.dtype) and
self._shape.is_compatible_with(spec_or_value.shape))
def __repr__(self):
return "{}(shape={}, dtype={}, name={})".format(
type(self).__name__, self.shape, repr(self.dtype), repr(self.name))
def __hash__(self):
return hash((self._shape, self.dtype))
def __eq__(self, other):
# pylint: disable=protected-access
return (type(self) is type(other) and self._shape == other._shape and
self._dtype == other._dtype and self._name == other._name)
def __ne__(self, other):
return not self == other
def _serialize(self):
return (self._shape, self._dtype, self._name)
def _to_legacy_output_types(self):
return self._dtype
def _to_legacy_output_shapes(self):
return self._shape
def _to_legacy_output_classes(self):
return self.value_type
@tf_export("TensorSpec")
@type_spec_registry.register("tf.TensorSpec")
class TensorSpec(DenseSpec, type_spec.BatchableTypeSpec,
trace_type.Serializable, internal.TensorSpec):
"""Describes the type of a tf.Tensor.
>>> t = tf.constant([[1,2,3],[4,5,6]])
>>> tf.TensorSpec.from_tensor(t)
TensorSpec(shape=(2, 3), dtype=tf.int32, name=None)
Contains metadata for describing the nature of `tf.Tensor` objects
accepted or returned by some TensorFlow APIs.
For example, it can be used to constrain the type of inputs accepted by
a tf.function:
>>> @tf.function(input_signature=[tf.TensorSpec([1, None])])
... def constrained_foo(t):
... print("tracing...")
... return t
Now the `tf.function` is able to assume that `t` is always of the type
`tf.TensorSpec([1, None])` which will avoid retracing as well as enforce the
type restriction on inputs.
As a result, the following call with tensor of type `tf.TensorSpec([1, 2])`
triggers a trace and succeeds:
>>> constrained_foo(tf.constant([[1., 2]])).numpy()
tracing...
array([[1., 2.]], dtype=float32)
The following subsequent call with tensor of type `tf.TensorSpec([1, 4])`
does not trigger a trace and succeeds:
>>> constrained_foo(tf.constant([[1., 2, 3, 4]])).numpy()
array([[1., 2., 3., 4.], dtype=float32)
But the following call with tensor of type `tf.TensorSpec([2, 2])` fails:
>>> constrained_foo(tf.constant([[1., 2], [3, 4]])).numpy()
Traceback (most recent call last):
...
TypeError: Binding inputs to tf.function `constrained_foo` failed ...
"""
__slots__ = []
@classmethod
def experimental_type_proto(cls) -> Type[struct_pb2.TensorSpecProto]:
"""Returns the type of proto associated with TensorSpec serialization."""
return struct_pb2.TensorSpecProto
@classmethod
def experimental_from_proto(
cls, proto: struct_pb2.TensorSpecProto) -> "TensorSpec":
"""Returns a TensorSpec instance based on the serialized proto."""
return TensorSpec(
shape=tensor_shape.TensorShape.experimental_from_proto(proto.shape),
dtype=proto.dtype,
name=proto.name if proto.name else None)
def experimental_as_proto(self) -> struct_pb2.TensorSpecProto:
"""Returns a proto representation of the TensorSpec instance."""
return struct_pb2.TensorSpecProto(
shape=self.shape.experimental_as_proto(),
dtype=self.dtype.experimental_as_proto().datatype,
name=self.name)
def is_compatible_with(self, spec_or_tensor): # pylint:disable=useless-super-delegation,arguments-renamed
"""Returns True if spec_or_tensor is compatible with this TensorSpec.
Two tensors are considered compatible if they have the same dtype
and their shapes are compatible (see `tf.TensorShape.is_compatible_with`).
Args:
spec_or_tensor: A tf.TensorSpec or a tf.Tensor
Returns:
True if spec_or_tensor is compatible with self.
"""
return super(TensorSpec, self).is_compatible_with(spec_or_tensor)
def is_subtype_of(self, other):
if not isinstance(other, TensorSpec):
return False
return (
(not self.name or self.name == other.name)
and self.shape.is_subtype_of(other.shape)
and self.dtype.is_subtype_of(other.dtype)
)
def placeholder_value(self, placeholder_context):
"""Generates a graph placholder with the given TensorSpec information."""
if placeholder_context.unnest_only:
return self
name = self.name or placeholder_context.naming_scope
context_graph = placeholder_context.context_graph
if placeholder_context.with_none_control_dependencies:
# Note: setting ops.control_dependencies(None) ensures we always put
# capturing placeholders outside of any control flow context.
with context_graph.control_dependencies(None):
placeholder = self._graph_placeholder(context_graph, name=name)
else:
placeholder = self._graph_placeholder(context_graph, name=name)
if name is not None:
# Record the requested/user-specified name in case it's different than
# the uniquified name, for validation when exporting signatures.
placeholder.op._set_attr( # pylint: disable=protected-access
"_user_specified_name",
attr_value_pb2.AttrValue(s=compat.as_bytes(name)))
handle_data = self.dtype._handle_data # pylint: disable=protected-access
if (
handle_data is not None
and handle_data.shape_inference.is_set
and handle_data.shape_inference.shape_and_type
):
handle_data_util.set_handle_data(placeholder, handle_data.shape_inference)
# Record the composite device as an attribute to the placeholder.
# This attribute would be propagated into the arg_attr of the FunctionDef.
# Currently, a packed eager tensor is always placed on a CompositeDevice.
if placeholder_context.composite_device_name is not None:
placeholder.op._set_attr( # pylint: disable=protected-access
"_composite_device",
attr_value_pb2.AttrValue(s=compat.as_bytes(
placeholder_context.composite_device_name)))
return placeholder
def _graph_placeholder(self, graph, name=None):
"""Graph-only version of tf.compat.v1.placeholder(), for internal use only."""
dtype = self.dtype.base_dtype
shape = self.shape
dtype_value = attr_value_pb2.AttrValue(type=dtype.as_datatype_enum)
if isinstance(shape, (list, tuple)):
shape = tensor_shape.TensorShape(shape)
shape = attr_value_pb2.AttrValue(shape=shape.as_proto())
attrs = {"dtype": dtype_value, "shape": shape}
try:
op = graph._create_op_internal( # pylint: disable=protected-access
"Placeholder", [], [dtype], input_types=[],
attrs=attrs, name=name)
except ValueError as e:
# TODO(b/262413656) Sometimes parameter names are not valid op names, in
# which case an unnamed placeholder is created instead. Update this logic
# to sanitize the name instead of falling back on unnamed placeholders.
logging.warning(e)
op = graph._create_op_internal( # pylint: disable=protected-access
"Placeholder", [], [dtype], input_types=[], attrs=attrs)
(result,) = op.outputs
if op_callbacks.should_invoke_op_callbacks():
# TODO(b/147670703): Once the special-op creation code paths
# are unified. Remove this `if` block.
callback_outputs = op_callbacks.invoke_op_callbacks(
"Placeholder", tuple(), attrs, tuple(op.outputs),
op_name=name, graph=graph)
if callback_outputs is not None:
(result,) = callback_outputs
return result
def to_tensors(self, value):
value = self.cast(value, trace_type.InternalCastContext())
if not value.shape.is_subtype_of(self.shape):
raise TypeError(
f"Received tensor of shape {value.shape} instead of {self.shape}"
)
return [value]
def from_tensors(self, tensors):
tensor = next(tensors)
handle_data = self.dtype._handle_data # pylint: disable=protected-access
if handle_data:
handle_data_util.set_handle_data(tensor, handle_data.shape_inference)
return tensor
def flatten(self):
return [self]
def cast(self, value, casting_context):
"""Cast value to a tensor that is a subtype of this TensorSpec."""
# This method is mainly used to cast Python primitives to tensor.
# Currently, cast tensor to tensor with different types are not supported.
# For example, casting int32 to float32 would raise a ValueError.
if casting_context.allow_specs and isinstance(value, TensorSpec):
assert value.is_subtype_of(self), f"Can not cast {value!r} to {self!r}"
return self
if not isinstance(value, Tensor):
value = tensor_conversion_registry.convert(value, self.dtype)
value_spec = TensorSpec(value.shape, value.dtype, self.name)
if not value_spec.is_subtype_of(self):
if self.is_subtype_of(value_spec):
value.set_shape(self.shape)
else:
raise TypeError(f"Can not cast {value_spec!r} to {self!r}")
return value
def _alias_id(self):
"""Returns an id specifying identical tensors to avoid duplication."""
alias_id = None
if self.dtype._handle_data: # pylint: disable=protected-access
alias_id = self.dtype._handle_data.alias_id # pylint: disable=protected-access
return alias_id
@classmethod
def from_spec(cls, spec, name=None):
"""Returns a `TensorSpec` with the same shape and dtype as `spec`.
>>> spec = tf.TensorSpec(shape=[8, 3], dtype=tf.int32, name="OriginalName")
>>> tf.TensorSpec.from_spec(spec, "NewName")
TensorSpec(shape=(8, 3), dtype=tf.int32, name='NewName')
Args:
spec: The `TypeSpec` used to create the new `TensorSpec`.
name: The name for the new `TensorSpec`. Defaults to `spec.name`.
"""
return cls(spec.shape, spec.dtype, name or spec.name)
@classmethod
def from_tensor(cls, tensor, name=None):
"""Returns a `TensorSpec` that describes `tensor`.
>>> tf.TensorSpec.from_tensor(tf.constant([1, 2, 3]))
TensorSpec(shape=(3,), dtype=tf.int32, name=None)
Args:
tensor: The `tf.Tensor` that should be described.
name: A name for the `TensorSpec`. Defaults to `tensor.op.name`.
Returns:
A `TensorSpec` that describes `tensor`.
"""
if isinstance(tensor, core_tf_types.Value):
return TensorSpec(tensor.shape, tensor.dtype, name)
elif isinstance(tensor, core_tf_types.Symbol):
# TODO(b/249802365): Return a sanitized version of op name or no name.
return TensorSpec(tensor.shape, tensor.dtype, name or tensor.op.name)
else:
raise ValueError(
f"`tensor` should be a tf.Tensor, but got type {type(tensor)}.")
@property
def value_type(self):
"""The Python type for values that are compatible with this TypeSpec."""
return Tensor
def _to_components(self, value):
assert isinstance(value, core_tf_types.Tensor)
return value
def _from_components(self, components):
return components
def _from_compatible_tensor_list(self, tensor_list):
# TODO(b/112266545): It would be cleaner to create a new `ensure_shape()`
# op here and return that, instead of mutating the input's shape using
# `Tensor.set_shape()`. However, that would add extra ops, which could
# impact performance. When this bug is resolved, we should be able to add
# the `ensure_shape()` ops and optimize them away using contextual shape
# information.
assert len(tensor_list) == 1
tensor_list[0].set_shape(self._shape)
return tensor_list[0]
def _to_batchable_tensor_list(self, value, batched=False):
if batched and self._shape.merge_with(value.shape).ndims == 0:
raise ValueError("Unbatching a tensor is only supported for rank >= 1")
return self._to_components(value)
def _batch(self, batch_size):
return TensorSpec(
tensor_shape.TensorShape([batch_size]).concatenate(self._shape),
self._dtype)
def _unbatch(self):
if self._shape.ndims == 0:
raise ValueError("Unbatching a tensor is only supported for rank >= 1")
return TensorSpec(self._shape[1:], self._dtype)
@property
def _flat_tensor_specs(self):
return [self]
def _to_tensor_list(self, value):
return [self._to_components(value)]
def _to_batched_tensor_list(self, value):
return self._to_tensor_list(value)
# TODO(b/206014848): Helper function to support logic that does not consider
# Tensor name. Will be removed once load-bearing usages of Tensor name are
# fixed.
def _without_tensor_names(self) -> "TensorSpec":
"""Returns a version of `TensorSpec` with the name removed."""
if self.name is None:
return self
else:
return TensorSpec(self.shape, self.dtype)
trace_type.register_serializable(TensorSpec)
trace_type.register_tensor_type(TensorSpec)
class _TensorSpecCodec:
"""Codec for `TensorSpec`."""
def can_encode(self, pyobj):
# BoundedTensorSpec has its own decoder.
return (isinstance(pyobj, TensorSpec) and
not isinstance(pyobj, BoundedTensorSpec))
def do_encode(self, tensor_spec_value, encode_fn):
encoded_tensor_spec = struct_pb2.StructuredValue()
encoded_tensor_spec.tensor_spec_value.CopyFrom(
struct_pb2.TensorSpecProto(
shape=encode_fn(tensor_spec_value.shape).tensor_shape_value,
dtype=encode_fn(tensor_spec_value.dtype).tensor_dtype_value,
name=tensor_spec_value.name))
return encoded_tensor_spec
def can_decode(self, value):
return value.HasField("tensor_spec_value")
def do_decode(self, value, decode_fn):
name = value.tensor_spec_value.name
return TensorSpec(
shape=decode_fn(
struct_pb2.StructuredValue(
tensor_shape_value=value.tensor_spec_value.shape)),
dtype=decode_fn(
struct_pb2.StructuredValue(
tensor_dtype_value=value.tensor_spec_value.dtype)),
name=(name if name else None))
nested_structure_coder.register_codec(_TensorSpecCodec())
# TODO(b/133606651): Should is_compatible_with should check min/max bounds?
@type_spec_registry.register("tf.BoundedTensorSpec")
class BoundedTensorSpec(TensorSpec, trace_type.Serializable):
"""A `TensorSpec` that specifies minimum and maximum values.
Example usage:
```python
spec = tensor_spec.BoundedTensorSpec((1, 2, 3), tf.float32, 0, (5, 5, 5))
tf_minimum = tf.convert_to_tensor(spec.minimum, dtype=spec.dtype)
tf_maximum = tf.convert_to_tensor(spec.maximum, dtype=spec.dtype)
```
Bounds are meant to be inclusive. This is especially important for
integer types. The following spec will be satisfied by tensors
with values in the set {0, 1, 2}:
```python
spec = tensor_spec.BoundedTensorSpec((3, 5), tf.int32, 0, 2)
```
"""
__slots__ = ("_minimum", "_maximum")
def __init__(self, shape, dtype, minimum, maximum, name=None):
"""Initializes a new `BoundedTensorSpec`.
Args:
shape: Value convertible to `tf.TensorShape`. The shape of the tensor.
dtype: Value convertible to `tf.DType`. The type of the tensor values.
minimum: Number or sequence specifying the minimum element bounds
(inclusive). Must be broadcastable to `shape`.
maximum: Number or sequence specifying the maximum element bounds
(inclusive). Must be broadcastable to `shape`.
name: Optional string containing a semantic name for the corresponding
array. Defaults to `None`.
Raises:
ValueError: If `minimum` or `maximum` are not provided or not
broadcastable to `shape`.
TypeError: If the shape is not an iterable or if the `dtype` is an invalid
numpy dtype.
"""
super(BoundedTensorSpec, self).__init__(shape, dtype, name)
if minimum is None:
raise ValueError("`minimum` can not be None.")
if maximum is None:
raise ValueError("`maximum` can not be None.")
try:
minimum_shape = np.shape(minimum)
common_shapes.broadcast_shape(
tensor_shape.TensorShape(minimum_shape), self.shape)
except ValueError as exception:
raise ValueError(
f"`minimum` {minimum} is not compatible with shape {self.shape}."
) from exception
try:
maximum_shape = np.shape(maximum)
common_shapes.broadcast_shape(
tensor_shape.TensorShape(maximum_shape), self.shape)
except ValueError as exception:
raise ValueError(
f"`maximum` {maximum} is not compatible with shape {self.shape}."
) from exception
self._minimum = np.array(minimum, dtype=self.dtype.as_numpy_dtype)
self._minimum.setflags(write=False)
self._maximum = np.array(maximum, dtype=self.dtype.as_numpy_dtype)
self._maximum.setflags(write=False)
@classmethod
def experimental_type_proto(cls) -> Type[struct_pb2.BoundedTensorSpecProto]:
"""Returns the type of proto associated with BoundedTensorSpec serialization."""
return struct_pb2.BoundedTensorSpecProto
@classmethod
def experimental_from_proto(
cls, proto: struct_pb2.BoundedTensorSpecProto) -> "BoundedTensorSpec":
"""Returns a BoundedTensorSpec instance based on the serialized proto."""
return BoundedTensorSpec(
shape=tensor_shape.TensorShape.experimental_from_proto(proto.shape),
dtype=proto.dtype,
minimum=tensor_util.MakeNdarray(proto.minimum),
maximum=tensor_util.MakeNdarray(proto.maximum),
name=proto.name if proto.name else None)
def experimental_as_proto(self) -> struct_pb2.BoundedTensorSpecProto:
"""Returns a proto representation of the BoundedTensorSpec instance."""
return struct_pb2.BoundedTensorSpecProto(
shape=self.shape.experimental_as_proto(),
dtype=self.dtype.experimental_as_proto().datatype,
minimum=tensor_util.make_tensor_proto(self._minimum),
maximum=tensor_util.make_tensor_proto(self._maximum),
name=self.name)
@classmethod
def from_spec(cls, spec):
"""Returns a `TensorSpec` with the same shape and dtype as `spec`.
If `spec` is a `BoundedTensorSpec`, then the new spec's bounds are set to
`spec.minimum` and `spec.maximum`; otherwise, the bounds are set to
`spec.dtype.min` and `spec.dtype.max`.
>>> spec = tf.TensorSpec(shape=[8, 3], dtype=tf.int32, name="x")
>>> BoundedTensorSpec.from_spec(spec)
BoundedTensorSpec(shape=(8, 3), dtype=tf.int32, name='x',
minimum=array(-2147483648, dtype=int32),
maximum=array(2147483647, dtype=int32))
Args:
spec: The `TypeSpec` used to create the new `BoundedTensorSpec`.
"""
dtype = dtypes.as_dtype(spec.dtype)
minimum = getattr(spec, "minimum", dtype.min)
maximum = getattr(spec, "maximum", dtype.max)
return BoundedTensorSpec(spec.shape, dtype, minimum, maximum, spec.name)
@property
def minimum(self):
"""Returns a NumPy array specifying the minimum bounds (inclusive)."""
return self._minimum
@property
def maximum(self):
"""Returns a NumPy array specifying the maximum bounds (inclusive)."""
return self._maximum
def cast(self, value, casting_context):
if casting_context.allow_specs and isinstance(value, BoundedTensorSpec):
assert value.is_subtype_of(self), f"Can not cast {value!r} to {self!r}"
return self
actual_spec = TensorSpec(shape=self.shape, dtype=self.dtype, name=self.name)
return actual_spec.cast(value, casting_context) # pylint: disable=protected-access
def __repr__(self):
s = "BoundedTensorSpec(shape={}, dtype={}, name={}, minimum={}, maximum={})"
return s.format(self.shape, repr(self.dtype), repr(self.name),
repr(self.minimum), repr(self.maximum))
def __eq__(self, other):
tensor_spec_eq = super(BoundedTensorSpec, self).__eq__(other)
return (tensor_spec_eq and np.allclose(self.minimum, other.minimum) and
np.allclose(self.maximum, other.maximum))
def __hash__(self):
return hash((self._shape, self.dtype))
def __reduce__(self):
return BoundedTensorSpec, (self._shape, self._dtype, self._minimum,
self._maximum, self._name)
def _serialize(self):
return (self._shape, self._dtype, self._minimum, self._maximum, self._name)
class _BoundedTensorSpecCodec:
"""Codec for `BoundedTensorSpec`."""
def can_encode(self, pyobj):
return isinstance(pyobj, BoundedTensorSpec)
def do_encode(self, bounded_tensor_spec_value, encode_fn):
"""Returns an encoded proto for the given `tf.BoundedTensorSpec`."""
encoded_bounded_tensor_spec = struct_pb2.StructuredValue()
encoded_bounded_tensor_spec.bounded_tensor_spec_value.CopyFrom(
struct_pb2.BoundedTensorSpecProto(
shape=encode_fn(bounded_tensor_spec_value.shape).tensor_shape_value,
dtype=encode_fn(bounded_tensor_spec_value.dtype).tensor_dtype_value,
name=bounded_tensor_spec_value.name,
minimum=tensor_util.make_tensor_proto(
bounded_tensor_spec_value.minimum),
maximum=tensor_util.make_tensor_proto(
bounded_tensor_spec_value.maximum)))
return encoded_bounded_tensor_spec
def can_decode(self, value):
return value.HasField("bounded_tensor_spec_value")
def do_decode(self, value, decode_fn):
btsv = value.bounded_tensor_spec_value
name = btsv.name
return BoundedTensorSpec(
shape=decode_fn(
struct_pb2.StructuredValue(tensor_shape_value=btsv.shape)),
dtype=decode_fn(
struct_pb2.StructuredValue(tensor_dtype_value=btsv.dtype)),
minimum=tensor_util.MakeNdarray(btsv.minimum),
maximum=tensor_util.MakeNdarray(btsv.maximum),
name=(name if name else None))
nested_structure_coder.register_codec(_BoundedTensorSpecCodec())
trace_type.register_serializable(BoundedTensorSpec)
# Note: we do not include Tensor names when constructing TypeSpecs.
type_spec.register_type_spec_from_value_converter(
Tensor, lambda tensor: TensorSpec(tensor.shape, tensor.dtype))
type_spec.register_type_spec_from_value_converter(
np.ndarray, lambda array: TensorSpec(array.shape, array.dtype))