op

package
v0.8.2 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Jul 23, 2024 License: Apache-2.0 Imports: 3 Imported by: 9

Documentation

Overview

Package op defines functions for adding TensorFlow operations to a Graph.

Functions for adding an operation to a graph take a Scope object as the first argument. The Scope object encapsulates a graph and a set of properties (such as a name prefix) for all operations being added to the graph.

WARNING: The API in this package has not been finalized and can change without notice.

Example 
// This example creates a Graph that multiplies a constant matrix with
// a matrix to be provided during graph execution (via
// tensorflow.Session).
s := NewScope()
input := Placeholder(s, tf.Float) // Matrix to be provided to Session.Run
output := MatMul(s,
	Const(s, [][]float32{{10}, {20}}), // Constant 2x1 matrix
	input,
	MatMulTransposeB(true))
if s.Err() != nil {
	panic(s.Err())
}
// Shape of the product: The number of rows is fixed by m1, but the
// number of columns will depend on m2, which is unknown.
fmt.Println(output.Shape())

Output:

[2, ?]

Index

Examples

Constants

This section is empty.

Variables

This section is empty.

Functions

func Abort 

func Abort(scope *Scope, optional ...AbortAttr) (o *tf.Operation)

Raise a exception to abort the process when called.

If exit_without_error is true, the process will exit normally, otherwise it will exit with a SIGABORT signal.

Returns nothing but an exception.

Returns the created operation.

func Abs 

func Abs(scope *Scope, x tf.Output) (y tf.Output)

Computes the absolute value of a tensor.

Given a tensor `x`, this operation returns a tensor containing the absolute value of each element in `x`. For example, if x is an input element and y is an output element, this operation computes \\(y = |x|\\).

func AccumulateNV2 

func AccumulateNV2(scope *Scope, inputs []tf.Output, shape tf.Shape) (sum tf.Output)

Returns the element-wise sum of a list of tensors.

`tf.accumulate_n_v2` performs the same operation as `tf.add_n`, but does not wait for all of its inputs to be ready before beginning to sum. This can save memory if inputs are ready at different times, since minimum temporary storage is proportional to the output size rather than the inputs size.

Unlike the original `accumulate_n`, `accumulate_n_v2` is differentiable.

Returns a `Tensor` of same shape and type as the elements of `inputs`.

Arguments: 

inputs: A list of `Tensor` objects, each with same shape and type.
shape: Shape of elements of `inputs`.

func Acos 

func Acos(scope *Scope, x tf.Output) (y tf.Output)

Computes acos of x element-wise. 

Provided an input tensor, the `tf.math.acos` operation returns the inverse cosine of each element of the tensor. If `y = tf.math.cos(x)` then, `x = tf.math.acos(y)`.

Input range is `[-1, 1]` and the output has a range of `[0, pi]`.

func Acosh 

func Acosh(scope *Scope, x tf.Output) (y tf.Output)

Computes inverse hyperbolic cosine of x element-wise.

Given an input tensor, the function computes inverse hyperbolic cosine of every element. Input range is `[1, inf]`. It returns `nan` if the input lies outside the range.

```python x = tf.constant([-2, -0.5, 1, 1.2, 200, 10000, float("inf")]) tf.math.acosh(x) ==> [nan nan 0. 0.62236255 5.9914584 9.903487 inf] ```

func Add 

func Add(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x + y element-wise.

*NOTE*: `Add` supports broadcasting. `AddN` does not. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

Given two input tensors, the `tf.add` operation computes the sum for every element in the tensor.

Both input and output have a range `(-inf, inf)`.

func AddManySparseToTensorsMap 

func AddManySparseToTensorsMap(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...AddManySparseToTensorsMapAttr) (sparse_handles tf.Output)

Add an `N`-minibatch `SparseTensor` to a `SparseTensorsMap`, return `N` handles.

A `SparseTensor` of rank `R` is represented by three tensors: `sparse_indices`, `sparse_values`, and `sparse_shape`, where

```sparse_indices.shape[1] == sparse_shape.shape[0] == R```

An `N`-minibatch of `SparseTensor` objects is represented as a `SparseTensor` having a first `sparse_indices` column taking values between `[0, N)`, where the minibatch size `N == sparse_shape[0]`.

The input `SparseTensor` must have rank `R` greater than 1, and the first dimension is treated as the minibatch dimension. Elements of the `SparseTensor` must be sorted in increasing order of this first dimension. The stored `SparseTensor` objects pointed to by each row of the output `sparse_handles` will have rank `R-1`.

The `SparseTensor` values can then be read out as part of a minibatch by passing the given keys as vector elements to `TakeManySparseFromTensorsMap`. To ensure the correct `SparseTensorsMap` is accessed, ensure that the same `container` and `shared_name` are passed to that Op. If no `shared_name` is provided here, instead use the *name* of the Operation created by calling `AddManySparseToTensorsMap` as the `shared_name` passed to `TakeManySparseFromTensorsMap`. Ensure the Operations are colocated.

Arguments: 

sparse_indices: 2-D.  The `indices` of the minibatch `SparseTensor`.

`sparse_indices[:, 0]` must be ordered values in `[0, N)`. 

sparse_values: 1-D.  The `values` of the minibatch `SparseTensor`.
sparse_shape: 1-D.  The `shape` of the minibatch `SparseTensor`.

The minibatch size `N == sparse_shape[0]`.

Returns 1-D. The handles of the `SparseTensor` now stored in the `SparseTensorsMap`. Shape: `[N]`.

func AddN 

func AddN(scope *Scope, inputs []tf.Output) (sum tf.Output)

Add all input tensors element wise. 

Inputs must be of same size and shape.

```python
x = [9, 7, 10]
tf.math.add_n(x) ==> 26
```

func AddSparseToTensorsMap 

func AddSparseToTensorsMap(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...AddSparseToTensorsMapAttr) (sparse_handle tf.Output)

Add a `SparseTensor` to a `SparseTensorsMap` return its handle.

A `SparseTensor` is represented by three tensors: `sparse_indices`, `sparse_values`, and `sparse_shape`.

This operator takes the given `SparseTensor` and adds it to a container object (a `SparseTensorsMap`). A unique key within this container is generated in the form of an `int64`, and this is the value that is returned.

The `SparseTensor` can then be read out as part of a minibatch by passing the key as a vector element to `TakeManySparseFromTensorsMap`. To ensure the correct `SparseTensorsMap` is accessed, ensure that the same `container` and `shared_name` are passed to that Op. If no `shared_name` is provided here, instead use the *name* of the Operation created by calling `AddSparseToTensorsMap` as the `shared_name` passed to `TakeManySparseFromTensorsMap`. Ensure the Operations are colocated.

Arguments: 

sparse_indices: 2-D.  The `indices` of the `SparseTensor`.
sparse_values: 1-D.  The `values` of the `SparseTensor`.
sparse_shape: 1-D.  The `shape` of the `SparseTensor`.

Returns 0-D. The handle of the `SparseTensor` now stored in the `SparseTensorsMap`.

func AddV2 

func AddV2(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x + y element-wise.

*NOTE*: `Add` supports broadcasting. `AddN` does not. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func AdjustContrast 

func AdjustContrast(scope *Scope, images tf.Output, contrast_factor tf.Output, min_value tf.Output, max_value tf.Output) (output tf.Output)

Deprecated. Disallowed in GraphDef version >= 2.

DEPRECATED at GraphDef version 2: Use AdjustContrastv2 instead

func AdjustContrastv2 

func AdjustContrastv2(scope *Scope, images tf.Output, contrast_factor tf.Output) (output tf.Output)

Adjust the contrast of one or more images.

`images` is a tensor of at least 3 dimensions. The last 3 dimensions are interpreted as `[height, width, channels]`. The other dimensions only represent a collection of images, such as `[batch, height, width, channels].`

Contrast is adjusted independently for each channel of each image.

For each channel, the Op first computes the mean of the image pixels in the channel and then adjusts each component of each pixel to `(x - mean) * contrast_factor + mean`.

Arguments: 

images: Images to adjust.  At least 3-D.
contrast_factor: A float multiplier for adjusting contrast.

Returns The contrast-adjusted image or images.

func AdjustHue 

func AdjustHue(scope *Scope, images tf.Output, delta tf.Output) (output tf.Output)

Adjust the hue of one or more images.

`images` is a tensor of at least 3 dimensions. The last dimension is interpreted as channels, and must be three.

The input image is considered in the RGB colorspace. Conceptually, the RGB colors are first mapped into HSV. A delta is then applied all the hue values, and then remapped back to RGB colorspace.

Arguments: 

images: Images to adjust.  At least 3-D.
delta: A float delta to add to the hue.

Returns The hue-adjusted image or images.

func AdjustSaturation 

func AdjustSaturation(scope *Scope, images tf.Output, scale tf.Output) (output tf.Output)

Adjust the saturation of one or more images.

`images` is a tensor of at least 3 dimensions. The last dimension is interpreted as channels, and must be three.

The input image is considered in the RGB colorspace. Conceptually, the RGB colors are first mapped into HSV. A scale is then applied all the saturation values, and then remapped back to RGB colorspace.

Arguments: 

images: Images to adjust.  At least 3-D.
scale: A float scale to add to the saturation.

Returns The hue-adjusted image or images.

func All 

func All(scope *Scope, input tf.Output, axis tf.Output, optional ...AllAttr) (output tf.Output)

Computes the "logical and" of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func AllCandidateSampler 

func AllCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, optional ...AllCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a learned unigram distribution.

See explanations of candidate sampling and the data formats at go/candidate-sampling.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to produce.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities.

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func AllToAll 

func AllToAll(scope *Scope, input tf.Output, group_assignment tf.Output, concat_dimension int64, split_dimension int64, split_count int64) (output tf.Output)

An Op to exchange data across TPU replicas.

On each replica, the input is split into `split_count` blocks along `split_dimension` and send to the other replicas given group_assignment. After receiving `split_count` - 1 blocks from other replicas, we concatenate the blocks along `concat_dimension` as the output.

For example, suppose there are 2 TPU replicas: replica 0 receives input: `[[A, B]]` replica 1 receives input: `[[C, D]]`

group_assignment=`[[0, 1]]` concat_dimension=0 split_dimension=1 split_count=2

replica 0's output: `[[A], [C]]` replica 1's output: `[[B], [D]]`

Arguments: 

input: The local input to the sum.
group_assignment: An int32 tensor with shape

[num_groups, num_replicas_per_group]. `group_assignment[i]` represents the replica ids in the ith subgroup. 

concat_dimension: The dimension number to concatenate.
split_dimension: The dimension number to split.
split_count: The number of splits, this number must equal to the sub-group

size(group_assignment.get_shape()[1])

Returns The exchanged result.

func Angle 

func Angle(scope *Scope, input tf.Output, optional ...AngleAttr) (output tf.Output)

Returns the argument of a complex number.

Given a tensor `input` of complex numbers, this operation returns a tensor of type `float` that is the argument of each element in `input`. All elements in `input` must be complex numbers of the form \\(a + bj\\), where *a* is the real part and *b* is the imaginary part.

The argument returned by this operation is of the form \\(atan2(b, a)\\).

For example:

``` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.math.angle(input) ==> [2.0132, 1.056] ```

@compatibility(numpy) Equivalent to np.angle. @end_compatibility

func AnonymousHashTable 

func AnonymousHashTable(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType) (table_handle tf.Output)

Creates a uninitialized anonymous hash table.

This op creates a new anonymous hash table (as a resource) everytime it is executed, with the specified dtype of its keys and values, returning the resource handle. Before using the table you will have to initialize it. After initialization the table will be immutable. The table is anonymous in the sense that it can only be accessed by the returned resource handle (e.g. it cannot be looked up by a name in a resource manager). The table will be automatically deleted when all resource handles pointing to it are gone.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns The resource handle to the newly created hash-table resource.

func AnonymousIterator 

func AnonymousIterator(scope *Scope, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A container for an iterator resource.

Returns A handle to the iterator that can be passed to a "MakeIterator" or "IteratorGetNext" op. In contrast to Iterator, AnonymousIterator prevents resource sharing by name, and does not keep a reference to the resource container.

func AnonymousIteratorV2 

func AnonymousIteratorV2(scope *Scope, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output, deleter tf.Output)

A container for an iterator resource.

Returns: 

handle: A handle to the iterator that can be passed to a "MakeIterator" or

"IteratorGetNext" op. In contrast to Iterator, AnonymousIterator prevents resource sharing by name, and does not keep a reference to the resource container. 

deleter: A variant deleter that should be passed into the op that deletes the iterator.

func AnonymousIteratorV3 

func AnonymousIteratorV3(scope *Scope, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A container for an iterator resource.

Returns A handle to the iterator that can be passed to a "MakeIterator" or "IteratorGetNext" op. In contrast to Iterator, AnonymousIterator prevents resource sharing by name, and does not keep a reference to the resource container.

func AnonymousMultiDeviceIterator 

func AnonymousMultiDeviceIterator(scope *Scope, devices []string, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output, deleter tf.Output)

A container for a multi device iterator resource.

Returns: 

handle: A handle to a multi device iterator that can be passed to a

"MultiDeviceIteratorGetNextFromShard" op. In contrast to MultiDeviceIterator, AnonymousIterator prevents resource sharing by name, and does not keep a reference to the resource container. 

deleter: A variant deleter that should be passed into the op that deletes the iterator.

func AnonymousMultiDeviceIteratorV3 

func AnonymousMultiDeviceIteratorV3(scope *Scope, devices []string, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A container for a multi device iterator resource.

Returns A handle to a multi device iterator that can be passed to a "MultiDeviceIteratorGetNextFromShard" op. In contrast to MultiDeviceIterator, AnonymousIterator prevents resource sharing by name, and does not keep a reference to the resource container.

func AnonymousMutableDenseHashTable 

func AnonymousMutableDenseHashTable(scope *Scope, empty_key tf.Output, deleted_key tf.Output, value_dtype tf.DataType, optional ...AnonymousMutableDenseHashTableAttr) (table_handle tf.Output)

Creates an empty anonymous mutable hash table that uses tensors as the backing store.

This op creates a new anonymous mutable hash table (as a resource) everytime it is executed, with the specified dtype of its keys and values, returning the resource handle. Each value must be a scalar. Data can be inserted into the table using the insert operations. It does not support the initialization operation.

It uses "open addressing" with quadratic reprobing to resolve collisions.

The table is anonymous in the sense that it can only be accessed by the returned resource handle (e.g. it cannot be looked up by a name in a resource manager). The table will be automatically deleted when all resource handles pointing to it are gone.

Arguments: 

empty_key: The key used to represent empty key buckets internally. Must not

be used in insert or lookup operations. 

value_dtype: Type of the table values.

Returns The resource handle to the newly created hash-table resource.

func AnonymousMutableHashTable 

func AnonymousMutableHashTable(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType) (table_handle tf.Output)

Creates an empty anonymous mutable hash table.

This op creates a new anonymous mutable hash table (as a resource) everytime it is executed, with the specified dtype of its keys and values, returning the resource handle. Each value must be a scalar. Data can be inserted into the table using the insert operations. It does not support the initialization operation. The table is anonymous in the sense that it can only be accessed by the returned resource handle (e.g. it cannot be looked up by a name in a resource manager). The table will be automatically deleted when all resource handles pointing to it are gone.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns The resource handle to the newly created hash-table resource.

func AnonymousMutableHashTableOfTensors 

func AnonymousMutableHashTableOfTensors(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType, optional ...AnonymousMutableHashTableOfTensorsAttr) (table_handle tf.Output)

Creates an empty anonymous mutable hash table of vector values.

This op creates a new anonymous mutable hash table (as a resource) everytime it is executed, with the specified dtype of its keys and values, returning the resource handle. Each value must be a vector. Data can be inserted into the table using the insert operations. It does not support the initialization operation. The table is anonymous in the sense that it can only be accessed by the returned resource handle (e.g. it cannot be looked up by a name in a resource manager). The table will be automatically deleted when all resource handles pointing to it are gone.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns The resource handle to the newly created hash-table resource.

func Any 

func Any(scope *Scope, input tf.Output, axis tf.Output, optional ...AnyAttr) (output tf.Output)

Computes the "logical or" of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func ApproxTopK added in v0.2.0 

func ApproxTopK(scope *Scope, input tf.Output, k int64, optional ...ApproxTopKAttr) (values tf.Output, indices tf.Output)

Returns min/max k values and their indices of the input operand in an approximate manner.

See https://arxiv.org/abs/2206.14286 for the algorithm details. This op is only optimized on TPU currently.

Arguments: 

input: Array to search. Must be at least 1-D of the floating type
k: Specifies the number of min/max-k.

Returns: 

values: The min/max k values along the `reduction_dimension` of the `input` operand.

The dimension are the same as the `input` operand except for the `reduction_dimension`: when `aggregate_to_topk` is true, the reduction dimension is `k`; otherwise, it is greater equals to `k` where the size is implementation-defined. 

indices: The indices of `values` along the `reduction_dimension` of the `input` operand.

func ApproximateEqual 

func ApproximateEqual(scope *Scope, x tf.Output, y tf.Output, optional ...ApproximateEqualAttr) (z tf.Output)

Returns the truth value of abs(x-y) < tolerance element-wise.

func ArgMax 

func ArgMax(scope *Scope, input tf.Output, dimension tf.Output, optional ...ArgMaxAttr) (output tf.Output)

Returns the index with the largest value across dimensions of a tensor.

Note that in case of ties the identity of the return value is not guaranteed.

Usage: 

```python
import tensorflow as tf
a = [1, 10, 26.9, 2.8, 166.32, 62.3]
b = tf.math.argmax(input = a)
c = tf.keras.backend.eval(b)
# c = 4
# here a[4] = 166.32 which is the largest element of a across axis 0
```

Arguments: 

dimension: int16, int32 or int64, must be in the range `[-rank(input), rank(input))`.

Describes which dimension of the input Tensor to reduce across. For vectors, use dimension = 0.

func ArgMin 

func ArgMin(scope *Scope, input tf.Output, dimension tf.Output, optional ...ArgMinAttr) (output tf.Output)

Returns the index with the smallest value across dimensions of a tensor.

Note that in case of ties the identity of the return value is not guaranteed.

Usage: 

```python
import tensorflow as tf
a = [1, 10, 26.9, 2.8, 166.32, 62.3]
b = tf.math.argmin(input = a)
c = tf.keras.backend.eval(b)
# c = 0
# here a[0] = 1 which is the smallest element of a across axis 0
```

Arguments: 

dimension: int32 or int64, must be in the range `[-rank(input), rank(input))`.

Describes which dimension of the input Tensor to reduce across. For vectors, use dimension = 0.

func AsString 

func AsString(scope *Scope, input tf.Output, optional ...AsStringAttr) (output tf.Output)

Converts each entry in the given tensor to strings.

Supports many numeric types and boolean.

For Unicode, see the [https://www.tensorflow.org/tutorials/representation/unicode](Working with Unicode text) tutorial.

Examples:

>>> tf.strings.as_string([3, 2]) <tf.Tensor: shape=(2,), dtype=string, numpy=array([b'3', b'2'], dtype=object)> >>> tf.strings.as_string([3.1415926, 2.71828], precision=2).numpy() array([b'3.14', b'2.72'], dtype=object)

func Asin 

func Asin(scope *Scope, x tf.Output) (y tf.Output)

Computes the trignometric inverse sine of x element-wise.

The `tf.math.asin` operation returns the inverse of `tf.math.sin`, such that if `y = tf.math.sin(x)` then, `x = tf.math.asin(y)`.

**Note**: The output of `tf.math.asin` will lie within the invertible range of sine, i.e [-pi/2, pi/2].

For example:

```python # Note: [1.047, 0.785] ~= [(pi/3), (pi/4)] x = tf.constant([1.047, 0.785]) y = tf.math.sin(x) # [0.8659266, 0.7068252]

tf.math.asin(y) # [1.047, 0.785] = x ```

func Asinh 

func Asinh(scope *Scope, x tf.Output) (y tf.Output)

Computes inverse hyperbolic sine of x element-wise. 

Given an input tensor, this function computes inverse hyperbolic sine
for every element in the tensor. Both input and output has a range of
`[-inf, inf]`.

```python
x = tf.constant([-float("inf"), -2, -0.5, 1, 1.2, 200, 10000, float("inf")])
tf.math.asinh(x) ==> [-inf -1.4436355 -0.4812118 0.8813736 1.0159732 5.991471 9.903487 inf]
```

func Assert 

func Assert(scope *Scope, condition tf.Output, data []tf.Output, optional ...AssertAttr) (o *tf.Operation)

Asserts that the given condition is true.

If `condition` evaluates to false, print the list of tensors in `data`. `summarize` determines how many entries of the tensors to print.

Arguments: 

condition: The condition to evaluate.
data: The tensors to print out when condition is false.

Returns the created operation.

func AssertNextDataset 

func AssertNextDataset(scope *Scope, input_dataset tf.Output, transformations tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A transformation that asserts which transformations happen next.

This transformation checks whether the camel-case names (i.e. "FlatMap", not "flat_map") of the transformations following this transformation match the list of names in the `transformations` argument. If there is a mismatch, the transformation raises an exception.

The check occurs when iterating over the contents of the dataset, which means that the check happens *after* any static optimizations are applied to the dataset graph.

Arguments: 

input_dataset: A variant tensor representing the input dataset.

`AssertNextDataset` passes through the outputs of its input dataset. 

transformations: A `tf.string` vector `tf.Tensor` identifying the transformations that are

expected to happen next.

func AssertPrevDataset 

func AssertPrevDataset(scope *Scope, input_dataset tf.Output, transformations tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A transformation that asserts which transformations happened previously.

This transformation checks the names and, optionally, the attribute name-value pairs in the `transformations` argument against those of the transformations that preceded this transformation. If there is a mismatch, the transformation raises an exception.

The check occurs when iterating over the contents of the dataset, which means that the check happens *after* any static optimizations are applied to the dataset graph.

Arguments: 

input_dataset: A variant tensor representing the input dataset.

`AssertPrevDataset` passes through the outputs of its input dataset. 

transformations: A `tf.string` vector `tf.Tensor` identifying the transformations, with optional

attribute name-value pairs, that are expected to have happened previously.

func AssignAddVariableOp 

func AssignAddVariableOp(scope *Scope, resource tf.Output, value tf.Output) (o *tf.Operation)

Adds a value to the current value of a variable.

Any ReadVariableOp with a control dependency on this op is guaranteed to see the incremented value or a subsequent newer one.

Arguments: 

resource: handle to the resource in which to store the variable.
value: the value by which the variable will be incremented.

Returns the created operation.

func AssignSubVariableOp 

func AssignSubVariableOp(scope *Scope, resource tf.Output, value tf.Output) (o *tf.Operation)

Subtracts a value from the current value of a variable.

Any ReadVariableOp with a control dependency on this op is guaranteed to see the decremented value or a subsequent newer one.

Arguments: 

resource: handle to the resource in which to store the variable.
value: the value by which the variable will be incremented.

Returns the created operation.

func AssignVariableOp 

func AssignVariableOp(scope *Scope, resource tf.Output, value tf.Output, optional ...AssignVariableOpAttr) (o *tf.Operation)

Assigns a new value to a variable.

Any ReadVariableOp with a control dependency on this op is guaranteed to return this value or a subsequent newer value of the variable.

Arguments: 

resource: handle to the resource in which to store the variable.
value: the value to set the new tensor to use.

Returns the created operation.

func AssignVariableXlaConcatND 

func AssignVariableXlaConcatND(scope *Scope, resource tf.Output, inputs []tf.Output, num_concats []int64, optional ...AssignVariableXlaConcatNDAttr) (o *tf.Operation)

Concats input tensor across all dimensions.

An op which merges slices the input tensor based on the given num_splits attribute, strips paddings optionally, and writes the merged tensor without paddings to the resource variable.

This op may be generated via the TPU bridge.

For example, with `input` tensor: ``` [[0, 1], 

[4, 5]]

[[2, 3], 

[6, 7]]

[[8, 9], 

[12, 13]]

[[10, 11], 

[14, 15]]

``` `num_splits`: ``` [2, 2] ``` and `paddings`: ``` [1, 1] ``` the expected `outputs` is: ``` [[0, 1, 2], 

[4, 5, 6],
[8, 9, 10]]

```

Arguments: 

resource: Resource variable for concatenated input tensors across all dimensions.
inputs: Input tensor slices in row-major order to merge across all dimensions. All

inputs must have the same shape. 

num_concats: Number of ways to merge per dimension.

Returns the created operation.

func Atan 

func Atan(scope *Scope, x tf.Output) (y tf.Output)

Computes the trignometric inverse tangent of x element-wise.

The `tf.math.atan` operation returns the inverse of `tf.math.tan`, such that if `y = tf.math.tan(x)` then, `x = tf.math.atan(y)`.

**Note**: The output of `tf.math.atan` will lie within the invertible range of tan, i.e (-pi/2, pi/2).

For example:

```python # Note: [1.047, 0.785] ~= [(pi/3), (pi/4)] x = tf.constant([1.047, 0.785]) y = tf.math.tan(x) # [1.731261, 0.99920404]

tf.math.atan(y) # [1.047, 0.785] = x ```

func Atan2 

func Atan2(scope *Scope, y tf.Output, x tf.Output) (z tf.Output)

Computes arctangent of `y/x` element-wise, respecting signs of the arguments.

This is the angle \\( \theta \in [-\pi, \pi] \\) such that \\[ x = r \cos(\theta) \\] and \\[ y = r \sin(\theta) \\] where \\(r = \sqrt{x^2 + y^2} \\).

For example:

>>> x = [1., 1.] >>> y = [1., -1.] >>> print((tf.math.atan2(y,x) * (180 / np.pi)).numpy()) [ 45. -45.]

func Atanh 

func Atanh(scope *Scope, x tf.Output) (y tf.Output)

Computes inverse hyperbolic tangent of x element-wise. 

Given an input tensor, this function computes inverse hyperbolic tangent
for every element in the tensor. Input range is `[-1,1]` and output range is
`[-inf, inf]`. If input is `-1`, output will be `-inf` and if the
input is `1`, output will be `inf`. Values outside the range will have
`nan` as output.

```python
x = tf.constant([-float("inf"), -1, -0.5, 1, 0, 0.5, 10, float("inf")])
tf.math.atanh(x) ==> [nan -inf -0.54930615 inf  0. 0.54930615 nan nan]
```

func AudioSpectrogram 

func AudioSpectrogram(scope *Scope, input tf.Output, window_size int64, stride int64, optional ...AudioSpectrogramAttr) (spectrogram tf.Output)

Produces a visualization of audio data over time.

Spectrograms are a standard way of representing audio information as a series of slices of frequency information, one slice for each window of time. By joining these together into a sequence, they form a distinctive fingerprint of the sound over time.

This op expects to receive audio data as an input, stored as floats in the range -1 to 1, together with a window width in samples, and a stride specifying how far to move the window between slices. From this it generates a three dimensional output. The first dimension is for the channels in the input, so a stereo audio input would have two here for example. The second dimension is time, with successive frequency slices. The third dimension has an amplitude value for each frequency during that time slice.

This means the layout when converted and saved as an image is rotated 90 degrees clockwise from a typical spectrogram. Time is descending down the Y axis, and the frequency decreases from left to right.

Each value in the result represents the square root of the sum of the real and imaginary parts of an FFT on the current window of samples. In this way, the lowest dimension represents the power of each frequency in the current window, and adjacent windows are concatenated in the next dimension.

To get a more intuitive and visual look at what this operation does, you can run tensorflow/examples/wav_to_spectrogram to read in an audio file and save out the resulting spectrogram as a PNG image.

Arguments: 

input: Float representation of audio data.
window_size: How wide the input window is in samples. For the highest efficiency

this should be a power of two, but other values are accepted. 

stride: How widely apart the center of adjacent sample windows should be.

Returns 3D representation of the audio frequencies as an image.

func AudioSummary 

func AudioSummary(scope *Scope, tag tf.Output, tensor tf.Output, sample_rate float32, optional ...AudioSummaryAttr) (summary tf.Output)

Outputs a `Summary` protocol buffer with audio.

DEPRECATED at GraphDef version 15: Use AudioSummaryV2.

The summary has up to `max_outputs` summary values containing audio. The audio is built from `tensor` which must be 3-D with shape `[batch_size, frames, channels]` or 2-D with shape `[batch_size, frames]`. The values are assumed to be in the range of `[-1.0, 1.0]` with a sample rate of `sample_rate`.

The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:

  • If `max_outputs` is 1, the summary value tag is '*tag*/audio'.
  • If `max_outputs` is greater than 1, the summary value tags are generated sequentially as '*tag*/audio/0', '*tag*/audio/1', etc.

Arguments: 

tag: Scalar. Used to build the `tag` attribute of the summary values.
tensor: 2-D of shape `[batch_size, frames]`.
sample_rate: The sample rate of the signal in hertz.

Returns Scalar. Serialized `Summary` protocol buffer.

func AudioSummaryV2 

func AudioSummaryV2(scope *Scope, tag tf.Output, tensor tf.Output, sample_rate tf.Output, optional ...AudioSummaryV2Attr) (summary tf.Output)

Outputs a `Summary` protocol buffer with audio.

The summary has up to `max_outputs` summary values containing audio. The audio is built from `tensor` which must be 3-D with shape `[batch_size, frames, channels]` or 2-D with shape `[batch_size, frames]`. The values are assumed to be in the range of `[-1.0, 1.0]` with a sample rate of `sample_rate`.

The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:

  • If `max_outputs` is 1, the summary value tag is '*tag*/audio'.
  • If `max_outputs` is greater than 1, the summary value tags are generated sequentially as '*tag*/audio/0', '*tag*/audio/1', etc.

Arguments: 

tag: Scalar. Used to build the `tag` attribute of the summary values.
tensor: 2-D of shape `[batch_size, frames]`.
sample_rate: The sample rate of the signal in hertz.

Returns Scalar. Serialized `Summary` protocol buffer.

func AutoShardDataset 

func AutoShardDataset(scope *Scope, input_dataset tf.Output, num_workers tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...AutoShardDatasetAttr) (handle tf.Output)

Creates a dataset that shards the input dataset.

Creates a dataset that shards the input dataset by num_workers, returning a sharded dataset for the index-th worker. This attempts to automatically shard a dataset by examining the Dataset graph and inserting a shard op before the inputs to a reader Dataset (e.g. CSVDataset, TFRecordDataset).

This dataset will throw a NotFound error if we cannot shard the dataset automatically.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
num_workers: A scalar representing the number of workers to distribute this dataset across.
index: A scalar representing the index of the current worker out of num_workers.

func AvgPool 

func AvgPool(scope *Scope, value tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPoolAttr) (output tf.Output)

Performs average pooling on the input.

Each entry in `output` is the mean of the corresponding size `ksize` window in `value`.

Arguments: 

value: 4-D with shape `[batch, height, width, channels]`.
ksize: The size of the sliding window for each dimension of `value`.
strides: The stride of the sliding window for each dimension of `value`.
padding: The type of padding algorithm to use.

Returns The average pooled output tensor.

func AvgPool3D 

func AvgPool3D(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPool3DAttr) (output tf.Output)

Performs 3D average pooling on the input.

Each entry in `output` is the mean of the corresponding size `ksize` window in `value`.

Arguments: 

input: Shape `[batch, depth, rows, cols, channels]` tensor to pool over.
ksize: 1-D tensor of length 5. The size of the window for each dimension of

the input tensor. Must have `ksize[0] = ksize[4] = 1`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

Returns The average pooled output tensor.

func AvgPool3DGrad 

func AvgPool3DGrad(scope *Scope, orig_input_shape tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPool3DGradAttr) (output tf.Output)

Computes gradients of average pooling function.

Arguments: 

orig_input_shape: The original input dimensions.
grad: Output backprop of shape `[batch, depth, rows, cols, channels]`.
ksize: 1-D tensor of length 5. The size of the window for each dimension of

the input tensor. Must have `ksize[0] = ksize[4] = 1`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

Returns The backprop for input.

func AvgPoolGrad 

func AvgPoolGrad(scope *Scope, orig_input_shape tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPoolGradAttr) (output tf.Output)

Computes gradients of the average pooling function.

Arguments: 

orig_input_shape: 1-D.  Shape of the original input to `avg_pool`.
grad: 4-D with shape `[batch, height, width, channels]`.  Gradients w.r.t.

the output of `avg_pool`. 

ksize: The size of the sliding window for each dimension of the input.
strides: The stride of the sliding window for each dimension of the input.
padding: The type of padding algorithm to use.

Returns 4-D. Gradients w.r.t. the input of `avg_pool`.

func Batch 

func Batch(scope *Scope, in_tensors []tf.Output, num_batch_threads int64, max_batch_size int64, batch_timeout_micros int64, grad_timeout_micros int64, optional ...BatchAttr) (batched_tensors []tf.Output, batch_index tf.Output, id tf.Output)

Batches all input tensors nondeterministically.

When many instances of this Op are being run concurrently with the same container/shared_name in the same device, some will output zero-shaped Tensors and others will output Tensors of size up to max_batch_size.

All Tensors in in_tensors are batched together (so, for example, labels and features should be batched with a single instance of this operation.

Each invocation of batch emits an `id` scalar which will be used to identify this particular invocation when doing unbatch or its gradient.

Each op which emits a non-empty batch will also emit a non-empty batch_index Tensor, which, is a [K, 3] matrix where each row contains the invocation's id, start, and length of elements of each set of Tensors present in batched_tensors.

Batched tensors are concatenated along the first dimension, and all tensors in in_tensors must have the first dimension of the same size.

in_tensors: The tensors to be batched. num_batch_threads: Number of scheduling threads for processing batches of work. 

Determines the number of batches processed in parallel.

max_batch_size: Batch sizes will never be bigger than this. batch_timeout_micros: Maximum number of microseconds to wait before outputting 

an incomplete batch.

allowed_batch_sizes: Optional list of allowed batch sizes. If left empty, does 

nothing. Otherwise, supplies a list of batch sizes, causing the op to pad
batches up to one of those sizes. The entries must increase monotonically, and
the final entry must equal max_batch_size.

grad_timeout_micros: The timeout to use for the gradient. See Unbatch. batched_tensors: Either empty tensors or a batch of concatenated Tensors. batch_index: If out_tensors is non-empty, has information to invert it. container: Controls the scope of sharing of this batch. id: always contains a scalar with a unique ID for this invocation of Batch. shared_name: Concurrently running instances of batch in the same device with the 

same container and shared_name will batch their elements together. If left
empty, the op name will be used as the shared name.

T: the types of tensors to be batched.

func BatchDataset 

func BatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...BatchDatasetAttr) (handle tf.Output)

Creates a dataset that batches `batch_size` elements from `input_dataset`.

Arguments: 

batch_size: A scalar representing the number of elements to accumulate in a

batch.

func BatchDatasetV2 

func BatchDatasetV2(scope *Scope, input_dataset tf.Output, batch_size tf.Output, drop_remainder tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...BatchDatasetV2Attr) (handle tf.Output)

Creates a dataset that batches `batch_size` elements from `input_dataset`.

Arguments: 

batch_size: A scalar representing the number of elements to accumulate in a batch.
drop_remainder: A scalar representing whether the last batch should be dropped in case its size

is smaller than desired.

func BatchMatMul 

func BatchMatMul(scope *Scope, x tf.Output, y tf.Output, optional ...BatchMatMulAttr) (output tf.Output)

Multiplies slices of two tensors in batches.

Multiplies all slices of `Tensor` `x` and `y` (each slice can be viewed as an element of a batch), and arranges the individual results in a single output tensor of the same batch size. Each of the individual slices can optionally be adjointed (to adjoint a matrix means to transpose and conjugate it) before multiplication by setting the `adj_x` or `adj_y` flag to `True`, which are by default `False`.

The input tensors `x` and `y` are 2-D or higher with shape `[..., r_x, c_x]` and `[..., r_y, c_y]`.

The output tensor is 2-D or higher with shape `[..., r_o, c_o]`, where: 

r_o = c_x if adj_x else r_x
c_o = r_y if adj_y else c_y

It is computed as: 

output[..., :, :] = matrix(x[..., :, :]) * matrix(y[..., :, :])

Arguments: 

x: 2-D or higher with shape `[..., r_x, c_x]`.
y: 2-D or higher with shape `[..., r_y, c_y]`.

Returns 3-D or higher with shape `[..., r_o, c_o]`

func BatchMatMulV2 

func BatchMatMulV2(scope *Scope, x tf.Output, y tf.Output, optional ...BatchMatMulV2Attr) (output tf.Output)

Multiplies slices of two tensors in batches.

Multiplies all slices of `Tensor` `x` and `y` (each slice can be viewed as an element of a batch), and arranges the individual results in a single output tensor of the same batch size. Each of the individual slices can optionally be adjointed (to adjoint a matrix means to transpose and conjugate it) before multiplication by setting the `adj_x` or `adj_y` flag to `True`, which are by default `False`.

The input tensors `x` and `y` are 2-D or higher with shape `[..., r_x, c_x]` and `[..., r_y, c_y]`.

The output tensor is 2-D or higher with shape `[..., r_o, c_o]`, where: 

r_o = c_x if adj_x else r_x
c_o = r_y if adj_y else c_y

It is computed as: 

output[..., :, :] = matrix(x[..., :, :]) * matrix(y[..., :, :])

*NOTE*: `BatchMatMulV2` supports broadcasting in the batch dimensions. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html).

Arguments: 

x: 2-D or higher with shape `[..., r_x, c_x]`.
y: 2-D or higher with shape `[..., r_y, c_y]`.

Returns 3-D or higher with shape `[..., r_o, c_o]`

func BatchMatMulV3 

func BatchMatMulV3(scope *Scope, x tf.Output, y tf.Output, Tout tf.DataType, optional ...BatchMatMulV3Attr) (output tf.Output)

Multiplies slices of two tensors in batches.

Multiplies all slices of `Tensor` `x` and `y` (each slice can be viewed as an element of a batch), and arranges the individual results in a single output tensor of the same batch size. Each of the individual slices can optionally be adjointed (to adjoint a matrix means to transpose and conjugate it) before multiplication by setting the `adj_x` or `adj_y` flag to `True`, which are by default `False`.

The input tensors `x` and `y` are 2-D or higher with shape `[..., r_x, c_x]` and `[..., r_y, c_y]`.

The output tensor is 2-D or higher with shape `[..., r_o, c_o]`, where: 

r_o = c_x if adj_x else r_x
c_o = r_y if adj_y else c_y

It is computed as: 

output[..., :, :] = matrix(x[..., :, :]) * matrix(y[..., :, :])

*NOTE*: `BatchMatMulV3` supports broadcasting in the batch dimensions. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html).

Arguments: 

x: 2-D or higher with shape `[..., r_x, c_x]`.
y: 2-D or higher with shape `[..., r_y, c_y]`.
Tout: If not spcified, Tout is the same type to input type.

Returns 3-D or higher with shape `[..., r_o, c_o]`

func BatchNormWithGlobalNormalization 

func BatchNormWithGlobalNormalization(scope *Scope, t tf.Output, m tf.Output, v tf.Output, beta tf.Output, gamma tf.Output, variance_epsilon float32, scale_after_normalization bool) (result tf.Output)

Batch normalization.

DEPRECATED at GraphDef version 9: Use tf.nn.batch_normalization()

This op is deprecated. Prefer `tf.nn.batch_normalization`.

Arguments: 

t: A 4D input Tensor.
m: A 1D mean Tensor with size matching the last dimension of t.

This is the first output from tf.nn.moments, or a saved moving average thereof. 

v: A 1D variance Tensor with size matching the last dimension of t.

This is the second output from tf.nn.moments, or a saved moving average thereof. 

beta: A 1D beta Tensor with size matching the last dimension of t.

An offset to be added to the normalized tensor. 

gamma: A 1D gamma Tensor with size matching the last dimension of t.

If "scale_after_normalization" is true, this tensor will be multiplied with the normalized tensor. 

variance_epsilon: A small float number to avoid dividing by 0.
scale_after_normalization: A bool indicating whether the resulted tensor

needs to be multiplied with gamma.

func BatchNormWithGlobalNormalizationGrad 

func BatchNormWithGlobalNormalizationGrad(scope *Scope, t tf.Output, m tf.Output, v tf.Output, gamma tf.Output, backprop tf.Output, variance_epsilon float32, scale_after_normalization bool) (dx tf.Output, dm tf.Output, dv tf.Output, db tf.Output, dg tf.Output)

Gradients for batch normalization.

DEPRECATED at GraphDef version 9: Use tf.nn.batch_normalization()

This op is deprecated. See `tf.nn.batch_normalization`.

Arguments: 

t: A 4D input Tensor.
m: A 1D mean Tensor with size matching the last dimension of t.

This is the first output from tf.nn.moments, or a saved moving average thereof. 

v: A 1D variance Tensor with size matching the last dimension of t.

This is the second output from tf.nn.moments, or a saved moving average thereof. 

gamma: A 1D gamma Tensor with size matching the last dimension of t.

If "scale_after_normalization" is true, this Tensor will be multiplied with the normalized Tensor. 

backprop: 4D backprop Tensor.
variance_epsilon: A small float number to avoid dividing by 0.
scale_after_normalization: A bool indicating whether the resulted tensor

needs to be multiplied with gamma.

Returns: 

dx: 4D backprop tensor for input.
dm: 1D backprop tensor for mean.
dv: 1D backprop tensor for variance.
db: 1D backprop tensor for beta.
dg: 1D backprop tensor for gamma.

func BatchToSpace 

func BatchToSpace(scope *Scope, input tf.Output, crops tf.Output, block_size int64) (output tf.Output)

BatchToSpace for 4-D tensors of type T.

This is a legacy version of the more general BatchToSpaceND.

Rearranges (permutes) data from batch into blocks of spatial data, followed by cropping. This is the reverse transformation of SpaceToBatch. More specifically, this op outputs a copy of the input tensor where values from the `batch` dimension are moved in spatial blocks to the `height` and `width` dimensions, followed by cropping along the `height` and `width` dimensions.

Arguments: 

input: 4-D tensor with shape

`[batch*block_size*block_size, height_pad/block_size, width_pad/block_size, 

depth]`. Note that the batch size of the input tensor must be divisible by

`block_size * block_size`. 

crops: 2-D tensor of non-negative integers with shape `[2, 2]`. It specifies

how many elements to crop from the intermediate result across the spatial dimensions as follows: 

crops = [[crop_top, crop_bottom], [crop_left, crop_right]]

Returns 4-D with shape `[batch, height, width, depth]`, where: 

height = height_pad - crop_top - crop_bottom
width = width_pad - crop_left - crop_right

The attr `block_size` must be greater than one. It indicates the block size.

Some examples:

(1) For the following input of shape `[4, 1, 1, 1]` and block_size of 2:

``` [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] ```

The output tensor has shape `[1, 2, 2, 1]` and value:

``` x = [[[[1], [2]], [[3], [4]]]] ```

(2) For the following input of shape `[4, 1, 1, 3]` and block_size of 2:

``` [[[[1, 2, 3]]], [[[4, 5, 6]]], [[[7, 8, 9]]], [[[10, 11, 12]]]] ```

The output tensor has shape `[1, 2, 2, 3]` and value:

``` x = [[[[1, 2, 3], [4, 5, 6]], 

[[7, 8, 9], [10, 11, 12]]]]

```

(3) For the following input of shape `[4, 2, 2, 1]` and block_size of 2:

``` x = [[[[1], [3]], [[9], [11]]], 

[[[2], [4]], [[10], [12]]],
[[[5], [7]], [[13], [15]]],
[[[6], [8]], [[14], [16]]]]

```

The output tensor has shape `[1, 4, 4, 1]` and value:

``` x = [[[[1], [2], [3], [4]], 

[[5],   [6],  [7],  [8]],
[[9],  [10], [11],  [12]],
[[13], [14], [15],  [16]]]]

```

(4) For the following input of shape `[8, 1, 2, 1]` and block_size of 2:

``` x = [[[[1], [3]]], [[[9], [11]]], [[[2], [4]]], [[[10], [12]]], 

[[[5], [7]]], [[[13], [15]]], [[[6], [8]]], [[[14], [16]]]]

```

The output tensor has shape `[2, 2, 4, 1]` and value:

``` x = [[[[1], [3]], [[5], [7]]], 

[[[2], [4]], [[10], [12]]],
[[[5], [7]], [[13], [15]]],
[[[6], [8]], [[14], [16]]]]

```

func BatchToSpaceND 

func BatchToSpaceND(scope *Scope, input tf.Output, block_shape tf.Output, crops tf.Output) (output tf.Output)

BatchToSpace for N-D tensors of type T.

This operation reshapes the "batch" dimension 0 into `M + 1` dimensions of shape `block_shape + [batch]`, interleaves these blocks back into the grid defined by the spatial dimensions `[1, ..., M]`, to obtain a result with the same rank as the input. The spatial dimensions of this intermediate result are then optionally cropped according to `crops` to produce the output. This is the reverse of SpaceToBatch. See below for a precise description.

Arguments: 

input: N-D with shape `input_shape = [batch] + spatial_shape + remaining_shape`,

where spatial_shape has M dimensions. 

	block_shape: 1-D with shape `[M]`, all values must be >= 1.
	crops: 2-D with shape `[M, 2]`, all values must be >= 0.
  `crops[i] = [crop_start, crop_end]` specifies the amount to crop from input
  dimension `i + 1`, which corresponds to spatial dimension `i`.  It is
  required that
  `crop_start[i] + crop_end[i] <= block_shape[i] * input_shape[i + 1]`.

This operation is equivalent to the following steps:

  1. Reshape `input` to `reshaped` of shape: [block_shape[0], ..., block_shape[M-1], batch / prod(block_shape), input_shape[1], ..., input_shape[N-1]]

  2. Permute dimensions of `reshaped` to produce `permuted` of shape [batch / prod(block_shape),

    input_shape[1], block_shape[0], ..., input_shape[M], block_shape[M-1],

    input_shape[M+1], ..., input_shape[N-1]]

  3. Reshape `permuted` to produce `reshaped_permuted` of shape [batch / prod(block_shape),

    input_shape[1] * block_shape[0], ..., input_shape[M] * block_shape[M-1],

    input_shape[M+1], ..., input_shape[N-1]]

  4. Crop the start and end of dimensions `[1, ..., M]` of `reshaped_permuted` according to `crops` to produce the output of shape: [batch / prod(block_shape),

    input_shape[1] * block_shape[0] - crops[0,0] - crops[0,1], ..., input_shape[M] * block_shape[M-1] - crops[M-1,0] - crops[M-1,1],

    input_shape[M+1], ..., input_shape[N-1]]

Some examples:

(1) For the following input of shape `[4, 1, 1, 1]`, `block_shape = [2, 2]`, and 

`crops = [[0, 0], [0, 0]]`:

``` [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] ```

The output tensor has shape `[1, 2, 2, 1]` and value:

``` x = [[[[1], [2]], [[3], [4]]]] ```

(2) For the following input of shape `[4, 1, 1, 3]`, `block_shape = [2, 2]`, and 

`crops = [[0, 0], [0, 0]]`:

``` [[[[1, 2, 3]]], [[[4, 5, 6]]], [[[7, 8, 9]]], [[[10, 11, 12]]]] ```

The output tensor has shape `[1, 2, 2, 3]` and value:

``` x = [[[[1, 2, 3], [4, 5, 6]], 

[[7, 8, 9], [10, 11, 12]]]]

```

(3) For the following input of shape `[4, 2, 2, 1]`, `block_shape = [2, 2]`, and 

`crops = [[0, 0], [0, 0]]`:

``` x = [[[[1], [3]], [[9], [11]]], 

[[[2], [4]], [[10], [12]]],
[[[5], [7]], [[13], [15]]],
[[[6], [8]], [[14], [16]]]]

```

The output tensor has shape `[1, 4, 4, 1]` and value:

``` x = [[[[1], [2], [3], [4]], 

[[5],   [6],  [7],  [8]],
[[9],  [10], [11],  [12]],
[[13], [14], [15],  [16]]]]

```

(4) For the following input of shape `[8, 1, 3, 1]`, `block_shape = [2, 2]`, and 

`crops = [[0, 0], [2, 0]]`:

``` x = [[[[0], [1], [3]]], [[[0], [9], [11]]], 

[[[0], [2], [4]]], [[[0], [10], [12]]],
[[[0], [5], [7]]], [[[0], [13], [15]]],
[[[0], [6], [8]]], [[[0], [14], [16]]]]

```

The output tensor has shape `[2, 2, 4, 1]` and value:

``` x = [[[[1], [2], [3], [4]], 

 [[5],   [6],  [7],  [8]]],
[[[9],  [10], [11],  [12]],
 [[13], [14], [15],  [16]]]]

```

func Betainc 

func Betainc(scope *Scope, a tf.Output, b tf.Output, x tf.Output) (z tf.Output)

Compute the regularized incomplete beta integral \\(I_x(a, b)\\).

The regularized incomplete beta integral is defined as:

\\(I_x(a, b) = \frac{B(x; a, b)}{B(a, b)}\\)

where

\\(B(x; a, b) = \int_0^x t^{a-1} (1 - t)^{b-1} dt\\)

is the incomplete beta function and \\(B(a, b)\\) is the *complete* beta function.

func BiasAdd 

func BiasAdd(scope *Scope, value tf.Output, bias tf.Output, optional ...BiasAddAttr) (output tf.Output)

Adds `bias` to `value`.

This is a special case of `tf.add` where `bias` is restricted to be 1-D. Broadcasting is supported, so `value` may have any number of dimensions.

Arguments: 

value: Any number of dimensions.
bias: 1-D with size the last dimension of `value`.

Returns Broadcasted sum of `value` and `bias`.

func BiasAddGrad 

func BiasAddGrad(scope *Scope, out_backprop tf.Output, optional ...BiasAddGradAttr) (output tf.Output)

The backward operation for "BiasAdd" on the "bias" tensor.

It accumulates all the values from out_backprop into the feature dimension. For NHWC data format, the feature dimension is the last. For NCHW data format, the feature dimension is the third-to-last.

Arguments: 

out_backprop: Any number of dimensions.

Returns 1-D with size the feature dimension of `out_backprop`.

func BiasAddV1 

func BiasAddV1(scope *Scope, value tf.Output, bias tf.Output) (output tf.Output)

Adds `bias` to `value`.

This is a deprecated version of BiasAdd and will be soon removed.

This is a special case of `tf.add` where `bias` is restricted to be 1-D. Broadcasting is supported, so `value` may have any number of dimensions.

Arguments: 

value: Any number of dimensions.
bias: 1-D with size the last dimension of `value`.

Returns Broadcasted sum of `value` and `bias`.

func Bincount 

func Bincount(scope *Scope, arr tf.Output, size tf.Output, weights tf.Output) (bins tf.Output)

Counts the number of occurrences of each value in an integer array.

Outputs a vector with length `size` and the same dtype as `weights`. If `weights` are empty, then index `i` stores the number of times the value `i` is counted in `arr`. If `weights` are non-empty, then index `i` stores the sum of the value in `weights` at each index where the corresponding value in `arr` is `i`.

Values in `arr` outside of the range [0, size) are ignored.

Arguments: 

arr: int32 `Tensor`.
size: non-negative int32 scalar `Tensor`.
weights: is an int32, int64, float32, or float64 `Tensor` with the same

shape as `arr`, or a length-0 `Tensor`, in which case it acts as all weights equal to 1.

Returns 1D `Tensor` with length equal to `size`. The counts or summed weights for each value in the range [0, size).

func Bitcast 

func Bitcast(scope *Scope, input tf.Output, type_ tf.DataType) (output tf.Output)

Bitcasts a tensor from one type to another without copying data.

Given a tensor `input`, this operation returns a tensor that has the same buffer data as `input` with datatype `type`.

If the input datatype `T` is larger than the output datatype `type` then the shape changes from [...] to [..., sizeof(`T`)/sizeof(`type`)].

If `T` is smaller than `type`, the operator requires that the rightmost dimension be equal to sizeof(`type`)/sizeof(`T`). The shape then goes from [..., sizeof(`type`)/sizeof(`T`)] to [...].

tf.bitcast() and tf.cast() work differently when real dtype is casted as a complex dtype (e.g. tf.complex64 or tf.complex128) as tf.cast() make imaginary part 0 while tf.bitcast() gives module error. For example,

Example 1:

>>> a = [1., 2., 3.] >>> equality_bitcast = tf.bitcast(a, tf.complex128) Traceback (most recent call last): ... InvalidArgumentError: Cannot bitcast from 1 to 18 [Op:Bitcast] >>> equality_cast = tf.cast(a, tf.complex128) >>> print(equality_cast) tf.Tensor([1.+0.j 2.+0.j 3.+0.j], shape=(3,), dtype=complex128)

Example 2:

>>> tf.bitcast(tf.constant(0xffffffff, dtype=tf.uint32), tf.uint8) <tf.Tensor: shape=(4,), dtype=uint8, numpy=array([255, 255, 255, 255], dtype=uint8)>

Example 3:

>>> x = [1., 2., 3.] >>> y = [0., 2., 3.] >>> equality= tf.equal(x,y) >>> equality_cast = tf.cast(equality,tf.float32) >>> equality_bitcast = tf.bitcast(equality_cast,tf.uint8) >>> print(equality) tf.Tensor([False True True], shape=(3,), dtype=bool) >>> print(equality_cast) tf.Tensor([0. 1. 1.], shape=(3,), dtype=float32) >>> print(equality_bitcast) tf.Tensor( 

[[  0   0   0   0]
 [  0   0 128  63]
 [  0   0 128  63]], shape=(3, 4), dtype=uint8)

*NOTE*: Bitcast is implemented as a low-level cast, so machines with different endian orderings will give different results. A copy from input buffer to output buffer is made on BE machines when types are of different sizes in order to get the same casting results as on LE machines.

func BitwiseAnd 

func BitwiseAnd(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Elementwise computes the bitwise AND of `x` and `y`.

The result will have those bits set, that are set in both `x` and `y`. The computation is performed on the underlying representations of `x` and `y`.

For example:

```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops dtype_list = [tf.int8, tf.int16, tf.int32, tf.int64, 

tf.uint8, tf.uint16, tf.uint32, tf.uint64]

for dtype in dtype_list: 

lhs = tf.constant([0, 5, 3, 14], dtype=dtype)
rhs = tf.constant([5, 0, 7, 11], dtype=dtype)
exp = tf.constant([0, 0, 3, 10], dtype=tf.float32)

res = bitwise_ops.bitwise_and(lhs, rhs)
tf.assert_equal(tf.cast(res, tf.float32), exp) # TRUE

```

func BitwiseOr 

func BitwiseOr(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Elementwise computes the bitwise OR of `x` and `y`.

The result will have those bits set, that are set in `x`, `y` or both. The computation is performed on the underlying representations of `x` and `y`.

For example:

```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops dtype_list = [tf.int8, tf.int16, tf.int32, tf.int64, 

tf.uint8, tf.uint16, tf.uint32, tf.uint64]

for dtype in dtype_list: 

lhs = tf.constant([0, 5, 3, 14], dtype=dtype)
rhs = tf.constant([5, 0, 7, 11], dtype=dtype)
exp = tf.constant([5, 5, 7, 15], dtype=tf.float32)

res = bitwise_ops.bitwise_or(lhs, rhs)
tf.assert_equal(tf.cast(res,  tf.float32), exp)  # TRUE

```

func BitwiseXor 

func BitwiseXor(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Elementwise computes the bitwise XOR of `x` and `y`.

The result will have those bits set, that are different in `x` and `y`. The computation is performed on the underlying representations of `x` and `y`.

For example:

```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops dtype_list = [tf.int8, tf.int16, tf.int32, tf.int64, 

tf.uint8, tf.uint16, tf.uint32, tf.uint64]

for dtype in dtype_list: 

lhs = tf.constant([0, 5, 3, 14], dtype=dtype)
rhs = tf.constant([5, 0, 7, 11], dtype=dtype)
exp = tf.constant([5, 5, 4, 5],  dtype=tf.float32)

res = bitwise_ops.bitwise_xor(lhs, rhs)
tf.assert_equal(tf.cast(res, tf.float32), exp) # TRUE

```

func BlockLSTM 

func BlockLSTM(scope *Scope, seq_len_max tf.Output, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, optional ...BlockLSTMAttr) (i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, h tf.Output)

Computes the LSTM cell forward propagation for all the time steps.

This is equivalent to applying LSTMBlockCell in a loop, like so:

```python for x1 in unpack(x): 

i1, cs1, f1, o1, ci1, co1, h1 = LSTMBlock(
  x1, cs_prev, h_prev, w, wci, wcf, wco, b)
cs_prev = cs1
h_prev = h1
i.append(i1)
cs.append(cs1)
f.append(f1)
o.append(o1)
ci.append(ci1)
co.append(co1)
h.append(h1)

return pack(i), pack(cs), pack(f), pack(o), pack(ci), pack(ch), pack(h) ```

Arguments: 

seq_len_max: Maximum time length actually used by this input. Outputs are padded

with zeros beyond this length. 

x: The sequence input to the LSTM, shape (timelen, batch_size, num_inputs).
cs_prev: Value of the initial cell state.
h_prev: Initial output of cell (to be used for peephole).
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.

Returns: 

i: The input gate over the whole time sequence.
cs: The cell state before the tanh over the whole time sequence.
f: The forget gate over the whole time sequence.
o: The output gate over the whole time sequence.
ci: The cell input over the whole time sequence.
co: The cell after the tanh over the whole time sequence.
h: The output h vector over the whole time sequence.

func BlockLSTMGrad 

func BlockLSTMGrad(scope *Scope, seq_len_max tf.Output, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, h tf.Output, cs_grad tf.Output, h_grad tf.Output, use_peephole bool) (x_grad tf.Output, cs_prev_grad tf.Output, h_prev_grad tf.Output, w_grad tf.Output, wci_grad tf.Output, wcf_grad tf.Output, wco_grad tf.Output, b_grad tf.Output)

Computes the LSTM cell backward propagation for the entire time sequence.

This implementation is to be used in conjunction of LSTMBlock.

Arguments: 

seq_len_max: Maximum time length actually used by this input. Outputs are padded

with zeros beyond this length. 

x: The sequence input to the LSTM, shape (timelen, batch_size, num_inputs).
cs_prev: Value of the initial cell state.
h_prev: Initial output of cell (to be used for peephole).
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.
i: The input gate over the whole time sequence.
cs: The cell state before the tanh over the whole time sequence.
f: The forget gate over the whole time sequence.
o: The output gate over the whole time sequence.
ci: The cell input over the whole time sequence.
co: The cell after the tanh over the whole time sequence.
h: The output h vector over the whole time sequence.
cs_grad: The current gradient of cs.
h_grad: The gradient of h vector.
use_peephole: Whether to use peephole weights.

Returns: 

x_grad: The gradient of x to be back-propped.
cs_prev_grad: The gradient of cs_prev to be back-propped.
h_prev_grad: The gradient of h_prev to be back-propped.
w_grad: The gradient for w to be back-propped.
wci_grad: The gradient for wci to be back-propped.
wcf_grad: The gradient for wcf to be back-propped.
wco_grad: The gradient for wco to be back-propped.
b_grad: The gradient for w to be back-propped.

func BlockLSTMGradV2 

func BlockLSTMGradV2(scope *Scope, seq_len_max tf.Output, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, h tf.Output, cs_grad tf.Output, h_grad tf.Output, use_peephole bool) (x_grad tf.Output, cs_prev_grad tf.Output, h_prev_grad tf.Output, w_grad tf.Output, wci_grad tf.Output, wcf_grad tf.Output, wco_grad tf.Output, b_grad tf.Output)

Computes the LSTM cell backward propagation for the entire time sequence.

This implementation is to be used in conjunction of BlockLSTMV2.

Arguments: 

seq_len_max: Maximum time length actually used by this input. Outputs are padded

with zeros beyond this length. 

x: The sequence input to the LSTM, shape (timelen, batch_size, num_inputs).
cs_prev: Value of the initial cell state.
h_prev: Initial output of cell (to be used for peephole).
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.
i: The input gate over the whole time sequence.
cs: The cell state before the tanh over the whole time sequence.
f: The forget gate over the whole time sequence.
o: The output gate over the whole time sequence.
ci: The cell input over the whole time sequence.
co: The cell after the tanh over the whole time sequence.
h: The output h vector over the whole time sequence.
cs_grad: The current gradient of cs.
h_grad: The gradient of h vector.
use_peephole: Whether to use peephole weights.

Returns: 

x_grad: The gradient of x to be back-propped.
cs_prev_grad: The gradient of cs_prev to be back-propped.
h_prev_grad: The gradient of h_prev to be back-propped.
w_grad: The gradient for w to be back-propped.
wci_grad: The gradient for wci to be back-propped.
wcf_grad: The gradient for wcf to be back-propped.
wco_grad: The gradient for wco to be back-propped.
b_grad: The gradient for w to be back-propped.

func BlockLSTMV2 

func BlockLSTMV2(scope *Scope, seq_len_max tf.Output, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, optional ...BlockLSTMV2Attr) (i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, h tf.Output)

Computes the LSTM cell forward propagation for all the time steps.

This is equivalent to applying LSTMBlockCell in a loop, like so:

```python for x1 in unpack(x): 

i1, cs1, f1, o1, ci1, co1, h1 = LSTMBlock(
  x1, cs_prev, h_prev, w, wci, wcf, wco, b)
cs_prev = cs1
h_prev = h1
i.append(i1)
cs.append(cs1)
f.append(f1)
o.append(o1)
ci.append(ci1)
co.append(co1)
h.append(h1)

return pack(i), pack(cs), pack(f), pack(o), pack(ci), pack(ch), pack(h)

Note that unlike LSTMBlockCell (and BlockLSTM) which uses ICFO gate layout, this op uses IFCO. So in order for the following snippet to be equivalent all gate-related outputs should be reordered. ```

Arguments: 

seq_len_max: Maximum time length actually used by this input. Outputs are padded

with zeros beyond this length. 

x: The sequence input to the LSTM, shape (timelen, batch_size, num_inputs).
cs_prev: Value of the initial cell state.
h_prev: Initial output of cell (to be used for peephole).
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.

Returns: 

i: The input gate over the whole time sequence.
cs: The cell state before the tanh over the whole time sequence.
f: The forget gate over the whole time sequence.
o: The output gate over the whole time sequence.
ci: The cell input over the whole time sequence.
co: The cell after the tanh over the whole time sequence.
h: The output h vector over the whole time sequence.

func BoostedTreesAggregateStats 

func BoostedTreesAggregateStats(scope *Scope, node_ids tf.Output, gradients tf.Output, hessians tf.Output, feature tf.Output, max_splits int64, num_buckets int64) (stats_summary tf.Output)

Aggregates the summary of accumulated stats for the batch.

The summary stats contains gradients and hessians accumulated for each node, feature dimension id and bucket.

Arguments: 

node_ids: int32; Rank 1 Tensor containing node ids for each example, shape [batch_size].
gradients: float32; Rank 2 Tensor (shape=[batch_size, logits_dimension]) with gradients for each example.
hessians: float32; Rank 2 Tensor (shape=[batch_size, hessian_dimension]) with hessians for each example.
feature: int32; Rank 2 feature Tensors (shape=[batch_size, feature_dimension]).
max_splits: int; the maximum number of splits possible in the whole tree.
num_buckets: int; equals to the maximum possible value of bucketized feature.

Returns output Rank 4 Tensor (shape=[splits, feature_dimension, buckets, logits_dimension + hessian_dimension]) containing accumulated stats for each node, feature dimension and bucket.

func BoostedTreesBucketize 

func BoostedTreesBucketize(scope *Scope, float_values []tf.Output, bucket_boundaries []tf.Output) (buckets []tf.Output)

Bucketize each feature based on bucket boundaries.

An op that returns a list of float tensors, where each tensor represents the bucketized values for a single feature.

Arguments: 

float_values: float; List of Rank 1 Tensor each containing float values for a single feature.
bucket_boundaries: float; List of Rank 1 Tensors each containing the bucket boundaries for a single

feature.

Returns int; List of Rank 1 Tensors each containing the bucketized values for a single feature.

func BoostedTreesCalculateBestFeatureSplit 

func BoostedTreesCalculateBestFeatureSplit(scope *Scope, node_id_range tf.Output, stats_summary tf.Output, l1 tf.Output, l2 tf.Output, tree_complexity tf.Output, min_node_weight tf.Output, logits_dimension int64, optional ...BoostedTreesCalculateBestFeatureSplitAttr) (node_ids tf.Output, gains tf.Output, feature_dimensions tf.Output, thresholds tf.Output, left_node_contribs tf.Output, right_node_contribs tf.Output, split_with_default_directions tf.Output)

Calculates gains for each feature and returns the best possible split information for the feature.

The split information is the best threshold (bucket id), gains and left/right node contributions per node for each feature.

It is possible that not all nodes can be split on each feature. Hence, the list of possible nodes can differ between the features. Therefore, we return `node_ids_list` for each feature, containing the list of nodes that this feature can be used to split.

In this manner, the output is the best split per features and per node, so that it needs to be combined later to produce the best split for each node (among all possible features).

The output shapes are compatible in a way that the first dimension of all tensors are the same and equal to the number of possible split nodes for each feature.

Arguments: 

node_id_range: A Rank 1 tensor (shape=[2]) to specify the range [first, last) of node ids to process within `stats_summary_list`. The nodes are iterated between the two nodes specified by the tensor, as like `for node_id in range(node_id_range[0], node_id_range[1])` (Note that the last index node_id_range[1] is exclusive).
stats_summary: A Rank 4 tensor (#shape=[max_splits, feature_dims, bucket, stats_dims]) for accumulated stats summary (gradient/hessian) per node, per dimension, per buckets for each feature.

The first dimension of the tensor is the maximum number of splits, and thus not all elements of it will be used, but only the indexes specified by node_ids will be used. 

l1: l1 regularization factor on leaf weights, per instance based.
l2: l2 regularization factor on leaf weights, per instance based.
tree_complexity: adjustment to the gain, per leaf based.
min_node_weight: minimum avg of hessians in a node before required for the node to be considered for splitting.
logits_dimension: The dimension of logit, i.e., number of classes.

Returns: 

node_ids: A Rank 1 tensors indicating possible split node ids for each feature. The length of the list is num_features, but each tensor has different size as each feature provides different possible nodes. See above for details like shapes and sizes.
gains: A Rank 1 tensors indicating the best gains for each feature to split for certain nodes. See above for details like shapes and sizes.
feature_dimensions: A Rank 1 tensors indicating the best feature dimension for each feature to split for certain nodes if the feature is multi-dimension. See above for details like shapes and sizes.
thresholds: A Rank 1 tensors indicating the bucket id to compare with (as a threshold) for split in each node. See above for details like shapes and sizes.
left_node_contribs: A Rank 2 tensors indicating the contribution of the left nodes when branching from parent nodes (given by the tensor element in the output node_ids_list) to the left direction by the given threshold for each feature. This value will be used to make the left node value by adding to the parent node value. Second dimension size is 1 for 1-dimensional logits, but would be larger for multi-class problems. See above for details like shapes and sizes.
right_node_contribs: A Rank 2 tensors, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.
split_with_default_directions: A Rank 1 tensors indicating the which direction to go if data is missing. See above for details like shapes and sizes.

Inequality with default left returns 0, inequality with default right returns 1, equality with default right returns 2.

func BoostedTreesCalculateBestFeatureSplitV2 

func BoostedTreesCalculateBestFeatureSplitV2(scope *Scope, node_id_range tf.Output, stats_summaries_list []tf.Output, split_types tf.Output, candidate_feature_ids tf.Output, l1 tf.Output, l2 tf.Output, tree_complexity tf.Output, min_node_weight tf.Output, logits_dimension int64) (node_ids tf.Output, gains tf.Output, feature_ids tf.Output, feature_dimensions tf.Output, thresholds tf.Output, left_node_contribs tf.Output, right_node_contribs tf.Output, split_with_default_directions tf.Output)

Calculates gains for each feature and returns the best possible split information for each node. However, if no split is found, then no split information is returned for that node.

The split information is the best threshold (bucket id), gains and left/right node contributions per node for each feature.

It is possible that not all nodes can be split on each feature. Hence, the list of possible nodes can differ between the features. Therefore, we return `node_ids_list` for each feature, containing the list of nodes that this feature can be used to split.

In this manner, the output is the best split per features and per node, so that it needs to be combined later to produce the best split for each node (among all possible features).

The output shapes are compatible in a way that the first dimension of all tensors are the same and equal to the number of possible split nodes for each feature.

Arguments: 

node_id_range: A Rank 1 tensor (shape=[2]) to specify the range [first, last) of node ids to process within `stats_summary_list`. The nodes are iterated between the two nodes specified by the tensor, as like `for node_id in range(node_id_range[0], node_id_range[1])` (Note that the last index node_id_range[1] is exclusive).
stats_summaries_list: A list of Rank 4 tensor (#shape=[max_splits, feature_dims, bucket, stats_dims]) for accumulated stats summary (gradient/hessian) per node, per dimension, per buckets for each feature.

The first dimension of the tensor is the maximum number of splits, and thus not all elements of it will be used, but only the indexes specified by node_ids will be used. 

split_types: A Rank 1 tensor indicating if this Op should perform inequality split or equality split per feature.
candidate_feature_ids: Rank 1 tensor with ids for each feature. This is the real id of the feature.
l1: l1 regularization factor on leaf weights, per instance based.
l2: l2 regularization factor on leaf weights, per instance based.
tree_complexity: adjustment to the gain, per leaf based.
min_node_weight: minimum avg of hessians in a node before required for the node to be considered for splitting.
logits_dimension: The dimension of logit, i.e., number of classes.

Returns: 

node_ids: A Rank 1 tensors indicating possible split node ids for each feature. The length of the list is num_features, but each tensor has different size as each feature provides different possible nodes. See above for details like shapes and sizes.
gains: A Rank 1 tensor indicating the best gains for each feature to split for certain nodes. See above for details like shapes and sizes.
feature_ids: A Rank 1 tensors indicating the best feature id for each node. See above for details like shapes and sizes.
feature_dimensions: A Rank 1 tensors indicating the best feature dimension for each feature to split for certain nodes if the feature is multi-dimension. See above for details like shapes and sizes.
thresholds: A Rank 1 tensors indicating the bucket id to compare with (as a threshold) for split in each node. See above for details like shapes and sizes.
left_node_contribs: A Rank 2 tensors indicating the contribution of the left nodes when branching from parent nodes (given by the tensor element in the output node_ids_list) to the left direction by the given threshold for each feature. This value will be used to make the left node value by adding to the parent node value. Second dimension size is 1 for 1-dimensional logits, but would be larger for multi-class problems. See above for details like shapes and sizes.
right_node_contribs: A Rank 2 tensors, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.
split_with_default_directions: A Rank 1 tensors indicating the which direction to go if data is missing. See above for details like shapes and sizes.

Inequality with default left returns 0, inequality with default right returns 1, equality with default right returns 2.

func BoostedTreesCalculateBestGainsPerFeature 

func BoostedTreesCalculateBestGainsPerFeature(scope *Scope, node_id_range tf.Output, stats_summary_list []tf.Output, l1 tf.Output, l2 tf.Output, tree_complexity tf.Output, min_node_weight tf.Output, max_splits int64) (node_ids_list []tf.Output, gains_list []tf.Output, thresholds_list []tf.Output, left_node_contribs_list []tf.Output, right_node_contribs_list []tf.Output)

Calculates gains for each feature and returns the best possible split information for the feature.

The split information is the best threshold (bucket id), gains and left/right node contributions per node for each feature.

It is possible that not all nodes can be split on each feature. Hence, the list of possible nodes can differ between the features. Therefore, we return `node_ids_list` for each feature, containing the list of nodes that this feature can be used to split.

In this manner, the output is the best split per features and per node, so that it needs to be combined later to produce the best split for each node (among all possible features).

The length of output lists are all of the same length, `num_features`. The output shapes are compatible in a way that the first dimension of all tensors of all lists are the same and equal to the number of possible split nodes for each feature.

Arguments: 

node_id_range: A Rank 1 tensor (shape=[2]) to specify the range [first, last) of node ids to process within `stats_summary_list`. The nodes are iterated between the two nodes specified by the tensor, as like `for node_id in range(node_id_range[0], node_id_range[1])` (Note that the last index node_id_range[1] is exclusive).
stats_summary_list: A list of Rank 3 tensor (#shape=[max_splits, bucket, 2]) for accumulated stats summary (gradient/hessian) per node per buckets for each feature. The first dimension of the tensor is the maximum number of splits, and thus not all elements of it will be used, but only the indexes specified by node_ids will be used.
l1: l1 regularization factor on leaf weights, per instance based.
l2: l2 regularization factor on leaf weights, per instance based.
tree_complexity: adjustment to the gain, per leaf based.
min_node_weight: minimum avg of hessians in a node before required for the node to be considered for splitting.
max_splits: the number of nodes that can be split in the whole tree. Used as a dimension of output tensors.

Returns: 

node_ids_list: An output list of Rank 1 tensors indicating possible split node ids for each feature. The length of the list is num_features, but each tensor has different size as each feature provides different possible nodes. See above for details like shapes and sizes.
gains_list: An output list of Rank 1 tensors indicating the best gains for each feature to split for certain nodes. See above for details like shapes and sizes.
thresholds_list: An output list of Rank 1 tensors indicating the bucket id to compare with (as a threshold) for split in each node. See above for details like shapes and sizes.
left_node_contribs_list: A list of Rank 2 tensors indicating the contribution of the left nodes when branching from parent nodes (given by the tensor element in the output node_ids_list) to the left direction by the given threshold for each feature. This value will be used to make the left node value by adding to the parent node value. Second dimension size is 1 for 1-dimensional logits, but would be larger for multi-class problems. See above for details like shapes and sizes.
right_node_contribs_list: A list of Rank 2 tensors, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.

func BoostedTreesCenterBias 

func BoostedTreesCenterBias(scope *Scope, tree_ensemble_handle tf.Output, mean_gradients tf.Output, mean_hessians tf.Output, l1 tf.Output, l2 tf.Output) (continue_centering tf.Output)

Calculates the prior from the training data (the bias) and fills in the first node with the logits' prior. Returns a boolean indicating whether to continue centering.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble.
mean_gradients: A tensor with shape=[logits_dimension] with mean of gradients for a first node.
mean_hessians: A tensor with shape=[logits_dimension] mean of hessians for a first node.
l1: l1 regularization factor on leaf weights, per instance based.
l2: l2 regularization factor on leaf weights, per instance based.

Returns Bool, whether to continue bias centering.

func BoostedTreesCreateEnsemble 

func BoostedTreesCreateEnsemble(scope *Scope, tree_ensemble_handle tf.Output, stamp_token tf.Output, tree_ensemble_serialized tf.Output) (o *tf.Operation)

Creates a tree ensemble model and returns a handle to it.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble resource to be created.
stamp_token: Token to use as the initial value of the resource stamp.
tree_ensemble_serialized: Serialized proto of the tree ensemble.

Returns the created operation.

func BoostedTreesCreateQuantileStreamResource 

func BoostedTreesCreateQuantileStreamResource(scope *Scope, quantile_stream_resource_handle tf.Output, epsilon tf.Output, num_streams tf.Output, optional ...BoostedTreesCreateQuantileStreamResourceAttr) (o *tf.Operation)

Create the Resource for Quantile Streams.

Arguments: 

quantile_stream_resource_handle: resource; Handle to quantile stream resource.
epsilon: float; The required approximation error of the stream resource.
num_streams: int; The number of streams managed by the resource that shares the same epsilon.

Returns the created operation.

func BoostedTreesDeserializeEnsemble 

func BoostedTreesDeserializeEnsemble(scope *Scope, tree_ensemble_handle tf.Output, stamp_token tf.Output, tree_ensemble_serialized tf.Output) (o *tf.Operation)

Deserializes a serialized tree ensemble config and replaces current tree

ensemble.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble.
stamp_token: Token to use as the new value of the resource stamp.
tree_ensemble_serialized: Serialized proto of the ensemble.

Returns the created operation.

func BoostedTreesEnsembleResourceHandleOp 

func BoostedTreesEnsembleResourceHandleOp(scope *Scope, optional ...BoostedTreesEnsembleResourceHandleOpAttr) (resource tf.Output)

Creates a handle to a BoostedTreesEnsembleResource

func BoostedTreesExampleDebugOutputs 

func BoostedTreesExampleDebugOutputs(scope *Scope, tree_ensemble_handle tf.Output, bucketized_features []tf.Output, logits_dimension int64) (examples_debug_outputs_serialized tf.Output)

Debugging/model interpretability outputs for each example.

It traverses all the trees and computes debug metrics for individual examples, such as getting split feature ids and logits after each split along the decision path used to compute directional feature contributions.

Arguments: 

bucketized_features: A list of rank 1 Tensors containing bucket id for each

feature. 

logits_dimension: scalar, dimension of the logits, to be used for constructing the protos in

examples_debug_outputs_serialized.

Returns Output rank 1 Tensor containing a proto serialized as a string for each example.

func BoostedTreesFlushQuantileSummaries 

func BoostedTreesFlushQuantileSummaries(scope *Scope, quantile_stream_resource_handle tf.Output, num_features int64) (summaries []tf.Output)

Flush the quantile summaries from each quantile stream resource.

An op that outputs a list of quantile summaries of a quantile stream resource. Each summary Tensor is rank 2, containing summaries (value, weight, min_rank, max_rank) for a single feature.

Arguments: 

quantile_stream_resource_handle: resource handle referring to a QuantileStreamResource.

func BoostedTreesGetEnsembleStates 

func BoostedTreesGetEnsembleStates(scope *Scope, tree_ensemble_handle tf.Output) (stamp_token tf.Output, num_trees tf.Output, num_finalized_trees tf.Output, num_attempted_layers tf.Output, last_layer_nodes_range tf.Output)

Retrieves the tree ensemble resource stamp token, number of trees and growing statistics.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble.

Returns: 

stamp_token: Stamp token of the tree ensemble resource.
num_trees: The number of trees in the tree ensemble resource.
num_finalized_trees: The number of trees that were finished successfully.
num_attempted_layers: The number of layers we attempted to build (but not necessarily succeeded).
last_layer_nodes_range: Rank size 2 tensor that contains start and end ids of the nodes in the latest

layer.

func BoostedTreesMakeQuantileSummaries 

func BoostedTreesMakeQuantileSummaries(scope *Scope, float_values []tf.Output, example_weights tf.Output, epsilon tf.Output) (summaries []tf.Output)

Makes the summary of quantiles for the batch.

An op that takes a list of tensors (one tensor per feature) and outputs the quantile summaries for each tensor.

Arguments: 

float_values: float; List of Rank 1 Tensors each containing values for a single feature.
example_weights: float; Rank 1 Tensor with weights per instance.
epsilon: float; The required maximum approximation error.

Returns float; List of Rank 2 Tensors each containing the quantile summary (value, weight, min_rank, max_rank) of a single feature.

func BoostedTreesMakeStatsSummary 

func BoostedTreesMakeStatsSummary(scope *Scope, node_ids tf.Output, gradients tf.Output, hessians tf.Output, bucketized_features_list []tf.Output, max_splits int64, num_buckets int64) (stats_summary tf.Output)

Makes the summary of accumulated stats for the batch.

The summary stats contains gradients and hessians accumulated into the corresponding node and bucket for each example.

Arguments: 

node_ids: int32 Rank 1 Tensor containing node ids, which each example falls into for the requested layer.
gradients: float32; Rank 2 Tensor (shape=[#examples, 1]) for gradients.
hessians: float32; Rank 2 Tensor (shape=[#examples, 1]) for hessians.
bucketized_features_list: int32 list of Rank 1 Tensors, each containing the bucketized feature (for each feature column).
max_splits: int; the maximum number of splits possible in the whole tree.
num_buckets: int; equals to the maximum possible value of bucketized feature.

Returns output Rank 4 Tensor (shape=[#features, #splits, #buckets, 2]) containing accumulated stats put into the corresponding node and bucket. The first index of 4th dimension refers to gradients, and the second to hessians.

func BoostedTreesPredict 

func BoostedTreesPredict(scope *Scope, tree_ensemble_handle tf.Output, bucketized_features []tf.Output, logits_dimension int64) (logits tf.Output)

Runs multiple additive regression ensemble predictors on input instances and

computes the logits. It is designed to be used during prediction. It traverses all the trees and calculates the final score for each instance.

Arguments: 

bucketized_features: A list of rank 1 Tensors containing bucket id for each

feature. 

logits_dimension: scalar, dimension of the logits, to be used for partial logits

shape.

Returns Output rank 2 Tensor containing logits for each example.

func BoostedTreesQuantileStreamResourceAddSummaries 

func BoostedTreesQuantileStreamResourceAddSummaries(scope *Scope, quantile_stream_resource_handle tf.Output, summaries []tf.Output) (o *tf.Operation)

Add the quantile summaries to each quantile stream resource.

An op that adds a list of quantile summaries to a quantile stream resource. Each summary Tensor is rank 2, containing summaries (value, weight, min_rank, max_rank) for a single feature.

Arguments: 

quantile_stream_resource_handle: resource handle referring to a QuantileStreamResource.
summaries: string; List of Rank 2 Tensor each containing the summaries for a single feature.

Returns the created operation.

func BoostedTreesQuantileStreamResourceDeserialize 

func BoostedTreesQuantileStreamResourceDeserialize(scope *Scope, quantile_stream_resource_handle tf.Output, bucket_boundaries []tf.Output) (o *tf.Operation)

Deserialize bucket boundaries and ready flag into current QuantileAccumulator.

An op that deserializes bucket boundaries and are boundaries ready flag into current QuantileAccumulator.

Arguments: 

quantile_stream_resource_handle: resource handle referring to a QuantileStreamResource.
bucket_boundaries: float; List of Rank 1 Tensors each containing the bucket boundaries for a feature.

Returns the created operation.

func BoostedTreesQuantileStreamResourceFlush 

func BoostedTreesQuantileStreamResourceFlush(scope *Scope, quantile_stream_resource_handle tf.Output, num_buckets tf.Output, optional ...BoostedTreesQuantileStreamResourceFlushAttr) (o *tf.Operation)

Flush the summaries for a quantile stream resource.

An op that flushes the summaries for a quantile stream resource.

Arguments: 

quantile_stream_resource_handle: resource handle referring to a QuantileStreamResource.
num_buckets: int; approximate number of buckets unless using generate_quantiles.

Returns the created operation.

func BoostedTreesQuantileStreamResourceGetBucketBoundaries 

func BoostedTreesQuantileStreamResourceGetBucketBoundaries(scope *Scope, quantile_stream_resource_handle tf.Output, num_features int64) (bucket_boundaries []tf.Output)

Generate the bucket boundaries for each feature based on accumulated summaries.

An op that returns a list of float tensors for a quantile stream resource. Each tensor is Rank 1 containing bucket boundaries for a single feature.

Arguments: 

quantile_stream_resource_handle: resource handle referring to a QuantileStreamResource.
num_features: inferred int; number of features to get bucket boundaries for.

Returns float; List of Rank 1 Tensors each containing the bucket boundaries for a feature.

func BoostedTreesQuantileStreamResourceHandleOp 

func BoostedTreesQuantileStreamResourceHandleOp(scope *Scope, optional ...BoostedTreesQuantileStreamResourceHandleOpAttr) (resource tf.Output)

Creates a handle to a BoostedTreesQuantileStreamResource.

func BoostedTreesSerializeEnsemble 

func BoostedTreesSerializeEnsemble(scope *Scope, tree_ensemble_handle tf.Output) (stamp_token tf.Output, tree_ensemble_serialized tf.Output)

Serializes the tree ensemble to a proto.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble.

Returns: 

stamp_token: Stamp token of the tree ensemble resource.
tree_ensemble_serialized: Serialized proto of the ensemble.

func BoostedTreesSparseAggregateStats 

func BoostedTreesSparseAggregateStats(scope *Scope, node_ids tf.Output, gradients tf.Output, hessians tf.Output, feature_indices tf.Output, feature_values tf.Output, feature_shape tf.Output, max_splits int64, num_buckets int64) (stats_summary_indices tf.Output, stats_summary_values tf.Output, stats_summary_shape tf.Output)

Aggregates the summary of accumulated stats for the batch.

The summary stats contains gradients and hessians accumulated for each node, bucket and dimension id.

Arguments: 

node_ids: int32; Rank 1 Tensor containing node ids for each example, shape [batch_size].
gradients: float32; Rank 2 Tensor (shape=[batch_size, logits_dimension]) with gradients for each example.
hessians: float32; Rank 2 Tensor (shape=[batch_size, hessian_dimension]) with hessians for each example.
feature_indices: int32; Rank 2 indices of feature sparse Tensors (shape=[number of sparse entries, 2]).

Number of sparse entries across all instances from the batch. The first value is the index of the instance, the second is dimension of the feature. The second axis can only have 2 values, i.e., the input dense version of Tensor can only be matrix. 

feature_values: int32; Rank 1 values of feature sparse Tensors (shape=[number of sparse entries]).

Number of sparse entries across all instances from the batch. The first value is the index of the instance, the second is dimension of the feature. 

feature_shape: int32; Rank 1 dense shape of feature sparse Tensors (shape=[2]).

The first axis can only have 2 values, [batch_size, feature_dimension]. 

max_splits: int; the maximum number of splits possible in the whole tree.
num_buckets: int; equals to the maximum possible value of bucketized feature + 1.

Returns: 

stats_summary_indices: int32; Rank 2 indices of summary sparse Tensors (shape=[number of non zero statistics, 4])

The second axis can only be 4 including node id, feature dimension, bucket id, and statistics_dimension. statistics_dimension = logits_dimension + hessian_dimension. 

stats_summary_values: output Rank 1 Tensor (shape=[number of non zero statistics])
stats_summary_shape: output Rank 1 Tensor (shape=[4])

The tensor has following 4 values: [max_splits, feature_dimension, num_buckets, statistics_dimension], where statistics_dimension = gradient_dimension + hessian_dimension. gradient_dimension is the same as label_dimension, i.e., the output space. hessian_dimension can be the same as logits dimension when diagonal hessian is used, or label_dimension^2 when full hessian is used.

func BoostedTreesSparseCalculateBestFeatureSplit 

func BoostedTreesSparseCalculateBestFeatureSplit(scope *Scope, node_id_range tf.Output, stats_summary_indices tf.Output, stats_summary_values tf.Output, stats_summary_shape tf.Output, l1 tf.Output, l2 tf.Output, tree_complexity tf.Output, min_node_weight tf.Output, logits_dimension int64, optional ...BoostedTreesSparseCalculateBestFeatureSplitAttr) (node_ids tf.Output, gains tf.Output, feature_dimensions tf.Output, thresholds tf.Output, left_node_contribs tf.Output, right_node_contribs tf.Output, split_with_default_directions tf.Output)

Calculates gains for each feature and returns the best possible split information for the feature.

The split information is the best threshold (bucket id), gains and left/right node contributions per node for each feature.

It is possible that not all nodes can be split on each feature. Hence, the list of possible nodes can differ between the features. Therefore, we return `node_ids_list` for each feature, containing the list of nodes that this feature can be used to split.

In this manner, the output is the best split per features and per node, so that it needs to be combined later to produce the best split for each node (among all possible features).

The output shapes are compatible in a way that the first dimension of all tensors are the same and equal to the number of possible split nodes for each feature.

Arguments: 

node_id_range: A Rank 1 tensor (shape=[2]) to specify the range [first, last) of node ids to process within `stats_summary_list`. The nodes are iterated between the two nodes specified by the tensor, as like `for node_id in range(node_id_range[0], node_id_range[1])` (Note that the last index node_id_range[1] is exclusive).
stats_summary_indices: A Rank 2 int64 tensor of dense shape [N, 4] (N specifies the number of non-zero values) for accumulated stats summary (gradient/hessian) per node per bucket for each feature. The second dimension contains node id, feature dimension, bucket id, and stats dim.

stats dim is the sum of logits dimension and hessian dimension, hessian dimension can either be logits dimension if diagonal hessian is used, or logits dimension^2 if full hessian is used. 

stats_summary_values: A Rank 1 float tensor of dense shape [N] (N specifies the number of non-zero values), which supplies the values for each element in summary_indices.
stats_summary_shape: A Rank 1 float tensor of dense shape [4], which specifies the dense shape of the sparse tensor, which is [num tree nodes, feature dimensions, num buckets, stats dim].
l1: l1 regularization factor on leaf weights, per instance based.
l2: l2 regularization factor on leaf weights, per instance based.
tree_complexity: adjustment to the gain, per leaf based.
min_node_weight: minimum avg of hessians in a node before required for the node to be considered for splitting.
logits_dimension: The dimension of logit, i.e., number of classes.

Returns: 

node_ids: A Rank 1 tensor indicating possible node ids that can be split.
gains: A Rank 1 tensor indicating the best gains to split each node.
feature_dimensions: A Rank 1 tensor indicating the best feature dimension for each feature to split for each node.
thresholds: A Rank 1 tensor indicating the bucket id to compare with (as a threshold) for split in each node.
left_node_contribs: A Rank 2 tensor indicating the contribution of the left nodes when branching from parent nodes to the left direction by the given threshold for each feature.

This value will be used to make the left node value by adding to the parent node value. Second dimension size is logits dimension. 

right_node_contribs: A Rank 2 tensor, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.
split_with_default_directions: A Rank 1 tensor indicating which direction to go if data is missing.

Inequality with default left returns 0, inequality with default right returns 1, equality with default right returns 2.

func BoostedTreesTrainingPredict 

func BoostedTreesTrainingPredict(scope *Scope, tree_ensemble_handle tf.Output, cached_tree_ids tf.Output, cached_node_ids tf.Output, bucketized_features []tf.Output, logits_dimension int64) (partial_logits tf.Output, tree_ids tf.Output, node_ids tf.Output)

Runs multiple additive regression ensemble predictors on input instances and

computes the update to cached logits. It is designed to be used during training. It traverses the trees starting from cached tree id and cached node id and calculates the updates to be pushed to the cache.

Arguments: 

cached_tree_ids: Rank 1 Tensor containing cached tree ids which is the starting

tree of prediction. 

cached_node_ids: Rank 1 Tensor containing cached node id which is the starting

node of prediction. 

bucketized_features: A list of rank 1 Tensors containing bucket id for each

feature. 

logits_dimension: scalar, dimension of the logits, to be used for partial logits

shape.

Returns: 

partial_logits: Rank 2 Tensor containing logits update (with respect to cached

values stored) for each example. 

tree_ids: Rank 1 Tensor containing new tree ids for each example.
node_ids: Rank 1 Tensor containing new node ids in the new tree_ids.

func BoostedTreesUpdateEnsemble 

func BoostedTreesUpdateEnsemble(scope *Scope, tree_ensemble_handle tf.Output, feature_ids tf.Output, node_ids []tf.Output, gains []tf.Output, thresholds []tf.Output, left_node_contribs []tf.Output, right_node_contribs []tf.Output, max_depth tf.Output, learning_rate tf.Output, pruning_mode int64) (o *tf.Operation)

Updates the tree ensemble by either adding a layer to the last tree being grown

or by starting a new tree.

Arguments: 

tree_ensemble_handle: Handle to the ensemble variable.
feature_ids: Rank 1 tensor with ids for each feature. This is the real id of

the feature that will be used in the split. 

node_ids: List of rank 1 tensors representing the nodes for which this feature

has a split. 

gains: List of rank 1 tensors representing the gains for each of the feature's

split. 

thresholds: List of rank 1 tensors representing the thesholds for each of the

feature's split. 

left_node_contribs: List of rank 2 tensors with left leaf contribs for each of

the feature's splits. Will be added to the previous node values to constitute the values of the left nodes. 

right_node_contribs: List of rank 2 tensors with right leaf contribs for each

of the feature's splits. Will be added to the previous node values to constitute the values of the right nodes. 

max_depth: Max depth of the tree to build.
learning_rate: shrinkage const for each new tree.
pruning_mode: 0-No pruning, 1-Pre-pruning, 2-Post-pruning.

Returns the created operation.

func BoostedTreesUpdateEnsembleV2 

func BoostedTreesUpdateEnsembleV2(scope *Scope, tree_ensemble_handle tf.Output, feature_ids []tf.Output, dimension_ids []tf.Output, node_ids []tf.Output, gains []tf.Output, thresholds []tf.Output, left_node_contribs []tf.Output, right_node_contribs []tf.Output, split_types []tf.Output, max_depth tf.Output, learning_rate tf.Output, pruning_mode tf.Output, optional ...BoostedTreesUpdateEnsembleV2Attr) (o *tf.Operation)

Updates the tree ensemble by adding a layer to the last tree being grown

or by starting a new tree.

Arguments: 

tree_ensemble_handle: Handle to the ensemble variable.
feature_ids: Rank 1 tensor with ids for each feature. This is the real id of

the feature that will be used in the split. 

dimension_ids: List of rank 1 tensors representing the dimension in each feature.
node_ids: List of rank 1 tensors representing the nodes for which this feature

has a split. 

gains: List of rank 1 tensors representing the gains for each of the feature's

split. 

thresholds: List of rank 1 tensors representing the thesholds for each of the

feature's split. 

left_node_contribs: List of rank 2 tensors with left leaf contribs for each of

the feature's splits. Will be added to the previous node values to constitute the values of the left nodes. 

right_node_contribs: List of rank 2 tensors with right leaf contribs for each

of the feature's splits. Will be added to the previous node values to constitute the values of the right nodes. 

split_types: List of rank 1 tensors representing the split type for each feature.
max_depth: Max depth of the tree to build.
learning_rate: shrinkage const for each new tree.
pruning_mode: 0-No pruning, 1-Pre-pruning, 2-Post-pruning.

Returns the created operation.

func BroadcastArgs 

func BroadcastArgs(scope *Scope, s0 tf.Output, s1 tf.Output) (r0 tf.Output)

Return the shape of s0 op s1 with broadcast.

Given `s0` and `s1`, tensors that represent shapes, compute `r0`, the broadcasted shape. `s0`, `s1` and `r0` are all integer vectors.

func BroadcastGradientArgs 

func BroadcastGradientArgs(scope *Scope, s0 tf.Output, s1 tf.Output) (r0 tf.Output, r1 tf.Output)

Return the reduction indices for computing gradients of s0 op s1 with broadcast.

This is typically used by gradient computations for a broadcasting operation.

func BroadcastTo 

func BroadcastTo(scope *Scope, input tf.Output, shape tf.Output) (output tf.Output)

Broadcast an array for a compatible shape.

Broadcasting is the process of making arrays to have compatible shapes for arithmetic operations. Two shapes are compatible if for each dimension pair they are either equal or one of them is one.

For example:

>>> x = tf.constant([[1, 2, 3]]) # Shape (1, 3,) >>> y = tf.broadcast_to(x, [2, 3]) >>> print(y) tf.Tensor( 

[[1 2 3]
 [1 2 3]], shape=(2, 3), dtype=int32)

In the above example, the input Tensor with the shape of `[1, 3]` is broadcasted to output Tensor with shape of `[2, 3]`.

When broadcasting, if a tensor has fewer axes than necessary its shape is padded on the left with ones. So this gives the same result as the previous example:

>>> x = tf.constant([1, 2, 3]) # Shape (3,) >>> y = tf.broadcast_to(x, [2, 3])

When doing broadcasted operations such as multiplying a tensor by a scalar, broadcasting (usually) confers some time or space benefit, as the broadcasted tensor is never materialized.

However, `broadcast_to` does not carry with it any such benefits. The newly-created tensor takes the full memory of the broadcasted shape. (In a graph context, `broadcast_to` might be fused to subsequent operation and then be optimized away, however.)

Arguments: 

input: A Tensor to broadcast.
shape: An 1-D `int` Tensor. The shape of the desired output.

Returns A Tensor.

func Bucketize 

func Bucketize(scope *Scope, input tf.Output, boundaries []float32) (output tf.Output)

Bucketizes 'input' based on 'boundaries'.

For example, if the inputs are 

boundaries = [0, 10, 100]
input = [[-5, 10000]
         [150,   10]
         [5,    100]]

then the output will be 

output = [[0, 3]
          [3, 2]
          [1, 3]]

Arguments: 

input: Any shape of Tensor contains with int or float type.
boundaries: A sorted list of floats gives the boundary of the buckets.

Returns Same shape with 'input', each value of input replaced with bucket index.

@compatibility(numpy) Equivalent to np.digitize. @end_compatibility

func BytesProducedStatsDataset 

func BytesProducedStatsDataset(scope *Scope, input_dataset tf.Output, tag tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Records the bytes size of each element of `input_dataset` in a StatsAggregator.

func CSRSparseMatrixComponents 

func CSRSparseMatrixComponents(scope *Scope, csr_sparse_matrix tf.Output, index tf.Output, type_ tf.DataType) (row_ptrs tf.Output, col_inds tf.Output, values tf.Output)

Reads out the CSR components at batch `index`.

This op is meant only for debugging / testing, and its interface is not expected to be stable.

Arguments: 

csr_sparse_matrix: A batched CSRSparseMatrix.
index: The index in `csr_sparse_matrix`'s batch.

Returns: 

row_ptrs: An array containing CSR matrix row pointers.
col_inds: An array containing CSR matrix column indices.
values: An array containing CSR matrix nonzero values.

func CSRSparseMatrixToDense 

func CSRSparseMatrixToDense(scope *Scope, sparse_input tf.Output, type_ tf.DataType) (dense_output tf.Output)

Convert a (possibly batched) CSRSparseMatrix to dense.

Arguments: 

sparse_input: A batched CSRSparseMatrix.

Returns A dense tensor.

func CSRSparseMatrixToSparseTensor 

func CSRSparseMatrixToSparseTensor(scope *Scope, sparse_matrix tf.Output, type_ tf.DataType) (indices tf.Output, values tf.Output, dense_shape tf.Output)

Converts a (possibly batched) CSRSparesMatrix to a SparseTensor.

Arguments: 

sparse_matrix: A (possibly batched) CSRSparseMatrix.

Returns: 

indices: SparseTensor indices.
values: SparseTensor values.
dense_shape: SparseTensor dense shape.

func CTCBeamSearchDecoder 

func CTCBeamSearchDecoder(scope *Scope, inputs tf.Output, sequence_length tf.Output, beam_width int64, top_paths int64, optional ...CTCBeamSearchDecoderAttr) (decoded_indices []tf.Output, decoded_values []tf.Output, decoded_shape []tf.Output, log_probability tf.Output)

Performs beam search decoding on the logits given in input.

A note about the attribute merge_repeated: For the beam search decoder, this means that if consecutive entries in a beam are the same, only the first of these is emitted. That is, when the top path is "A B B B B", "A B" is returned if merge_repeated = True but "A B B B B" is returned if merge_repeated = False.

Arguments: 

inputs: 3-D, shape: `(max_time x batch_size x num_classes)`, the logits.
sequence_length: A vector containing sequence lengths, size `(batch)`.
beam_width: A scalar >= 0 (beam search beam width).
top_paths: A scalar >= 0, <= beam_width (controls output size).

Returns: 

decoded_indices: A list (length: top_paths) of indices matrices.  Matrix j,

size `(total_decoded_outputs[j] x 2)`, has indices of a `SparseTensor<int64, 2>`. The rows store: [batch, time]. 

decoded_values: A list (length: top_paths) of values vectors.  Vector j,

size `(length total_decoded_outputs[j])`, has the values of a `SparseTensor<int64, 2>`. The vector stores the decoded classes for beam j. 

decoded_shape: A list (length: top_paths) of shape vector.  Vector j,

size `(2)`, stores the shape of the decoded `SparseTensor[j]`. Its values are: `[batch_size, max_decoded_length[j]]`. 

log_probability: A matrix, shaped: `(batch_size x top_paths)`.  The

sequence log-probabilities.

func CTCGreedyDecoder 

func CTCGreedyDecoder(scope *Scope, inputs tf.Output, sequence_length tf.Output, optional ...CTCGreedyDecoderAttr) (decoded_indices tf.Output, decoded_values tf.Output, decoded_shape tf.Output, log_probability tf.Output)

Performs greedy decoding on the logits given in inputs.

A note about the attribute merge_repeated: if enabled, when consecutive logits' maximum indices are the same, only the first of these is emitted. Labeling the blank '*', the sequence "A B B * B B" becomes "A B B" if merge_repeated = True and "A B B B B" if merge_repeated = False.

Regardless of the value of merge_repeated, if the maximum index of a given time and batch corresponds to the blank, index `(num_classes - 1)`, no new element is emitted.

Arguments: 

inputs: 3-D, shape: `(max_time x batch_size x num_classes)`, the logits.
sequence_length: A vector containing sequence lengths, size `(batch_size)`.

Returns: 

decoded_indices: Indices matrix, size `(total_decoded_outputs x 2)`,

of a `SparseTensor<int64, 2>`. The rows store: [batch, time]. 

decoded_values: Values vector, size: `(total_decoded_outputs)`,

of a `SparseTensor<int64, 2>`. The vector stores the decoded classes. 

decoded_shape: Shape vector, size `(2)`, of the decoded SparseTensor.

Values are: `[batch_size, max_decoded_length]`. 

log_probability: Matrix, size `(batch_size x 1)`, containing sequence

log-probabilities.

func CTCLoss 

func CTCLoss(scope *Scope, inputs tf.Output, labels_indices tf.Output, labels_values tf.Output, sequence_length tf.Output, optional ...CTCLossAttr) (loss tf.Output, gradient tf.Output)

Calculates the CTC Loss (log probability) for each batch entry. Also calculates

the gradient. This class performs the softmax operation for you, so inputs should be e.g. linear projections of outputs by an LSTM.

Arguments: 

inputs: 3-D, shape: `(max_time x batch_size x num_classes)`, the logits.
labels_indices: The indices of a `SparseTensor<int32, 2>`.

`labels_indices(i, :) == [b, t]` means `labels_values(i)` stores the id for `(batch b, time t)`. 

labels_values: The values (labels) associated with the given batch and time.
sequence_length: A vector containing sequence lengths (batch).

Returns: 

loss: A vector (batch) containing log-probabilities.
gradient: The gradient of `loss`.  3-D, shape:

`(max_time x batch_size x num_classes)`.

func CTCLossV2 

func CTCLossV2(scope *Scope, inputs tf.Output, labels_indices tf.Output, labels_values tf.Output, sequence_length tf.Output, optional ...CTCLossV2Attr) (loss tf.Output, gradient tf.Output)

Calculates the CTC Loss (log probability) for each batch entry. Also calculates

the gradient. This class performs the softmax operation for you, so inputs should be e.g. linear projections of outputs by an LSTM.

Arguments: 

inputs: 3-D, shape: `(max_time x batch_size x num_classes)`, the logits. Default blank

label is 0 rather num_classes - 1. 

labels_indices: The indices of a `SparseTensor<int32, 2>`.

`labels_indices(i, :) == [b, t]` means `labels_values(i)` stores the id for `(batch b, time t)`. 

labels_values: The values (labels) associated with the given batch and time.
sequence_length: A vector containing sequence lengths (batch).

Returns: 

loss: A vector (batch) containing log-probabilities.
gradient: The gradient of `loss`.  3-D, shape:

`(max_time x batch_size x num_classes)`.

func CacheDataset 

func CacheDataset(scope *Scope, input_dataset tf.Output, filename tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...CacheDatasetAttr) (handle tf.Output)

Creates a dataset that caches elements from `input_dataset`.

A CacheDataset will iterate over the input_dataset, and store tensors. If the cache already exists, the cache will be used. If the cache is inappropriate (e.g. cannot be opened, contains tensors of the wrong shape / size), an error will the returned when used.

Arguments: 

filename: A path on the filesystem where we should cache the dataset. Note: this

will be a directory.

func Cast 

func Cast(scope *Scope, x tf.Output, DstT tf.DataType, optional ...CastAttr) (y tf.Output)

Cast x of type SrcT to y of DstT.

func Ceil 

func Ceil(scope *Scope, x tf.Output) (y tf.Output)

Returns element-wise smallest integer not less than x.

func CheckNumerics 

func CheckNumerics(scope *Scope, tensor tf.Output, message string) (output tf.Output)

Checks a tensor for NaN and Inf values.

When run, reports an `InvalidArgument` error if `tensor` has any values that are not a number (NaN) or infinity (Inf). Otherwise, returns the input tensor.

Example usage:

``` python a = tf.Variable(1.0) tf.debugging.check_numerics(a, message=”)

b = tf.Variable(np.nan) try: 

tf.debugging.check_numerics(b, message='Checking b')

except Exception as e: 

assert "Checking b : Tensor had NaN values" in e.message

c = tf.Variable(np.inf) try: 

tf.debugging.check_numerics(c, message='Checking c')

except Exception as e: 

assert "Checking c : Tensor had Inf values" in e.message

```

Arguments: 

message: Prefix of the error message.

func CheckNumericsV2 

func CheckNumericsV2(scope *Scope, tensor tf.Output, message string) (output tf.Output)

Checks a tensor for NaN, -Inf and +Inf values.

When run, reports an `InvalidArgument` error if `tensor` has any values that are not a number (NaN) or infinity (Inf). Otherwise, returns the input tensor. Unlike CheckNumerics (V1), CheckNumericsV2 distinguishes -Inf and +Inf in the errors it throws.

Arguments: 

message: Prefix of the error message.

func Cholesky 

func Cholesky(scope *Scope, input tf.Output) (output tf.Output)

Computes the Cholesky decomposition of one or more square matrices.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices.

The input has to be symmetric and positive definite. Only the lower-triangular part of the input will be used for this operation. The upper-triangular part will not be read.

The output is a tensor of the same shape as the input containing the Cholesky decompositions for all input submatrices `[..., :, :]`.

**Note**: The gradient computation on GPU is faster for large matrices but not for large batch dimensions when the submatrices are small. In this case it might be faster to use the CPU.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[..., M, M]`.

func CholeskyGrad 

func CholeskyGrad(scope *Scope, l tf.Output, grad tf.Output) (output tf.Output)

Computes the reverse mode backpropagated gradient of the Cholesky algorithm.

For an explanation see "Differentiation of the Cholesky algorithm" by Iain Murray http://arxiv.org/abs/1602.07527.

Arguments: 

l: Output of batch Cholesky algorithm l = cholesky(A). Shape is `[..., M, M]`.

Algorithm depends only on lower triangular part of the innermost matrices of this tensor. 

grad: df/dl where f is some scalar function. Shape is `[..., M, M]`.

Algorithm depends only on lower triangular part of the innermost matrices of this tensor.

Returns Symmetrized version of df/dA . Shape is `[..., M, M]`

func ClipByValue 

func ClipByValue(scope *Scope, t tf.Output, clip_value_min tf.Output, clip_value_max tf.Output) (output tf.Output)

Clips tensor values to a specified min and max.

Given a tensor `t`, this operation returns a tensor of the same type and shape as `t` with its values clipped to `clip_value_min` and `clip_value_max`. Any values less than `clip_value_min` are set to `clip_value_min`. Any values greater than `clip_value_max` are set to `clip_value_max`.

Arguments: 

t: A `Tensor`.
clip_value_min: A 0-D (scalar) `Tensor`, or a `Tensor` with the same shape

as `t`. The minimum value to clip by. 

clip_value_max: A 0-D (scalar) `Tensor`, or a `Tensor` with the same shape

as `t`. The maximum value to clip by.

Returns A clipped `Tensor` with the same shape as input 't'.

func CollateTPUEmbeddingMemory added in v0.2.0 

func CollateTPUEmbeddingMemory(scope *Scope, memory_configs []tf.Output) (merged_memory_config tf.Output)

An op that merges the string-encoded memory config protos from all hosts.

Arguments: 

memory_configs: String-encoded memory config protos containing metadata about

the memory allocations reserved for TPUEmbedding across all hosts.

func CollectiveAllToAllV2 added in v0.5.0 

func CollectiveAllToAllV2(scope *Scope, input tf.Output, group_size tf.Output, group_key tf.Output, instance_key tf.Output, ordering_token []tf.Output, optional ...CollectiveAllToAllV2Attr) (data tf.Output)

Mutually exchanges multiple tensors of identical type and shape.

`is_stateless` means each op does not need control dependencies to other collective ops. In this case, keys that are unique at runtime (e.g. `instance_key`) should be used to distinguish collective groups.

func CollectiveAllToAllV3 

func CollectiveAllToAllV3(scope *Scope, input tf.Output, communicator tf.Output, group_assignment tf.Output, optional ...CollectiveAllToAllV3Attr) (data tf.Output)

Mutually exchanges multiple tensors of identical type and shape.

func CollectiveAssignGroupV2 

func CollectiveAssignGroupV2(scope *Scope, group_assignment tf.Output, device_index tf.Output, base_key tf.Output) (group_size tf.Output, group_key tf.Output)

Assign group keys based on group assignment.

func CollectiveBcastRecv 

func CollectiveBcastRecv(scope *Scope, T tf.DataType, group_size int64, group_key int64, instance_key int64, shape tf.Shape, optional ...CollectiveBcastRecvAttr) (data tf.Output)

Receives a tensor value broadcast from another device.

func CollectiveBcastRecvV2 

func CollectiveBcastRecvV2(scope *Scope, group_size tf.Output, group_key tf.Output, instance_key tf.Output, shape tf.Output, T tf.DataType, optional ...CollectiveBcastRecvV2Attr) (data tf.Output)

Receives a tensor value broadcast from another device.

func CollectiveBcastSend 

func CollectiveBcastSend(scope *Scope, input tf.Output, group_size int64, group_key int64, instance_key int64, shape tf.Shape, optional ...CollectiveBcastSendAttr) (data tf.Output)

Broadcasts a tensor value to one or more other devices.

func CollectiveBcastSendV2 

func CollectiveBcastSendV2(scope *Scope, input tf.Output, group_size tf.Output, group_key tf.Output, instance_key tf.Output, optional ...CollectiveBcastSendV2Attr) (data tf.Output)

Broadcasts a tensor value to one or more other devices.

func CollectiveGather 

func CollectiveGather(scope *Scope, input tf.Output, group_size int64, group_key int64, instance_key int64, shape tf.Shape, optional ...CollectiveGatherAttr) (data tf.Output)

Mutually accumulates multiple tensors of identical type and shape.

func CollectiveGatherV2 

func CollectiveGatherV2(scope *Scope, input tf.Output, group_size tf.Output, group_key tf.Output, instance_key tf.Output, ordering_token []tf.Output, optional ...CollectiveGatherV2Attr) (data tf.Output)

Mutually accumulates multiple tensors of identical type and shape.

`is_stateless` means each op does not need control dependencies to other collective ops. In this case, keys that are unique at runtime (e.g. `instance_key`) should be used to distinguish collective groups.

func CollectiveInitializeCommunicator 

func CollectiveInitializeCommunicator(scope *Scope, group_key tf.Output, rank tf.Output, group_size tf.Output, optional ...CollectiveInitializeCommunicatorAttr) (communicator tf.Output)

Initializes a group for collective operations.

func CollectivePermute 

func CollectivePermute(scope *Scope, input tf.Output, source_target_pairs tf.Output) (output tf.Output)

An Op to permute tensors across replicated TPU instances.

Each instance supplies its own input.

For example, suppose there are 4 TPU instances: `[A, B, C, D]`. Passing source_target_pairs=`[[0,1],[1,2],[2,3],[3,0]]` gets the outputs: `[D, A, B, C]`.

Arguments: 

input: The local input to be permuted. Currently only supports float and

bfloat16. 

source_target_pairs: A tensor with shape [num_pairs, 2].

Returns The permuted input.

func CollectiveReduce 

func CollectiveReduce(scope *Scope, input tf.Output, group_size int64, group_key int64, instance_key int64, merge_op string, final_op string, subdiv_offsets []int64, optional ...CollectiveReduceAttr) (data tf.Output)

Mutually reduces multiple tensors of identical type and shape.

func CollectiveReduceScatterV2 added in v0.4.0 

func CollectiveReduceScatterV2(scope *Scope, input tf.Output, group_size tf.Output, group_key tf.Output, instance_key tf.Output, ordering_token []tf.Output, merge_op string, final_op string, optional ...CollectiveReduceScatterV2Attr) (data tf.Output)

Mutually reduces multiple tensors of identical type and shape and scatters the result.

`is_stateless` means each op does not need control dependencies to other collective ops. In this case, keys that are unique at runtime (e.g. `instance_key`) should be used to distinguish collective groups.

func CollectiveReduceV2 

func CollectiveReduceV2(scope *Scope, input tf.Output, group_size tf.Output, group_key tf.Output, instance_key tf.Output, ordering_token []tf.Output, merge_op string, final_op string, optional ...CollectiveReduceV2Attr) (data tf.Output)

Mutually reduces multiple tensors of identical type and shape.

`is_stateless` means each op does not need control dependencies to other collective ops. In this case, keys that are unique at runtime (e.g. `instance_key`) should be used to distinguish collective groups.

func CollectiveReduceV3 

func CollectiveReduceV3(scope *Scope, input tf.Output, communicator tf.Output, group_assignment tf.Output, reduction string, optional ...CollectiveReduceV3Attr) (data tf.Output)

Mutually reduces multiple tensors of identical type and shape.

func CombinedNonMaxSuppression 

func CombinedNonMaxSuppression(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size_per_class tf.Output, max_total_size tf.Output, iou_threshold tf.Output, score_threshold tf.Output, optional ...CombinedNonMaxSuppressionAttr) (nmsed_boxes tf.Output, nmsed_scores tf.Output, nmsed_classes tf.Output, valid_detections tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

This operation performs non_max_suppression on the inputs per batch, across all classes. Prunes away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system. Also note that this algorithm is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm. The output of this operation is the final boxes, scores and classes tensor returned after performing non_max_suppression.

Arguments: 

boxes: A 4-D float tensor of shape `[batch_size, num_boxes, q, 4]`. If `q` is 1 then

same boxes are used for all classes otherwise, if `q` is equal to number of classes, class-specific boxes are used. 

scores: A 3-D float tensor of shape `[batch_size, num_boxes, num_classes]`

representing a single score corresponding to each box (each row of boxes). 

max_output_size_per_class: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression per class 

max_total_size: An int32 scalar representing the maximum number of boxes retained over all

classes. Note that setting this value to a large number may result in OOM error depending on the system workload. 

iou_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too much with respect to IOU. 

score_threshold: A 0-D float tensor representing the threshold for deciding when to remove

boxes based on score.

Returns: 

nmsed_boxes: A [batch_size, max_detections, 4] float32 tensor

containing the non-max suppressed boxes. 

nmsed_scores: A [batch_size, max_detections] float32 tensor

containing the scores for the boxes. 

nmsed_classes: A [batch_size, max_detections] float32 tensor

containing the classes for the boxes. 

valid_detections: A [batch_size] int32 tensor indicating the number of

valid detections per batch item. Only the top num_detections[i] entries in nms_boxes[i], nms_scores[i] and nms_class[i] are valid. The rest of the entries are zero paddings.

func Complex 

func Complex(scope *Scope, real tf.Output, imag tf.Output, optional ...ComplexAttr) (out tf.Output)

Converts two real numbers to a complex number.

Given a tensor `real` representing the real part of a complex number, and a tensor `imag` representing the imaginary part of a complex number, this operation returns complex numbers elementwise of the form \\(a + bj\\), where *a* represents the `real` part and *b* represents the `imag` part.

The input tensors `real` and `imag` must have the same shape.

For example:

``` # tensor 'real' is [2.25, 3.25] # tensor `imag` is [4.75, 5.75] tf.complex(real, imag) ==> [[2.25 + 4.75j], [3.25 + 5.75j]] ```

func ComplexAbs 

func ComplexAbs(scope *Scope, x tf.Output, optional ...ComplexAbsAttr) (y tf.Output)

Computes the complex absolute value of a tensor.

Given a tensor `x` of complex numbers, this operation returns a tensor of type `float` or `double` that is the absolute value of each element in `x`. All elements in `x` must be complex numbers of the form \\(a + bj\\). The absolute value is computed as \\( \sqrt{a^2 + b^2}\\).

For example:

>>> x = tf.complex(3.0, 4.0) >>> print((tf.raw_ops.ComplexAbs(x=x, Tout=tf.dtypes.float32, name=None)).numpy()) 5.0

func CompositeTensorVariantFromComponents 

func CompositeTensorVariantFromComponents(scope *Scope, components []tf.Output, metadata string) (encoded tf.Output)

Encodes an `ExtensionType` value into a `variant` scalar Tensor.

Returns a scalar variant tensor containing a single `CompositeTensorVariant` with the specified Tensor components and TypeSpec.

Arguments: 

components: The component tensors for the extension type value.
metadata: String serialization for the TypeSpec.  (Note: the encoding for the TypeSpec

may change in future versions of TensorFlow.)

Returns A `variant` Tensor that containing the encoded value.

func CompositeTensorVariantToComponents 

func CompositeTensorVariantToComponents(scope *Scope, encoded tf.Output, metadata string, Tcomponents []tf.DataType) (components []tf.Output)

Decodes a `variant` scalar Tensor into an `ExtensionType` value.

Returns the Tensor components encoded in a `CompositeTensorVariant`.

Raises an error if `type_spec_proto` doesn't match the TypeSpec in `encoded`.

Arguments: 

encoded: A scalar `variant` Tensor containing an encoded ExtensionType value.
metadata: String serialization for the TypeSpec.  Must be compatible with the

`TypeSpec` contained in `encoded`. (Note: the encoding for the TypeSpec may change in future versions of TensorFlow.) 

Tcomponents: Expected dtypes for components.

Returns The component tensors for the ExtensionType value in `encoded`.

func CompressElement 

func CompressElement(scope *Scope, components []tf.Output) (compressed tf.Output)

Compresses a dataset element.

func ComputeAccidentalHits 

func ComputeAccidentalHits(scope *Scope, true_classes tf.Output, sampled_candidates tf.Output, num_true int64, optional ...ComputeAccidentalHitsAttr) (indices tf.Output, ids tf.Output, weights tf.Output)

Computes the ids of the positions in sampled_candidates that match true_labels.

When doing log-odds NCE, the result of this op should be passed through a SparseToDense op, then added to the logits of the sampled candidates. This has the effect of 'removing' the sampled labels that match the true labels by making the classifier sure that they are sampled labels.

Arguments: 

true_classes: The true_classes output of UnpackSparseLabels.
sampled_candidates: The sampled_candidates output of CandidateSampler.
num_true: Number of true labels per context.

Returns: 

indices: A vector of indices corresponding to rows of true_candidates.
ids: A vector of IDs of positions in sampled_candidates that match a true_label

for the row with the corresponding index in indices. 

weights: A vector of the same length as indices and ids, in which each element

is -FLOAT_MAX.

func ComputeBatchSize 

func ComputeBatchSize(scope *Scope, input_dataset tf.Output) (batch_size tf.Output)

Computes the static batch size of a dataset sans partial batches.

func ComputeDedupDataSize added in v0.8.0 

func ComputeDedupDataSize(scope *Scope, config string) (num_elements tf.Output)

An op computes the size of the deduplication data from embedding core and returns the updated config.

This op is to compute size of the deduplication data so to provide this information to the op that computes the tuple mask of deduplication data can have static output shape.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.

Returns The size of the deduplicated data from infeed.

func ComputeDedupDataSizeV2 added in v0.8.2 

func ComputeDedupDataSizeV2(scope *Scope, config string, embedding_partitions string, hbm_buffers_config string, tpu_topology string) (num_elements tf.Output)

An op computes the size of the deduplication data from embedding core and returns the updated config.

This op is to compute size of the deduplication data so to provide this information to the op that computes the tuple mask of deduplication data can have static output shape.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.
embedding_partitions: Serialized EmbeddingPartitionsProto proto.
hbm_buffers_config: Serialized HbmBuffersConfig proto.
tpu_topology: Serialized TpuTopologyArgsProto proto.

Returns The size of the deduplicated data from infeed.

func ComputeDedupDataTupleMask added in v0.4.0 

func ComputeDedupDataTupleMask(scope *Scope, config string) (output_shape tf.Output)

An op computes tuple mask of deduplication data from embedding core.

The deduplication data receiving from embedding core is a Tensor with type=DT_VARIANT. The tensor itself is an XLA nested tuple, whose elements are rank 1 tensors. This op is to represents types and length of these elements.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.

Returns A 2-D int tensor represent mask of deduplication data tuple generated by `XlaRecvTPUEmbeddingDeduplicationData`. The tuple has several integer and float type 1-D tensor tuple elements. The first dimenion of this output_shape 2-D tensor is tensor type of tuple elements, `0` represents integer tensor, `1` represents float tensor. The second dimension of `output_shape` gives length of each tuple element.

func ComputeDedupDataTupleMaskV2 added in v0.8.2 

func ComputeDedupDataTupleMaskV2(scope *Scope, config string, embedding_partitions string, hbm_buffers_config string, tpu_topology string) (output_shape tf.Output)

An op computes tuple mask of deduplication data from embedding core.

The deduplication data receiving from embedding core is a Tensor with type=DT_VARIANT. The tensor itself is an XLA nested tuple, whose elements are rank 1 tensors. This op is to represents types and length of these elements.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.
embedding_partitions: Serialized EmbeddingPartitionsProto proto.
hbm_buffers_config: Serialized HbmBuffersConfig proto.
tpu_topology: Serialized TpuTopologyArgsProto proto.

Returns A 2-D int tensor represent mask of deduplication data tuple generated by `XlaRecvTPUEmbeddingDeduplicationData`. The tuple has several integer and float type 1-D tensor tuple elements. The first dimenion of this output_shape 2-D tensor is tensor type of tuple elements, `0` represents integer tensor, `1` represents float tensor. The second dimension of `output_shape` gives length of each tuple element.

func Concat 

func Concat(scope *Scope, concat_dim tf.Output, values []tf.Output) (output tf.Output)

Concatenates tensors along one dimension.

Arguments: 

concat_dim: 0-D.  The dimension along which to concatenate.  Must be in the

range [0, rank(values)). 

values: The `N` Tensors to concatenate. Their ranks and types must match,

and their sizes must match in all dimensions except `concat_dim`.

Returns A `Tensor` with the concatenation of values stacked along the `concat_dim` dimension. This tensor's shape matches that of `values` except in `concat_dim` where it has the sum of the sizes.

func ConcatOffset 

func ConcatOffset(scope *Scope, concat_dim tf.Output, shape []tf.Output) (offset []tf.Output)

Computes offsets of concat inputs within its output.

For example:

>>> x = [2, 2, 7] >>> y = [2, 3, 7] >>> z = [2, 9, 7] >>> offsets = concat_offset(1, [x, y, z]) >>> [list(off.numpy()) for off in offsets] [[0, 0, 0], [0, 2, 0], [0, 5, 0]]

This is typically used by gradient computations for a concat operation.

Arguments: 

concat_dim: The dimension along which to concatenate.
shape: The `N` int32 or int64 vectors representing shape of tensors being concatenated.

Returns The `N` vectors representing the starting offset of input tensors within the concatenated output with type matching `shape`.

func ConcatV2 

func ConcatV2(scope *Scope, values []tf.Output, axis tf.Output) (output tf.Output)

Concatenates tensors along one dimension.

Arguments: 

values: List of `N` Tensors to concatenate. Their ranks and types must match,

and their sizes must match in all dimensions except `concat_dim`. 

axis: 0-D.  The dimension along which to concatenate.  Must be in the

range [-rank(values), rank(values)).

Returns A `Tensor` with the concatenation of values stacked along the `concat_dim` dimension. This tensor's shape matches that of `values` except in `concat_dim` where it has the sum of the sizes.

func ConcatenateDataset 

func ConcatenateDataset(scope *Scope, input_dataset tf.Output, another_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ConcatenateDatasetAttr) (handle tf.Output)

Creates a dataset that concatenates `input_dataset` with `another_dataset`.

func ConfigureAndInitializeGlobalTPU added in v0.2.0 

func ConfigureAndInitializeGlobalTPU(scope *Scope, optional ...ConfigureAndInitializeGlobalTPUAttr) (output tf.Output)

An op that sets up the centralized structures for a distributed TPU system.

Returns A vector containing the global TPU id of each TPU on the host.

func ConfigureDistributedTPU 

func ConfigureDistributedTPU(scope *Scope, optional ...ConfigureDistributedTPUAttr) (topology tf.Output)

Sets up the centralized structures for a distributed TPU system.

Returns A serialized tensorflow.tpu.TopologyProto that describes the TPU topology.

func ConfigureTPUEmbedding 

func ConfigureTPUEmbedding(scope *Scope, config string) (o *tf.Operation)

Sets up TPUEmbedding in a distributed TPU system.

Arguments: 

config: Serialized tensorflow.tpu.TPUEmbeddingConfiguration that

describes the embedding lookups of the program.

Returns the created operation.

func ConfigureTPUEmbeddingHost added in v0.2.0 

func ConfigureTPUEmbeddingHost(scope *Scope, common_config tf.Output, memory_config tf.Output, config string) (network_config tf.Output)

An op that configures the TPUEmbedding software on a host.

Arguments: 

common_config: A string-encoded common configuration proto containing metadata

about the TPUEmbedding partitioner output. 

memory_config: A string-encoded memory config proto containing metadata about

the memory allocations reserved for TPUEmbedding. 

config: An TPUEmbeddingConfiguration proto serialized to a string,

describing the desired TPUEmbedding configuration.

Returns A string containing metadata about the hostname and RPC port used for communication with this host.

func ConfigureTPUEmbeddingMemory added in v0.2.0 

func ConfigureTPUEmbeddingMemory(scope *Scope, common_config tf.Output) (memory_config tf.Output)

An op that configures the TPUEmbedding software on a host.

Arguments: 

common_config: A string-encoded CommonConfiguration proto containing metadata

about the TPUEmbedding partitioner output and the HBM size (in bytes) required for operation.

Returns A string-encoded memory configuration containing metadata about the memory allocations reserved for TPUEmbedding.

func Conj 

func Conj(scope *Scope, input tf.Output) (output tf.Output)

Returns the complex conjugate of a complex number.

Given a tensor `input` of complex numbers, this operation returns a tensor of complex numbers that are the complex conjugate of each element in `input`. The complex numbers in `input` must be of the form \\(a + bj\\), where *a* is the real part and *b* is the imaginary part.

The complex conjugate returned by this operation is of the form \\(a - bj\\).

For example:

``` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.conj(input) ==> [-2.25 - 4.75j, 3.25 - 5.75j] ```

func ConjugateTranspose 

func ConjugateTranspose(scope *Scope, x tf.Output, perm tf.Output) (y tf.Output)

Shuffle dimensions of x according to a permutation and conjugate the result.

The output `y` has the same rank as `x`. The shapes of `x` and `y` satisfy: 

`y.shape[i] == x.shape[perm[i]] for i in [0, 1, ..., rank(x) - 1]`
`y[i,j,k,...,s,t,u] == conj(x[perm[i], perm[j], perm[k],...,perm[s], perm[t], perm[u]])`

func ConnectTPUEmbeddingHosts added in v0.2.0 

func ConnectTPUEmbeddingHosts(scope *Scope, network_configs []tf.Output) (o *tf.Operation)

An op that sets up communication between TPUEmbedding host software instances

after ConfigureTPUEmbeddingHost has been called on each host.

Arguments: 

network_configs: Strings containing metadata about the hostname and RPC port

used for communication with all hosts.

Returns the created operation.

func Const 

func Const(scope *Scope, value interface{}) (output tf.Output)

Const adds an operation to graph that produces value as output.

func ConsumeMutexLock 

func ConsumeMutexLock(scope *Scope, mutex_lock tf.Output) (o *tf.Operation)

This op consumes a lock created by `MutexLock`.

This op exists to consume a tensor created by `MutexLock` (other than direct control dependencies). It should be the only that consumes the tensor, and will raise an error if it is not. Its only purpose is to keep the mutex lock tensor alive until it is consumed by this op.

**NOTE**: This operation must run on the same device as its input. This may be enforced via the `colocate_with` mechanism.

Arguments: 

mutex_lock: A tensor returned by `MutexLock`.

Returns the created operation.

func ControlTrigger 

func ControlTrigger(scope *Scope) (o *tf.Operation)

Does nothing. Serves as a control trigger for scheduling.

Only useful as a placeholder for control edges.

Returns the created operation.

func Conv added in v0.6.0 

func Conv(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...ConvAttr) (output tf.Output)

Computes a N-D convolution given (N+1+batch_dims)-D `input` and (N+2)-D `filter` tensors.

General function for computing a N-D convolution. It is required that `1 <= N <= 3`.

Arguments: 

input: Tensor of type T and shape `batch_shape + spatial_shape + [in_channels]` in the

case that `channels_last_format = true` or shape `batch_shape + [in_channels] + spatial_shape` if `channels_last_format = false`. spatial_shape is N-dimensional with `N=2` or `N=3`. Also note that `batch_shape` is dictated by the parameter `batch_dims` and defaults to 1. 

filter: An `(N+2)-D` Tensor with the same type as `input` and shape

`spatial_filter_shape + [in_channels, out_channels]`, where spatial_filter_shape is N-dimensional with `N=2` or `N=3`. 

strides: 1-D tensor of length `N+2`. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[N+1] = 1`. 

padding: The type of padding algorithm to use.

Returns A (N+1+batch_dims)-D tensor. The dimension order is determined by the value of `channels_last_format`, see below for details.

func Conv2D 

func Conv2D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...Conv2DAttr) (output tf.Output)

Computes a 2-D convolution given 4-D `input` and `filter` tensors.

Given an input tensor of shape `[batch, in_height, in_width, in_channels]` and a filter / kernel tensor of shape `[filter_height, filter_width, in_channels, out_channels]`, this op performs the following:

  1. Flattens the filter to a 2-D matrix with shape `[filter_height * filter_width * in_channels, output_channels]`.
  2. Extracts image patches from the input tensor to form a *virtual* tensor of shape `[batch, out_height, out_width, filter_height * filter_width * in_channels]`.
  3. For each patch, right-multiplies the filter matrix and the image patch vector.

In detail, with the default NHWC format, 

output[b, i, j, k] =
    sum_{di, dj, q} input[b, strides[1] * i + di, strides[2] * j + dj, q] *
                    filter[di, dj, q, k]

Must have `strides[0] = strides[3] = 1`. For the most common case of the same horizontal and vertices strides, `strides = [1, stride, stride, 1]`.

Arguments: 

input: A 4-D tensor. The dimension order is interpreted according to the value

of `data_format`, see below for details. 

filter: A 4-D tensor of shape

`[filter_height, filter_width, in_channels, out_channels]` 

strides: 1-D tensor of length 4.  The stride of the sliding window for each

dimension of `input`. The dimension order is determined by the value of `data_format`, see below for details. 

padding: The type of padding algorithm to use.

Returns A 4-D tensor. The dimension order is determined by the value of `data_format`, see below for details.

func Conv2DBackpropFilter 

func Conv2DBackpropFilter(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropFilterAttr) (output tf.Output)

Computes the gradients of convolution with respect to the filter.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, in_channels]`.
filter_sizes: An integer vector representing the tensor shape of `filter`,

where `filter` is a 4-D `[filter_height, filter_width, in_channels, out_channels]` tensor. 

out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.

Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`. Gradient w.r.t. the `filter` input of the convolution.

func Conv2DBackpropFilterV2 added in v0.4.0 

func Conv2DBackpropFilterV2(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropFilterV2Attr) (output tf.Output)

Computes the gradients of convolution with respect to the filter.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, in_channels]`.
filter: 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`.

Only shape of tensor is used. 

out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.

Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`. Gradient w.r.t. the `filter` input of the convolution.

func Conv2DBackpropInput 

func Conv2DBackpropInput(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropInputAttr) (output tf.Output)

Computes the gradients of convolution with respect to the input.

Arguments: 

input_sizes: An integer vector representing the shape of `input`,

where `input` is a 4-D `[batch, height, width, channels]` tensor. 

filter: 4-D with shape

`[filter_height, filter_width, in_channels, out_channels]`. 

out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.

Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[batch, in_height, in_width, in_channels]`. Gradient w.r.t. the input of the convolution.

func Conv2DBackpropInputV2 added in v0.4.0 

func Conv2DBackpropInputV2(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropInputV2Attr) (output tf.Output)

Computes the gradients of convolution with respect to the input.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, in_channels]`.

Only shape of tensor is used. 

filter: 4-D with shape

`[filter_height, filter_width, in_channels, out_channels]`. 

out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.

Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[batch, in_height, in_width, in_channels]`. Gradient w.r.t. the input of the convolution.

func Conv3D 

func Conv3D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...Conv3DAttr) (output tf.Output)

Computes a 3-D convolution given 5-D `input` and `filter` tensors.

In signal processing, cross-correlation is a measure of similarity of two waveforms as a function of a time-lag applied to one of them. This is also known as a sliding dot product or sliding inner-product.

Our Conv3D implements a form of cross-correlation.

Arguments: 

input: Shape `[batch, in_depth, in_height, in_width, in_channels]`.
filter: Shape `[filter_depth, filter_height, filter_width, in_channels,

out_channels]`. `in_channels` must match between `input` and `filter`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func Conv3DBackpropFilter 

func Conv3DBackpropFilter(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv3DBackpropFilterAttr) (output tf.Output)

Computes the gradients of 3-D convolution with respect to the filter.

DEPRECATED at GraphDef version 10: Use Conv3DBackpropFilterV2

Arguments: 

input: Shape `[batch, depth, rows, cols, in_channels]`.
filter: Shape `[depth, rows, cols, in_channels, out_channels]`.

`in_channels` must match between `input` and `filter`. 

out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,

out_channels]`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func Conv3DBackpropFilterV2 

func Conv3DBackpropFilterV2(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv3DBackpropFilterV2Attr) (output tf.Output)

Computes the gradients of 3-D convolution with respect to the filter.

Arguments: 

input: Shape `[batch, depth, rows, cols, in_channels]`.
filter_sizes: An integer vector representing the tensor shape of `filter`,

where `filter` is a 5-D `[filter_depth, filter_height, filter_width, in_channels, out_channels]` tensor. 

out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,

out_channels]`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func Conv3DBackpropInput 

func Conv3DBackpropInput(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv3DBackpropInputAttr) (output tf.Output)

Computes the gradients of 3-D convolution with respect to the input.

DEPRECATED at GraphDef version 10: Use Conv3DBackpropInputV2

Arguments: 

input: Shape `[batch, depth, rows, cols, in_channels]`.
filter: Shape `[depth, rows, cols, in_channels, out_channels]`.

`in_channels` must match between `input` and `filter`. 

out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,

out_channels]`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func Conv3DBackpropInputV2 

func Conv3DBackpropInputV2(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv3DBackpropInputV2Attr) (output tf.Output)

Computes the gradients of 3-D convolution with respect to the input.

Arguments: 

input_sizes: An integer vector representing the tensor shape of `input`,

where `input` is a 5-D `[batch, depth, rows, cols, in_channels]` tensor. 

filter: Shape `[depth, rows, cols, in_channels, out_channels]`.

`in_channels` must match between `input` and `filter`. 

out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,

out_channels]`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func Copy 

func Copy(scope *Scope, input tf.Output, optional ...CopyAttr) (output tf.Output)

Copy a tensor from CPU-to-CPU or GPU-to-GPU.

Performs CPU-to-CPU or GPU-to-GPU deep-copying of tensor, depending on the device on which the tensor is allocated. N.B.: If the all downstream attached debug ops are disabled given the current gRPC gating status, the output will simply forward the input tensor without deep-copying. See the documentation of Debug* ops for more details.

Unlike the CopyHost Op, this op does not have HostMemory constraint on its input or output.

Arguments: 

input: Input tensor.

func CopyHost 

func CopyHost(scope *Scope, input tf.Output, optional ...CopyHostAttr) (output tf.Output)

Copy a tensor to host.

Performs CPU-to-CPU deep-copying of tensor. N.B.: If the all downstream attached debug ops are disabled given the current gRPC gating status, the output will simply forward the input tensor without deep-copying. See the documentation of Debug* ops for more details.

Unlike the Copy Op, this op has HostMemory constraint on its input or output.

Arguments: 

input: Input tensor.

func Cos 

func Cos(scope *Scope, x tf.Output) (y tf.Output)

Computes cos of x element-wise. 

Given an input tensor, this function computes cosine of every
element in the tensor. Input range is `(-inf, inf)` and
output range is `[-1,1]`. If input lies outside the boundary, `nan`
is returned.

```python
x = tf.constant([-float("inf"), -9, -0.5, 1, 1.2, 200, 10000, float("inf")])
tf.math.cos(x) ==> [nan -0.91113025 0.87758255 0.5403023 0.36235774 0.48718765 -0.95215535 nan]
```

func Cosh 

func Cosh(scope *Scope, x tf.Output) (y tf.Output)

Computes hyperbolic cosine of x element-wise. 

Given an input tensor, this function computes hyperbolic cosine of every
element in the tensor. Input range is `[-inf, inf]` and output range
is `[1, inf]`.

```python
x = tf.constant([-float("inf"), -9, -0.5, 1, 1.2, 2, 10, float("inf")])
tf.math.cosh(x) ==> [inf 4.0515420e+03 1.1276259e+00 1.5430807e+00 1.8106556e+00 3.7621956e+00 1.1013233e+04 inf]
```

func CropAndResize 

func CropAndResize(scope *Scope, image tf.Output, boxes tf.Output, box_ind tf.Output, crop_size tf.Output, optional ...CropAndResizeAttr) (crops tf.Output)

Extracts crops from the input image tensor and resizes them.

Extracts crops from the input image tensor and resizes them using bilinear sampling or nearest neighbor sampling (possibly with aspect ratio change) to a common output size specified by `crop_size`. This is more general than the `crop_to_bounding_box` op which extracts a fixed size slice from the input image and does not allow resizing or aspect ratio change.

Returns a tensor with `crops` from the input `image` at positions defined at the bounding box locations in `boxes`. The cropped boxes are all resized (with bilinear or nearest neighbor interpolation) to a fixed `size = [crop_height, crop_width]`. The result is a 4-D tensor `[num_boxes, crop_height, crop_width, depth]`. The resizing is corner aligned. In particular, if `boxes = [[0, 0, 1, 1]]`, the method will give identical results to using `tf.image.resize_bilinear()` or `tf.image.resize_nearest_neighbor()`(depends on the `method` argument) with `align_corners=True`.

Arguments: 

image: A 4-D tensor of shape `[batch, image_height, image_width, depth]`.

Both `image_height` and `image_width` need to be positive. 

boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor

specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1]` in image height coordinates. We do allow `y1` > `y2`, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values. 

box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.

The value of `box_ind[i]` specifies the image that the `i`-th box refers to. 

crop_size: A 1-D tensor of 2 elements, `size = [crop_height, crop_width]`. All

cropped image patches are resized to this size. The aspect ratio of the image content is not preserved. Both `crop_height` and `crop_width` need to be positive.

Returns A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.

func CropAndResizeGradBoxes 

func CropAndResizeGradBoxes(scope *Scope, grads tf.Output, image tf.Output, boxes tf.Output, box_ind tf.Output, optional ...CropAndResizeGradBoxesAttr) (output tf.Output)

Computes the gradient of the crop_and_resize op wrt the input boxes tensor.

Arguments: 

grads: A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.
image: A 4-D tensor of shape `[batch, image_height, image_width, depth]`.

Both `image_height` and `image_width` need to be positive. 

boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor

specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values. 

box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.

The value of `box_ind[i]` specifies the image that the `i`-th box refers to.

Returns A 2-D tensor of shape `[num_boxes, 4]`.

func CropAndResizeGradImage 

func CropAndResizeGradImage(scope *Scope, grads tf.Output, boxes tf.Output, box_ind tf.Output, image_size tf.Output, T tf.DataType, optional ...CropAndResizeGradImageAttr) (output tf.Output)

Computes the gradient of the crop_and_resize op wrt the input image tensor.

Arguments: 

grads: A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.
boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor

specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values. 

box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.

The value of `box_ind[i]` specifies the image that the `i`-th box refers to. 

image_size: A 1-D tensor with value `[batch, image_height, image_width, depth]`

containing the original image size. Both `image_height` and `image_width` need to be positive.

Returns A 4-D tensor of shape `[batch, image_height, image_width, depth]`.

func Cross 

func Cross(scope *Scope, a tf.Output, b tf.Output) (product tf.Output)

Compute the pairwise cross product.

`a` and `b` must be the same shape; they can either be simple 3-element vectors, or any shape where the innermost dimension is 3. In the latter case, each pair of corresponding 3-element vectors is cross-multiplied independently.

Arguments: 

a: A tensor containing 3-element vectors.
b: Another tensor, of same type and shape as `a`.

Returns Pairwise cross product of the vectors in `a` and `b`.

func CrossReplicaSum 

func CrossReplicaSum(scope *Scope, input tf.Output, group_assignment tf.Output) (output tf.Output)

An Op to sum inputs across replicated TPU instances.

Each instance supplies its own input.

For example, suppose there are 8 TPU instances: `[A, B, C, D, E, F, G, H]`. Passing group_assignment=`[[0,2,4,6],[1,3,5,7]]` sets `A, C, E, G` as group 0, and `B, D, F, H` as group 1. Thus we get the outputs: `[A+C+E+G, B+D+F+H, A+C+E+G, B+D+F+H, A+C+E+G, B+D+F+H, A+C+E+G, B+D+F+H]`.

Arguments: 

input: The local input to the sum.
group_assignment: An int32 tensor with shape

[num_groups, num_replicas_per_group]. `group_assignment[i]` represents the replica ids in the ith subgroup.

Returns The sum of all the distributed inputs.

func CudnnRNN 

func CudnnRNN(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, optional ...CudnnRNNAttr) (output tf.Output, output_h tf.Output, output_c tf.Output, reserve_space tf.Output)

A RNN backed by cuDNN.

Computes the RNN from the input and initial states, with respect to the params buffer.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: A 3-D tensor with the shape of [seq_length, batch_size, input_size]. input_h: A 3-D tensor with the shape of [num_layer * dir, batch_size, 

num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

output: A 3-D tensor with the shape of [seq_length, batch_size, 

dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. is_training: Indicates whether this operation is used for inference or 

training.

reserve_space: An opaque tensor that can be used in backprop calculation. It 

is only produced if is_training is false.

func CudnnRNNBackprop 

func CudnnRNNBackprop(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, output tf.Output, output_h tf.Output, output_c tf.Output, output_backprop tf.Output, output_h_backprop tf.Output, output_c_backprop tf.Output, reserve_space tf.Output, optional ...CudnnRNNBackpropAttr) (input_backprop tf.Output, input_h_backprop tf.Output, input_c_backprop tf.Output, params_backprop tf.Output)

Backprop step of CudnnRNN.

Compute the backprop of both data and weights in a RNN.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: A 3-D tensor with the shape of [seq_length, batch_size, input_size]. input_h: A 3-D tensor with the shape of [num_layer * dir, batch_size, 

num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

output: A 3-D tensor with the shape of [seq_length, batch_size, 

dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. output_backprop: A 3-D tensor with the same shape as output in the forward pass. output_h_backprop: A 3-D tensor with the same shape as output_h in the forward 

pass.

output_c_backprop: A 3-D tensor with the same shape as output_c in the forward 

pass.

reserve_space: The same reserve_space produced in for forward operation. input_backprop: The backprop to input in the forward pass. Has the same shape 

as input.

input_h_backprop: The backprop to input_h in the forward pass. Has the same 

shape as input_h.

input_c_backprop: The backprop to input_c in the forward pass. Has the same 

shape as input_c.

params_backprop: The backprop to the params buffer in the forward pass. Has the 

same shape as params.

func CudnnRNNBackpropV2 

func CudnnRNNBackpropV2(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, output tf.Output, output_h tf.Output, output_c tf.Output, output_backprop tf.Output, output_h_backprop tf.Output, output_c_backprop tf.Output, reserve_space tf.Output, host_reserved tf.Output, optional ...CudnnRNNBackpropV2Attr) (input_backprop tf.Output, input_h_backprop tf.Output, input_c_backprop tf.Output, params_backprop tf.Output)

Backprop step of CudnnRNN.

Compute the backprop of both data and weights in a RNN. Takes an extra 

"host_reserved" inupt than CudnnRNNBackprop, which is used to determine RNN
cudnnRNNAlgo_t and cudnnMathType_t.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicates whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: A 3-D tensor with the shape of [seq_length, batch_size, input_size]. input_h: A 3-D tensor with the shape of [num_layer * dir, batch_size, 

num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

output: A 3-D tensor with the shape of [seq_length, batch_size, 

dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. output_backprop: A 3-D tensor with the same shape as output in the forward pass. output_h_backprop: A 3-D tensor with the same shape as output_h in the forward 

pass.

output_c_backprop: A 3-D tensor with the same shape as output_c in the forward 

pass.

reserve_space: The same reserve_space produced in the forward operation. host_reserved: The same host_reserved produced in the forward operation. input_backprop: The backprop to input in the forward pass. Has the same shape 

as input.

input_h_backprop: The backprop to input_h in the forward pass. Has the same 

shape as input_h.

input_c_backprop: The backprop to input_c in the forward pass. Has the same 

shape as input_c.

params_backprop: The backprop to the params buffer in the forward pass. Has the 

same shape as params.

func CudnnRNNBackpropV3 

func CudnnRNNBackpropV3(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, sequence_lengths tf.Output, output tf.Output, output_h tf.Output, output_c tf.Output, output_backprop tf.Output, output_h_backprop tf.Output, output_c_backprop tf.Output, reserve_space tf.Output, host_reserved tf.Output, optional ...CudnnRNNBackpropV3Attr) (input_backprop tf.Output, input_h_backprop tf.Output, input_c_backprop tf.Output, params_backprop tf.Output)

Backprop step of CudnnRNNV3.

Compute the backprop of both data and weights in a RNN. Takes an extra 

"sequence_lengths" input than CudnnRNNBackprop.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicates whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: If time_major is true, this is a 3-D tensor with the shape of 

[seq_length, batch_size, input_size]. If time_major is false, the shape is
[batch_size, seq_length, input_size].

input_h: If time_major is true, this is a 3-D tensor with the shape of 

[num_layer * dir, batch_size, num_units]. If time_major is false, the shape
is [batch_size, num_layer * dir, num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

sequence_lengths: a vector of lengths of each input sequence. output: If time_major is true, this is a 3-D tensor with the shape of 

[seq_length, batch_size, dir * num_units]. If time_major is false, the
shape is [batch_size, seq_length, dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. output_backprop: A 3-D tensor with the same shape as output in the forward pass. output_h_backprop: A 3-D tensor with the same shape as output_h in the forward 

pass.

output_c_backprop: A 3-D tensor with the same shape as output_c in the forward 

pass.

time_major: Indicates whether the input/output format is time major or batch 

major.

reserve_space: The same reserve_space produced in the forward operation. input_backprop: The backprop to input in the forward pass. Has the same shape 

as input.

input_h_backprop: The backprop to input_h in the forward pass. Has the same 

shape as input_h.

input_c_backprop: The backprop to input_c in the forward pass. Has the same 

shape as input_c.

params_backprop: The backprop to the params buffer in the forward pass. Has the 

same shape as params.

func CudnnRNNCanonicalToParams 

func CudnnRNNCanonicalToParams(scope *Scope, num_layers tf.Output, num_units tf.Output, input_size tf.Output, weights []tf.Output, biases []tf.Output, optional ...CudnnRNNCanonicalToParamsAttr) (params tf.Output)

Converts CudnnRNN params from canonical form to usable form.

Writes a set of weights into the opaque params buffer so they can be used in upcoming training or inferences.

Note that the params buffer may not be compatible across different GPUs. So any save and restoration should be converted to and from the canonical weights and biases.

num_layers: Specifies the number of layers in the RNN model. num_units: Specifies the size of the hidden state. input_size: Specifies the size of the input state. weights: the canonical form of weights that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

biases: the canonical form of biases that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

num_params: number of parameter sets for all layers. 

Each layer may contain multiple parameter sets, with each set consisting of
a weight matrix and a bias vector.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

The actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. 

dir = (direction == bidirectional) ? 2 : 1

dropout: dropout probability. When set to 0., dropout is disabled. seed: the 1st part of a seed to initialize dropout. seed2: the 2nd part of a seed to initialize dropout.

func CudnnRNNCanonicalToParamsV2 

func CudnnRNNCanonicalToParamsV2(scope *Scope, num_layers tf.Output, num_units tf.Output, input_size tf.Output, weights []tf.Output, biases []tf.Output, optional ...CudnnRNNCanonicalToParamsV2Attr) (params tf.Output)

Converts CudnnRNN params from canonical form to usable form. It supports the projection in LSTM.

Writes a set of weights into the opaque params buffer so they can be used in upcoming training or inferences.

Note that the params buffer may not be compatible across different GPUs. So any save and restoration should be converted to and from the canonical weights and biases.

num_layers: Specifies the number of layers in the RNN model. num_units: Specifies the size of the hidden state. input_size: Specifies the size of the input state. weights: the canonical form of weights that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

biases: the canonical form of biases that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

num_params_weights: number of weight parameter matrix for all layers. num_params_biases: number of bias parameter vector for all layers. rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

The actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. 

dir = (direction == bidirectional) ? 2 : 1

dropout: dropout probability. When set to 0., dropout is disabled. seed: the 1st part of a seed to initialize dropout. seed2: the 2nd part of a seed to initialize dropout. num_proj: The output dimensionality for the projection matrices. If None or 0, 

no projection is performed.

func CudnnRNNParamsSize 

func CudnnRNNParamsSize(scope *Scope, num_layers tf.Output, num_units tf.Output, input_size tf.Output, T tf.DataType, S tf.DataType, optional ...CudnnRNNParamsSizeAttr) (params_size tf.Output)

Computes size of weights that can be used by a Cudnn RNN model.

Return the params size that can be used by the Cudnn RNN model. Subsequent weight allocation and initialization should use this size.

num_layers: Specifies the number of layers in the RNN model. num_units: Specifies the size of the hidden state. input_size: Specifies the size of the input state. rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

The actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. 

dir = (direction == bidirectional) ? 2 : 1

dropout: dropout probability. When set to 0., dropout is disabled. seed: the 1st part of a seed to initialize dropout. seed2: the 2nd part of a seed to initialize dropout. params_size: The size of the params buffer that should be allocated and 

initialized for this RNN model. Note that this params buffer may not be
compatible across GPUs. Please use CudnnRNNParamsWeights and
CudnnRNNParamsBiases to save and restore them in a way that is compatible
across different runs.

func CudnnRNNParamsToCanonical 

func CudnnRNNParamsToCanonical(scope *Scope, num_layers tf.Output, num_units tf.Output, input_size tf.Output, params tf.Output, num_params int64, optional ...CudnnRNNParamsToCanonicalAttr) (weights []tf.Output, biases []tf.Output)

Retrieves CudnnRNN params in canonical form.

Retrieves a set of weights from the opaque params buffer that can be saved and restored in a way compatible with future runs.

Note that the params buffer may not be compatible across different GPUs. So any save and restoration should be converted to and from the canonical weights and biases.

num_layers: Specifies the number of layers in the RNN model. num_units: Specifies the size of the hidden state. input_size: Specifies the size of the input state. num_params: number of parameter sets for all layers. 

Each layer may contain multiple parameter sets, with each set consisting of
a weight matrix and a bias vector.

weights: the canonical form of weights that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

biases: the canonical form of biases that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

The actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. 

dir = (direction == bidirectional) ? 2 : 1

dropout: dropout probability. When set to 0., dropout is disabled. seed: the 1st part of a seed to initialize dropout. seed2: the 2nd part of a seed to initialize dropout.

func CudnnRNNParamsToCanonicalV2 

func CudnnRNNParamsToCanonicalV2(scope *Scope, num_layers tf.Output, num_units tf.Output, input_size tf.Output, params tf.Output, num_params_weights int64, num_params_biases int64, optional ...CudnnRNNParamsToCanonicalV2Attr) (weights []tf.Output, biases []tf.Output)

Retrieves CudnnRNN params in canonical form. It supports the projection in LSTM.

Retrieves a set of weights from the opaque params buffer that can be saved and restored in a way compatible with future runs.

Note that the params buffer may not be compatible across different GPUs. So any save and restoration should be converted to and from the canonical weights and biases.

num_layers: Specifies the number of layers in the RNN model. num_units: Specifies the size of the hidden state. input_size: Specifies the size of the input state. num_params_weights: number of weight parameter matrix for all layers. num_params_biases: number of bias parameter vector for all layers. weights: the canonical form of weights that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

biases: the canonical form of biases that can be used for saving 

and restoration. They are more likely to be compatible across different
generations.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicate whether there is a linear projection between the input and 

The actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. 

dir = (direction == bidirectional) ? 2 : 1

dropout: dropout probability. When set to 0., dropout is disabled. seed: the 1st part of a seed to initialize dropout. seed2: the 2nd part of a seed to initialize dropout. num_proj: The output dimensionality for the projection matrices. If None or 0, 

no projection is performed.

func CudnnRNNV2 

func CudnnRNNV2(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, optional ...CudnnRNNV2Attr) (output tf.Output, output_h tf.Output, output_c tf.Output, reserve_space tf.Output, host_reserved tf.Output)

A RNN backed by cuDNN.

Computes the RNN from the input and initial states, with respect to the params buffer. Produces one extra output "host_reserved" than CudnnRNN.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicates whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: A 3-D tensor with the shape of [seq_length, batch_size, input_size]. input_h: A 3-D tensor with the shape of [num_layer * dir, batch_size, 

num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

output: A 3-D tensor with the shape of [seq_length, batch_size, 

dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. is_training: Indicates whether this operation is used for inference or 

training.

reserve_space: An opaque tensor that can be used in backprop calculation. It 

is only produced if is_training is true.

host_reserved: An opaque tensor that can be used in backprop calculation. It is 

only produced if is_training is true. It is output on host memory rather than
device memory.

func CudnnRNNV3 

func CudnnRNNV3(scope *Scope, input tf.Output, input_h tf.Output, input_c tf.Output, params tf.Output, sequence_lengths tf.Output, optional ...CudnnRNNV3Attr) (output tf.Output, output_h tf.Output, output_c tf.Output, reserve_space tf.Output, host_reserved tf.Output)

A RNN backed by cuDNN.

Computes the RNN from the input and initial states, with respect to the params buffer. Accepts one extra input "sequence_lengths" than CudnnRNN.

rnn_mode: Indicates the type of the RNN model. input_mode: Indicates whether there is a linear projection between the input and 

the actual computation before the first layer. 'skip_input' is only allowed
when input_size == num_units; 'auto_select' implies 'skip_input' when
input_size == num_units; otherwise, it implies 'linear_input'.

direction: Indicates whether a bidirectional model will be used. Should be 

"unidirectional" or "bidirectional".

dropout: Dropout probability. When set to 0., dropout is disabled. seed: The 1st part of a seed to initialize dropout. seed2: The 2nd part of a seed to initialize dropout. input: If time_major is true, this is a 3-D tensor with the shape of 

[seq_length, batch_size, input_size]. If time_major is false, the shape is
[batch_size, seq_length, input_size].

input_h: If time_major is true, this is a 3-D tensor with the shape of 

[num_layer * dir, batch_size, num_units]. If time_major is false, the shape
is [batch_size, num_layer * dir, num_units].

input_c: For LSTM, a 3-D tensor with the shape of 

[num_layer * dir, batch, num_units]. For other models, it is ignored.

params: A 1-D tensor that contains the weights and biases in an opaque layout. 

The size must be created through CudnnRNNParamsSize, and initialized
separately. Note that they might not be compatible across different
generations. So it is a good idea to save and restore

sequence_lengths: a vector of lengths of each input sequence. output: If time_major is true, this is a 3-D tensor with the shape of 

[seq_length, batch_size, dir * num_units]. If time_major is false, the
shape is [batch_size, seq_length, dir * num_units].

output_h: The same shape has input_h. output_c: The same shape as input_c for LSTM. An empty tensor for other models. is_training: Indicates whether this operation is used for inference or 

training.

time_major: Indicates whether the input/output format is time major or batch 

major.

reserve_space: An opaque tensor that can be used in backprop calculation. It 

is only produced if is_training is true.

func Cumprod 

func Cumprod(scope *Scope, x tf.Output, axis tf.Output, optional ...CumprodAttr) (out tf.Output)

Compute the cumulative product of the tensor `x` along `axis`.

By default, this op performs an inclusive cumprod, which means that the first element of the input is identical to the first element of the output:

```python tf.cumprod([a, b, c]) # => [a, a * b, a * b * c] ```

By setting the `exclusive` kwarg to `True`, an exclusive cumprod is performed instead:

```python tf.cumprod([a, b, c], exclusive=True) # => [1, a, a * b] ```

By setting the `reverse` kwarg to `True`, the cumprod is performed in the opposite direction:

```python tf.cumprod([a, b, c], reverse=True) # => [a * b * c, b * c, c] ```

This is more efficient than using separate `tf.reverse` ops.

The `reverse` and `exclusive` kwargs can also be combined:

```python tf.cumprod([a, b, c], exclusive=True, reverse=True) # => [b * c, c, 1] ```

Arguments: 

x: A `Tensor`. Must be one of the following types: `float32`, `float64`,

`int64`, `int32`, `uint8`, `uint16`, `int16`, `int8`, `complex64`, `complex128`, `qint8`, `quint8`, `qint32`, `half`. 

axis: A `Tensor` of type `int32` (default: 0). Must be in the range

`[-rank(x), rank(x))`.

func Cumsum 

func Cumsum(scope *Scope, x tf.Output, axis tf.Output, optional ...CumsumAttr) (out tf.Output)

Compute the cumulative sum of the tensor `x` along `axis`.

By default, this op performs an inclusive cumsum, which means that the first element of the input is identical to the first element of the output:

```python tf.cumsum([a, b, c]) # => [a, a + b, a + b + c] ```

By setting the `exclusive` kwarg to `True`, an exclusive cumsum is performed instead:

```python tf.cumsum([a, b, c], exclusive=True) # => [0, a, a + b] ```

By setting the `reverse` kwarg to `True`, the cumsum is performed in the opposite direction:

```python tf.cumsum([a, b, c], reverse=True) # => [a + b + c, b + c, c] ```

This is more efficient than using separate `tf.reverse` ops.

The `reverse` and `exclusive` kwargs can also be combined:

```python tf.cumsum([a, b, c], exclusive=True, reverse=True) # => [b + c, c, 0] ```

Arguments: 

x: A `Tensor`. Must be one of the following types: `float32`, `float64`,

`int64`, `int32`, `uint8`, `uint16`, `int16`, `int8`, `complex64`, `complex128`, `qint8`, `quint8`, `qint32`, `half`. 

axis: A `Tensor` of type `int32` (default: 0). Must be in the range

`[-rank(x), rank(x))`.

func CumulativeLogsumexp 

func CumulativeLogsumexp(scope *Scope, x tf.Output, axis tf.Output, optional ...CumulativeLogsumexpAttr) (out tf.Output)

Compute the cumulative product of the tensor `x` along `axis`.

By default, this op performs an inclusive cumulative log-sum-exp, which means that the first element of the input is identical to the first element of the output: ```python tf.math.cumulative_logsumexp([a, b, c]) # => [a, log(exp(a) + exp(b)), log(exp(a) + exp(b) + exp(c))] ```

By setting the `exclusive` kwarg to `True`, an exclusive cumulative log-sum-exp is performed instead: ```python tf.cumulative_logsumexp([a, b, c], exclusive=True) # => [-inf, a, log(exp(a) * exp(b))] ``` Note that the neutral element of the log-sum-exp operation is `-inf`, however, for performance reasons, the minimal value representable by the floating point type is used instead.

By setting the `reverse` kwarg to `True`, the cumulative log-sum-exp is performed in the opposite direction.

Arguments: 

x: A `Tensor`. Must be one of the following types: `float16`, `float32`, `float64`.
axis: A `Tensor` of type `int32` (default: 0). Must be in the range

`[-rank(x), rank(x))`.

func DTensorSetGlobalTPUArray added in v0.2.0 

func DTensorSetGlobalTPUArray(scope *Scope, topology tf.Output) (o *tf.Operation)

An op that informs a host of the global ids of all the of TPUs in the system.

Arguments: 

topology: A serialized tensorflow.tpu.TopologyProto that describes the TPU topology.

Returns the created operation.

func DataFormatDimMap 

func DataFormatDimMap(scope *Scope, x tf.Output, optional ...DataFormatDimMapAttr) (y tf.Output)

Returns the dimension index in the destination data format given the one in

the source data format.

Arguments: 

x: A Tensor with each element as a dimension index in source data format.

Must be in the range [-4, 4).

Returns A Tensor with each element as a dimension index in destination data format.

func DataFormatVecPermute 

func DataFormatVecPermute(scope *Scope, x tf.Output, optional ...DataFormatVecPermuteAttr) (y tf.Output)

Permute input tensor from `src_format` to `dst_format`.

Given source and destination format strings of length n=4 or 5, the input tensor must be a vector of size n or n-2, or a 2D tensor of shape (n, 2) or (n-2, 2).

If the first dimension of the input tensor is n-2, it is assumed that non-spatial dimensions are omitted (i.e `N`, `C`).

For example, with `src_format` of `NHWC`, `dst_format` of `NCHW`, and input: ``` [1, 2, 3, 4] ``` , the output will be: ``` [1, 4, 2, 3] ``` With `src_format` of `NDHWC`, `dst_format` of `NCDHW`, and input: ``` [[1, 6], [2, 7], [3, 8], [4, 9], [5, 10]] ``` , the output will be: ``` [[1, 6], [5, 10], [2, 7], [3, 8], [4, 9]] ``` With `src_format` of `NHWC`, `dst_format` of `NCHW`, and input: ``` [1, 2] ``` , the output will be: ``` [1, 2] ```

Arguments: 

x: Tensor of rank 1 or 2 in source data format.

Returns Tensor of rank 1 or 2 in destination data format.

func DataServiceDataset 

func DataServiceDataset(scope *Scope, dataset_id tf.Output, processing_mode tf.Output, address tf.Output, protocol tf.Output, job_name tf.Output, max_outstanding_requests tf.Output, iteration_counter tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...DataServiceDatasetAttr) (handle tf.Output)

Creates a dataset that reads data from the tf.data service.

func DataServiceDatasetV2 

func DataServiceDatasetV2(scope *Scope, dataset_id tf.Output, processing_mode tf.Output, address tf.Output, protocol tf.Output, job_name tf.Output, consumer_index tf.Output, num_consumers tf.Output, max_outstanding_requests tf.Output, iteration_counter tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...DataServiceDatasetV2Attr) (handle tf.Output)

Creates a dataset that reads data from the tf.data service.

func DatasetCardinality 

func DatasetCardinality(scope *Scope, input_dataset tf.Output, optional ...DatasetCardinalityAttr) (cardinality tf.Output)

Returns the cardinality of `input_dataset`.

Returns the cardinality of `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the dataset to return cardinality for.

Returns The cardinality of `input_dataset`. Named constants are used to represent infinite and unknown cardinality.

func DatasetFingerprint added in v0.8.0 

func DatasetFingerprint(scope *Scope, input_dataset tf.Output) (fingerprint tf.Output)

Returns the fingerprint of `input_dataset`.

Returns the fingerprint of `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the dataset to return fingerprint for.

Returns The fingerprint of `input_dataset` in `uint64`

func DatasetFromGraph 

func DatasetFromGraph(scope *Scope, graph_def tf.Output) (handle tf.Output)

Creates a dataset from the given `graph_def`.

Creates a dataset from the provided `graph_def`.

Arguments: 

graph_def: The graph representation of the dataset (as serialized GraphDef).

Returns A variant tensor representing the dataset.

func DatasetToGraph 

func DatasetToGraph(scope *Scope, input_dataset tf.Output, optional ...DatasetToGraphAttr) (graph tf.Output)

Returns a serialized GraphDef representing `input_dataset`.

Returns a graph representation for `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the dataset to return the graph representation for.

Returns The graph representation of the dataset (as serialized GraphDef).

func DatasetToGraphV2 

func DatasetToGraphV2(scope *Scope, input_dataset tf.Output, optional ...DatasetToGraphV2Attr) (graph tf.Output)

Returns a serialized GraphDef representing `input_dataset`.

Returns a graph representation for `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the dataset to return the graph representation for.

Returns The graph representation of the dataset (as serialized GraphDef).

func DatasetToSingleElement 

func DatasetToSingleElement(scope *Scope, dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...DatasetToSingleElementAttr) (components []tf.Output)

Outputs the single element from the given dataset.

Arguments: 

dataset: A handle to a dataset that contains a single element.

Returns The components of the single element of `input`.

func DatasetToTFRecord 

func DatasetToTFRecord(scope *Scope, input_dataset tf.Output, filename tf.Output, compression_type tf.Output) (o *tf.Operation)

Writes the given dataset to the given file using the TFRecord format.

Arguments: 

input_dataset: A variant tensor representing the dataset to write.
filename: A scalar string tensor representing the filename to use.
compression_type: A scalar string tensor containing either (i) the empty string (no

compression), (ii) "ZLIB", or (iii) "GZIP".

Returns the created operation.

func DebugGradientIdentity 

func DebugGradientIdentity(scope *Scope, input tf.Output) (output tf.Output)

Identity op for gradient debugging.

This op is hidden from public in Python. It is used by TensorFlow Debugger to register gradient tensors for gradient debugging. This op operates on non-reference-type tensors.

func DebugIdentity 

func DebugIdentity(scope *Scope, input tf.Output, optional ...DebugIdentityAttr) (output tf.Output)

Provides an identity mapping of the non-Ref type input tensor for debugging.

Provides an identity mapping of the non-Ref type input tensor for debugging.

Arguments: 

input: Input tensor, non-Reference type

func DebugIdentityV2 

func DebugIdentityV2(scope *Scope, input tf.Output, optional ...DebugIdentityV2Attr) (output tf.Output)

Debug Identity V2 Op.

Provides an identity mapping from input to output, while writing the content of the input tensor by calling DebugEventsWriter.

The semantics of the input tensor depends on tensor_debug_mode. In typical usage, the input tensor comes directly from the user computation only when graph_debug_mode is FULL_TENSOR (see protobuf/debug_event.proto for a list of all the possible values of graph_debug_mode). For the other debug modes, the input tensor should be produced by an additional op or subgraph that computes summary information about one or more tensors.

Arguments: 

input: Input tensor, non-Reference type

func DebugIdentityV3 added in v0.5.0 

func DebugIdentityV3(scope *Scope, input tf.Output, optional ...DebugIdentityV3Attr) (output tf.Output)

Provides an identity mapping of the non-Ref type input tensor for debugging.

Provides an identity mapping of the non-Ref type input tensor for debugging.

Arguments: 

input: Input tensor, non-Reference type

func DebugNanCount 

func DebugNanCount(scope *Scope, input tf.Output, optional ...DebugNanCountAttr) (output tf.Output)

Debug NaN Value Counter Op.

Counts number of NaNs in the input tensor, for debugging.

Arguments: 

input: Input tensor, non-Reference type.

func DebugNumericSummary 

func DebugNumericSummary(scope *Scope, input tf.Output, optional ...DebugNumericSummaryAttr) (output tf.Output)

Debug Numeric Summary Op.

Provide a basic summary of numeric value types, range and distribution.

output: A double tensor of shape [14 + nDimensions], where nDimensions is the 

number of dimensions of the tensor's shape. The elements of output are:
[0]: is initialized (1.0) or not (0.0).
[1]: total number of elements
[2]: NaN element count
[3]: generalized -inf count: elements <= lower_bound. lower_bound is -inf by
  default.
[4]: negative element count (excluding -inf), if lower_bound is the default
  -inf. Otherwise, this is the count of elements > lower_bound and < 0.
[5]: zero element count
[6]: positive element count (excluding +inf), if upper_bound is the default
  +inf. Otherwise, this is the count of elements < upper_bound and > 0.
[7]: generalized +inf count, elements >= upper_bound. upper_bound is +inf by
  default.

Output elements [1:8] are all zero, if the tensor is uninitialized. 

[8]: minimum of all non-inf and non-NaN elements.
     If uninitialized or no such element exists: +inf.
[9]: maximum of all non-inf and non-NaN elements.
     If uninitialized or no such element exists: -inf.
[10]: mean of all non-inf and non-NaN elements.
      If uninitialized or no such element exists: NaN.
[11]: variance of all non-inf and non-NaN elements.
      If uninitialized or no such element exists: NaN.
[12]: Data type of the tensor encoded as an enum integer. See the DataType
      proto for more details.
[13]: Number of dimensions of the tensor (ndims).
[14+]: Sizes of the dimensions.

Arguments: 

input: Input tensor, non-Reference type.

func DebugNumericSummaryV2 

func DebugNumericSummaryV2(scope *Scope, input tf.Output, optional ...DebugNumericSummaryV2Attr) (output tf.Output)

Debug Numeric Summary V2 Op.

Computes a numeric summary of the input tensor. The shape of the output depends on the tensor_debug_mode attribute. This op is used internally by TensorFlow Debugger (tfdbg) v2.

Arguments: 

input: Input tensor, to be summarized by the op.

func DecodeAndCropJpeg 

func DecodeAndCropJpeg(scope *Scope, contents tf.Output, crop_window tf.Output, optional ...DecodeAndCropJpegAttr) (image tf.Output)

Decode and Crop a JPEG-encoded image to a uint8 tensor.

The attr `channels` indicates the desired number of color channels for the decoded image.

Accepted values are:

* 0: Use the number of channels in the JPEG-encoded image. * 1: output a grayscale image. * 3: output an RGB image.

If needed, the JPEG-encoded image is transformed to match the requested number of color channels.

The attr `ratio` allows downscaling the image by an integer factor during decoding. Allowed values are: 1, 2, 4, and 8. This is much faster than downscaling the image later.

It is equivalent to a combination of decode and crop, but much faster by only decoding partial jpeg image.

Arguments: 

contents: 0-D.  The JPEG-encoded image.
crop_window: 1-D.  The crop window: [crop_y, crop_x, crop_height, crop_width].

Returns 3-D with shape `[height, width, channels]`..

func DecodeBase64 

func DecodeBase64(scope *Scope, input tf.Output) (output tf.Output)

Decode web-safe base64-encoded strings.

Input may or may not have padding at the end. See EncodeBase64(https://www.tensorflow.org/api_docs/python/tf/io/encode_base64) for padding. Web-safe means that input must use - and _ instead of + and /.

Arguments: 

input: Base64 strings to decode.

Returns Decoded strings.

func DecodeBmp 

func DecodeBmp(scope *Scope, contents tf.Output, optional ...DecodeBmpAttr) (image tf.Output)

Decode the first frame of a BMP-encoded image to a uint8 tensor.

The attr `channels` indicates the desired number of color channels for the decoded image.

Accepted values are:

* 0: Use the number of channels in the BMP-encoded image. * 3: output an RGB image. * 4: output an RGBA image.

Arguments: 

contents: 0-D.  The BMP-encoded image.

Returns 3-D with shape `[height, width, channels]`. RGB order

func DecodeCSV 

func DecodeCSV(scope *Scope, records tf.Output, record_defaults []tf.Output, optional ...DecodeCSVAttr) (output []tf.Output)

Convert CSV records to tensors. Each column maps to one tensor.

RFC 4180 format is expected for the CSV records. (https://tools.ietf.org/html/rfc4180) Note that we allow leading and trailing spaces with int or float field.

Arguments: 

records: Each string is a record/row in the csv and all records should have

the same format. 

record_defaults: One tensor per column of the input record, with either a

scalar default value for that column or an empty vector if the column is required.

Returns Each tensor will have the same shape as records.

func DecodeCompressed 

func DecodeCompressed(scope *Scope, bytes tf.Output, optional ...DecodeCompressedAttr) (output tf.Output)

Decompress strings.

This op decompresses each element of the `bytes` input `Tensor`, which is assumed to be compressed using the given `compression_type`.

The `output` is a string `Tensor` of the same shape as `bytes`, each element containing the decompressed data from the corresponding element in `bytes`.

Arguments: 

bytes: A Tensor of string which is compressed.

Returns A Tensor with the same shape as input `bytes`, uncompressed from bytes.

func DecodeGif 

func DecodeGif(scope *Scope, contents tf.Output) (image tf.Output)

Decode the frame(s) of a GIF-encoded image to a uint8 tensor.

GIF images with frame or transparency compression are not supported. On Linux and MacOS systems, convert animated GIFs from compressed to uncompressed by running: 

convert $src.gif -coalesce $dst.gif

This op also supports decoding JPEGs and PNGs, though it is cleaner to use `tf.io.decode_image`.

Arguments: 

contents: 0-D.  The GIF-encoded image.

Returns 4-D with shape `[num_frames, height, width, 3]`. RGB channel order.

func DecodeImage 

func DecodeImage(scope *Scope, contents tf.Output, optional ...DecodeImageAttr) (image tf.Output)

Function for decode_bmp, decode_gif, decode_jpeg, and decode_png.

Detects whether an image is a BMP, GIF, JPEG, or PNG, and performs the appropriate operation to convert the input bytes string into a Tensor of type dtype.

*NOTE*: decode_gif returns a 4-D array [num_frames, height, width, 3], as opposed to decode_bmp, decode_jpeg and decode_png, which return 3-D arrays [height, width, num_channels]. Make sure to take this into account when constructing your graph if you are intermixing GIF files with BMP, JPEG, and/or PNG files. Alternately, set the expand_animations argument of this function to False, in which case the op will return 3-dimensional tensors and will truncate animated GIF files to the first frame.

*NOTE*: If the first frame of an animated GIF does not occupy the entire canvas (maximum frame width x maximum frame height), then it fills the unoccupied areas (in the first frame) with zeros (black). For frames after the first frame that does not occupy the entire canvas, it uses the previous frame to fill the unoccupied areas.

Arguments: 

contents: 0-D. The encoded image bytes.

Returns 3-D with shape `[height, width, channels]` or 4-D with shape `[frame, height, width, channels]`..

func DecodeJSONExample 

func DecodeJSONExample(scope *Scope, json_examples tf.Output) (binary_examples tf.Output)

Convert JSON-encoded Example records to binary protocol buffer strings.

Note: This is **not** a general purpose JSON parsing op.

This op converts JSON-serialized `tf.train.Example` (created with `json_format.MessageToJson`, following the [standard JSON mapping](https://developers.google.com/protocol-buffers/docs/proto3#json)) to a binary-serialized `tf.train.Example` (equivalent to `Example.SerializeToString()`) suitable for conversion to tensors with `tf.io.parse_example`.

Arguments: 

json_examples: Each string is a JSON object serialized according to the JSON

mapping of the Example proto.

Returns Each string is a binary Example protocol buffer corresponding to the respective element of `json_examples`.

func DecodeJpeg 

func DecodeJpeg(scope *Scope, contents tf.Output, optional ...DecodeJpegAttr) (image tf.Output)

Decode a JPEG-encoded image to a uint8 tensor.

The attr `channels` indicates the desired number of color channels for the decoded image.

Accepted values are:

* 0: Use the number of channels in the JPEG-encoded image. * 1: output a grayscale image. * 3: output an RGB image.

If needed, the JPEG-encoded image is transformed to match the requested number of color channels.

The attr `ratio` allows downscaling the image by an integer factor during decoding. Allowed values are: 1, 2, 4, and 8. This is much faster than downscaling the image later.

This op also supports decoding PNGs and non-animated GIFs since the interface is the same, though it is cleaner to use `tf.io.decode_image`.

Arguments: 

contents: 0-D.  The JPEG-encoded image.

Returns 3-D with shape `[height, width, channels]`..

func DecodePaddedRaw 

func DecodePaddedRaw(scope *Scope, input_bytes tf.Output, fixed_length tf.Output, out_type tf.DataType, optional ...DecodePaddedRawAttr) (output tf.Output)

Reinterpret the bytes of a string as a vector of numbers.

Arguments: 

input_bytes: Tensor of string to be decoded.
fixed_length: Length in bytes for each element of the decoded output. Must be a multiple

of the size of the output type.

Returns A Tensor with one more dimension than the input `bytes`. The added dimension will have size equal to the length of the elements of `bytes` divided by the number of bytes to represent `out_type`.

func DecodePng 

func DecodePng(scope *Scope, contents tf.Output, optional ...DecodePngAttr) (image tf.Output)

Decode a PNG-encoded image to a uint8 or uint16 tensor.

The attr `channels` indicates the desired number of color channels for the decoded image.

Accepted values are:

* 0: Use the number of channels in the PNG-encoded image. * 1: output a grayscale image. * 3: output an RGB image. * 4: output an RGBA image.

If needed, the PNG-encoded image is transformed to match the requested number of color channels.

This op also supports decoding JPEGs and non-animated GIFs since the interface is the same, though it is cleaner to use `tf.io.decode_image`.

Arguments: 

contents: 0-D.  The PNG-encoded image.

Returns 3-D with shape `[height, width, channels]`.

func DecodeProtoV2 

func DecodeProtoV2(scope *Scope, bytes tf.Output, message_type string, field_names []string, output_types []tf.DataType, optional ...DecodeProtoV2Attr) (sizes tf.Output, values []tf.Output)

The op extracts fields from a serialized protocol buffers message into tensors.

Note: This API is designed for orthogonality rather than human-friendliness. It can be used to parse input protos by hand, but it is intended for use in generated code.

The `decode_proto` op extracts fields from a serialized protocol buffers message into tensors. The fields in `field_names` are decoded and converted to the corresponding `output_types` if possible.

A `message_type` name must be provided to give context for the field names. The actual message descriptor can be looked up either in the linked-in descriptor pool or a filename provided by the caller using the `descriptor_source` attribute.

Each output tensor is a dense tensor. This means that it is padded to hold the largest number of repeated elements seen in the input minibatch. (The shape is also padded by one to prevent zero-sized dimensions). The actual repeat counts for each example in the minibatch can be found in the `sizes` output. In many cases the output of `decode_proto` is fed immediately into tf.squeeze if missing values are not a concern. When using tf.squeeze, always pass the squeeze dimension explicitly to avoid surprises.

For the most part, the mapping between Proto field types and TensorFlow dtypes is straightforward. However, there are a few special cases:

- A proto field that contains a submessage or group can only be converted to `DT_STRING` (the serialized submessage). This is to reduce the complexity of the API. The resulting string can be used as input to another instance of the decode_proto op.

- TensorFlow lacks support for unsigned integers. The ops represent uint64 types as a `DT_INT64` with the same twos-complement bit pattern (the obvious way). Unsigned int32 values can be represented exactly by specifying type `DT_INT64`, or using twos-complement if the caller specifies `DT_INT32` in the `output_types` attribute.

- `map` fields are not directly decoded. They are treated as `repeated` fields, of the appropriate entry type. The proto-compiler defines entry types for each map field. The type-name is the field name, converted to "CamelCase" with "Entry" appended. The `tf.train.Features.FeatureEntry` message is an example of one of these implicit `Entry` types.

- `enum` fields should be read as int32.

Both binary and text proto serializations are supported, and can be chosen using the `format` attribute.

The `descriptor_source` attribute selects the source of protocol descriptors to consult when looking up `message_type`. This may be:

- An empty string or "local://", in which case protocol descriptors are created for C++ (not Python) proto definitions linked to the binary.

- A file, in which case protocol descriptors are created from the file, which is expected to contain a `FileDescriptorSet` serialized as a string. NOTE: You can build a `descriptor_source` file using the `--descriptor_set_out` and `--include_imports` options to the protocol compiler `protoc`.

- A "bytes://<bytes>", in which protocol descriptors are created from `<bytes>`, which is expected to be a `FileDescriptorSet` serialized as a string.

Arguments: 

bytes: Tensor of serialized protos with shape `batch_shape`.
message_type: Name of the proto message type to decode.
field_names: List of strings containing proto field names. An extension field can be decoded

by using its full name, e.g. EXT_PACKAGE.EXT_FIELD_NAME. 

output_types: List of TF types to use for the respective field in field_names.

Returns: 

sizes: Tensor of int32 with shape `[batch_shape, len(field_names)]`.

Each entry is the number of values found for the corresponding field. Optional fields may have 0 or 1 values. 

values: List of tensors containing values for the corresponding field.

`values[i]` has datatype `output_types[i]` and shape `[batch_shape, max(sizes[...,i])]`.

func DecodeRaw 

func DecodeRaw(scope *Scope, bytes tf.Output, out_type tf.DataType, optional ...DecodeRawAttr) (output tf.Output)

Reinterpret the bytes of a string as a vector of numbers.

Arguments: 

bytes: All the elements must have the same length.

Returns A Tensor with one more dimension than the input `bytes`. The added dimension will have size equal to the length of the elements of `bytes` divided by the number of bytes to represent `out_type`.

func DecodeWav 

func DecodeWav(scope *Scope, contents tf.Output, optional ...DecodeWavAttr) (audio tf.Output, sample_rate tf.Output)

Decode a 16-bit PCM WAV file to a float tensor.

The -32768 to 32767 signed 16-bit values will be scaled to -1.0 to 1.0 in float.

When desired_channels is set, if the input contains fewer channels than this then the last channel will be duplicated to give the requested number, else if the input has more channels than requested then the additional channels will be ignored.

If desired_samples is set, then the audio will be cropped or padded with zeroes to the requested length.

The first output contains a Tensor with the content of the audio samples. The lowest dimension will be the number of channels, and the second will be the number of samples. For example, a ten-sample-long stereo WAV file should give an output shape of [10, 2].

Arguments: 

contents: The WAV-encoded audio, usually from a file.

Returns: 

audio: 2-D with shape `[length, channels]`.
sample_rate: Scalar holding the sample rate found in the WAV header.

func DeepCopy 

func DeepCopy(scope *Scope, x tf.Output) (y tf.Output)

Makes a copy of `x`.

Arguments: 

x: The source tensor of type `T`.

Returns y: A `Tensor` of type `T`. A copy of `x`. Guaranteed that `y` 

is not an alias of `x`.

func DeleteIterator 

func DeleteIterator(scope *Scope, handle tf.Output, deleter tf.Output) (o *tf.Operation)

A container for an iterator resource.

Arguments: 

handle: A handle to the iterator to delete.
deleter: A variant deleter.

Returns the created operation.

func DeleteMultiDeviceIterator 

func DeleteMultiDeviceIterator(scope *Scope, multi_device_iterator tf.Output, iterators []tf.Output, deleter tf.Output) (o *tf.Operation)

A container for an iterator resource.

Arguments: 

multi_device_iterator: A handle to the multi device iterator to delete.
iterators: A list of iterator handles (unused). This is added so that automatic control dependencies get added during function tracing that ensure this op runs after all the dependent iterators are deleted.
deleter: A variant deleter.

Returns the created operation.

func DeleteSessionTensor 

func DeleteSessionTensor(scope *Scope, handle tf.Output) (o *tf.Operation)

Delete the tensor specified by its handle in the session.

Arguments: 

handle: The handle for a tensor stored in the session state.

Returns the created operation.

func DenseBincount 

func DenseBincount(scope *Scope, input tf.Output, size tf.Output, weights tf.Output, optional ...DenseBincountAttr) (output tf.Output)

Counts the number of occurrences of each value in an integer array.

Outputs a vector with length `size` and the same dtype as `weights`. If `weights` are empty, then index `i` stores the number of times the value `i` is counted in `arr`. If `weights` are non-empty, then index `i` stores the sum of the value in `weights` at each index where the corresponding value in `arr` is `i`.

Values in `arr` outside of the range [0, size) are ignored.

Arguments: 

input: 1D or 2D int `Tensor`.
size: non-negative int scalar `Tensor`.
weights: is an int32, int64, float32, or float64 `Tensor` with the same

shape as `arr`, or a length-0 `Tensor`, in which case it acts as all weights equal to 1.

Returns 1D `Tensor` with length equal to `size` or 2D `Tensor` with [batch_size, `size`]. The counts or summed weights for each value in the range [0, size).

func DenseCountSparseOutput 

func DenseCountSparseOutput(scope *Scope, values tf.Output, weights tf.Output, binary_output bool, optional ...DenseCountSparseOutputAttr) (output_indices tf.Output, output_values tf.Output, output_dense_shape tf.Output)

Performs sparse-output bin counting for a tf.tensor input. 

Counts the number of times each value occurs in the input.

Arguments: 

values: Tensor containing data to count.
weights: A Tensor of the same shape as indices containing per-index weight values. May

also be the empty tensor if no weights are used. 

binary_output: Whether to output the number of occurrences of each value or 1.

Returns: 

output_indices: Indices tensor for the resulting sparse tensor object.
output_values: Values tensor for the resulting sparse tensor object.
output_dense_shape: Shape tensor for the resulting sparse tensor object.

func DenseToCSRSparseMatrix 

func DenseToCSRSparseMatrix(scope *Scope, dense_input tf.Output, indices tf.Output) (sparse_output tf.Output)

Converts a dense tensor to a (possibly batched) CSRSparseMatrix.

Arguments: 

dense_input: A Dense tensor.
indices: Indices of nonzero elements.

Returns A (possibly batched) CSRSparseMatrix.

func DenseToDenseSetOperation 

func DenseToDenseSetOperation(scope *Scope, set1 tf.Output, set2 tf.Output, set_operation string, optional ...DenseToDenseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)

Applies set operation along last dimension of 2 `Tensor` inputs.

See SetOperationOp::SetOperationFromContext for values of `set_operation`.

Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.

Arguments: 

set1: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set2`.

Dimension `n` contains values in a set, duplicates are allowed but ignored. 

set2: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set1`.

Dimension `n` contains values in a set, duplicates are allowed but ignored.

Returns: 

result_indices: 2D indices of a `SparseTensor`.
result_values: 1D values of a `SparseTensor`.
result_shape: 1D `Tensor` shape of a `SparseTensor`. `result_shape[0...n-1]` is

the same as the 1st `n-1` dimensions of `set1` and `set2`, `result_shape[n]` is the max result set size across all `0...n-1` dimensions.

func DenseToSparseBatchDataset 

func DenseToSparseBatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, row_shape tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that batches input elements into a SparseTensor.

Arguments: 

input_dataset: A handle to an input dataset. Must have a single component.
batch_size: A scalar representing the number of elements to accumulate in a

batch. 

row_shape: A vector representing the dense shape of each row in the produced

SparseTensor. The shape may be partially specified, using `-1` to indicate that a particular dimension should use the maximum size of all batch elements.

func DenseToSparseSetOperation 

func DenseToSparseSetOperation(scope *Scope, set1 tf.Output, set2_indices tf.Output, set2_values tf.Output, set2_shape tf.Output, set_operation string, optional ...DenseToSparseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)

Applies set operation along last dimension of `Tensor` and `SparseTensor`.

See SetOperationOp::SetOperationFromContext for values of `set_operation`.

Input `set2` is a `SparseTensor` represented by `set2_indices`, `set2_values`, and `set2_shape`. For `set2` ranked `n`, 1st `n-1` dimensions must be the same as `set1`. Dimension `n` contains values in a set, duplicates are allowed but ignored.

If `validate_indices` is `True`, this op validates the order and range of `set2` indices.

Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.

Arguments: 

set1: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set2`.

Dimension `n` contains values in a set, duplicates are allowed but ignored. 

set2_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major

order. 

set2_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major

order. 

set2_shape: 1D `Tensor`, shape of a `SparseTensor`. `set2_shape[0...n-1]` must

be the same as the 1st `n-1` dimensions of `set1`, `result_shape[n]` is the max set size across `n-1` dimensions.

Returns: 

result_indices: 2D indices of a `SparseTensor`.
result_values: 1D values of a `SparseTensor`.
result_shape: 1D `Tensor` shape of a `SparseTensor`. `result_shape[0...n-1]` is

the same as the 1st `n-1` dimensions of `set1` and `set2`, `result_shape[n]` is the max result set size across all `0...n-1` dimensions.

func DepthToSpace 

func DepthToSpace(scope *Scope, input tf.Output, block_size int64, optional ...DepthToSpaceAttr) (output tf.Output)

DepthToSpace for tensors of type T.

Rearranges data from depth into blocks of spatial data. This is the reverse transformation of SpaceToDepth. More specifically, this op outputs a copy of the input tensor where values from the `depth` dimension are moved in spatial blocks to the `height` and `width` dimensions. The attr `block_size` indicates the input block size and how the data is moved.

  • Chunks of data of size `block_size * block_size` from depth are rearranged into non-overlapping blocks of size `block_size x block_size`
  • The width of the output tensor is `input_depth * block_size`, whereas the height is `input_height * block_size`.
  • The Y, X coordinates within each block of the output image are determined by the high order component of the input channel index.
  • The depth of the input tensor must be divisible by `block_size * block_size`.

The `data_format` attr specifies the layout of the input and output tensors with the following options: 

"NHWC": `[ batch, height, width, channels ]`
"NCHW": `[ batch, channels, height, width ]`
"NCHW_VECT_C":
    `qint8 [ batch, channels / 4, height, width, 4 ]`

It is useful to consider the operation as transforming a 6-D Tensor. e.g. for data_format = NHWC, 

Each element in the input tensor can be specified via 6 coordinates,
ordered by decreasing memory layout significance as:
n,iY,iX,bY,bX,oC  (where n=batch index, iX, iY means X or Y coordinates
                   within the input image, bX, bY means coordinates
                   within the output block, oC means output channels).
The output would be the input transposed to the following layout:
n,iY,bY,iX,bX,oC

This operation is useful for resizing the activations between convolutions (but keeping all data), e.g. instead of pooling. It is also useful for training purely convolutional models.

For example, given an input of shape `[1, 1, 1, 4]`, data_format = "NHWC" and block_size = 2:

``` x = [[[[1, 2, 3, 4]]]]

```

This operation will output a tensor of shape `[1, 2, 2, 1]`:

``` 

[[[[1], [2]],
  [[3], [4]]]]

```

Here, the input has a batch of 1 and each batch element has shape `[1, 1, 4]`, the corresponding output will have 2x2 elements and will have a depth of 1 channel (1 = `4 / (block_size * block_size)`). The output element shape is `[2, 2, 1]`.

For an input tensor with larger depth, here of shape `[1, 1, 1, 12]`, e.g.

``` x = [[[[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]]]] ```

This operation, for block size of 2, will return the following tensor of shape `[1, 2, 2, 3]`

``` 

[[[[1, 2, 3], [4, 5, 6]],
  [[7, 8, 9], [10, 11, 12]]]]

```

Similarly, for the following input of shape `[1 2 2 4]`, and a block size of 2:

``` x = [[[[1, 2, 3, 4], 

 [5, 6, 7, 8]],
[[9, 10, 11, 12],
 [13, 14, 15, 16]]]]

```

the operator will return the following tensor of shape `[1 4 4 1]`:

``` x = [[[ [1], [2], [5], [6]], 

[ [3],   [4],  [7],  [8]],
[ [9],  [10], [13],  [14]],
[ [11], [12], [15],  [16]]]]

```

Arguments: 

block_size: The size of the spatial block, same as in Space2Depth.

func DepthwiseConv2dNative 

func DepthwiseConv2dNative(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeAttr) (output tf.Output)

Computes a 2-D depthwise convolution given 4-D `input` and `filter` tensors.

Given an input tensor of shape `[batch, in_height, in_width, in_channels]` and a filter / kernel tensor of shape `[filter_height, filter_width, in_channels, channel_multiplier]`, containing `in_channels` convolutional filters of depth 1, `depthwise_conv2d` applies a different filter to each input channel (expanding from 1 channel to `channel_multiplier` channels for each), then concatenates the results together. Thus, the output has `in_channels * channel_multiplier` channels.

``` for k in 0..in_channels-1 

for q in 0..channel_multiplier-1
  output[b, i, j, k * channel_multiplier + q] =
    sum_{di, dj} input[b, strides[1] * i + di, strides[2] * j + dj, k] *
                      filter[di, dj, k, q]

```

Must have `strides[0] = strides[3] = 1`. For the most common case of the same horizontal and vertices strides, `strides = [1, stride, stride, 1]`.

Arguments: 

strides: 1-D of length 4.  The stride of the sliding window for each dimension

of `input`. 

padding: The type of padding algorithm to use.

func DepthwiseConv2dNativeBackpropFilter 

func DepthwiseConv2dNativeBackpropFilter(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeBackpropFilterAttr) (output tf.Output)

Computes the gradients of depthwise convolution with respect to the filter.

Arguments: 

input: 4-D with shape based on `data_format`.  For example, if

`data_format` is 'NHWC' then `input` is a 4-D `[batch, in_height, in_width, in_channels]` tensor. 

filter_sizes: An integer vector representing the tensor shape of `filter`,

where `filter` is a 4-D `[filter_height, filter_width, in_channels, depthwise_multiplier]` tensor. 

out_backprop: 4-D with shape  based on `data_format`.

For example, if `data_format` is 'NHWC' then out_backprop shape is `[batch, out_height, out_width, out_channels]`. Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`. Gradient w.r.t. the `filter` input of the convolution.

func DepthwiseConv2dNativeBackpropInput 

func DepthwiseConv2dNativeBackpropInput(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeBackpropInputAttr) (output tf.Output)

Computes the gradients of depthwise convolution with respect to the input.

Arguments: 

input_sizes: An integer vector representing the shape of `input`, based

on `data_format`. For example, if `data_format` is 'NHWC' then 

 `input` is a 4-D `[batch, height, width, channels]` tensor.
	filter: 4-D with shape

`[filter_height, filter_width, in_channels, depthwise_multiplier]`. 

out_backprop: 4-D with shape  based on `data_format`.

For example, if `data_format` is 'NHWC' then out_backprop shape is `[batch, out_height, out_width, out_channels]`. Gradients w.r.t. the output of the convolution. 

strides: The stride of the sliding window for each dimension of the input

of the convolution. 

padding: The type of padding algorithm to use.

Returns 4-D with shape according to `data_format`. For example, if `data_format` is 'NHWC', output shape is `[batch, in_height, in_width, in_channels]`. Gradient w.r.t. the input of the convolution.

func Dequantize 

func Dequantize(scope *Scope, input tf.Output, min_range tf.Output, max_range tf.Output, optional ...DequantizeAttr) (output tf.Output)

Dequantize the 'input' tensor into a float or bfloat16 Tensor.

[min_range, max_range] are scalar floats that specify the range for the output. The 'mode' attribute controls exactly which calculations are used to convert the float values to their quantized equivalents.

In 'MIN_COMBINED' mode, each value of the tensor will undergo the following:

``` if T == qint8: in[i] += (range(T) + 1)/ 2.0 out[i] = min_range + (in[i]* (max_range - min_range) / range(T)) ``` here `range(T) = numeric_limits<T>::max() - numeric_limits<T>::min()`

*MIN_COMBINED Mode Example*

If the input comes from a QuantizedRelu6, the output type is quint8 (range of 0-255) but the possible range of QuantizedRelu6 is 0-6. The min_range and max_range values are therefore 0.0 and 6.0. Dequantize on quint8 will take each value, cast to float, and multiply by 6 / 255. Note that if quantizedtype is qint8, the operation will additionally add each value by 128 prior to casting.

If the mode is 'MIN_FIRST', then this approach is used:

```c++ num_discrete_values = 1 << (# of bits in T) range_adjust = num_discrete_values / (num_discrete_values - 1) range = (range_max - range_min) * range_adjust range_scale = range / num_discrete_values const double offset_input = static_cast<double>(input) - lowest_quantized; result = range_min + ((input - numeric_limits<T>::min()) * range_scale) ```

If the mode is `SCALED`, dequantization is performed by multiplying each input value by a scaling_factor. (Thus an input of 0 always maps to 0.0).

The scaling_factor is determined from `min_range`, `max_range`, and `narrow_range` in a way that is compatible with `QuantizeAndDequantize{V2|V3}` and `QuantizeV2`, using the following algorithm:

```c++ 

const int min_expected_T = std::numeric_limits<T>::min() +
  (narrow_range ? 1 : 0);
const int max_expected_T = std::numeric_limits<T>::max();
const float max_expected_T = std::numeric_limits<float>::max();

const float scale_factor =
  (std::numeric_limits<T>::min() == 0) ? (max_range / max_expected_T)
                                       : std::max(min_range / min_expected_T,
                                                  max_range / max_expected_T);

```

Arguments: 

min_range: The minimum scalar value possibly produced for the input.
max_range: The maximum scalar value possibly produced for the input.

func DeserializeIterator 

func DeserializeIterator(scope *Scope, resource_handle tf.Output, serialized tf.Output) (o *tf.Operation)

Converts the given variant tensor to an iterator and stores it in the given resource.

Arguments: 

resource_handle: A handle to an iterator resource.
serialized: A variant tensor storing the state of the iterator contained in the

resource.

Returns the created operation.

func DeserializeManySparse 

func DeserializeManySparse(scope *Scope, serialized_sparse tf.Output, dtype tf.DataType) (sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output)

Deserialize and concatenate `SparseTensors` from a serialized minibatch.

The input `serialized_sparse` must be a string matrix of shape `[N x 3]` where `N` is the minibatch size and the rows correspond to packed outputs of `SerializeSparse`. The ranks of the original `SparseTensor` objects must all match. When the final `SparseTensor` is created, it has rank one higher than the ranks of the incoming `SparseTensor` objects (they have been concatenated along a new row dimension).

The output `SparseTensor` object's shape values for all dimensions but the first are the max across the input `SparseTensor` objects' shape values for the corresponding dimensions. Its first shape value is `N`, the minibatch size.

The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, after this step run `SparseReorder` to restore index ordering.

For example, if the serialized input is a `[2 x 3]` matrix representing two original `SparseTensor` objects: 

index = [ 0]
        [10]
        [20]
values = [1, 2, 3]
shape = [50]

and 

index = [ 2]
        [10]
values = [4, 5]
shape = [30]

then the final deserialized `SparseTensor` will be: 

index = [0  0]
        [0 10]
        [0 20]
        [1  2]
        [1 10]
values = [1, 2, 3, 4, 5]
shape = [2 50]

Arguments: 

serialized_sparse: 2-D, The `N` serialized `SparseTensor` objects.

Must have 3 columns. 

dtype: The `dtype` of the serialized `SparseTensor` objects.

func DeserializeSparse 

func DeserializeSparse(scope *Scope, serialized_sparse tf.Output, dtype tf.DataType) (sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output)

Deserialize `SparseTensor` objects.

The input `serialized_sparse` must have the shape `[?, ?, ..., ?, 3]` where the last dimension stores serialized `SparseTensor` objects and the other N dimensions (N >= 0) correspond to a batch. The ranks of the original `SparseTensor` objects must all match. When the final `SparseTensor` is created, its rank is the rank of the incoming `SparseTensor` objects plus N; the sparse tensors have been concatenated along new dimensions, one for each batch.

The output `SparseTensor` object's shape values for the original dimensions are the max across the input `SparseTensor` objects' shape values for the corresponding dimensions. The new dimensions match the size of the batch.

The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, after this step run `SparseReorder` to restore index ordering.

For example, if the serialized input is a `[2 x 3]` matrix representing two original `SparseTensor` objects: 

index = [ 0]
        [10]
        [20]
values = [1, 2, 3]
shape = [50]

and 

index = [ 2]
        [10]
values = [4, 5]
shape = [30]

then the final deserialized `SparseTensor` will be: 

index = [0  0]
        [0 10]
        [0 20]
        [1  2]
        [1 10]
values = [1, 2, 3, 4, 5]
shape = [2 50]

Arguments: 

serialized_sparse: The serialized `SparseTensor` objects. The last dimension

must have 3 columns. 

dtype: The `dtype` of the serialized `SparseTensor` objects.

func DestroyResourceOp 

func DestroyResourceOp(scope *Scope, resource tf.Output, optional ...DestroyResourceOpAttr) (o *tf.Operation)

Deletes the resource specified by the handle.

All subsequent operations using the resource will result in a NotFound error status.

Arguments: 

resource: handle to the resource to delete.

Returns the created operation.

func DeviceIndex 

func DeviceIndex(scope *Scope, device_names []string) (index tf.Output)

Return the index of device the op runs.

Given a list of device names, this operation returns the index of the device this op runs. The length of the list is returned in two cases: (1) Device does not exist in the given device list. (2) It is in XLA compilation.

func Diag 

func Diag(scope *Scope, diagonal tf.Output) (output tf.Output)

Returns a diagonal tensor with a given diagonal values.

Given a `diagonal`, this operation returns a tensor with the `diagonal` and everything else padded with zeros. The diagonal is computed as follows:

Assume `diagonal` has dimensions [D1,..., Dk], then the output is a tensor of rank 2k with dimensions [D1,..., Dk, D1,..., Dk] where:

`output[i1,..., ik, i1,..., ik] = diagonal[i1, ..., ik]` and 0 everywhere else.

For example:

``` # 'diagonal' is [1, 2, 3, 4] tf.diag(diagonal) ==> [[1, 0, 0, 0] 

[0, 2, 0, 0]
[0, 0, 3, 0]
[0, 0, 0, 4]]

```

Arguments: 

diagonal: Rank k tensor where k is at most 1.

func DiagPart 

func DiagPart(scope *Scope, input tf.Output) (diagonal tf.Output)

Returns the diagonal part of the tensor.

This operation returns a tensor with the `diagonal` part of the `input`. The `diagonal` part is computed as follows:

Assume `input` has dimensions `[D1,..., Dk, D1,..., Dk]`, then the output is a tensor of rank `k` with dimensions `[D1,..., Dk]` where:

`diagonal[i1,..., ik] = input[i1, ..., ik, i1,..., ik]`.

For example:

``` # 'input' is [[1, 0, 0, 0] 

[0, 2, 0, 0]
[0, 0, 3, 0]
[0, 0, 0, 4]]

tf.diag_part(input) ==> [1, 2, 3, 4] ```

Arguments: 

input: Rank k tensor where k is even and not zero.

Returns The extracted diagonal.

func Digamma 

func Digamma(scope *Scope, x tf.Output) (y tf.Output)

Computes Psi, the derivative of Lgamma (the log of the absolute value of

`Gamma(x)`), element-wise.

func Dilation2D 

func Dilation2D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, rates []int64, padding string) (output tf.Output)

Computes the grayscale dilation of 4-D `input` and 3-D `filter` tensors.

The `input` tensor has shape `[batch, in_height, in_width, depth]` and the `filter` tensor has shape `[filter_height, filter_width, depth]`, i.e., each input channel is processed independently of the others with its own structuring function. The `output` tensor has shape `[batch, out_height, out_width, depth]`. The spatial dimensions of the output tensor depend on the `padding` algorithm. We currently only support the default "NHWC" `data_format`.

In detail, the grayscale morphological 2-D dilation is the max-sum correlation (for consistency with `conv2d`, we use unmirrored filters): 

output[b, y, x, c] =
   max_{dy, dx} input[b,
                      strides[1] * y + rates[1] * dy,
                      strides[2] * x + rates[2] * dx,
                      c] +
                filter[dy, dx, c]

Max-pooling is a special case when the filter has size equal to the pooling kernel size and contains all zeros.

Note on duality: The dilation of `input` by the `filter` is equal to the negation of the erosion of `-input` by the reflected `filter`.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, depth]`.
filter: 3-D with shape `[filter_height, filter_width, depth]`.
strides: The stride of the sliding window for each dimension of the input

tensor. Must be: `[1, stride_height, stride_width, 1]`. 

rates: The input stride for atrous morphological dilation. Must be:

`[1, rate_height, rate_width, 1]`. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[batch, out_height, out_width, depth]`.

func Dilation2DBackpropFilter 

func Dilation2DBackpropFilter(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, rates []int64, padding string) (filter_backprop tf.Output)

Computes the gradient of morphological 2-D dilation with respect to the filter.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, depth]`.
filter: 3-D with shape `[filter_height, filter_width, depth]`.
out_backprop: 4-D with shape `[batch, out_height, out_width, depth]`.
strides: 1-D of length 4. The stride of the sliding window for each dimension of

the input tensor. Must be: `[1, stride_height, stride_width, 1]`. 

rates: 1-D of length 4. The input stride for atrous morphological dilation.

Must be: `[1, rate_height, rate_width, 1]`. 

padding: The type of padding algorithm to use.

Returns 3-D with shape `[filter_height, filter_width, depth]`.

func Dilation2DBackpropInput 

func Dilation2DBackpropInput(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, rates []int64, padding string) (in_backprop tf.Output)

Computes the gradient of morphological 2-D dilation with respect to the input.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, depth]`.
filter: 3-D with shape `[filter_height, filter_width, depth]`.
out_backprop: 4-D with shape `[batch, out_height, out_width, depth]`.
strides: 1-D of length 4. The stride of the sliding window for each dimension of

the input tensor. Must be: `[1, stride_height, stride_width, 1]`. 

rates: 1-D of length 4. The input stride for atrous morphological dilation.

Must be: `[1, rate_height, rate_width, 1]`. 

padding: The type of padding algorithm to use.

Returns 4-D with shape `[batch, in_height, in_width, depth]`.

func DirectedInterleaveDataset 

func DirectedInterleaveDataset(scope *Scope, selector_input_dataset tf.Output, data_input_datasets []tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...DirectedInterleaveDatasetAttr) (handle tf.Output)

A substitute for `InterleaveDataset` on a fixed list of `N` datasets.

Arguments: 

selector_input_dataset: A dataset of scalar `DT_INT64` elements that determines which of the

`N` data inputs should produce the next output element. 

data_input_datasets: `N` datasets with the same type that will be interleaved according to

the values of `selector_input_dataset`.

func DisableCopyOnRead added in v0.2.0 

func DisableCopyOnRead(scope *Scope, resource tf.Output) (o *tf.Operation)

Turns off the copy-on-read mode.

Turns off the copy-on-read mode of a resource variable. If the variable is not in copy-on-read mode, this op has no effect.

Arguments: 

resource: The resource handle of the resource variable.

Returns the created operation.

func Div 

func Div(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x / y element-wise.

*NOTE*: `Div` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func DivNoNan 

func DivNoNan(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns 0 if the denominator is zero.

*NOTE*: `DivNoNan` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func DrawBoundingBoxes 

func DrawBoundingBoxes(scope *Scope, images tf.Output, boxes tf.Output) (output tf.Output)

Draw bounding boxes on a batch of images.

Outputs a copy of `images` but draws on top of the pixels zero or more bounding boxes specified by the locations in `boxes`. The coordinates of the each bounding box in `boxes` are encoded as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.

For example, if an image is 100 x 200 pixels (height x width) and the bounding box is `[0.1, 0.2, 0.5, 0.9]`, the upper-left and bottom-right coordinates of the bounding box will be `(40, 10)` to `(180, 50)` (in (x,y) coordinates).

Parts of the bounding box may fall outside the image.

Arguments: 

images: 4-D with shape `[batch, height, width, depth]`. A batch of images.
boxes: 3-D with shape `[batch, num_bounding_boxes, 4]` containing bounding

boxes.

Returns 4-D with the same shape as `images`. The batch of input images with bounding boxes drawn on the images.

func DrawBoundingBoxesV2 

func DrawBoundingBoxesV2(scope *Scope, images tf.Output, boxes tf.Output, colors tf.Output) (output tf.Output)

Draw bounding boxes on a batch of images.

Outputs a copy of `images` but draws on top of the pixels zero or more bounding boxes specified by the locations in `boxes`. The coordinates of the each bounding box in `boxes` are encoded as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.

For example, if an image is 100 x 200 pixels (height x width) and the bounding box is `[0.1, 0.2, 0.5, 0.9]`, the upper-left and bottom-right coordinates of the bounding box will be `(40, 10)` to `(100, 50)` (in (x,y) coordinates).

Parts of the bounding box may fall outside the image.

Arguments: 

images: 4-D with shape `[batch, height, width, depth]`. A batch of images.
boxes: 3-D with shape `[batch, num_bounding_boxes, 4]` containing bounding

boxes. 

colors: 2-D. A list of RGBA colors to cycle through for the boxes.

Returns 4-D with the same shape as `images`. The batch of input images with bounding boxes drawn on the images.

func DynamicEnqueueTPUEmbeddingArbitraryTensorBatch 

func DynamicEnqueueTPUEmbeddingArbitraryTensorBatch(scope *Scope, sample_indices_or_row_splits []tf.Output, embedding_indices []tf.Output, aggregation_weights []tf.Output, mode_override tf.Output, device_ordinal tf.Output, optional ...DynamicEnqueueTPUEmbeddingArbitraryTensorBatchAttr) (o *tf.Operation)

Eases the porting of code that uses tf.nn.embedding_lookup_sparse().

embedding_indices[i] and aggregation_weights[i] correspond to the ith feature.

The tensors at corresponding positions in the three input lists (sample_indices, embedding_indices and aggregation_weights) must have the same shape, i.e. rank 1 with dim_size() equal to the total number of lookups into the table described by the corresponding feature.

Arguments: 

sample_indices_or_row_splits: A list of rank 2 Tensors specifying the training example to which the

corresponding embedding_indices and aggregation_weights values belong. If the size of its first dimension is 0, we assume each embedding_indices belongs to a different sample. Both int32 and int64 are allowed and will be converted to int32 internally.

Or a list of rank 1 Tensors specifying the row splits for splitting embedding_indices and aggregation_weights into rows. It corresponds to ids.row_splits in embedding_lookup(), when ids is a RaggedTensor. When enqueuing N-D ragged tensor, only the last dimension is allowed to be ragged. the row splits is 1-D dense tensor. When empty, we assume a dense tensor is passed to the op Both int32 and int64 are allowed and will be converted to int32 internally. 

embedding_indices: A list of rank 1 Tensors, indices into the embedding

tables. Both int32 and int64 are allowed and will be converted to int32 internally. 

aggregation_weights: A list of rank 1 Tensors containing per training

example aggregation weights. Both float32 and float64 are allowed and will be converted to float32 internally. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used. 

device_ordinal: The TPU device to use. Should be >= 0 and less than the number

of TPU cores in the task on which the node is placed.

Returns the created operation.

func DynamicPartition 

func DynamicPartition(scope *Scope, data tf.Output, partitions tf.Output, num_partitions int64) (outputs []tf.Output)

Partitions `data` into `num_partitions` tensors using indices from `partitions`.

For each index tuple `js` of size `partitions.ndim`, the slice `data[js, ...]` becomes part of `outputs[partitions[js]]`. The slices with `partitions[js] = i` are placed in `outputs[i]` in lexicographic order of `js`, and the first dimension of `outputs[i]` is the number of entries in `partitions` equal to `i`. In detail,

```python 

outputs[i].shape = [sum(partitions == i)] + data.shape[partitions.ndim:]

outputs[i] = pack([data[js, ...] for js if partitions[js] == i])

```

`data.shape` must start with `partitions.shape`.

For example:

```python 

# Scalar partitions.
partitions = 1
num_partitions = 2
data = [10, 20]
outputs[0] = []  # Empty with shape [0, 2]
outputs[1] = [[10, 20]]

# Vector partitions.
partitions = [0, 0, 1, 1, 0]
num_partitions = 2
data = [10, 20, 30, 40, 50]
outputs[0] = [10, 20, 50]
outputs[1] = [30, 40]

```

See `dynamic_stitch` for an example on how to merge partitions back.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/DynamicPartition.png" alt> </div>

Raises:

  • `InvalidArgumentError` in following cases:
  • If partitions is not in range `[0, num_partiions)`
  • If `partitions.shape` does not match prefix of `data.shape` argument.

Arguments: 

partitions: Any shape.  Indices in the range `[0, num_partitions)`.
num_partitions: The number of partitions to output.

func DynamicStitch 

func DynamicStitch(scope *Scope, indices []tf.Output, data []tf.Output) (merged tf.Output)

Interleave the values from the `data` tensors into a single tensor.

Builds a merged tensor such that

```python 

merged[indices[m][i, ..., j], ...] = data[m][i, ..., j, ...]

```

For example, if each `indices[m]` is scalar or vector, we have

```python 

# Scalar indices:
merged[indices[m], ...] = data[m][...]

# Vector indices:
merged[indices[m][i], ...] = data[m][i, ...]

```

Each `data[i].shape` must start with the corresponding `indices[i].shape`, and the rest of `data[i].shape` must be constant w.r.t. `i`. That is, we must have `data[i].shape = indices[i].shape + constant`. In terms of this `constant`, the output shape is 

merged.shape = [max(indices) + 1] + constant

Values are merged in order, so if an index appears in both `indices[m][i]` and `indices[n][j]` for `(m,i) < (n,j)` the slice `data[n][j]` will appear in the merged result. If you do not need this guarantee, ParallelDynamicStitch might perform better on some devices.

For example:

```python 

indices[0] = 6
indices[1] = [4, 1]
indices[2] = [[5, 2], [0, 3]]
data[0] = [61, 62]
data[1] = [[41, 42], [11, 12]]
data[2] = [[[51, 52], [21, 22]], [[1, 2], [31, 32]]]
merged = [[1, 2], [11, 12], [21, 22], [31, 32], [41, 42],
          [51, 52], [61, 62]]

```

This method can be used to merge partitions created by `dynamic_partition` as illustrated on the following example:

```python 

# Apply function (increments x_i) on elements for which a certain condition
# apply (x_i != -1 in this example).
x=tf.constant([0.1, -1., 5.2, 4.3, -1., 7.4])
condition_mask=tf.not_equal(x,tf.constant(-1.))
partitioned_data = tf.dynamic_partition(
    x, tf.cast(condition_mask, tf.int32) , 2)
partitioned_data[1] = partitioned_data[1] + 1.0
condition_indices = tf.dynamic_partition(
    tf.range(tf.shape(x)[0]), tf.cast(condition_mask, tf.int32) , 2)
x = tf.dynamic_stitch(condition_indices, partitioned_data)
# Here x=[1.1, -1., 6.2, 5.3, -1, 8.4], the -1. values remain
# unchanged.

```

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/DynamicStitch.png" alt> </div>

func EagerPyFunc 

func EagerPyFunc(scope *Scope, input []tf.Output, token string, Tout []tf.DataType, optional ...EagerPyFuncAttr) (output []tf.Output)

Eagerly executes a python function to compute func(input)->output. The

semantics of the input, output, and attributes are the same as those for PyFunc.

func EditDistance 

func EditDistance(scope *Scope, hypothesis_indices tf.Output, hypothesis_values tf.Output, hypothesis_shape tf.Output, truth_indices tf.Output, truth_values tf.Output, truth_shape tf.Output, optional ...EditDistanceAttr) (output tf.Output)

Computes the (possibly normalized) Levenshtein Edit Distance.

The inputs are variable-length sequences provided by SparseTensors 

(hypothesis_indices, hypothesis_values, hypothesis_shape)

and 

(truth_indices, truth_values, truth_shape).

The inputs are:

Arguments: 

hypothesis_indices: The indices of the hypothesis list SparseTensor.

This is an N x R int64 matrix. 

hypothesis_values: The values of the hypothesis list SparseTensor.

This is an N-length vector. 

hypothesis_shape: The shape of the hypothesis list SparseTensor.

This is an R-length vector. 

truth_indices: The indices of the truth list SparseTensor.

This is an M x R int64 matrix. 

truth_values: The values of the truth list SparseTensor.

This is an M-length vector. 

truth_shape: truth indices, vector.

Returns A dense float tensor with rank R - 1.

For the example input: 

// hypothesis represents a 2x1 matrix with variable-length values:
//   (0,0) = ["a"]
//   (1,0) = ["b"]
hypothesis_indices = [[0, 0, 0],
                      [1, 0, 0]]
hypothesis_values = ["a", "b"]
hypothesis_shape = [2, 1, 1]

// truth represents a 2x2 matrix with variable-length values:
//   (0,0) = []
//   (0,1) = ["a"]
//   (1,0) = ["b", "c"]
//   (1,1) = ["a"]
truth_indices = [[0, 1, 0],
                 [1, 0, 0],
                 [1, 0, 1],
                 [1, 1, 0]]
truth_values = ["a", "b", "c", "a"]
truth_shape = [2, 2, 2]
normalize = true

The output will be: 

// output is a 2x2 matrix with edit distances normalized by truth lengths.
output = [[inf, 1.0],  // (0,0): no truth, (0,1): no hypothesis
          [0.5, 1.0]]  // (1,0): addition, (1,1): no hypothesis

func Eig 

func Eig(scope *Scope, input tf.Output, Tout tf.DataType, optional ...EigAttr) (e tf.Output, v tf.Output)

Computes the eigen decomposition of one or more square matrices.

Computes the eigenvalues and (optionally) right eigenvectors of each inner matrix in `input` such that `input[..., :, :] = v[..., :, :] * diag(e[..., :])`. The eigenvalues are sorted in non-decreasing order.

```python # a is a tensor. # e is a tensor of eigenvalues. # v is a tensor of eigenvectors. e, v = eig(a) e = eig(a, compute_v=False) ```

Arguments: 

input: `Tensor` input of shape `[N, N]`.

Returns: 

e: Eigenvalues. Shape is `[N]`.
v: Eigenvectors. Shape is `[N, N]`.

func Einsum 

func Einsum(scope *Scope, inputs []tf.Output, equation string) (output tf.Output)

Tensor contraction according to Einstein summation convention.

Implements generalized Tensor contraction and reduction. Each input Tensor must have a corresponding input subscript appearing in the comma-separated left-hand side of the equation. The right-hand side of the equation consists of the output subscript. The input subscripts and the output subscript should consist of zero or more named axis labels and at most one ellipsis (`...`).

The named axis labels may be any single character other than those having special meaning, namely `,.->`. The behavior of this Op is undefined if it receives an ill-formatted equation; since the validation is done at graph-building time, we omit format validation checks at runtime.

Note: This Op is *not* intended to be called by the user; instead users should call `tf.einsum` directly. It is a hidden Op used by `tf.einsum`.

Operations are applied to the input(s) according to the following rules: 

(a) Generalized Diagonals: For input dimensions corresponding to axis labels
    appearing more than once in the same input subscript, we take the
    generalized (`k`-dimensional) diagonal.
    For example, in the equation `iii->i` with input shape `[3, 3, 3]`, the
    generalized diagonal would consist of `3` elements at indices `(0, 0, 0)`,
    `(1, 1, 1)` and `(2, 2, 2)` to create a Tensor of shape `[3]`.

(b) Reduction: Axes corresponding to labels appearing only in one input
    subscript but not in the output subscript are summed over prior to Tensor
    contraction.
    For example, in the equation `ab,bc->b`, the axis labels `a` and `c` are
    the reduction axis labels.

(c) Batch Dimensions: Axes corresponding to labels appearing in each of the
    input subscripts and also in the output subscript make up the batch
    dimensions in Tensor contraction. Unnamed axis labels corresponding to
    ellipsis (`...`) also correspond to batch dimensions.
    For example, for the equation denoting batch matrix multiplication,
    `bij,bjk->bik`, the axis label `b` corresponds to a batch dimension.

(d) Contraction: In case of binary einsum, axes corresponding to labels
    appearing in two different inputs (and not in the output) are contracted
    against each other.
    Considering the batch matrix multiplication equation again
    (`bij,bjk->bik`), the contracted axis label is `j`.

(e) Expand Diagonal: If the output subscripts contain repeated (explicit) axis
    labels, the opposite operation of (a) is applied. For example, in the
    equation `i->iii`, and input shape `[3]`, the output of shape `[3, 3, 3]`
    are all zeros, except for the (generalized) diagonal which is populated
    with values from the input.
    Note: This operation is not supported by `np.einsum` or `tf.einsum`; it is
    provided to enable computing the symbolic gradient of `tf.einsum`.

The output subscripts must contain only labels appearing in at least one of the input subscripts. Furthermore, all dimensions mapping to the same axis label must be equal.

Any of the input and output subscripts may contain at most a single ellipsis (`...`). These ellipsis are mapped against dimensions not corresponding to any named axis label. If two inputs contain ellipsis, then they are broadcasted according to standard NumPy broadcasting [rules](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html).

The broadcasted dimensions are placed in the corresponding location of the ellipsis in the output subscript. If the broadcasted dimensions are non-empty and the output subscripts do not contain ellipsis, then an InvalidArgument error is raised.

@compatibility(numpy) Similar to [`numpy.einsum`](https://docs.scipy.org/doc/numpy/reference/generated/numpy.einsum.html).

Comparison with `numpy.einsum`:

  • This Op only supports unary and binary forms of `numpy.einsum`.
  • This Op does not support implicit form. (i.e. equations without `->`).
  • This Op also supports repeated indices in the output subscript, which is not supported by `numpy.einsum`.

@end_compatibility

Arguments: 

inputs: List of 1 or 2 Tensors.
equation: String describing the Einstein Summation operation; in the format of np.einsum.

Returns Output Tensor with shape depending upon `equation`.

func Elu 

func Elu(scope *Scope, features tf.Output) (activations tf.Output)

Computes the exponential linear function.

The ELU function is defined as:

  • $ e ^ x - 1 $ if $ x < 0 $
  • $ x $ if $ x >= 0 $

Examples:

>>> tf.nn.elu(1.0) <tf.Tensor: shape=(), dtype=float32, numpy=1.0> >>> tf.nn.elu(0.0) <tf.Tensor: shape=(), dtype=float32, numpy=0.0> >>> tf.nn.elu(-1000.0) <tf.Tensor: shape=(), dtype=float32, numpy=-1.0>

See [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) ](http://arxiv.org/abs/1511.07289)

func EluGrad 

func EluGrad(scope *Scope, gradients tf.Output, outputs tf.Output) (backprops tf.Output)

Computes gradients for the exponential linear (Elu) operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding Elu operation.
outputs: The outputs of the corresponding Elu operation.

Returns The gradients: `gradients * (outputs + 1)` if outputs < 0, `gradients` otherwise.

func Empty 

func Empty(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...EmptyAttr) (output tf.Output)

Creates a tensor with the given shape.

This operation creates a tensor of `shape` and `dtype`.

Arguments: 

shape: 1-D. Represents the shape of the output tensor.

Returns A `Tensor` of type `T`.

func EmptyTensorList 

func EmptyTensorList(scope *Scope, element_shape tf.Output, max_num_elements tf.Output, element_dtype tf.DataType) (handle tf.Output)

Creates and returns an empty tensor list.

All list elements must be tensors of dtype element_dtype and shape compatible with element_shape.

handle: an empty tensor list. element_dtype: the type of elements in the list. element_shape: a shape compatible with that of elements in the list.

func EmptyTensorMap 

func EmptyTensorMap(scope *Scope) (handle tf.Output)

Creates and returns an empty tensor map.

handle: an empty tensor map

func EncodeBase64 

func EncodeBase64(scope *Scope, input tf.Output, optional ...EncodeBase64Attr) (output tf.Output)

Encode strings into web-safe base64 format.

Refer to [this article](https://en.wikipedia.org/wiki/Base64) for more information on base64 format. Base64 strings may have padding with '=' at the end so that the encoded has length multiple of 4. See Padding section of the link above.

Web-safe means that the encoder uses - and _ instead of + and /.

Arguments: 

input: Strings to be encoded.

Returns Input strings encoded in base64.

func EncodeJpeg 

func EncodeJpeg(scope *Scope, image tf.Output, optional ...EncodeJpegAttr) (contents tf.Output)

JPEG-encode an image.

`image` is a 3-D uint8 Tensor of shape `[height, width, channels]`.

The attr `format` can be used to override the color format of the encoded output. Values can be:

  • `”`: Use a default format based on the number of channels in the image.
  • `grayscale`: Output a grayscale JPEG image. The `channels` dimension of `image` must be 1.
  • `rgb`: Output an RGB JPEG image. The `channels` dimension of `image` must be 3.

If `format` is not specified or is the empty string, a default format is picked in function of the number of channels in `image`:

* 1: Output a grayscale image. * 3: Output an RGB image.

Arguments: 

image: 3-D with shape `[height, width, channels]`.

Returns 0-D. JPEG-encoded image.

func EncodeJpegVariableQuality 

func EncodeJpegVariableQuality(scope *Scope, images tf.Output, quality tf.Output) (contents tf.Output)

JPEG encode input image with provided compression quality.

`image` is a 3-D uint8 Tensor of shape `[height, width, channels]`. `quality` is an int32 jpeg compression quality value between 0 and 100.

Arguments: 

images: Images to adjust.  At least 3-D.
quality: An int quality to encode to.

Returns 0-D. JPEG-encoded image.

func EncodePng 

func EncodePng(scope *Scope, image tf.Output, optional ...EncodePngAttr) (contents tf.Output)

PNG-encode an image.

`image` is a 3-D uint8 or uint16 Tensor of shape `[height, width, channels]` where `channels` is:

* 1: for grayscale. * 2: for grayscale + alpha. * 3: for RGB. * 4: for RGBA.

The ZLIB compression level, `compression`, can be -1 for the PNG-encoder default or a value from 0 to 9. 9 is the highest compression level, generating the smallest output, but is slower.

Arguments: 

image: 3-D with shape `[height, width, channels]`.

Returns 0-D. PNG-encoded image.

func EncodeProto 

func EncodeProto(scope *Scope, sizes tf.Output, values []tf.Output, field_names []string, message_type string, optional ...EncodeProtoAttr) (bytes tf.Output)

The op serializes protobuf messages provided in the input tensors.

The types of the tensors in `values` must match the schema for the fields specified in `field_names`. All the tensors in `values` must have a common shape prefix, *batch_shape*.

The `sizes` tensor specifies repeat counts for each field. The repeat count (last dimension) of a each tensor in `values` must be greater than or equal to corresponding repeat count in `sizes`.

A `message_type` name must be provided to give context for the field names. The actual message descriptor can be looked up either in the linked-in descriptor pool or a filename provided by the caller using the `descriptor_source` attribute.

For the most part, the mapping between Proto field types and TensorFlow dtypes is straightforward. However, there are a few special cases:

- A proto field that contains a submessage or group can only be converted to `DT_STRING` (the serialized submessage). This is to reduce the complexity of the API. The resulting string can be used as input to another instance of the decode_proto op.

- TensorFlow lacks support for unsigned integers. The ops represent uint64 types as a `DT_INT64` with the same twos-complement bit pattern (the obvious way). Unsigned int32 values can be represented exactly by specifying type `DT_INT64`, or using twos-complement if the caller specifies `DT_INT32` in the `output_types` attribute.

The `descriptor_source` attribute selects the source of protocol descriptors to consult when looking up `message_type`. This may be:

- An empty string or "local://", in which case protocol descriptors are created for C++ (not Python) proto definitions linked to the binary.

- A file, in which case protocol descriptors are created from the file, which is expected to contain a `FileDescriptorSet` serialized as a string. NOTE: You can build a `descriptor_source` file using the `--descriptor_set_out` and `--include_imports` options to the protocol compiler `protoc`.

- A "bytes://<bytes>", in which protocol descriptors are created from `<bytes>`, which is expected to be a `FileDescriptorSet` serialized as a string.

Arguments: 

sizes: Tensor of int32 with shape `[batch_shape, len(field_names)]`.
values: List of tensors containing values for the corresponding field.
field_names: List of strings containing proto field names.
message_type: Name of the proto message type to decode.

Returns Tensor of serialized protos with shape `batch_shape`.

func EncodeWav 

func EncodeWav(scope *Scope, audio tf.Output, sample_rate tf.Output) (contents tf.Output)

Encode audio data using the WAV file format.

This operation will generate a string suitable to be saved out to create a .wav audio file. It will be encoded in the 16-bit PCM format. It takes in float values in the range -1.0f to 1.0f, and any outside that value will be clamped to that range.

`audio` is a 2-D float Tensor of shape `[length, channels]`. `sample_rate` is a scalar Tensor holding the rate to use (e.g. 44100).

Arguments: 

audio: 2-D with shape `[length, channels]`.
sample_rate: Scalar containing the sample frequency.

Returns 0-D. WAV-encoded file contents.

func EnqueueTPUEmbeddingArbitraryTensorBatch 

func EnqueueTPUEmbeddingArbitraryTensorBatch(scope *Scope, sample_indices_or_row_splits []tf.Output, embedding_indices []tf.Output, aggregation_weights []tf.Output, mode_override tf.Output, optional ...EnqueueTPUEmbeddingArbitraryTensorBatchAttr) (o *tf.Operation)

Eases the porting of code that uses tf.nn.embedding_lookup_sparse().

embedding_indices[i] and aggregation_weights[i] correspond to the ith feature.

The tensors at corresponding positions in the three input lists (sample_indices, embedding_indices and aggregation_weights) must have the same shape, i.e. rank 1 with dim_size() equal to the total number of lookups into the table described by the corresponding feature.

Arguments: 

sample_indices_or_row_splits: A list of rank 2 Tensors specifying the training example to which the

corresponding embedding_indices and aggregation_weights values belong. If the size of its first dimension is 0, we assume each embedding_indices belongs to a different sample. Both int32 and int64 are allowed and will be converted to int32 internally.

Or a list of rank 1 Tensors specifying the row splits for splitting embedding_indices and aggregation_weights into rows. It corresponds to ids.row_splits in embedding_lookup(), when ids is a RaggedTensor. When enqueuing N-D ragged tensor, only the last dimension is allowed to be ragged. the row splits is 1-D dense tensor. When empty, we assume a dense tensor is passed to the op Both int32 and int64 are allowed and will be converted to int32 internally. 

embedding_indices: A list of rank 1 Tensors, indices into the embedding

tables. Both int32 and int64 are allowed and will be converted to int32 internally. 

aggregation_weights: A list of rank 1 Tensors containing per training

example aggregation weights. Both float32 and float64 are allowed and will be converted to float32 internally. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used.

Returns the created operation.

func EnqueueTPUEmbeddingBatch 

func EnqueueTPUEmbeddingBatch(scope *Scope, batch []tf.Output, mode_override tf.Output, optional ...EnqueueTPUEmbeddingBatchAttr) (o *tf.Operation)

An op that enqueues a list of input batch tensors to TPUEmbedding.

An op that enqueues a list of input batch tensors to TPUEmbedding.

Arguments: 

batch: A list of 1D tensors, one for each embedding table, containing the

batch inputs encoded as dist_belief.SparseFeatures protos. If the weight field in the SparseFeatures proto is not populated for an ID, a weight of 1.0 is assumed. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used.

Returns the created operation.

func EnqueueTPUEmbeddingIntegerBatch 

func EnqueueTPUEmbeddingIntegerBatch(scope *Scope, batch []tf.Output, mode_override tf.Output, optional ...EnqueueTPUEmbeddingIntegerBatchAttr) (o *tf.Operation)

An op that enqueues a list of input batch tensors to TPUEmbedding.

Arguments: 

batch: A list of 1D tensors, one for each embedding table, containing the

indices into the tables. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used.

Returns the created operation.

func EnqueueTPUEmbeddingRaggedTensorBatch 

func EnqueueTPUEmbeddingRaggedTensorBatch(scope *Scope, sample_splits []tf.Output, embedding_indices []tf.Output, aggregation_weights []tf.Output, mode_override tf.Output, table_ids []int64, optional ...EnqueueTPUEmbeddingRaggedTensorBatchAttr) (o *tf.Operation)

Eases the porting of code that uses tf.nn.embedding_lookup().

sample_splits[i], embedding_indices[i] and aggregation_weights[i] correspond to the ith feature. table_ids[i] indicates which embedding table to look up ith feature.

The tensors at corresponding positions in two of the input lists, embedding_indices and aggregation_weights, must have the same shape, i.e. rank 1 with dim_size() equal to the total number of lookups into the table described by the corresponding feature.

Arguments: 

sample_splits: A list of rank 1 Tensors specifying the break points for splitting

embedding_indices and aggregation_weights into rows. It corresponds to ids.row_splits in embedding_lookup(), when ids is a RaggedTensor. 

embedding_indices: A list of rank 1 Tensors, indices into the embedding tables.

It corresponds to ids.values in embedding_lookup(), when ids is a RaggedTensor. 

aggregation_weights: A list of rank 1 Tensors containing per training example

aggregation weights. It corresponds to the values field of a RaggedTensor with the same row_splits as ids in embedding_lookup(), when ids is a RaggedTensor. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used. 

table_ids: A list of integers specifying the identifier of the embedding table

(offset of TableDescriptor in the TPUEmbeddingConfiguration) to lookup the corresponding input. The ith input is looked up using table_ids[i]. The size of the table_ids list must be equal to that of sample_indices, embedding_indices and aggregation_weights.

Returns the created operation.

func EnqueueTPUEmbeddingSparseBatch 

func EnqueueTPUEmbeddingSparseBatch(scope *Scope, sample_indices []tf.Output, embedding_indices []tf.Output, aggregation_weights []tf.Output, mode_override tf.Output, optional ...EnqueueTPUEmbeddingSparseBatchAttr) (o *tf.Operation)

An op that enqueues TPUEmbedding input indices from a SparseTensor.

This Op eases the porting of code that uses embedding_lookup_sparse(), although some Python preprocessing of the SparseTensor arguments to embedding_lookup_sparse() is required to produce the arguments to this Op, since only a single EnqueueTPUEmbeddingSparseBatch Op is allowed per training step.

The tensors at corresponding positions in the three input lists must have the same shape, i.e. rank 1 with dim_size() equal to the total number of lookups into the table described by the corresponding table_id.

Arguments: 

sample_indices: A list of rank 1 Tensors specifying the training example and

feature to which the corresponding embedding_indices and aggregation_weights values belong. sample_indices[i] must equal b * nf + f, where nf is the number of features from the corresponding table, f is in [0, nf), and b is in [0, batch size). 

embedding_indices: A list of rank 1 Tensors, indices into the embedding tables.
aggregation_weights: A list of rank 1 Tensors containing per sample -- i.e. per

(training example, feature) -- aggregation weights. 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used.

Returns the created operation.

func EnqueueTPUEmbeddingSparseTensorBatch 

func EnqueueTPUEmbeddingSparseTensorBatch(scope *Scope, sample_indices []tf.Output, embedding_indices []tf.Output, aggregation_weights []tf.Output, mode_override tf.Output, table_ids []int64, optional ...EnqueueTPUEmbeddingSparseTensorBatchAttr) (o *tf.Operation)

Eases the porting of code that uses tf.nn.embedding_lookup_sparse().

sample_indices[i], embedding_indices[i] and aggregation_weights[i] correspond to the ith feature. table_ids[i] indicates which embedding table to look up ith feature.

The tensors at corresponding positions in the three input lists (sample_indices, embedding_indices and aggregation_weights) must have the same shape, i.e. rank 1 with dim_size() equal to the total number of lookups into the table described by the corresponding feature.

Arguments: 

sample_indices: A list of rank 1 Tensors specifying the training example to

which the corresponding embedding_indices and aggregation_weights values belong. It corresponds to sp_ids.indices[:,0] in embedding_lookup_sparse(). 

embedding_indices: A list of rank 1 Tensors, indices into the embedding tables.

It corresponds to sp_ids.values in embedding_lookup_sparse(). 

aggregation_weights: A list of rank 1 Tensors containing per training example

aggregation weights. It corresponds to sp_weights.values in embedding_lookup_sparse(). 

mode_override: A string input that overrides the mode specified in the

TPUEmbeddingConfiguration. Supported values are {'unspecified', 'inference', 'training', 'backward_pass_only'}. When set to 'unspecified', the mode set in TPUEmbeddingConfiguration is used, otherwise mode_override is used. 

table_ids: A list of integers specifying the identifier of the embedding table

(offset of TableDescriptor in the TPUEmbeddingConfiguration) to lookup the corresponding input. The ith input is looked up using table_ids[i]. The size of the table_ids list must be equal to that of sample_indices, embedding_indices and aggregation_weights.

Returns the created operation.

func EnsureShape 

func EnsureShape(scope *Scope, input tf.Output, shape tf.Shape) (output tf.Output)

Ensures that the tensor's shape matches the expected shape.

Raises an error if the input tensor's shape does not match the specified shape. Returns the input tensor otherwise.

Arguments: 

input: A tensor, whose shape is to be validated.
shape: The expected (possibly partially specified) shape of the input tensor.

Returns A tensor with the same shape and contents as the input tensor or value.

func Enter 

func Enter(scope *Scope, data tf.Output, frame_name string, optional ...EnterAttr) (output tf.Output)

Creates or finds a child frame, and makes `data` available to the child frame.

This op is used together with `Exit` to create loops in the graph. The unique `frame_name` is used by the `Executor` to identify frames. If `is_constant` is true, `output` is a constant in the child frame; otherwise it may be changed in the child frame. At most `parallel_iterations` iterations are run in parallel in the child frame.

Arguments: 

data: The tensor to be made available to the child frame.
frame_name: The name of the child frame.

Returns The same tensor as `data`.

func Equal 

func Equal(scope *Scope, x tf.Output, y tf.Output, optional ...EqualAttr) (z tf.Output)

Returns the truth value of (x == y) element-wise.

*NOTE*: `Equal` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

```python x = tf.constant([2, 4]) y = tf.constant(2) tf.math.equal(x, y) ==> array([True, False])

x = tf.constant([2, 4]) y = tf.constant([2, 4]) tf.math.equal(x, y) ==> array([True, True]) ```

func Erf 

func Erf(scope *Scope, x tf.Output) (y tf.Output)

Computes the [Gauss error function](https://en.wikipedia.org/wiki/Error_function) of `x` element-wise. In statistics, for non-negative values of $x$, the error function has the following interpretation: for a random variable $Y$ that is normally distributed with mean 0 and variance $1/\sqrt{2}$, $erf(x)$ is the probability that $Y$ falls in the range $[−x, x]$.

func Erfc 

func Erfc(scope *Scope, x tf.Output) (y tf.Output)

Computes the complementary error function of `x` element-wise.

func EuclideanNorm 

func EuclideanNorm(scope *Scope, input tf.Output, axis tf.Output, optional ...EuclideanNormAttr) (output tf.Output)

Computes the euclidean norm of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func ExecuteTPUEmbeddingPartitioner added in v0.2.0 

func ExecuteTPUEmbeddingPartitioner(scope *Scope, config string) (common_config tf.Output)

An op that executes the TPUEmbedding partitioner on the central configuration

device and computes the HBM size (in bytes) required for TPUEmbedding operation.

Arguments: 

config: An TPUEmbeddingConfiguration proto serialized to a string,

describing the desired TPUEmbedding configuration.

Returns A string-encoded common configuration proto containing metadata about the TPUEmbedding partitioner output and the HBM size (in bytes) required for operation.

func Exit 

func Exit(scope *Scope, data tf.Output) (output tf.Output)

Exits the current frame to its parent frame.

Exit makes its input `data` available to the parent frame.

Arguments: 

data: The tensor to be made available to the parent frame.

Returns The same tensor as `data`.

func Exp 

func Exp(scope *Scope, x tf.Output) (y tf.Output)

Computes exponential of x element-wise. \\(y = e^x\\). 

This function computes the exponential of every element in the input tensor.
i.e. `exp(x)` or `e^(x)`, where `x` is the input tensor.
`e` denotes Euler's number and is approximately equal to 2.718281.
Output is positive for any real input.

```python
x = tf.constant(2.0)
tf.math.exp(x) ==> 7.389056

x = tf.constant([2.0, 8.0])
tf.math.exp(x) ==> array([7.389056, 2980.958], dtype=float32)
```

For complex numbers, the exponential value is calculated as follows:

```
e^(x+iy) = e^x * e^iy = e^x * (cos y + i sin y)
```

Let's consider complex number 1+1j as an example.
e^1 * (cos 1 + i sin 1) = 2.7182818284590 * (0.54030230586+0.8414709848j)

```python
x = tf.constant(1 + 1j)
tf.math.exp(x) ==> 1.4686939399158851+2.2873552871788423j
```

func ExpandDims 

func ExpandDims(scope *Scope, input tf.Output, axis tf.Output) (output tf.Output)

Inserts a dimension of 1 into a tensor's shape.

Given a tensor `input`, this operation inserts a dimension of 1 at the dimension index `axis` of `input`'s shape. The dimension index `axis` starts at zero; if you specify a negative number for `axis` it is counted backward from the end.

This operation is useful if you want to add a batch dimension to a single element. For example, if you have a single image of shape `[height, width, channels]`, you can make it a batch of 1 image with `expand_dims(image, 0)`, which will make the shape `[1, height, width, channels]`.

Other examples:

``` # 't' is a tensor of shape [2] shape(expand_dims(t, 0)) ==> [1, 2] shape(expand_dims(t, 1)) ==> [2, 1] shape(expand_dims(t, -1)) ==> [2, 1]

# 't2' is a tensor of shape [2, 3, 5] shape(expand_dims(t2, 0)) ==> [1, 2, 3, 5] shape(expand_dims(t2, 2)) ==> [2, 3, 1, 5] shape(expand_dims(t2, 3)) ==> [2, 3, 5, 1] ```

This operation requires that:

`-1-input.dims() <= dim <= input.dims()`

This operation is related to `squeeze()`, which removes dimensions of size 1.

Arguments: 

axis: 0-D (scalar). Specifies the dimension index at which to

expand the shape of `input`. Must be in the range `[-rank(input) - 1, rank(input)]`.

Returns Contains the same data as `input`, but its shape has an additional dimension of size 1 added.

func ExperimentalAutoShardDataset 

func ExperimentalAutoShardDataset(scope *Scope, input_dataset tf.Output, num_workers tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ExperimentalAutoShardDatasetAttr) (handle tf.Output)

Creates a dataset that shards the input dataset.

Creates a dataset that shards the input dataset by num_workers, returning a sharded dataset for the index-th worker. This attempts to automatically shard a dataset by examining the Dataset graph and inserting a shard op before the inputs to a reader Dataset (e.g. CSVDataset, TFRecordDataset).

This dataset will throw a NotFound error if we cannot shard the dataset automatically.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
num_workers: A scalar representing the number of workers to distribute this dataset across.
index: A scalar representing the index of the current worker out of num_workers.

func ExperimentalBytesProducedStatsDataset 

func ExperimentalBytesProducedStatsDataset(scope *Scope, input_dataset tf.Output, tag tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Records the bytes size of each element of `input_dataset` in a StatsAggregator.

func ExperimentalDatasetCardinality 

func ExperimentalDatasetCardinality(scope *Scope, input_dataset tf.Output) (cardinality tf.Output)

Returns the cardinality of `input_dataset`.

Returns the cardinality of `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the dataset to return cardinality for.

Returns The cardinality of `input_dataset`. Named constants are used to represent infinite and unknown cardinality.

func ExperimentalDatasetToTFRecord 

func ExperimentalDatasetToTFRecord(scope *Scope, input_dataset tf.Output, filename tf.Output, compression_type tf.Output) (o *tf.Operation)

Writes the given dataset to the given file using the TFRecord format.

Arguments: 

input_dataset: A variant tensor representing the dataset to write.
filename: A scalar string tensor representing the filename to use.
compression_type: A scalar string tensor containing either (i) the empty string (no

compression), (ii) "ZLIB", or (iii) "GZIP".

Returns the created operation.

func ExperimentalDenseToSparseBatchDataset 

func ExperimentalDenseToSparseBatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, row_shape tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that batches input elements into a SparseTensor.

Arguments: 

input_dataset: A handle to an input dataset. Must have a single component.
batch_size: A scalar representing the number of elements to accumulate in a

batch. 

row_shape: A vector representing the dense shape of each row in the produced

SparseTensor. The shape may be partially specified, using `-1` to indicate that a particular dimension should use the maximum size of all batch elements.

func ExperimentalDirectedInterleaveDataset 

func ExperimentalDirectedInterleaveDataset(scope *Scope, selector_input_dataset tf.Output, data_input_datasets []tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A substitute for `InterleaveDataset` on a fixed list of `N` datasets.

Arguments: 

selector_input_dataset: A dataset of scalar `DT_INT64` elements that determines which of the

`N` data inputs should produce the next output element. 

data_input_datasets: `N` datasets with the same type that will be interleaved according to

the values of `selector_input_dataset`.

func ExperimentalIgnoreErrorsDataset 

func ExperimentalIgnoreErrorsDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ExperimentalIgnoreErrorsDatasetAttr) (handle tf.Output)

Creates a dataset that contains the elements of `input_dataset` ignoring errors.

func ExperimentalIteratorGetDevice 

func ExperimentalIteratorGetDevice(scope *Scope, resource tf.Output) (device tf.Output)

Returns the name of the device on which `resource` has been placed.

func ExperimentalLatencyStatsDataset 

func ExperimentalLatencyStatsDataset(scope *Scope, input_dataset tf.Output, tag tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Records the latency of producing `input_dataset` elements in a StatsAggregator.

func ExperimentalMaxIntraOpParallelismDataset 

func ExperimentalMaxIntraOpParallelismDataset(scope *Scope, input_dataset tf.Output, max_intra_op_parallelism tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that overrides the maximum intra-op parallelism.

Arguments: 

max_intra_op_parallelism: Identifies the maximum intra-op parallelism to use.

func ExperimentalParseExampleDataset 

func ExperimentalParseExampleDataset(scope *Scope, input_dataset tf.Output, num_parallel_calls tf.Output, dense_defaults []tf.Output, sparse_keys []string, dense_keys []string, sparse_types []tf.DataType, dense_shapes []tf.Shape, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ExperimentalParseExampleDatasetAttr) (handle tf.Output)

Transforms `input_dataset` containing `Example` protos as vectors of DT_STRING into a dataset of `Tensor` or `SparseTensor` objects representing the parsed features.

Arguments: 

dense_defaults: A dict mapping string keys to `Tensor`s.

The keys of the dict must match the dense_keys of the feature. 

sparse_keys: A list of string keys in the examples features.

The results for these keys will be returned as `SparseTensor` objects. 

dense_keys: A list of Ndense string Tensors (scalars).

The keys expected in the Examples features associated with dense values. 

sparse_types: A list of `DTypes` of the same length as `sparse_keys`.

Only `tf.float32` (`FloatList`), `tf.int64` (`Int64List`), and `tf.string` (`BytesList`) are supported. 

dense_shapes: List of tuples with the same length as `dense_keys`.

The shape of the data for each dense feature referenced by `dense_keys`. Required for any input tensors identified by `dense_keys`. Must be either fully defined, or may contain an unknown first dimension. An unknown first dimension means the feature is treated as having a variable number of blocks, and the output shape along this dimension is considered unknown at graph build time. Padding is applied for minibatch elements smaller than the maximum number of blocks for the given feature along this dimension. 

output_types: The type list for the return values.
output_shapes: The list of shapes being produced.

func ExperimentalPrivateThreadPoolDataset 

func ExperimentalPrivateThreadPoolDataset(scope *Scope, input_dataset tf.Output, num_threads tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

num_threads: Identifies the number of threads to use for the private threadpool.

func ExperimentalRandomDataset 

func ExperimentalRandomDataset(scope *Scope, seed tf.Output, seed2 tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a Dataset that returns pseudorandom numbers.

Arguments: 

seed: A scalar seed for the random number generator. If either seed or

seed2 is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. 

seed2: A second scalar seed to avoid seed collision.

func ExperimentalRebatchDataset 

func ExperimentalRebatchDataset(scope *Scope, input_dataset tf.Output, num_replicas tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ExperimentalRebatchDatasetAttr) (handle tf.Output)

Creates a dataset that changes the batch size.

Creates a dataset that changes the batch size of the dataset to current batch size // num_replicas.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
num_replicas: A scalar representing the number of replicas to distribute this batch across. As

a result of this transformation the current batch size would end up being divided by this parameter.

func ExperimentalSlidingWindowDataset 

func ExperimentalSlidingWindowDataset(scope *Scope, input_dataset tf.Output, window_size tf.Output, window_shift tf.Output, window_stride tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that passes a sliding window over `input_dataset`.

Arguments: 

window_size: A scalar representing the number of elements in the

sliding window. 

window_shift: A scalar representing the steps moving the sliding window

forward in one iteration. It must be positive. 

window_stride: A scalar representing the stride of the input elements of the sliding window.

It must be positive.

func ExperimentalSqlDataset 

func ExperimentalSqlDataset(scope *Scope, driver_name tf.Output, data_source_name tf.Output, query tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that executes a SQL query and emits rows of the result set.

Arguments: 

driver_name: The database type. Currently, the only supported type is 'sqlite'.
data_source_name: A connection string to connect to the database.
query: A SQL query to execute.

func ExperimentalStatsAggregatorHandle 

func ExperimentalStatsAggregatorHandle(scope *Scope, optional ...ExperimentalStatsAggregatorHandleAttr) (handle tf.Output)

Creates a statistics manager resource.

func ExperimentalStatsAggregatorSummary 

func ExperimentalStatsAggregatorSummary(scope *Scope, iterator tf.Output) (summary tf.Output)

Produces a summary of any statistics recorded by the given statistics manager.

func ExperimentalThreadPoolDataset 

func ExperimentalThreadPoolDataset(scope *Scope, input_dataset tf.Output, thread_pool tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

thread_pool: A resource produced by the ThreadPoolHandle op.

func ExperimentalThreadPoolHandle 

func ExperimentalThreadPoolHandle(scope *Scope, num_threads int64, display_name string, optional ...ExperimentalThreadPoolHandleAttr) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

num_threads: The number of threads in the thread pool.
display_name: A human-readable name for the threads that may be visible in some

visualizations. threadpool.

Returns A resource that can be consumed by one or more ExperimentalThreadPoolDataset ops.

func ExperimentalUnbatchDataset 

func ExperimentalUnbatchDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A dataset that splits the elements of its input into multiple elements.

func ExperimentalUniqueDataset 

func ExperimentalUniqueDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that contains the unique elements of `input_dataset`.

func Expm1 

func Expm1(scope *Scope, x tf.Output) (y tf.Output)

Computes `exp(x) - 1` element-wise. 

i.e. `exp(x) - 1` or `e^(x) - 1`, where `x` is the input tensor.
`e` denotes Euler's number and is approximately equal to 2.718281.

```python
x = tf.constant(2.0)
tf.math.expm1(x) ==> 6.389056

x = tf.constant([2.0, 8.0])
tf.math.expm1(x) ==> array([6.389056, 2979.958], dtype=float32)

x = tf.constant(1 + 1j)
tf.math.expm1(x) ==> (0.46869393991588515+2.2873552871788423j)
```

func ExtractGlimpse 

func ExtractGlimpse(scope *Scope, input tf.Output, size tf.Output, offsets tf.Output, optional ...ExtractGlimpseAttr) (glimpse tf.Output)

Extracts a glimpse from the input tensor.

Returns a set of windows called glimpses extracted at location `offsets` from the input tensor. If the windows only partially overlaps the inputs, the non overlapping areas will be filled with random noise.

The result is a 4-D tensor of shape `[batch_size, glimpse_height, glimpse_width, channels]`. The channels and batch dimensions are the same as that of the input tensor. The height and width of the output windows are specified in the `size` parameter.

The argument `normalized` and `centered` controls how the windows are built:

  • If the coordinates are normalized but not centered, 0.0 and 1.0 correspond to the minimum and maximum of each height and width dimension.
  • If the coordinates are both normalized and centered, they range from -1.0 to 1.0. The coordinates (-1.0, -1.0) correspond to the upper left corner, the lower right corner is located at (1.0, 1.0) and the center is at (0, 0).
  • If the coordinates are not normalized they are interpreted as numbers of pixels.

Arguments: 

input: A 4-D float tensor of shape `[batch_size, height, width, channels]`.
size: A 1-D tensor of 2 elements containing the size of the glimpses

to extract. The glimpse height must be specified first, following by the glimpse width. 

offsets: A 2-D integer tensor of shape `[batch_size, 2]` containing

the y, x locations of the center of each window.

Returns A tensor representing the glimpses `[batch_size, glimpse_height, glimpse_width, channels]`.

func ExtractGlimpseV2 

func ExtractGlimpseV2(scope *Scope, input tf.Output, size tf.Output, offsets tf.Output, optional ...ExtractGlimpseV2Attr) (glimpse tf.Output)

Extracts a glimpse from the input tensor.

Returns a set of windows called glimpses extracted at location `offsets` from the input tensor. If the windows only partially overlaps the inputs, the non overlapping areas will be filled with random noise.

The result is a 4-D tensor of shape `[batch_size, glimpse_height, glimpse_width, channels]`. The channels and batch dimensions are the same as that of the input tensor. The height and width of the output windows are specified in the `size` parameter.

The argument `normalized` and `centered` controls how the windows are built:

  • If the coordinates are normalized but not centered, 0.0 and 1.0 correspond to the minimum and maximum of each height and width dimension.
  • If the coordinates are both normalized and centered, they range from -1.0 to 1.0. The coordinates (-1.0, -1.0) correspond to the upper left corner, the lower right corner is located at (1.0, 1.0) and the center is at (0, 0).
  • If the coordinates are not normalized they are interpreted as numbers of pixels.

Arguments: 

input: A 4-D float tensor of shape `[batch_size, height, width, channels]`.
size: A 1-D tensor of 2 elements containing the size of the glimpses

to extract. The glimpse height must be specified first, following by the glimpse width. 

offsets: A 2-D integer tensor of shape `[batch_size, 2]` containing

the y, x locations of the center of each window.

Returns A tensor representing the glimpses `[batch_size, glimpse_height, glimpse_width, channels]`.

func ExtractImagePatches 

func ExtractImagePatches(scope *Scope, images tf.Output, ksizes []int64, strides []int64, rates []int64, padding string) (patches tf.Output)

Extract `patches` from `images` and put them in the "depth" output dimension.

Arguments: 

images: 4-D Tensor with shape `[batch, in_rows, in_cols, depth]`.
ksizes: The size of the sliding window for each dimension of `images`.
strides: How far the centers of two consecutive patches are in

the images. Must be: `[1, stride_rows, stride_cols, 1]`. 

rates: Must be: `[1, rate_rows, rate_cols, 1]`. This is the

input stride, specifying how far two consecutive patch samples are in the input. Equivalent to extracting patches with `patch_sizes_eff = patch_sizes + (patch_sizes - 1) * (rates - 1)`, followed by subsampling them spatially by a factor of `rates`. This is equivalent to `rate` in dilated (a.k.a. Atrous) convolutions. 

padding: The type of padding algorithm to use.

Returns 4-D Tensor with shape `[batch, out_rows, out_cols, ksize_rows * ksize_cols * depth]` containing image patches with size `ksize_rows x ksize_cols x depth` vectorized in the "depth" dimension. Note `out_rows` and `out_cols` are the dimensions of the output patches.

func ExtractJpegShape 

func ExtractJpegShape(scope *Scope, contents tf.Output, optional ...ExtractJpegShapeAttr) (image_shape tf.Output)

Extract the shape information of a JPEG-encoded image.

This op only parses the image header, so it is much faster than DecodeJpeg.

Arguments: 

contents: 0-D. The JPEG-encoded image.

Returns 1-D. The image shape with format [height, width, channels].

func ExtractVolumePatches 

func ExtractVolumePatches(scope *Scope, input tf.Output, ksizes []int64, strides []int64, padding string) (patches tf.Output)

Extract `patches` from `input` and put them in the `"depth"` output dimension. 3D extension of `extract_image_patches`.

Arguments: 

input: 5-D Tensor with shape `[batch, in_planes, in_rows, in_cols, depth]`.
ksizes: The size of the sliding window for each dimension of `input`.
strides: 1-D of length 5. How far the centers of two consecutive patches are in

`input`. Must be: `[1, stride_planes, stride_rows, stride_cols, 1]`. 

padding: The type of padding algorithm to use.

The size-related attributes are specified as follows:

```python ksizes = [1, ksize_planes, ksize_rows, ksize_cols, 1] strides = [1, stride_planes, strides_rows, strides_cols, 1] ```

Returns 5-D Tensor with shape `[batch, out_planes, out_rows, out_cols, ksize_planes * ksize_rows * ksize_cols * depth]` containing patches with size `ksize_planes x ksize_rows x ksize_cols x depth` vectorized in the "depth" dimension. Note `out_planes`, `out_rows` and `out_cols` are the dimensions of the output patches.

func FFT 

func FFT(scope *Scope, input tf.Output) (output tf.Output)

Fast Fourier transform.

Computes the 1-dimensional discrete Fourier transform over the inner-most dimension of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 

dimension of `input` is replaced with its 1D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.fft @end_compatibility

func FFT2D 

func FFT2D(scope *Scope, input tf.Output) (output tf.Output)

2D fast Fourier transform.

Computes the 2-dimensional discrete Fourier transform over the inner-most 2 dimensions of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 2 

dimensions of `input` are replaced with their 2D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.fft2 @end_compatibility

func FFT3D 

func FFT3D(scope *Scope, input tf.Output) (output tf.Output)

3D fast Fourier transform.

Computes the 3-dimensional discrete Fourier transform over the inner-most 3 dimensions of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 3 

dimensions of `input` are replaced with their 3D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.fftn with 3 dimensions. @end_compatibility

func FFTND added in v0.7.0 

func FFTND(scope *Scope, input tf.Output, fft_length tf.Output, axes tf.Output) (output tf.Output)

ND fast Fourier transform.

Computes the n-dimensional discrete Fourier transform over designated dimensions of `input`. The designated dimensions of `input` are assumed to be the result of `FFTND`.

If fft_length[i]<shape(input)[i], the input is cropped. If fft_length[i]>shape(input)[i], the input is padded with zeros. If fft_length is not given, the default shape(input) is used.

Axes mean the dimensions to perform the transform on. Default is to perform on all axes.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor. The FFT length for each dimension.
axes: An int32 tensor with a same shape as fft_length. Axes to perform the transform.

Returns A complex tensor of the same shape as `input`. The designated dimensions of `input` are replaced with their Fourier transforms.

@compatibility(numpy) Equivalent to np.fft.fftn. @end_compatibility

func FIFOQueueV2 

func FIFOQueueV2(scope *Scope, component_types []tf.DataType, optional ...FIFOQueueV2Attr) (handle tf.Output)

A queue that produces elements in first-in first-out order.

Arguments: 

component_types: The type of each component in a value.

Returns The handle to the queue.

func Fact 

func Fact(scope *Scope) (fact tf.Output)

Output a fact about factorials.

func FakeParam 

func FakeParam(scope *Scope, dtype tf.DataType, shape tf.Shape) (output tf.Output)
This op is used as a placeholder in If branch functions. It doesn't provide a
valid output when run, so must either be removed (e.g. replaced with a
function input) or guaranteed not to be used (e.g. if mirroring an
intermediate output needed for the gradient computation of the other branch).

Arguments: 

	dtype: The type of the output.
	shape:     The purported shape of the output. This is only used for shape inference;
    the output will not necessarily have this shape. Can be a partial shape.

Returns \"Fake\" output value. This should not be consumed by another op.

func FakeQuantWithMinMaxArgs 

func FakeQuantWithMinMaxArgs(scope *Scope, inputs tf.Output, optional ...FakeQuantWithMinMaxArgsAttr) (outputs tf.Output)

Fake-quantize the 'inputs' tensor, type float to 'outputs' tensor of same shape and type. 

Quantization is called fake since the output is still in floating point.
The API converts inputs into values within the range [min and max] and returns
as output.

Attributes

* `[min; max]` define the clamping range for the `inputs` data. * `inputs` values are quantized into the quantization range ( `[0; 2^num_bits - 1]` when `narrow_range` is false and `[1; 2^num_bits - 1]` when it is true) and then de-quantized and output as floats in `[min; max]` interval. * `num_bits` is the bitwidth of the quantization; between 2 and 16, inclusive.

Before quantization, `min` and `max` values are adjusted with the following logic. It is suggested to have `min <= 0 <= max`. If `0` is not in the range of values, the behavior can be unexpected:

* If `0 < min < max`: `min_adj = 0` and `max_adj = max - min`. * If `min < max < 0`: `min_adj = min - max` and `max_adj = 0`. * If `min <= 0 <= max`: `scale = (max - min) / (2^num_bits - 1) `, `min_adj = scale * round(min / scale)` and `max_adj = max + min_adj - min`.

Examples

```python

inp = tf.constant ([10.03, -10.23, 3]) out = tf.quantization.fake_quant_with_min_max_args(inp, min=-5, max=5, 

num_bits=16)

print(out)

# Output: # tf.Tensor([ 4.9999237 -5.0000763 3.0000763], shape=(3,), dtype=float32) ```

Raises:

  • InvalidArgumentError:
  • If num_bits are outside of range [2, 16].
  • If min >= max.
  • ValueError: If `inputs` are of any other type than float32.

func FakeQuantWithMinMaxArgsGradient 

func FakeQuantWithMinMaxArgsGradient(scope *Scope, gradients tf.Output, inputs tf.Output, optional ...FakeQuantWithMinMaxArgsGradientAttr) (backprops tf.Output)

Compute gradients for a FakeQuantWithMinMaxArgs operation.

Arguments: 

gradients: Backpropagated gradients above the FakeQuantWithMinMaxArgs operation.
inputs: Values passed as inputs to the FakeQuantWithMinMaxArgs operation.

Returns Backpropagated gradients below the FakeQuantWithMinMaxArgs operation: `gradients * (inputs >= min && inputs <= max)`.

``` import tensorflow as tf

# Define some sample data gradients = tf.random.uniform((2, 3), minval=-5.0, maxval=5.0, dtype=tf.float32) inputs = tf.random.uniform((2, 3), minval=-10.0, maxval=10.0, dtype=tf.float32)

# Define quantization parameters (adjust as needed) min_val = -2.0 max_val = 8.0 num_bits = 4 # Number of bits for quantization

# Calculate gradients for fake quantization with specified parameters output_gradients = tf.quantization.fake_quant_with_min_max_args_gradient( 

gradients=gradients, inputs=inputs, min=min_val, max=max_val, num_bits=num_bits, narrow_range = False, name=None

)

# Print the original gradients and the gradients after the fake-quant operation print("Original Gradients:") print(gradients) print("\nGradients after Fake-Quantization:") print(output_gradients)

``` #Original Gradients: #tf.Tensor( #[[ 1.242547 3.217492 3.568469 ] #[-0.55371046 0.23130894 2.608243 ]], shape=(2, 3), dtype=float32)

#Gradients after Fake-Quantization: #tf.Tensor( #[[ 0. 3.217492 3.568469 ] # [-0.55371046 0.23130894 2.608243 ]], shape=(2, 3), dtype=float32)

func FakeQuantWithMinMaxVars 

func FakeQuantWithMinMaxVars(scope *Scope, inputs tf.Output, min tf.Output, max tf.Output, optional ...FakeQuantWithMinMaxVarsAttr) (outputs tf.Output)

Fake-quantize the 'inputs' tensor of type float via global float scalars

Fake-quantize the `inputs` tensor of type float via global float scalars `min` and `max` to `outputs` tensor of same shape as `inputs`.

Attributes

* `[min; max]` define the clamping range for the `inputs` data. * `inputs` values are quantized into the quantization range ( `[0; 2^num_bits - 1]` when `narrow_range` is false and `[1; 2^num_bits - 1]` when it is true) and then de-quantized and output as floats in `[min; max]` interval. * `num_bits` is the bitwidth of the quantization; between 2 and 16, inclusive.

Before quantization, `min` and `max` values are adjusted with the following logic. It is suggested to have `min <= 0 <= max`. If `0` is not in the range of values, the behavior can be unexpected:

* If `0 < min < max`: `min_adj = 0` and `max_adj = max - min`. * If `min < max < 0`: `min_adj = min - max` and `max_adj = 0`. * If `min <= 0 <= max`: `scale = (max - min) / (2^num_bits - 1) `, `min_adj = scale * round(min / scale)` and `max_adj = max + min_adj - min`.

This operation has a gradient and thus allows for training `min` and `max` values.

>>> constant_input = tf.constant([[1.2, -0.3, 0.7], [2.1, 0.5, -1.0]], dtype=tf.float32) >>> >>> min_val = -0.5 >>> max_val = 0.8 >>> num_bits = 8 >>> narrow_range = False #False:for the quantization range [0; 2^num_bits - 1] >>> >>> quantized_data = tf.quantization.fake_quant_with_min_max_vars( ... inputs=constant_input, min=min_val, max=max_val, num_bits=num_bits, narrow_range=narrow_range ... ) >>> >>> print("Input:\n", constant_input.numpy()) Input: [[ 1.2 -0.3 0.7] [ 2.1 0.5 -1. ]] >>> print("Output:\n", quantized_data.numpy()) Output: [[ 0.8003921 -0.3007843 0.6984313] [ 0.8003921 0.4996078 -0.4996078]]

func FakeQuantWithMinMaxVarsGradient 

func FakeQuantWithMinMaxVarsGradient(scope *Scope, gradients tf.Output, inputs tf.Output, min tf.Output, max tf.Output, optional ...FakeQuantWithMinMaxVarsGradientAttr) (backprops_wrt_input tf.Output, backprop_wrt_min tf.Output, backprop_wrt_max tf.Output)

Compute gradients for a FakeQuantWithMinMaxVars operation.

Arguments: 

gradients: Backpropagated gradients above the FakeQuantWithMinMaxVars operation.
inputs: Values passed as inputs to the FakeQuantWithMinMaxVars operation.

min, max: Quantization interval, scalar floats.

Returns: 

backprops_wrt_input: Backpropagated gradients w.r.t. inputs:

`gradients * (inputs >= min && inputs <= max)`. 

backprop_wrt_min: Backpropagated gradients w.r.t. min parameter:

`sum(gradients * (inputs < min))`. 

backprop_wrt_max: Backpropagated gradients w.r.t. max parameter:

`sum(gradients * (inputs > max))`.

func FakeQuantWithMinMaxVarsPerChannel 

func FakeQuantWithMinMaxVarsPerChannel(scope *Scope, inputs tf.Output, min tf.Output, max tf.Output, optional ...FakeQuantWithMinMaxVarsPerChannelAttr) (outputs tf.Output)

Fake-quantize the 'inputs' tensor of type float via per-channel floats

Fake-quantize the `inputs` tensor of type float per-channel and one of the shapes: `[d]`, `[b, d]` `[b, h, w, d]` via per-channel floats `min` and `max` of shape `[d]` to `outputs` tensor of same shape as `inputs`.

Attributes

* `[min; max]` define the clamping range for the `inputs` data. * `inputs` values are quantized into the quantization range ( `[0; 2^num_bits - 1]` when `narrow_range` is false and `[1; 2^num_bits - 1]` when it is true) and then de-quantized and output as floats in `[min; max]` interval. * `num_bits` is the bitwidth of the quantization; between 2 and 16, inclusive.

Before quantization, `min` and `max` values are adjusted with the following logic. It is suggested to have `min <= 0 <= max`. If `0` is not in the range of values, the behavior can be unexpected:

* If `0 < min < max`: `min_adj = 0` and `max_adj = max - min`. * If `min < max < 0`: `min_adj = min - max` and `max_adj = 0`. * If `min <= 0 <= max`: `scale = (max - min) / (2^num_bits - 1) `, `min_adj = scale * round(min / scale)` and `max_adj = max + min_adj - min`.

This operation has a gradient and thus allows for training `min` and `max` values.

func FakeQuantWithMinMaxVarsPerChannelGradient 

func FakeQuantWithMinMaxVarsPerChannelGradient(scope *Scope, gradients tf.Output, inputs tf.Output, min tf.Output, max tf.Output, optional ...FakeQuantWithMinMaxVarsPerChannelGradientAttr) (backprops_wrt_input tf.Output, backprop_wrt_min tf.Output, backprop_wrt_max tf.Output)

Compute gradients for a FakeQuantWithMinMaxVarsPerChannel operation.

Arguments: 

gradients: Backpropagated gradients above the FakeQuantWithMinMaxVars operation,

shape one of: `[d]`, `[b, d]`, `[b, h, w, d]`. 

	inputs: Values passed as inputs to the FakeQuantWithMinMaxVars operation, shape
  same as `gradients`.

min, max: Quantization interval, floats of shape `[d]`.

Returns: 

backprops_wrt_input: Backpropagated gradients w.r.t. inputs, shape same as

`inputs`: 

  `gradients * (inputs >= min && inputs <= max)`.
	backprop_wrt_min: Backpropagated gradients w.r.t. min parameter, shape `[d]`:

`sum_per_d(gradients * (inputs < min))`. 

backprop_wrt_max: Backpropagated gradients w.r.t. max parameter, shape `[d]`:

`sum_per_d(gradients * (inputs > max))`.

func FileSystemSetConfiguration 

func FileSystemSetConfiguration(scope *Scope, scheme tf.Output, key tf.Output, value tf.Output) (o *tf.Operation)

Set configuration of the file system.

Arguments: 

scheme: File system scheme.
key: The name of the configuration option.
value: The value of the configuration option.

Returns the created operation.

func Fill 

func Fill(scope *Scope, dims tf.Output, value tf.Output) (output tf.Output)

Creates a tensor filled with a scalar value.

This operation creates a tensor of shape `dims` and fills it with `value`.

For example:

``` # Output tensor has shape [2, 3]. fill([2, 3], 9) ==> [[9, 9, 9] 

[9, 9, 9]]

```

`tf.fill` differs from `tf.constant` in a few ways:

  • `tf.fill` only supports scalar contents, whereas `tf.constant` supports Tensor values.
  • `tf.fill` creates an Op in the computation graph that constructs the actual Tensor value at runtime. This is in contrast to `tf.constant` which embeds the entire Tensor into the graph with a `Const` node.
  • Because `tf.fill` evaluates at graph runtime, it supports dynamic shapes based on other runtime Tensors, unlike `tf.constant`.

Arguments: 

dims: 1-D. Represents the shape of the output tensor.
value: 0-D (scalar). Value to fill the returned tensor.

@compatibility(numpy) Equivalent to np.full @end_compatibility

func FilterByLastComponentDataset 

func FilterByLastComponentDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (output tf.Output)

Creates a dataset containing elements of first component of `input_dataset` having true in the last component.

func FinalizeDataset 

func FinalizeDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...FinalizeDatasetAttr) (handle tf.Output)

Creates a dataset by applying `tf.data.Options` to `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the input dataset.

func FinalizeTPUEmbedding added in v0.2.0 

func FinalizeTPUEmbedding(scope *Scope, common_config tf.Output, memory_config tf.Output) (o *tf.Operation)

An op that finalizes the TPUEmbedding configuration.

Arguments: 

common_config: A string-encoded common configuration proto containing metadata

about the TPUEmbedding partitioner output and the HBM size (in bytes) required for operation. 

memory_config: A string-encoded memory config proto containing metadata about

the memory allocations reserved for TPUEmbedding.

Returns the created operation.

func FinalizeTPUEmbeddingV2 added in v0.8.2 

func FinalizeTPUEmbeddingV2(scope *Scope, common_config tf.Output, memory_config tf.Output) (embedding_partitions tf.Output, hbm_buffers_config tf.Output)

An op that finalizes the TPUEmbedding configuration.

Arguments: 

common_config: A string-encoded common configuration proto containing metadata

about the TPUEmbedding partitioner output and the HBM size (in bytes) required for operation. 

memory_config: A string-encoded memory config proto containing metadata about

the memory allocations reserved for TPUEmbedding.

Returns: 

embedding_partitions: A string-encoded embedding partitions proto describing how embedding tables are

partitioned along their feature and ID. 

hbm_buffers_config: A string-encoded HBM buffers config proto specifies where HBM buffers are

located.

func Fingerprint 

func Fingerprint(scope *Scope, data tf.Output, method tf.Output) (fingerprint tf.Output)

Generates fingerprint values.

Generates fingerprint values of `data`.

Fingerprint op considers the first dimension of `data` as the batch dimension, and `output[i]` contains the fingerprint value generated from contents in `data[i, ...]` for all `i`.

Fingerprint op writes fingerprint values as byte arrays. For example, the default method `farmhash64` generates a 64-bit fingerprint value at a time. This 8-byte value is written out as an `uint8` array of size 8, in little-endian order.

For example, suppose that `data` has data type `DT_INT32` and shape (2, 3, 4), and that the fingerprint method is `farmhash64`. In this case, the output shape is (2, 8), where 2 is the batch dimension size of `data`, and 8 is the size of each fingerprint value in bytes. `output[0, :]` is generated from 12 integers in `data[0, :, :]` and similarly `output[1, :]` is generated from other 12 integers in `data[1, :, :]`.

Note that this op fingerprints the raw underlying buffer, and it does not fingerprint Tensor's metadata such as data type and/or shape. For example, the fingerprint values are invariant under reshapes and bitcasts as long as the batch dimension remain the same:

``` Fingerprint(data) == Fingerprint(Reshape(data, ...)) Fingerprint(data) == Fingerprint(Bitcast(data, ...)) ```

For string data, one should expect `Fingerprint(data) != Fingerprint(ReduceJoin(data))` in general.

Arguments: 

data: Must have rank 1 or higher.
method: Fingerprint method used by this op. Currently available method is

`farmhash::fingerprint64`.

Returns A two-dimensional `Tensor` of type `tf.uint8`. The first dimension equals to `data`'s first dimension, and the second dimension size depends on the fingerprint algorithm.

func FixedLengthRecordDataset 

func FixedLengthRecordDataset(scope *Scope, filenames tf.Output, header_bytes tf.Output, record_bytes tf.Output, footer_bytes tf.Output, buffer_size tf.Output, optional ...FixedLengthRecordDatasetAttr) (handle tf.Output)

Creates a dataset that emits the records from one or more binary files.

Arguments: 

filenames: A scalar or a vector containing the name(s) of the file(s) to be

read. 

header_bytes: A scalar representing the number of bytes to skip at the

beginning of a file. 

record_bytes: A scalar representing the number of bytes in each record.
footer_bytes: A scalar representing the number of bytes to skip at the end

of a file. 

buffer_size: A scalar representing the number of bytes to buffer. Must be > 0.

func FixedLengthRecordReaderV2 

func FixedLengthRecordReaderV2(scope *Scope, record_bytes int64, optional ...FixedLengthRecordReaderV2Attr) (reader_handle tf.Output)

A Reader that outputs fixed-length records from a file.

Arguments: 

record_bytes: Number of bytes in the record.

Returns The handle to reference the Reader.

func FixedUnigramCandidateSampler 

func FixedUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...FixedUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a learned unigram distribution.

A unigram sampler could use a fixed unigram distribution read from a file or passed in as an in-memory array instead of building up the distribution from data on the fly. There is also an option to skew the distribution by applying a distortion power to the weights.

The vocabulary file should be in CSV-like format, with the last field being the weight associated with the word.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to randomly sample.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities. 

range_max: The sampler will sample integers from the interval [0, range_max).

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func Floor 

func Floor(scope *Scope, x tf.Output) (y tf.Output)

Returns element-wise largest integer not greater than x.

func FloorDiv 

func FloorDiv(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x // y element-wise.

*NOTE*: `FloorDiv` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func FloorMod 

func FloorMod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns element-wise remainder of division.

This follows Python semantics in that the result here is consistent with a flooring divide. E.g. `floor(x / y) * y + floormod(x, y) = x`, regardless of the signs of x and y.

*NOTE*: `FloorMod` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func FractionalAvgPool 

func FractionalAvgPool(scope *Scope, value tf.Output, pooling_ratio []float32, optional ...FractionalAvgPoolAttr) (output tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output)

Performs fractional average pooling on the input.

Fractional average pooling is similar to Fractional max pooling in the pooling region generation step. The only difference is that after pooling regions are generated, a mean operation is performed instead of a max operation in each pooling region.

Arguments: 

value: 4-D with shape `[batch, height, width, channels]`.
pooling_ratio: Pooling ratio for each dimension of `value`, currently only

supports row and col dimension and should be >= 1.0. For example, a valid pooling ratio looks like [1.0, 1.44, 1.73, 1.0]. The first and last elements must be 1.0 because we don't allow pooling on batch and channels dimensions. 1.44 and 1.73 are pooling ratio on height and width dimensions respectively.

Returns: 

output: output tensor after fractional avg pooling.
row_pooling_sequence: row pooling sequence, needed to calculate gradient.
col_pooling_sequence: column pooling sequence, needed to calculate gradient.

func FractionalAvgPoolGrad 

func FractionalAvgPoolGrad(scope *Scope, orig_input_tensor_shape tf.Output, out_backprop tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output, optional ...FractionalAvgPoolGradAttr) (output tf.Output)

Computes gradient of the FractionalAvgPool function.

Unlike FractionalMaxPoolGrad, we don't need to find arg_max for FractionalAvgPoolGrad, we just need to evenly back-propagate each element of out_backprop to those indices that form the same pooling cell. Therefore, we just need to know the shape of original input tensor, instead of the whole tensor.

Arguments: 

orig_input_tensor_shape: Original input tensor shape for `fractional_avg_pool`
out_backprop: 4-D with shape `[batch, height, width, channels]`.  Gradients

w.r.t. the output of `fractional_avg_pool`. 

row_pooling_sequence: row pooling sequence, form pooling region with

col_pooling_sequence. 

col_pooling_sequence: column pooling sequence, form pooling region with

row_pooling sequence.

Returns 4-D. Gradients w.r.t. the input of `fractional_avg_pool`.

func FractionalMaxPool 

func FractionalMaxPool(scope *Scope, value tf.Output, pooling_ratio []float32, optional ...FractionalMaxPoolAttr) (output tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output)

Performs fractional max pooling on the input.

Fractional max pooling is slightly different than regular max pooling. In regular max pooling, you downsize an input set by taking the maximum value of smaller N x N subsections of the set (often 2x2), and try to reduce the set by a factor of N, where N is an integer. Fractional max pooling, as you might expect from the word "fractional", means that the overall reduction ratio N does not have to be an integer.

The sizes of the pooling regions are generated randomly but are fairly uniform. For example, let's look at the height dimension, and the constraints on the list of rows that will be pool boundaries.

First we define the following:

1. input_row_length : the number of rows from the input set 2. output_row_length : which will be smaller than the input 3. alpha = input_row_length / output_row_length : our reduction ratio 4. K = floor(alpha) 5. row_pooling_sequence : this is the result list of pool boundary rows

Then, row_pooling_sequence should satisfy:

1. a[0] = 0 : the first value of the sequence is 0 2. a[end] = input_row_length : the last value of the sequence is the size 3. K <= (a[i+1] - a[i]) <= K+1 : all intervals are K or K+1 size 4. length(row_pooling_sequence) = output_row_length+1

For more details on fractional max pooling, see this paper: [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071)

Arguments: 

value: 4-D with shape `[batch, height, width, channels]`.
pooling_ratio: Pooling ratio for each dimension of `value`, currently only

supports row and col dimension and should be >= 1.0. For example, a valid pooling ratio looks like [1.0, 1.44, 1.73, 1.0]. The first and last elements must be 1.0 because we don't allow pooling on batch and channels dimensions. 1.44 and 1.73 are pooling ratio on height and width dimensions respectively.

Returns: 

output: output tensor after fractional max pooling.
row_pooling_sequence: row pooling sequence, needed to calculate gradient.
col_pooling_sequence: column pooling sequence, needed to calculate gradient.

func FractionalMaxPoolGrad 

func FractionalMaxPoolGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, out_backprop tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output, optional ...FractionalMaxPoolGradAttr) (output tf.Output)

Computes gradient of the FractionalMaxPool function.

Arguments: 

orig_input: Original input for `fractional_max_pool`
orig_output: Original output for `fractional_max_pool`
out_backprop: 4-D with shape `[batch, height, width, channels]`.  Gradients

w.r.t. the output of `fractional_max_pool`. 

row_pooling_sequence: row pooling sequence, form pooling region with

col_pooling_sequence. 

col_pooling_sequence: column pooling sequence, form pooling region with

row_pooling sequence.

Returns 4-D. Gradients w.r.t. the input of `fractional_max_pool`.

func FusedBatchNorm 

func FusedBatchNorm(scope *Scope, x tf.Output, scale tf.Output, offset tf.Output, mean tf.Output, variance tf.Output, optional ...FusedBatchNormAttr) (y tf.Output, batch_mean tf.Output, batch_variance tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output)

Batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
offset: A 1D Tensor for offset, to shift to the normalized x.
mean: A 1D Tensor for population mean. Used for inference only;

must be empty for training. 

variance: A 1D Tensor for population variance. Used for inference only;

must be empty for training.

Returns: 

y: A 4D Tensor for output data.
batch_mean: A 1D Tensor for the computed batch mean, to be used by TensorFlow

to compute the running mean. 

batch_variance: A 1D Tensor for the computed batch variance, to be used by

TensorFlow to compute the running variance. 

reserve_space_1: A 1D Tensor for the computed batch mean, to be reused

in the gradient computation. 

reserve_space_2: A 1D Tensor for the computed batch variance (inverted variance

in the cuDNN case), to be reused in the gradient computation.

func FusedBatchNormGrad 

func FusedBatchNormGrad(scope *Scope, y_backprop tf.Output, x tf.Output, scale tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output, optional ...FusedBatchNormGradAttr) (x_backprop tf.Output, scale_backprop tf.Output, offset_backprop tf.Output, reserve_space_3 tf.Output, reserve_space_4 tf.Output)

Gradient for batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

y_backprop: A 4D Tensor for the gradient with respect to y.
x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
reserve_space_1: When is_training is True, a 1D Tensor for the computed batch

mean to be reused in gradient computation. When is_training is False, a 1D Tensor for the population mean to be reused in both 1st and 2nd order gradient computation. 

reserve_space_2: When is_training is True, a 1D Tensor for the computed batch

variance (inverted variance in the cuDNN case) to be reused in gradient computation. When is_training is False, a 1D Tensor for the population variance to be reused in both 1st and 2nd order gradient computation.

Returns: 

x_backprop: A 4D Tensor for the gradient with respect to x.
scale_backprop: A 1D Tensor for the gradient with respect to scale.
offset_backprop: A 1D Tensor for the gradient with respect to offset.
reserve_space_3: Unused placeholder to match the mean input in FusedBatchNorm.
reserve_space_4: Unused placeholder to match the variance input

in FusedBatchNorm.

func FusedBatchNormGradV2 

func FusedBatchNormGradV2(scope *Scope, y_backprop tf.Output, x tf.Output, scale tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output, optional ...FusedBatchNormGradV2Attr) (x_backprop tf.Output, scale_backprop tf.Output, offset_backprop tf.Output, reserve_space_3 tf.Output, reserve_space_4 tf.Output)

Gradient for batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

y_backprop: A 4D Tensor for the gradient with respect to y.
x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
reserve_space_1: When is_training is True, a 1D Tensor for the computed batch

mean to be reused in gradient computation. When is_training is False, a 1D Tensor for the population mean to be reused in both 1st and 2nd order gradient computation. 

reserve_space_2: When is_training is True, a 1D Tensor for the computed batch

variance (inverted variance in the cuDNN case) to be reused in gradient computation. When is_training is False, a 1D Tensor for the population variance to be reused in both 1st and 2nd order gradient computation.

Returns: 

x_backprop: A 4D Tensor for the gradient with respect to x.
scale_backprop: A 1D Tensor for the gradient with respect to scale.
offset_backprop: A 1D Tensor for the gradient with respect to offset.
reserve_space_3: Unused placeholder to match the mean input in FusedBatchNorm.
reserve_space_4: Unused placeholder to match the variance input

in FusedBatchNorm.

func FusedBatchNormGradV3 

func FusedBatchNormGradV3(scope *Scope, y_backprop tf.Output, x tf.Output, scale tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output, reserve_space_3 tf.Output, optional ...FusedBatchNormGradV3Attr) (x_backprop tf.Output, scale_backprop tf.Output, offset_backprop tf.Output, reserve_space_4 tf.Output, reserve_space_5 tf.Output)

Gradient for batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

y_backprop: A 4D Tensor for the gradient with respect to y.
x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
reserve_space_1: When is_training is True, a 1D Tensor for the computed batch

mean to be reused in gradient computation. When is_training is False, a 1D Tensor for the population mean to be reused in both 1st and 2nd order gradient computation. 

reserve_space_2: When is_training is True, a 1D Tensor for the computed batch

variance (inverted variance in the cuDNN case) to be reused in gradient computation. When is_training is False, a 1D Tensor for the population variance to be reused in both 1st and 2nd order gradient computation. 

reserve_space_3: When is_training is True, a 1D Tensor for some intermediate results to be reused

in gradient computation. When is_training is False, a dummy empty Tensor will be created.

Returns: 

x_backprop: A 4D Tensor for the gradient with respect to x.
scale_backprop: A 1D Tensor for the gradient with respect to scale.
offset_backprop: A 1D Tensor for the gradient with respect to offset.
reserve_space_4: Unused placeholder to match the mean input in FusedBatchNorm.
reserve_space_5: Unused placeholder to match the variance input

in FusedBatchNorm.

func FusedBatchNormV2 

func FusedBatchNormV2(scope *Scope, x tf.Output, scale tf.Output, offset tf.Output, mean tf.Output, variance tf.Output, optional ...FusedBatchNormV2Attr) (y tf.Output, batch_mean tf.Output, batch_variance tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output)

Batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
offset: A 1D Tensor for offset, to shift to the normalized x.
mean: A 1D Tensor for population mean. Used for inference only;

must be empty for training. 

variance: A 1D Tensor for population variance. Used for inference only;

must be empty for training.

Returns: 

y: A 4D Tensor for output data.
batch_mean: A 1D Tensor for the computed batch mean, to be used by TensorFlow

to compute the running mean. 

batch_variance: A 1D Tensor for the computed batch variance, to be used by

TensorFlow to compute the running variance. 

reserve_space_1: A 1D Tensor for the computed batch mean, to be reused

in the gradient computation. 

reserve_space_2: A 1D Tensor for the computed batch variance (inverted variance

in the cuDNN case), to be reused in the gradient computation.

func FusedBatchNormV3 

func FusedBatchNormV3(scope *Scope, x tf.Output, scale tf.Output, offset tf.Output, mean tf.Output, variance tf.Output, optional ...FusedBatchNormV3Attr) (y tf.Output, batch_mean tf.Output, batch_variance tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output, reserve_space_3 tf.Output)

Batch normalization.

Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.

Arguments: 

x: A 4D Tensor for input data.
scale: A 1D Tensor for scaling factor, to scale the normalized x.
offset: A 1D Tensor for offset, to shift to the normalized x.
mean: A 1D Tensor for population mean. Used for inference only;

must be empty for training. 

variance: A 1D Tensor for population variance. Used for inference only;

must be empty for training.

Returns: 

y: A 4D Tensor for output data.
batch_mean: A 1D Tensor for the computed batch mean, to be used by TensorFlow

to compute the running mean. 

batch_variance: A 1D Tensor for the computed batch variance, to be used by

TensorFlow to compute the running variance. 

reserve_space_1: A 1D Tensor for the computed batch mean, to be reused

in the gradient computation. 

reserve_space_2: A 1D Tensor for the computed batch variance (inverted variance

in the cuDNN case), to be reused in the gradient computation. 

reserve_space_3: A 1D Tensor for some intermediate results, to be reused in the gradient

computation for better efficiency.

func FusedPadConv2D 

func FusedPadConv2D(scope *Scope, input tf.Output, paddings tf.Output, filter tf.Output, mode string, strides []int64, padding string) (output tf.Output)

Performs a padding as a preprocess during a convolution.

Similar to FusedResizeAndPadConv2d, this op allows for an optimized implementation where the spatial padding transformation stage is fused with the im2col lookup, but in this case without the bilinear filtering required for resizing. Fusing the padding prevents the need to write out the intermediate results as whole tensors, reducing memory pressure, and we can get some latency gains by merging the transformation calculations. The data_format attribute for Conv2D isn't supported by this op, and 'NHWC' order is used instead. Internally this op uses a single per-graph scratch buffer, which means that it will block if multiple versions are being run in parallel. This is because this operator is primarily an optimization to minimize memory usage.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, in_channels]`.
paddings: A two-column matrix specifying the padding sizes. The number of

rows must be the same as the rank of `input`. 

filter: 4-D with shape

`[filter_height, filter_width, in_channels, out_channels]`. 

strides: 1-D of length 4.  The stride of the sliding window for each dimension

of `input`. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

func FusedResizeAndPadConv2D 

func FusedResizeAndPadConv2D(scope *Scope, input tf.Output, size tf.Output, paddings tf.Output, filter tf.Output, mode string, strides []int64, padding string, optional ...FusedResizeAndPadConv2DAttr) (output tf.Output)

Performs a resize and padding as a preprocess during a convolution.

It's often possible to do spatial transformations more efficiently as part of the packing stage of a convolution, so this op allows for an optimized implementation where these stages are fused together. This prevents the need to write out the intermediate results as whole tensors, reducing memory pressure, and we can get some latency gains by merging the transformation calculations. The data_format attribute for Conv2D isn't supported by this op, and defaults to 'NHWC' order. Internally this op uses a single per-graph scratch buffer, which means that it will block if multiple versions are being run in parallel. This is because this operator is primarily an optimization to minimize memory usage.

Arguments: 

input: 4-D with shape `[batch, in_height, in_width, in_channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images. 

paddings: A two-column matrix specifying the padding sizes. The number of

rows must be the same as the rank of `input`. 

filter: 4-D with shape

`[filter_height, filter_width, in_channels, out_channels]`. 

strides: 1-D of length 4.  The stride of the sliding window for each dimension

of `input`. Must be in the same order as the dimension specified with format. 

padding: The type of padding algorithm to use.

func GRUBlockCell 

func GRUBlockCell(scope *Scope, x tf.Output, h_prev tf.Output, w_ru tf.Output, w_c tf.Output, b_ru tf.Output, b_c tf.Output) (r tf.Output, u tf.Output, c tf.Output, h tf.Output)

Computes the GRU cell forward propagation for 1 time step.

Args 

x: Input to the GRU cell.
h_prev: State input from the previous GRU cell.
w_ru: Weight matrix for the reset and update gate.
w_c: Weight matrix for the cell connection gate.
b_ru: Bias vector for the reset and update gate.
b_c: Bias vector for the cell connection gate.

Returns 

r: Output of the reset gate.
u: Output of the update gate.
c: Output of the cell connection gate.
h: Current state of the GRU cell.

Note on notation of the variables:

Concatenation of a and b is represented by a_b Element-wise dot product of a and b is represented by ab Element-wise dot product is represented by \circ Matrix multiplication is represented by *

Biases are initialized with : `b_ru` - constant_initializer(1.0) `b_c` - constant_initializer(0.0)

This kernel op implements the following mathematical equations:

``` x_h_prev = [x, h_prev]

[r_bar u_bar] = x_h_prev * w_ru + b_ru

r = sigmoid(r_bar) u = sigmoid(u_bar)

h_prevr = h_prev \circ r

x_h_prevr = [x h_prevr]

c_bar = x_h_prevr * w_c + b_c c = tanh(c_bar)

h = (1-u) \circ c + u \circ h_prev ```

func GRUBlockCellGrad 

func GRUBlockCellGrad(scope *Scope, x tf.Output, h_prev tf.Output, w_ru tf.Output, w_c tf.Output, b_ru tf.Output, b_c tf.Output, r tf.Output, u tf.Output, c tf.Output, d_h tf.Output) (d_x tf.Output, d_h_prev tf.Output, d_c_bar tf.Output, d_r_bar_u_bar tf.Output)

Computes the GRU cell back-propagation for 1 time step.

Args 

x: Input to the GRU cell.
h_prev: State input from the previous GRU cell.
w_ru: Weight matrix for the reset and update gate.
w_c: Weight matrix for the cell connection gate.
b_ru: Bias vector for the reset and update gate.
b_c: Bias vector for the cell connection gate.
r: Output of the reset gate.
u: Output of the update gate.
c: Output of the cell connection gate.
d_h: Gradients of the h_new wrt to objective function.

Returns 

d_x: Gradients of the x wrt to objective function.
d_h_prev: Gradients of the h wrt to objective function.
d_c_bar Gradients of the c_bar wrt to objective function.
d_r_bar_u_bar Gradients of the r_bar & u_bar wrt to objective function.

This kernel op implements the following mathematical equations:

Note on notation of the variables:

Concatenation of a and b is represented by a_b Element-wise dot product of a and b is represented by ab Element-wise dot product is represented by \circ Matrix multiplication is represented by *

Additional notes for clarity:

`w_ru` can be segmented into 4 different matrices. ``` w_ru = [w_r_x w_u_x 

w_r_h_prev w_u_h_prev]

``` Similarly, `w_c` can be segmented into 2 different matrices. ``` w_c = [w_c_x w_c_h_prevr] ``` Same goes for biases. ``` b_ru = [b_ru_x b_ru_h] b_c = [b_c_x b_c_h] ``` Another note on notation: ``` d_x = d_x_component_1 + d_x_component_2

where d_x_component_1 = d_r_bar * w_r_x^T + d_u_bar * w_r_x^T and d_x_component_2 = d_c_bar * w_c_x^T

d_h_prev = d_h_prev_component_1 + d_h_prevr \circ r + d_h \circ u where d_h_prev_componenet_1 = d_r_bar * w_r_h_prev^T + d_u_bar * w_r_h_prev^T ```

Mathematics behind the Gradients below: ``` d_c_bar = d_h \circ (1-u) \circ (1-c \circ c) d_u_bar = d_h \circ (h-c) \circ u \circ (1-u)

d_r_bar_u_bar = [d_r_bar d_u_bar]

[d_x_component_1 d_h_prev_component_1] = d_r_bar_u_bar * w_ru^T

[d_x_component_2 d_h_prevr] = d_c_bar * w_c^T

d_x = d_x_component_1 + d_x_component_2

d_h_prev = d_h_prev_component_1 + d_h_prevr \circ r + u ``` Below calculation is performed in the python wrapper for the Gradients (not in the gradient kernel.) ``` d_w_ru = x_h_prevr^T * d_c_bar

d_w_c = x_h_prev^T * d_r_bar_u_bar

d_b_ru = sum of d_r_bar_u_bar along axis = 0

d_b_c = sum of d_c_bar along axis = 0 ```

func Gather 

func Gather(scope *Scope, params tf.Output, indices tf.Output, optional ...GatherAttr) (output tf.Output)

Gather slices from `params` according to `indices`.

`indices` must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape `indices.shape + params.shape[1:]` where:

```python 

# Scalar indices
output[:, ..., :] = params[indices, :, ... :]

# Vector indices
output[i, :, ..., :] = params[indices[i], :, ... :]

# Higher rank indices
output[i, ..., j, :, ... :] = params[indices[i, ..., j], :, ..., :]

```

If `indices` is a permutation and `len(indices) == params.shape[0]` then this operation will permute `params` accordingly.

`validate_indices`: DEPRECATED. If this operation is assigned to CPU, values in `indices` are always validated to be within range. If assigned to GPU, out-of-bound indices result in safe but unspecified behavior, which may include raising an error.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/Gather.png" alt> </div>

func GatherNd 

func GatherNd(scope *Scope, params tf.Output, indices tf.Output, optional ...GatherNdAttr) (output tf.Output)

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

`indices` is a K-dimensional integer tensor, best thought of as a (K-1)-dimensional tensor of indices into `params`, where each element defines a slice of `params`: 

output[\\(i_0, ..., i_{K-2}\\)] = params[indices[\\(i_0, ..., i_{K-2}\\)]]

Whereas in `tf.gather` `indices` defines slices into the `axis` dimension of `params`, in `tf.gather_nd`, `indices` defines slices into the first `N` dimensions of `params`, where `N = indices.shape[-1]`.

The last dimension of `indices` can be at most the rank of `params`: 

indices.shape[-1] <= params.rank

The last dimension of `indices` corresponds to elements (if `indices.shape[-1] == params.rank`) or slices (if `indices.shape[-1] < params.rank`) along dimension `indices.shape[-1]` of `params`. The output tensor has shape 

indices.shape[:-1] + params.shape[indices.shape[-1]:]

If `indices` contains any out-of-bound indices, depending on `bad_indices_policy`, the op will either return an error or ignore the out-of-bound indices. `bad_indices_policy` can be one of the following values:

  1. "" or "DEFAULT": raises on CPU and ignore on GPU. This is because historically on CPU and GPU we handle errors in different ways, and for backward compatibility we keep the default behavior.
  2. "ERROR": raises error; GPU does not support this value.
  3. "IGNORE": ignore error and set the corresponding output to 0; supported on both CPU and GPU.

Some examples below.

Simple indexing into a matrix:

```python 

indices = [[0, 0], [1, 1]]
params = [['a', 'b'], ['c', 'd']]
output = ['a', 'd']

```

Slice indexing into a matrix:

```python 

indices = [[1], [0]]
params = [['a', 'b'], ['c', 'd']]
output = [['c', 'd'], ['a', 'b']]

```

Indexing into a 3-tensor:

```python 

indices = [[1]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = [[['a1', 'b1'], ['c1', 'd1']]]

indices = [[0, 1], [1, 0]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = [['c0', 'd0'], ['a1', 'b1']]

indices = [[0, 0, 1], [1, 0, 1]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = ['b0', 'b1']

```

Batched indexing into a matrix:

```python 

indices = [[[0, 0]], [[0, 1]]]
params = [['a', 'b'], ['c', 'd']]
output = [['a'], ['b']]

```

Batched slice indexing into a matrix:

```python 

indices = [[[1]], [[0]]]
params = [['a', 'b'], ['c', 'd']]
output = [[['c', 'd']], [['a', 'b']]]

```

Batched indexing into a 3-tensor:

```python 

indices = [[[1]], [[0]]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = [[[['a1', 'b1'], ['c1', 'd1']]],
          [[['a0', 'b0'], ['c0', 'd0']]]]

indices = [[[0, 1], [1, 0]], [[0, 0], [1, 1]]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = [[['c0', 'd0'], ['a1', 'b1']],
          [['a0', 'b0'], ['c1', 'd1']]]

indices = [[[0, 0, 1], [1, 0, 1]], [[0, 1, 1], [1, 1, 0]]]
params = [[['a0', 'b0'], ['c0', 'd0']],
          [['a1', 'b1'], ['c1', 'd1']]]
output = [['b0', 'b1'], ['d0', 'c1']]

```

See also `tf.gather` and `tf.batch_gather`.

Arguments: 

params: The tensor from which to gather values.
indices: Index tensor.

Returns Values from `params` gathered from indices given by `indices`, with shape `indices.shape[:-1] + params.shape[indices.shape[-1]:]`.

func GatherV2 

func GatherV2(scope *Scope, params tf.Output, indices tf.Output, axis tf.Output, optional ...GatherV2Attr) (output tf.Output)

Gather slices from `params` axis `axis` according to `indices`.

`indices` must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape `params.shape[:axis] + indices.shape[batch_dims:] + params.shape[axis + 1:]` where:

```python 

# Scalar indices (output is rank(params) - 1).
output[a_0, ..., a_n, b_0, ..., b_n] =
  params[a_0, ..., a_n, indices, b_0, ..., b_n]

# Vector indices (output is rank(params)).
output[a_0, ..., a_n, i, b_0, ..., b_n] =
  params[a_0, ..., a_n, indices[i], b_0, ..., b_n]

# Higher rank indices (output is rank(params) + rank(indices) - 1).
output[a_0, ..., a_n, i, ..., j, b_0, ... b_n] =
  params[a_0, ..., a_n, indices[i, ..., j], b_0, ..., b_n]

```

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/Gather.png" alt> </div>

Note that on CPU, if an out of bound index is found, an error is returned. On GPU, if an out of bound index is found, a 0 is stored in the corresponding output value.

Note that on TPU, if any dimension of `params` is of size 0 then the output will be the expected shape filled with zeros. On CPU and GPU an error will be returned.

See also `tf.batch_gather` and `tf.gather_nd`.

Arguments: 

params: The tensor from which to gather values. Must be at least rank

`axis + 1`. 

indices: Index tensor. Must be in range `[0, params.shape[axis])`.
axis: The axis in `params` to gather `indices` from. Defaults to the first

dimension. Supports negative indexes.

Returns Values from `params` gathered from indices given by `indices`, with shape `params.shape[:axis] + indices.shape + params.shape[axis + 1:]`.

func GenerateBoundingBoxProposals 

func GenerateBoundingBoxProposals(scope *Scope, scores tf.Output, bbox_deltas tf.Output, image_info tf.Output, anchors tf.Output, nms_threshold tf.Output, pre_nms_topn tf.Output, min_size tf.Output, optional ...GenerateBoundingBoxProposalsAttr) (rois tf.Output, roi_probabilities tf.Output)

This op produces Region of Interests from given bounding boxes(bbox_deltas) encoded wrt anchors according to eq.2 in arXiv:1506.01497 

The op selects top `pre_nms_topn` scoring boxes, decodes them with respect to anchors,
applies non-maximal suppression on overlapping boxes with higher than
`nms_threshold` intersection-over-union (iou) value, discarding boxes where shorter
side is less than `min_size`.
Inputs:
`scores`: A 4D tensor of shape [Batch, Height, Width, Num Anchors] containing the scores per anchor at given position
`bbox_deltas`: is a tensor of shape [Batch, Height, Width, 4 x Num Anchors] boxes encoded to each anchor
`anchors`: A 1D tensor of shape [4 x Num Anchors], representing the anchors.
Outputs:
`rois`: output RoIs, a 3D tensor of shape [Batch, post_nms_topn, 4], padded by 0 if less than post_nms_topn candidates found.
`roi_probabilities`: probability scores of each roi in 'rois', a 2D tensor of shape [Batch,post_nms_topn], padded with 0 if needed, sorted by scores.

Arguments: 

scores: A 4-D float tensor of shape `[num_images, height, width, num_achors]` containing scores of the boxes for given anchors, can be unsorted.
bbox_deltas: A 4-D float tensor of shape `[num_images, height, width, 4 x num_anchors]`. encoding boxes with respec to each anchor.

Coordinates are given in the form [dy, dx, dh, dw]. 

image_info: A 2-D float tensor of shape `[num_images, 5]` containing image information Height, Width, Scale.
anchors: A 2-D float tensor of shape `[num_anchors, 4]` describing the anchor boxes. Boxes are formatted in the form [y1, x1, y2, x2].
nms_threshold: A scalar float tensor for non-maximal-suppression threshold.
pre_nms_topn: A scalar int tensor for the number of top scoring boxes to be used as input.
min_size: A scalar float tensor. Any box that has a smaller size than min_size will be discarded.

Returns: 

rois: A 3-D float tensor of shape `[num_images,post_nms_topn,4]` representing the selected

region of interest boxes. Sorted in descending order in scores. 

roi_probabilities: A 2-D float tensor of shape `[num_images, post_nms_topn]` representing the score of the

region of interest box in `rois` tensor at the same index.

func GenerateVocabRemapping 

func GenerateVocabRemapping(scope *Scope, new_vocab_file tf.Output, old_vocab_file tf.Output, new_vocab_offset int64, num_new_vocab int64, optional ...GenerateVocabRemappingAttr) (remapping tf.Output, num_present tf.Output)

Given a path to new and old vocabulary files, returns a remapping Tensor of

length `num_new_vocab`, where `remapping[i]` contains the row number in the old vocabulary that corresponds to row `i` in the new vocabulary (starting at line `new_vocab_offset` and up to `num_new_vocab` entities), or `-1` if entry `i` in the new vocabulary is not in the old vocabulary. The old vocabulary is constrained to the first `old_vocab_size` entries if `old_vocab_size` is not the default value of -1.

`num_vocab_offset` enables use in the partitioned variable case, and should generally be set through examining partitioning info. The format of the files should be a text file, with each line containing a single entity within the vocabulary.

For example, with `new_vocab_file` a text file containing each of the following elements on a single line: `[f0, f1, f2, f3]`, old_vocab_file = [f1, f0, f3], `num_new_vocab = 3, new_vocab_offset = 1`, the returned remapping would be `[0, -1, 2]`.

The op also returns a count of how many entries in the new vocabulary were present in the old vocabulary, which is used to calculate the number of values to initialize in a weight matrix remapping

This functionality can be used to remap both row vocabularies (typically, features) and column vocabularies (typically, classes) from TensorFlow checkpoints. Note that the partitioning logic relies on contiguous vocabularies corresponding to div-partitioned variables. Moreover, the underlying remapping uses an IndexTable (as opposed to an inexact CuckooTable), so client code should use the corresponding index_table_from_file() as the FeatureColumn framework does (as opposed to tf.feature_to_id(), which uses a CuckooTable).

Arguments: 

new_vocab_file: Path to the new vocab file.
old_vocab_file: Path to the old vocab file.
new_vocab_offset: How many entries into the new vocab file to start reading.
num_new_vocab: Number of entries in the new vocab file to remap.

Returns: 

remapping: A Tensor of length num_new_vocab where the element at index i

is equal to the old ID that maps to the new ID i. This element is -1 for any new ID that is not found in the old vocabulary. 

num_present: Number of new vocab entries found in old vocab.

func GetElementAtIndex 

func GetElementAtIndex(scope *Scope, dataset tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Gets the element at the specified index in a dataset.

func GetOptions 

func GetOptions(scope *Scope, input_dataset tf.Output) (serialized_options tf.Output)

Returns the `tf.data.Options` attached to `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the input dataset.

func GetSessionHandle 

func GetSessionHandle(scope *Scope, value tf.Output) (handle tf.Output)

Store the input tensor in the state of the current session.

Arguments: 

value: The tensor to be stored.

Returns The handle for the tensor stored in the session state, represented as a string.

func GetSessionHandleV2 

func GetSessionHandleV2(scope *Scope, value tf.Output) (handle tf.Output)

Store the input tensor in the state of the current session.

Arguments: 

value: The tensor to be stored.

Returns The handle for the tensor stored in the session state, represented as a ResourceHandle object.

func GetSessionTensor 

func GetSessionTensor(scope *Scope, handle tf.Output, dtype tf.DataType) (value tf.Output)

Get the value of the tensor specified by its handle.

Arguments: 

handle: The handle for a tensor stored in the session state.
dtype: The type of the output value.

Returns The tensor for the given handle.

func GetTpuTaskId added in v0.8.2 

func GetTpuTaskId(scope *Scope) (tpu_task_id tf.Output)

An op returns the TPU task ID from TPU topology.

This op is to return the TPU task ID from TPU topology.

Returns The TPU task ID from TPU topology.

func Gradients 

func Gradients(scope *Scope, y []tf.Output, x []tf.Output, dx ...tf.Output) (output []tf.Output)

Gradients adds gradients computation ops to the graph according to scope.

Arguments: 

y: output of the function to derive
x: inputs of the function for which partial derivatives are computed
dx: if not null, the partial derivatives of some loss function L w.r.t. y

return the partial derivatives

func Greater 

func Greater(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of (x > y) element-wise.

*NOTE*: `Greater` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

Example:

```python x = tf.constant([5, 4, 6]) y = tf.constant([5, 2, 5]) tf.math.greater(x, y) ==> [False, True, True]

x = tf.constant([5, 4, 6]) y = tf.constant([5]) tf.math.greater(x, y) ==> [False, False, True] ```

func GreaterEqual 

func GreaterEqual(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of (x >= y) element-wise.

*NOTE*: `GreaterEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

Example:

```python x = tf.constant([5, 4, 6, 7]) y = tf.constant([5, 2, 5, 10]) tf.math.greater_equal(x, y) ==> [True, True, True, False]

x = tf.constant([5, 4, 6, 7]) y = tf.constant([5]) tf.math.greater_equal(x, y) ==> [True, False, True, True] ```

func GuaranteeConst 

func GuaranteeConst(scope *Scope, input tf.Output) (output tf.Output)

Gives a guarantee to the TF runtime that the input tensor is a constant.

The runtime is then free to make optimizations based on this.

Only accepts value typed tensors as inputs and rejects resource variable handles as input.

Returns the input tensor without modification.

func HSVToRGB 

func HSVToRGB(scope *Scope, images tf.Output) (output tf.Output)

Convert one or more images from HSV to RGB.

Outputs a tensor of the same shape as the `images` tensor, containing the RGB value of the pixels. The output is only well defined if the value in `images` are in `[0,1]`.

See `rgb_to_hsv` for a description of the HSV encoding.

Arguments: 

images: 1-D or higher rank. HSV data to convert. Last dimension must be size 3.

Returns `images` converted to RGB.

func HashTableV2 

func HashTableV2(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType, optional ...HashTableV2Attr) (table_handle tf.Output)

Creates a non-initialized hash table.

This op creates a hash table, specifying the type of its keys and values. Before using the table you will have to initialize it. After initialization the table will be immutable.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns Handle to a table.

func HistogramFixedWidth 

func HistogramFixedWidth(scope *Scope, values tf.Output, value_range tf.Output, nbins tf.Output, optional ...HistogramFixedWidthAttr) (out tf.Output)

Return histogram of values.

Given the tensor `values`, this operation returns a rank 1 histogram counting the number of entries in `values` that fall into every bin. The bins are equal width and determined by the arguments `value_range` and `nbins`.

```python # Bins will be: (-inf, 1), [1, 2), [2, 3), [3, 4), [4, inf) nbins = 5 value_range = [0.0, 5.0] new_values = [-1.0, 0.0, 1.5, 2.0, 5.0, 15]

with tf.get_default_session() as sess: 

hist = tf.histogram_fixed_width(new_values, value_range, nbins=5)
variables.global_variables_initializer().run()
sess.run(hist) => [2, 1, 1, 0, 2]

```

Arguments: 

values: Numeric `Tensor`.
value_range: Shape [2] `Tensor` of same `dtype` as `values`.

values <= value_range[0] will be mapped to hist[0], values >= value_range[1] will be mapped to hist[-1]. 

nbins: Scalar `int32 Tensor`.  Number of histogram bins.

Returns A 1-D `Tensor` holding histogram of values.

func HistogramSummary 

func HistogramSummary(scope *Scope, tag tf.Output, values tf.Output) (summary tf.Output)

Outputs a `Summary` protocol buffer with a histogram.

The generated [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) has one summary value containing a histogram for `values`.

This op reports an `InvalidArgument` error if any value is not finite.

Arguments: 

tag: Scalar.  Tag to use for the `Summary.Value`.
values: Any shape. Values to use to build the histogram.

Returns Scalar. Serialized `Summary` protocol buffer.

func HostConst 

func HostConst(scope *Scope, value tf.Tensor, dtype tf.DataType) (output tf.Output)

Returns a constant tensor on the host. Only for writing C++ tests.

Arguments: 

value: Attr `value` is the tensor to return.

func IFFT 

func IFFT(scope *Scope, input tf.Output) (output tf.Output)

Inverse fast Fourier transform.

Computes the inverse 1-dimensional discrete Fourier transform over the inner-most dimension of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 

dimension of `input` is replaced with its inverse 1D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.ifft @end_compatibility

func IFFT2D 

func IFFT2D(scope *Scope, input tf.Output) (output tf.Output)

Inverse 2D fast Fourier transform.

Computes the inverse 2-dimensional discrete Fourier transform over the inner-most 2 dimensions of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 2 

dimensions of `input` are replaced with their inverse 2D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.ifft2 @end_compatibility

func IFFT3D 

func IFFT3D(scope *Scope, input tf.Output) (output tf.Output)

Inverse 3D fast Fourier transform.

Computes the inverse 3-dimensional discrete Fourier transform over the inner-most 3 dimensions of `input`.

Arguments: 

input: A complex tensor.

Returns A complex tensor of the same shape as `input`. The inner-most 3 

dimensions of `input` are replaced with their inverse 3D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.ifftn with 3 dimensions. @end_compatibility

func IFFTND added in v0.7.0 

func IFFTND(scope *Scope, input tf.Output, fft_length tf.Output, axes tf.Output) (output tf.Output)

ND inverse fast Fourier transform.

Computes the n-dimensional inverse discrete Fourier transform over designated dimensions of `input`. The designated dimensions of `input` are assumed to be the result of `IFFTND`.

If fft_length[i]<shape(input)[i], the input is cropped. If fft_length[i]>shape(input)[i], the input is padded with zeros. If fft_length is not given, the default shape(input) is used.

Axes mean the dimensions to perform the transform on. Default is to perform on all axes.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor. The FFT length for each dimension.
axes: An int32 tensor with a same shape as fft_length. Axes to perform the transform.

Returns A complex tensor of the same shape as `input`. The designated dimensions of `input` are replaced with their inverse Fourier transforms.

@compatibility(numpy) Equivalent to np.fft.fftn. @end_compatibility

func IRFFT 

func IRFFT(scope *Scope, input tf.Output, fft_length tf.Output, optional ...IRFFTAttr) (output tf.Output)

Inverse real-valued fast Fourier transform.

Computes the inverse 1-dimensional discrete Fourier transform of a real-valued signal over the inner-most dimension of `input`.

The inner-most dimension of `input` is assumed to be the result of `RFFT`: the `fft_length / 2 + 1` unique components of the DFT of a real-valued signal. If `fft_length` is not provided, it is computed from the size of the inner-most dimension of `input` (`fft_length = 2 * (inner - 1)`). If the FFT length used to compute `input` is odd, it should be provided since it cannot be inferred properly.

Along the axis `IRFFT` is computed on, if `fft_length / 2 + 1` is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor of shape [1]. The FFT length.

Returns A float32 tensor of the same rank as `input`. The inner-most 

dimension of `input` is replaced with the `fft_length` samples of its inverse
1D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.irfft @end_compatibility

func IRFFT2D 

func IRFFT2D(scope *Scope, input tf.Output, fft_length tf.Output, optional ...IRFFT2DAttr) (output tf.Output)

Inverse 2D real-valued fast Fourier transform.

Computes the inverse 2-dimensional discrete Fourier transform of a real-valued signal over the inner-most 2 dimensions of `input`.

The inner-most 2 dimensions of `input` are assumed to be the result of `RFFT2D`: The inner-most dimension contains the `fft_length / 2 + 1` unique components of the DFT of a real-valued signal. If `fft_length` is not provided, it is computed from the size of the inner-most 2 dimensions of `input`. If the FFT length used to compute `input` is odd, it should be provided since it cannot be inferred properly.

Along each axis `IRFFT2D` is computed on, if `fft_length` (or `fft_length / 2 + 1` for the inner-most dimension) is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor of shape [2]. The FFT length for each dimension.

Returns A float32 tensor of the same rank as `input`. The inner-most 2 

dimensions of `input` are replaced with the `fft_length` samples of their
inverse 2D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.irfft2 @end_compatibility

func IRFFT3D 

func IRFFT3D(scope *Scope, input tf.Output, fft_length tf.Output, optional ...IRFFT3DAttr) (output tf.Output)

Inverse 3D real-valued fast Fourier transform.

Computes the inverse 3-dimensional discrete Fourier transform of a real-valued signal over the inner-most 3 dimensions of `input`.

The inner-most 3 dimensions of `input` are assumed to be the result of `RFFT3D`: The inner-most dimension contains the `fft_length / 2 + 1` unique components of the DFT of a real-valued signal. If `fft_length` is not provided, it is computed from the size of the inner-most 3 dimensions of `input`. If the FFT length used to compute `input` is odd, it should be provided since it cannot be inferred properly.

Along each axis `IRFFT3D` is computed on, if `fft_length` (or `fft_length / 2 + 1` for the inner-most dimension) is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor of shape [3]. The FFT length for each dimension.

Returns A float32 tensor of the same rank as `input`. The inner-most 3 

dimensions of `input` are replaced with the `fft_length` samples of their
inverse 3D real Fourier transform.

@compatibility(numpy) Equivalent to np.irfftn with 3 dimensions. @end_compatibility

func IRFFTND added in v0.7.0 

func IRFFTND(scope *Scope, input tf.Output, fft_length tf.Output, axes tf.Output, optional ...IRFFTNDAttr) (output tf.Output)

ND inverse real fast Fourier transform.

Computes the n-dimensional inverse real discrete Fourier transform over designated dimensions of `input`. The designated dimensions of `input` are assumed to be the result of `IRFFTND`. The inner-most dimension contains the `fft_length / 2 + 1` unique components of the DFT of a real-valued signal.

If fft_length[i]<shape(input)[i], the input is cropped. If fft_length[i]>shape(input)[i], the input is padded with zeros. If fft_length is not given, the default shape(input) is used.

Axes mean the dimensions to perform the transform on. Default is to perform on all axes.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor. The FFT length for each dimension.
axes: An int32 tensor with a same shape as fft_length. Axes to perform the transform.

Returns A complex tensor of the same shape as `input`. The designated dimensions of `input` are replaced with their inverse real Fourier transforms.

@compatibility(numpy) Equivalent to np.fft.irfftn. @end_compatibility

func Identity 

func Identity(scope *Scope, input tf.Output) (output tf.Output)

Return a tensor with the same shape and contents as the input tensor or value.

func IdentityN 

func IdentityN(scope *Scope, input []tf.Output) (output []tf.Output)

Returns a list of tensors with the same shapes and contents as the input

tensors.

This op can be used to override the gradient for complicated functions. For example, suppose y = f(x) and we wish to apply a custom function g for backprop such that dx = g(dy). In Python,

```python with tf.get_default_graph().gradient_override_map( 

  {'IdentityN': 'OverrideGradientWithG'}):
y, _ = identity_n([f(x), x])

@tf.RegisterGradient('OverrideGradientWithG') def ApplyG(op, dy, _): 

return [None, g(dy)]  # Do not backprop to f(x).

```

func IdentityReaderV2 

func IdentityReaderV2(scope *Scope, optional ...IdentityReaderV2Attr) (reader_handle tf.Output)

A Reader that outputs the queued work as both the key and value.

To use, enqueue strings in a Queue. ReaderRead will take the front work string and output (work, work).

Returns The handle to reference the Reader.

func Igamma 

func Igamma(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)

Compute the lower regularized incomplete Gamma function `P(a, x)`.

The lower regularized incomplete Gamma function is defined as:

\\(P(a, x) = gamma(a, x) / Gamma(a) = 1 - Q(a, x)\\)

where

\\(gamma(a, x) = \\int_{0}^{x} t^{a-1} exp(-t) dt\\)

is the lower incomplete Gamma function.

Note, above `Q(a, x)` (`Igammac`) is the upper regularized complete Gamma function.

func IgammaGradA 

func IgammaGradA(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)

Computes the gradient of `igamma(a, x)` wrt `a`.

func Igammac 

func Igammac(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)

Compute the upper regularized incomplete Gamma function `Q(a, x)`.

The upper regularized incomplete Gamma function is defined as:

\\(Q(a, x) = Gamma(a, x) / Gamma(a) = 1 - P(a, x)\\)

where

\\(Gamma(a, x) = \int_{x}^{\infty} t^{a-1} exp(-t) dt\\)

is the upper incomplete Gamma function.

Note, above `P(a, x)` (`Igamma`) is the lower regularized complete Gamma function.

func IgnoreErrorsDataset 

func IgnoreErrorsDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...IgnoreErrorsDatasetAttr) (handle tf.Output)

Creates a dataset that contains the elements of `input_dataset` ignoring errors.

func Imag 

func Imag(scope *Scope, input tf.Output, optional ...ImagAttr) (output tf.Output)

Returns the imaginary part of a complex number.

Given a tensor `input` of complex numbers, this operation returns a tensor of type `float` that is the imaginary part of each element in `input`. All elements in `input` must be complex numbers of the form \\(a + bj\\), where *a* is the real part and *b* is the imaginary part returned by this operation.

For example:

``` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.imag(input) ==> [4.75, 5.75] ```

func ImageProjectiveTransformV2 

func ImageProjectiveTransformV2(scope *Scope, images tf.Output, transforms tf.Output, output_shape tf.Output, interpolation string, optional ...ImageProjectiveTransformV2Attr) (transformed_images tf.Output)

Applies the given transform to each of the images.

If one row of `transforms` is `[a0, a1, a2, b0, b1, b2, c0, c1]`, then it maps the *output* point `(x, y)` to a transformed *input* point `(x', y') = ((a0 x + a1 y + a2) / k, (b0 x + b1 y + b2) / k)`, where `k = c0 x + c1 y + 1`. If the transformed point lays outside of the input image, the output pixel is set to 0.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
transforms: 2-D Tensor, `[batch, 8]` or `[1, 8]` matrix, where each row corresponds to a 3 x 3

projective transformation matrix, with the last entry assumed to be 1. If there is one row, the same transformation will be applied to all images. 

output_shape: 1-D Tensor [new_height, new_width].
interpolation: Interpolation method, "NEAREST" or "BILINEAR".

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ImageProjectiveTransformV3 

func ImageProjectiveTransformV3(scope *Scope, images tf.Output, transforms tf.Output, output_shape tf.Output, fill_value tf.Output, interpolation string, optional ...ImageProjectiveTransformV3Attr) (transformed_images tf.Output)

Applies the given transform to each of the images.

If one row of `transforms` is `[a0, a1, a2, b0, b1, b2, c0, c1]`, then it maps the *output* point `(x, y)` to a transformed *input* point `(x', y') = ((a0 x + a1 y + a2) / k, (b0 x + b1 y + b2) / k)`, where `k = c0 x + c1 y + 1`. If the transformed point lays outside of the input image, the output pixel is set to fill_value.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
transforms: 2-D Tensor, `[batch, 8]` or `[1, 8]` matrix, where each row corresponds to a 3 x 3

projective transformation matrix, with the last entry assumed to be 1. If there is one row, the same transformation will be applied to all images. 

output_shape: 1-D Tensor [new_height, new_width].
fill_value: float, the value to be filled when fill_mode is constant".
interpolation: Interpolation method, "NEAREST" or "BILINEAR".

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ImageSummary 

func ImageSummary(scope *Scope, tag tf.Output, tensor tf.Output, optional ...ImageSummaryAttr) (summary tf.Output)

Outputs a `Summary` protocol buffer with images.

The summary has up to `max_images` summary values containing images. The images are built from `tensor` which must be 4-D with shape `[batch_size, height, width, channels]` and where `channels` can be:

* 1: `tensor` is interpreted as Grayscale. * 3: `tensor` is interpreted as RGB. * 4: `tensor` is interpreted as RGBA.

The images have the same number of channels as the input tensor. For float input, the values are normalized one image at a time to fit in the range `[0, 255]`. `uint8` values are unchanged. The op uses two different normalization algorithms:

  • If the input values are all positive, they are rescaled so the largest one is 255.

  • If any input value is negative, the values are shifted so input value 0.0 is at 127. They are then rescaled so that either the smallest value is 0, or the largest one is 255.

The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:

  • If `max_images` is 1, the summary value tag is '*tag*/image'.
  • If `max_images` is greater than 1, the summary value tags are generated sequentially as '*tag*/image/0', '*tag*/image/1', etc.

The `bad_color` argument is the color to use in the generated images for non-finite input values. It is a `uint8` 1-D tensor of length `channels`. Each element must be in the range `[0, 255]` (It represents the value of a pixel in the output image). Non-finite values in the input tensor are replaced by this tensor in the output image. The default value is the color red.

Arguments: 

tag: Scalar. Used to build the `tag` attribute of the summary values.
tensor: 4-D of shape `[batch_size, height, width, channels]` where

`channels` is 1, 3, or 4.

Returns Scalar. Serialized `Summary` protocol buffer.

func ImmutableConst 

func ImmutableConst(scope *Scope, dtype tf.DataType, shape tf.Shape, memory_region_name string) (tensor tf.Output)

Returns immutable tensor from memory region.

The current implementation memmaps the tensor from a file.

Arguments: 

dtype: Type of the returned tensor.
shape: Shape of the returned tensor.
memory_region_name: Name of readonly memory region used by the tensor, see

NewReadOnlyMemoryRegionFromFile in tensorflow::Env.

func InTopK 

func InTopK(scope *Scope, predictions tf.Output, targets tf.Output, k int64) (precision tf.Output)

Says whether the targets are in the top `K` predictions.

This outputs a `batch_size` bool array, an entry `out[i]` is `true` if the prediction for the target class is among the top `k` predictions among all predictions for example `i`. Note that the behavior of `InTopK` differs from the `TopK` op in its handling of ties; if multiple classes have the same prediction value and straddle the top-`k` boundary, all of those classes are considered to be in the top `k`.

More formally, let 

\\(predictions_i\\) be the predictions for all classes for example `i`,
\\(targets_i\\) be the target class for example `i`,
\\(out_i\\) be the output for example `i`,

$$out_i = predictions_{i, targets_i} \in TopKIncludingTies(predictions_i)$$

Arguments: 

predictions: A `batch_size` x `classes` tensor.
targets: A `batch_size` vector of class ids.
k: Number of top elements to look at for computing precision.

Returns Computed Precision at `k` as a `bool Tensor`.

func InTopKV2 

func InTopKV2(scope *Scope, predictions tf.Output, targets tf.Output, k tf.Output) (precision tf.Output)

Says whether the targets are in the top `K` predictions.

This outputs a `batch_size` bool array, an entry `out[i]` is `true` if the prediction for the target class is among the top `k` predictions among all predictions for example `i`. Note that the behavior of `InTopK` differs from the `TopK` op in its handling of ties; if multiple classes have the same prediction value and straddle the top-`k` boundary, all of those classes are considered to be in the top `k`.

More formally, let 

\\(predictions_i\\) be the predictions for all classes for example `i`,
\\(targets_i\\) be the target class for example `i`,
\\(out_i\\) be the output for example `i`,

$$out_i = predictions_{i, targets_i} \in TopKIncludingTies(predictions_i)$$

Arguments: 

predictions: A `batch_size` x `classes` tensor.
targets: A `batch_size` vector of class ids.
k: Number of top elements to look at for computing precision.

Returns Computed precision at `k` as a `bool Tensor`.

func InfeedDequeue 

func InfeedDequeue(scope *Scope, dtype tf.DataType, shape tf.Shape) (output tf.Output)

A placeholder op for a value that will be fed into the computation.

Arguments: 

dtype: The type of elements in the tensor.
shape: The shape of the tensor.

Returns A tensor that will be provided using the infeed mechanism.

func InfeedDequeueTuple 

func InfeedDequeueTuple(scope *Scope, dtypes []tf.DataType, shapes []tf.Shape) (outputs []tf.Output)

Fetches multiple values from infeed as an XLA tuple.

Arguments: 

dtypes: The element types of each element in `outputs`.
shapes: The shapes of each tensor in `outputs`.

Returns A list of tensors that will be provided using the infeed mechanism.

func InfeedEnqueue 

func InfeedEnqueue(scope *Scope, input tf.Output, optional ...InfeedEnqueueAttr) (o *tf.Operation)

An op which feeds a single Tensor value into the computation.

Arguments: 

input: A tensor that will be provided using the infeed mechanism.

Returns the created operation.

func InfeedEnqueuePrelinearizedBuffer 

func InfeedEnqueuePrelinearizedBuffer(scope *Scope, input tf.Output, optional ...InfeedEnqueuePrelinearizedBufferAttr) (o *tf.Operation)

An op which enqueues prelinearized buffer into TPU infeed.

Arguments: 

input: A variant tensor representing linearized output.

Returns the created operation.

func InfeedEnqueueTuple 

func InfeedEnqueueTuple(scope *Scope, inputs []tf.Output, shapes []tf.Shape, optional ...InfeedEnqueueTupleAttr) (o *tf.Operation)

Feeds multiple Tensor values into the computation as an XLA tuple.

Arguments: 

inputs: A list of tensors that will be provided using the infeed mechanism.
shapes: The shapes of each tensor in `inputs`.

Returns the created operation.

func InitializeTableFromTextFileV2 

func InitializeTableFromTextFileV2(scope *Scope, table_handle tf.Output, filename tf.Output, key_index int64, value_index int64, optional ...InitializeTableFromTextFileV2Attr) (o *tf.Operation)

Initializes a table from a text file.

It inserts one key-value pair into the table for each line of the file. The key and value is extracted from the whole line content, elements from the split line based on `delimiter` or the line number (starting from zero). Where to extract the key and value from a line is specified by `key_index` and `value_index`.

  • A value of -1 means use the line number(starting from zero), expects `int64`.
  • A value of -2 means use the whole line content, expects `string`.
  • A value >= 0 means use the index (starting at zero) of the split line based on `delimiter`.

Arguments: 

table_handle: Handle to a table which will be initialized.
filename: Filename of a vocabulary text file.
key_index: Column index in a line to get the table `key` values from.
value_index: Column index that represents information of a line to get the table

`value` values from.

Returns the created operation.

func InitializeTableV2 

func InitializeTableV2(scope *Scope, table_handle tf.Output, keys tf.Output, values tf.Output) (o *tf.Operation)

Table initializer that takes two tensors for keys and values respectively.

Arguments: 

table_handle: Handle to a table which will be initialized.
keys: Keys of type Tkey.
values: Values of type Tval.

Returns the created operation.

func InplaceAdd 

func InplaceAdd(scope *Scope, x tf.Output, i tf.Output, v tf.Output) (y tf.Output)

Adds v into specified rows of x. 

Computes y = x; y[i, :] += v; return y.

Arguments: 

x: A `Tensor` of type T.
i: A vector. Indices into the left-most dimension of `x`.
v: A `Tensor` of type T. Same dimension sizes as x except the first dimension, which must be the same as i's size.

Returns A `Tensor` of type T. An alias of `x`. The content of `y` is undefined if there are duplicates in `i`.

func InplaceSub 

func InplaceSub(scope *Scope, x tf.Output, i tf.Output, v tf.Output) (y tf.Output)
Subtracts `v` into specified rows of `x`.

Computes y = x; y[i, :] -= v; return y.

Arguments: 

x: A `Tensor` of type T.
i: A vector. Indices into the left-most dimension of `x`.
v: A `Tensor` of type T. Same dimension sizes as x except the first dimension, which must be the same as i's size.

Returns A `Tensor` of type T. An alias of `x`. The content of `y` is undefined if there are duplicates in `i`.

func InplaceUpdate 

func InplaceUpdate(scope *Scope, x tf.Output, i tf.Output, v tf.Output) (y tf.Output)

Updates specified rows 'i' with values 'v'.

Computes `x[i, :] = v; return x`.

Originally this function is mutative however for compilation we make this operation create / operate on a copy of `x`.

Arguments: 

x: A tensor of type `T`.
i: A vector. Indices into the left-most dimension of `x`.
v: A `Tensor` of type T. Same dimension sizes as x except the first dimension, which must be the same as i's size.

Returns A `Tensor` of type T. An alias of `x`. The content of `y` is undefined if there are duplicates in `i`.

func Inv 

func Inv(scope *Scope, x tf.Output) (y tf.Output)

Computes the reciprocal of x element-wise.

I.e., \\(y = 1 / x\\).

func InvGrad 

func InvGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient for the inverse of `x` wrt its input.

Specifically, `grad = -dy * y*y`, where `y = 1/x`, and `dy` is the corresponding input gradient.

func Invert 

func Invert(scope *Scope, x tf.Output) (y tf.Output)

Invert (flip) each bit of supported types; for example, type `uint8` value 01010101 becomes 10101010.

Flip each bit of supported types. For example, type `int8` (decimal 2) binary 00000010 becomes (decimal -3) binary 11111101. This operation is performed on each element of the tensor argument `x`.

Example: ```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops

# flip 2 (00000010) to -3 (11111101) tf.assert_equal(-3, bitwise_ops.invert(2))

dtype_list = [dtypes.int8, dtypes.int16, dtypes.int32, dtypes.int64, 

dtypes.uint8, dtypes.uint16, dtypes.uint32, dtypes.uint64]

inputs = [0, 5, 3, 14] for dtype in dtype_list: 

# Because of issues with negative numbers, let's test this indirectly.
# 1. invert(a) and a = 0
# 2. invert(a) or a = invert(0)
input_tensor = tf.constant([0, 5, 3, 14], dtype=dtype)
not_a_and_a, not_a_or_a, not_0 = [bitwise_ops.bitwise_and(
                                    input_tensor, bitwise_ops.invert(input_tensor)),
                                  bitwise_ops.bitwise_or(
                                    input_tensor, bitwise_ops.invert(input_tensor)),
                                  bitwise_ops.invert(
                                    tf.constant(0, dtype=dtype))]

expected = tf.constant([0, 0, 0, 0], dtype=tf.float32)
tf.assert_equal(tf.cast(not_a_and_a, tf.float32), expected)

expected = tf.cast([not_0] * 4, tf.float32)
tf.assert_equal(tf.cast(not_a_or_a, tf.float32), expected)

# For unsigned dtypes let's also check the result directly.
if dtype.is_unsigned:
  inverted = bitwise_ops.invert(input_tensor)
  expected = tf.constant([dtype.max - x for x in inputs], dtype=tf.float32)
  tf.assert_equal(tf.cast(inverted, tf.float32), tf.cast(expected, tf.float32))

```

func InvertPermutation 

func InvertPermutation(scope *Scope, x tf.Output) (y tf.Output)

Computes the inverse permutation of a tensor.

This operation computes the inverse of an index permutation. It takes a 1-D integer tensor `x`, which represents the indices of a zero-based array, and swaps each value with its index position. In other words, for an output tensor `y` and an input tensor `x`, this operation computes the following:

`y[x[i]] = i for i in [0, 1, ..., len(x) - 1]`

The values must include 0. There can be no duplicate values or negative values.

For example:

``` # tensor `x` is [3, 4, 0, 2, 1] invert_permutation(x) ==> [2, 4, 3, 0, 1] ```

Arguments: 

x: 1-D.

Returns 1-D.

func IsBoostedTreesEnsembleInitialized 

func IsBoostedTreesEnsembleInitialized(scope *Scope, tree_ensemble_handle tf.Output) (is_initialized tf.Output)

Checks whether a tree ensemble has been initialized.

Arguments: 

tree_ensemble_handle: Handle to the tree ensemble resource.

Returns output boolean on whether it is initialized or not.

func IsBoostedTreesQuantileStreamResourceInitialized 

func IsBoostedTreesQuantileStreamResourceInitialized(scope *Scope, quantile_stream_resource_handle tf.Output) (is_initialized tf.Output)

Checks whether a quantile stream has been initialized.

An Op that checks if quantile stream resource is initialized.

Arguments: 

quantile_stream_resource_handle: resource; The reference to quantile stream resource handle.

Returns bool; True if the resource is initialized, False otherwise.

func IsFinite 

func IsFinite(scope *Scope, x tf.Output) (y tf.Output)

Returns which elements of x are finite.

@compatibility(numpy) Equivalent to np.isfinite @end_compatibility

Example:

```python x = tf.constant([5.0, 4.8, 6.8, np.inf, np.nan]) tf.math.is_finite(x) ==> [True, True, True, False, False] ```

func IsInf 

func IsInf(scope *Scope, x tf.Output) (y tf.Output)

Returns which elements of x are Inf.

@compatibility(numpy) Equivalent to np.isinf @end_compatibility

Example:

```python x = tf.constant([5.0, np.inf, 6.8, np.inf]) tf.math.is_inf(x) ==> [False, True, False, True] ```

func IsNan 

func IsNan(scope *Scope, x tf.Output) (y tf.Output)

Returns which elements of x are NaN.

@compatibility(numpy) Equivalent to np.isnan @end_compatibility

Example:

```python x = tf.constant([5.0, np.nan, 6.8, np.nan, np.inf]) tf.math.is_nan(x) ==> [False, True, False, True, False] ```

func IsTPUEmbeddingInitialized 

func IsTPUEmbeddingInitialized(scope *Scope, optional ...IsTPUEmbeddingInitializedAttr) (is_tpu_embedding_initialized tf.Output)

Whether TPU Embedding is initialized in a distributed TPU system.

func IsotonicRegression 

func IsotonicRegression(scope *Scope, input tf.Output, optional ...IsotonicRegressionAttr) (output tf.Output, segments tf.Output)

Solves a batch of isotonic regression problems.

Arguments: 

input: A (batch_size, dim)-tensor holding a batch of inputs.

Returns: 

output: A (batch_size, dim)-tensor holding the per-batch element solutions.
segments: An int32 (batch_size, dim)-tensor with the segments.

func Iterator 

func Iterator(scope *Scope, shared_name string, container string, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

A container for an iterator resource.

Returns A handle to the iterator that can be passed to a "MakeIterator" or "IteratorGetNext" op.

func IteratorFromStringHandle 

func IteratorFromStringHandle(scope *Scope, string_handle tf.Output, optional ...IteratorFromStringHandleAttr) (resource_handle tf.Output)

Converts the given string representing a handle to an iterator to a resource.

Arguments: 

string_handle: A string representation of the given handle.

Returns A handle to an iterator resource.

func IteratorGetDevice 

func IteratorGetDevice(scope *Scope, resource tf.Output) (device tf.Output)

Returns the name of the device on which `resource` has been placed.

func IteratorGetModelProto added in v0.8.2 

func IteratorGetModelProto(scope *Scope, iterator tf.Output) (model_proto tf.Output)

Returns the serialized model proto of an iterator resource.

Returns the serialized model proto of an iterator resource.

Arguments: 

iterator: An resource from an dataset iterator.

Returns A serialized model proto.

func IteratorGetNext 

func IteratorGetNext(scope *Scope, iterator tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Gets the next output from the given iterator .

func IteratorGetNextAsOptional 

func IteratorGetNextAsOptional(scope *Scope, iterator tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (optional tf.Output)

Gets the next output from the given iterator as an Optional variant.

func IteratorGetNextSync 

func IteratorGetNextSync(scope *Scope, iterator tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Gets the next output from the given iterator.

This operation is a synchronous version IteratorGetNext. It should only be used in situations where the iterator does not block the calling thread, or where the calling thread is not a member of the thread pool used to execute parallel operations (e.g. in eager mode).

func IteratorToStringHandle 

func IteratorToStringHandle(scope *Scope, resource_handle tf.Output) (string_handle tf.Output)

Converts the given `resource_handle` representing an iterator to a string.

Arguments: 

resource_handle: A handle to an iterator resource.

Returns A string representation of the given handle.

func KMC2ChainInitialization 

func KMC2ChainInitialization(scope *Scope, distances tf.Output, seed tf.Output) (index tf.Output)

Returns the index of a data point that should be added to the seed set.

Entries in distances are assumed to be squared distances of candidate points to the already sampled centers in the seed set. The op constructs one Markov chain of the k-MC^2 algorithm and returns the index of one candidate point to be added as an additional cluster center.

Arguments: 

distances: Vector with squared distances to the closest previously sampled cluster center

for each candidate point. 

seed: Scalar. Seed for initializing the random number generator.

Returns Scalar with the index of the sampled point.

func KmeansPlusPlusInitialization 

func KmeansPlusPlusInitialization(scope *Scope, points tf.Output, num_to_sample tf.Output, seed tf.Output, num_retries_per_sample tf.Output) (samples tf.Output)

Selects num_to_sample rows of input using the KMeans++ criterion.

Rows of points are assumed to be input points. One row is selected at random. Subsequent rows are sampled with probability proportional to the squared L2 distance from the nearest row selected thus far till num_to_sample rows have been sampled.

Arguments: 

points: Matrix of shape (n, d). Rows are assumed to be input points.
num_to_sample: Scalar. The number of rows to sample. This value must not be larger than n.
seed: Scalar. Seed for initializing the random number generator.
num_retries_per_sample: Scalar. For each row that is sampled, this parameter

specifies the number of additional points to draw from the current distribution before selecting the best. If a negative value is specified, a heuristic is used to sample O(log(num_to_sample)) additional points.

Returns Matrix of shape (num_to_sample, d). The sampled rows.

func KthOrderStatistic 

func KthOrderStatistic(scope *Scope, input tf.Output, k int64) (output tf.Output)

Computes the Kth order statistic of a data set. The current

implementation uses a binary search requiring exactly 32 passes over the input data. The running time is linear with respect to input size. The median-of-medians algorithm is probably faster, but is difficult to implement efficiently in XLA. The implementation imposes a total ordering on floats. The ordering is consistent with the usual partial order. Positive NaNs are greater than positive infinity. Negative NaNs are less than negative infinity. NaNs with distinct payloads are treated as distinct. Subnormal numbers are preserved (not flushed to zero). Positive infinity is greater than all numbers. Negative infinity is less than all numbers. Positive is greater than negative zero. There are less than k values greater than the kth order statistic. There are at least k values greater than or equal to the Kth order statistic. The semantics are not the same as top_k_unique.

func L2Loss 

func L2Loss(scope *Scope, t tf.Output) (output tf.Output)

L2 Loss.

Computes half the L2 norm of a tensor without the `sqrt`: 

output = sum(t ** 2) / 2

Arguments: 

t: Typically 2-D, but may have any dimensions.

Returns 0-D.

func LMDBDataset 

func LMDBDataset(scope *Scope, filenames tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that emits the key-value pairs in one or more LMDB files.

The Lightning Memory-Mapped Database Manager, or LMDB, is an embedded binary key-value database. This dataset can read the contents of LMDB database files, the names of which generally have the `.mdb` suffix.

Each output element consists of a key-value pair represented as a pair of scalar string `Tensor`s, where the first `Tensor` contains the key and the second `Tensor` contains the value.

LMDB uses different file formats on big- and little-endian machines. `LMDBDataset` can only read files in the format of the host machine.

Arguments: 

filenames: A scalar or a vector containing the name(s) of the binary file(s) to be

read.

func LRN 

func LRN(scope *Scope, input tf.Output, optional ...LRNAttr) (output tf.Output)

Local Response Normalization.

The 4-D `input` tensor is treated as a 3-D array of 1-D vectors (along the last dimension), and each vector is normalized independently. Within a given vector, each component is divided by the weighted, squared sum of inputs within `depth_radius`. In detail, 

sqr_sum[a, b, c, d] =
    sum(input[a, b, c, d - depth_radius : d + depth_radius + 1] ** 2)
output = input / (bias + alpha * sqr_sum) ** beta

For details, see [Krizhevsky et al., ImageNet classification with deep convolutional neural networks (NIPS 2012)](http://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks).

Arguments: 

input: 4-D.

func LRNGrad 

func LRNGrad(scope *Scope, input_grads tf.Output, input_image tf.Output, output_image tf.Output, optional ...LRNGradAttr) (output tf.Output)

Gradients for Local Response Normalization.

Arguments: 

input_grads: 4-D with shape `[batch, height, width, channels]`.
input_image: 4-D with shape `[batch, height, width, channels]`.
output_image: 4-D with shape `[batch, height, width, channels]`.

Returns The gradients for LRN.

func LSTMBlockCell 

func LSTMBlockCell(scope *Scope, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, optional ...LSTMBlockCellAttr) (i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, h tf.Output)

Computes the LSTM cell forward propagation for 1 time step.

This implementation uses 1 weight matrix and 1 bias vector, and there's an optional peephole connection.

This kernel op implements the following mathematical equations:

```python xh = [x, h_prev] [i, f, ci, o] = xh * w + b f = f + forget_bias

if not use_peephole: 

wci = wcf = wco = 0

i = sigmoid(cs_prev * wci + i) f = sigmoid(cs_prev * wcf + f) ci = tanh(ci)

cs = ci .* i + cs_prev .* f cs = clip(cs, cell_clip)

o = sigmoid(cs * wco + o) co = tanh(cs) h = co .* o ```

Arguments: 

x: The input to the LSTM cell, shape (batch_size, num_inputs).
cs_prev: Value of the cell state at previous time step.
h_prev: Output of the previous cell at previous time step.
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.

Returns: 

i: The input gate.
cs: The cell state before the tanh.
f: The forget gate.
o: The output gate.
ci: The cell input.
co: The cell after the tanh.
h: The output h vector.

func LSTMBlockCellGrad 

func LSTMBlockCellGrad(scope *Scope, x tf.Output, cs_prev tf.Output, h_prev tf.Output, w tf.Output, wci tf.Output, wcf tf.Output, wco tf.Output, b tf.Output, i tf.Output, cs tf.Output, f tf.Output, o tf.Output, ci tf.Output, co tf.Output, cs_grad tf.Output, h_grad tf.Output, use_peephole bool) (cs_prev_grad tf.Output, dicfo tf.Output, wci_grad tf.Output, wcf_grad tf.Output, wco_grad tf.Output)

Computes the LSTM cell backward propagation for 1 timestep.

This implementation is to be used in conjunction of LSTMBlockCell.

Arguments: 

x: The input to the LSTM cell, shape (batch_size, num_inputs).
cs_prev: The previous cell state.
h_prev: The previous h state.
w: The weight matrix.
wci: The weight matrix for input gate peephole connection.
wcf: The weight matrix for forget gate peephole connection.
wco: The weight matrix for output gate peephole connection.
b: The bias vector.
i: The input gate.
cs: The cell state before the tanh.
f: The forget gate.
o: The output gate.
ci: The cell input.
co: The cell after the tanh.
cs_grad: The current gradient of cs.
h_grad: The gradient of h vector.
use_peephole: Whether the cell uses peephole connections.

Returns: 

cs_prev_grad: The gradient of cs to be back-propped.
dicfo: The derivative wrt to [i, cs, f, o].
wci_grad: The gradient for wci to be back-propped.
wcf_grad: The gradient for wcf to be back-propped.
wco_grad: The gradient for wco to be back-propped.

func LatencyStatsDataset 

func LatencyStatsDataset(scope *Scope, input_dataset tf.Output, tag tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Records the latency of producing `input_dataset` elements in a StatsAggregator.

func LeakyRelu 

func LeakyRelu(scope *Scope, features tf.Output, optional ...LeakyReluAttr) (activations tf.Output)

Computes rectified linear: `max(features, features * alpha)`.

func LeakyReluGrad 

func LeakyReluGrad(scope *Scope, gradients tf.Output, features tf.Output, optional ...LeakyReluGradAttr) (backprops tf.Output)

Computes rectified linear gradients for a LeakyRelu operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding LeakyRelu operation.
features: The features passed as input to the corresponding LeakyRelu operation,

OR the outputs of that operation (both work equivalently).

Returns `gradients * (features > 0) + alpha * gradients * (features <= 0)`.

func LearnedUnigramCandidateSampler 

func LearnedUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...LearnedUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a learned unigram distribution.

See explanations of candidate sampling and the data formats at go/candidate-sampling.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to randomly sample.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities. 

range_max: The sampler will sample integers from the interval [0, range_max).

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func LeftShift 

func LeftShift(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Elementwise computes the bitwise left-shift of `x` and `y`.

If `y` is negative, or greater than or equal to the width of `x` in bits the result is implementation defined.

Example:

```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops import numpy as np dtype_list = [tf.int8, tf.int16, tf.int32, tf.int64]

for dtype in dtype_list: 

lhs = tf.constant([-1, -5, -3, -14], dtype=dtype)
rhs = tf.constant([5, 0, 7, 11], dtype=dtype)

left_shift_result = bitwise_ops.left_shift(lhs, rhs)

print(left_shift_result)

# This will print: # tf.Tensor([ -32 -5 -128 0], shape=(4,), dtype=int8) # tf.Tensor([ -32 -5 -384 -28672], shape=(4,), dtype=int16) # tf.Tensor([ -32 -5 -384 -28672], shape=(4,), dtype=int32) # tf.Tensor([ -32 -5 -384 -28672], shape=(4,), dtype=int64)

lhs = np.array([-2, 64, 101, 32], dtype=np.int8) rhs = np.array([-1, -5, -3, -14], dtype=np.int8) bitwise_ops.left_shift(lhs, rhs) # <tf.Tensor: shape=(4,), dtype=int8, numpy=array([ -2, 64, 101, 32], dtype=int8)> ```

func Less 

func Less(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of (x < y) element-wise.

*NOTE*: `Less` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

Example:

```python x = tf.constant([5, 4, 6]) y = tf.constant([5]) tf.math.less(x, y) ==> [False, True, False]

x = tf.constant([5, 4, 6]) y = tf.constant([5, 6, 7]) tf.math.less(x, y) ==> [False, True, True] ```

func LessEqual 

func LessEqual(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of (x <= y) element-wise.

*NOTE*: `LessEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

Example:

```python x = tf.constant([5, 4, 6]) y = tf.constant([5]) tf.math.less_equal(x, y) ==> [True, True, False]

x = tf.constant([5, 4, 6]) y = tf.constant([5, 6, 6]) tf.math.less_equal(x, y) ==> [True, True, True] ```

func Lgamma 

func Lgamma(scope *Scope, x tf.Output) (y tf.Output)

Computes the log of the absolute value of `Gamma(x)` element-wise. 

For positive numbers, this function computes log((input - 1)!) for every element in the tensor.
`lgamma(5) = log((5-1)!) = log(4!) = log(24) = 3.1780539`

Example:

```python x = tf.constant([0, 0.5, 1, 4.5, -4, -5.6]) tf.math.lgamma(x) ==> [inf, 0.5723649, 0., 2.4537368, inf, -4.6477685] ```

func LinSpace 

func LinSpace(scope *Scope, start tf.Output, stop tf.Output, num tf.Output) (output tf.Output)

Generates values in an interval.

A sequence of `num` evenly-spaced values are generated beginning at `start`. If `num > 1`, the values in the sequence increase by `(stop - start) / (num - 1)`, so that the last one is exactly `stop`.

For example:

``` tf.linspace(10.0, 12.0, 3, name="linspace") => [ 10.0 11.0 12.0] ```

Arguments: 

start: 0-D tensor. First entry in the range.
stop: 0-D tensor. Last entry in the range.
num: 0-D tensor. Number of values to generate.

Returns 1-D. The generated values.

func ListDataset added in v0.2.0 

func ListDataset(scope *Scope, tensors []tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ListDatasetAttr) (handle tf.Output)

Creates a dataset that emits each of `tensors` once.

func ListDiff 

func ListDiff(scope *Scope, x tf.Output, y tf.Output, optional ...ListDiffAttr) (out tf.Output, idx tf.Output)

Computes the difference between two lists of numbers or strings.

Given a list `x` and a list `y`, this operation returns a list `out` that represents all values that are in `x` but not in `y`. The returned list `out` is sorted in the same order that the numbers appear in `x` (duplicates are preserved). This operation also returns a list `idx` that represents the position of each `out` element in `x`. In other words:

`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]`

For example, given this input:

``` x = [1, 2, 3, 4, 5, 6] y = [1, 3, 5] ```

This operation would return:

``` out ==> [2, 4, 6] idx ==> [1, 3, 5] ```

Arguments: 

x: 1-D. Values to keep.
y: 1-D. Values to remove.

Returns: 

out: 1-D. Values present in `x` but not in `y`.
idx: 1-D. Positions of `x` values preserved in `out`.

func LoadAllTPUEmbeddingParameters 

func LoadAllTPUEmbeddingParameters(scope *Scope, parameters []tf.Output, auxiliary1 []tf.Output, auxiliary2 []tf.Output, auxiliary3 []tf.Output, auxiliary4 []tf.Output, auxiliary5 []tf.Output, auxiliary6 []tf.Output, auxiliary7 []tf.Output, config string, num_shards int64, shard_id int64) (o *tf.Operation)

An op that loads optimization parameters into embedding memory.

An op that loads optimization parameters into embedding memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed. For Adagrad, auxiliary1 should be the accumulators. For SGD, all of the auxiliary* values should be empty. For FTRL, auxiliary1 should be the accumulators and auxiliary2 should be the linear terms. For ADAM, auxiliary1 should be the momenta and auxiliary2 should be the velocities.

Arguments: 

parameters: A list of tensors, one for each embedding table,

containing the initial embedding table parameters to use in embedding lookups. 

auxiliary1: A list of tensors, one for each embedding table, containing the

initial values of the first auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have at least one auxiliary parameter. 

auxiliary2: A list of tensors, one for each embedding table, containing the

initial values of the second auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have at least two auxiliary 

auxiliary3: A list of tensors, one for each embedding table, containing the

initial values of the third auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have three auxiliary parameters. 

auxiliary4: A list of tensors, one for each embedding table, containing the

initial values of the second auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have at least four auxiliary 

auxiliary5: A list of tensors, one for each embedding table, containing the

initial values of the third auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have five auxiliary parameters. 

auxiliary6: A list of tensors, one for each embedding table, containing the

initial values of the second auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have at least six auxiliary 

auxiliary7: A list of tensors, one for each embedding table, containing the

initial values of the third auxiliary optimization parameter to use in embedding training loop updates. The shape of each entry is ignored (and thus can be empty) for those tables whose optimization algorithms do not have sevan auxiliary parameters. 

config: An TPUEmbeddingConfiguration proto describing the

table parameters being loaded, serialized to a string. 

num_shards: Number of shards into which the embedding tables are divided.
shard_id: Identifier of shard for this operation.

Returns the created operation.

func LoadAndRemapMatrix 

func LoadAndRemapMatrix(scope *Scope, ckpt_path tf.Output, old_tensor_name tf.Output, row_remapping tf.Output, col_remapping tf.Output, initializing_values tf.Output, num_rows int64, num_cols int64, optional ...LoadAndRemapMatrixAttr) (output_matrix tf.Output)

Loads a 2-D (matrix) `Tensor` with name `old_tensor_name` from the checkpoint

at `ckpt_path` and potentially reorders its rows and columns using the specified remappings.

Most users should use one of the wrapper initializers (such as `tf.contrib.framework.load_and_remap_matrix_initializer`) instead of this function directly.

The remappings are 1-D tensors with the following properties:

  • `row_remapping` must have exactly `num_rows` entries. Row `i` of the output matrix will be initialized from the row corresponding to index `row_remapping[i]` in the old `Tensor` from the checkpoint.
  • `col_remapping` must have either 0 entries (indicating that no column reordering is needed) or `num_cols` entries. If specified, column `j` of the output matrix will be initialized from the column corresponding to index `col_remapping[j]` in the old `Tensor` from the checkpoint.
  • A value of -1 in either of the remappings signifies a "missing" entry. In that case, values from the `initializing_values` tensor will be used to fill that missing row or column. If `row_remapping` has `r` missing entries and `col_remapping` has `c` missing entries, then the following condition must be true:

`(r * num_cols) + (c * num_rows) - (r * c) == len(initializing_values)`

The remapping tensors can be generated using the GenerateVocabRemapping op.

As an example, with row_remapping = [1, 0, -1], col_remapping = [0, 2, -1], initializing_values = [0.5, -0.5, 0.25, -0.25, 42], and w(i, j) representing the value from row i, column j of the old tensor in the checkpoint, the output matrix will look like the following:

[[w(1, 0), w(1, 2), 0.5], 

[w(0, 0),  w(0, 2), -0.5],
[0.25,    -0.25,      42]]

Arguments: 

ckpt_path: Path to the TensorFlow checkpoint (version 2, `TensorBundle`) from

which the old matrix `Tensor` will be loaded. 

old_tensor_name: Name of the 2-D `Tensor` to load from checkpoint.
row_remapping: An int `Tensor` of row remappings (generally created by

`generate_vocab_remapping`). Even if no row remapping is needed, this must still be an index-valued Tensor (e.g. [0, 1, 2, ...]), or a shifted index-valued `Tensor` (e.g. [8, 9, 10, ...], for partitioned `Variables`). 

col_remapping: An int `Tensor` of column remappings (generally created by

`generate_vocab_remapping`). May be a size-0 `Tensor` if only row remapping is to be done (e.g. column ordering is the same). 

initializing_values: A float `Tensor` containing  values to fill in for cells

in the output matrix that are not loaded from the checkpoint. Length must be exactly the same as the number of missing / new cells. 

num_rows: Number of rows (length of the 1st dimension) in the output matrix.
num_cols: Number of columns (length of the 2nd dimension) in the output matrix.

Returns Output matrix containing existing values loaded from the checkpoint, and with any missing values filled in from initializing_values.

func LoadTPUEmbeddingADAMParameters 

func LoadTPUEmbeddingADAMParameters(scope *Scope, parameters tf.Output, momenta tf.Output, velocities tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingADAMParametersAttr) (o *tf.Operation)

Load ADAM embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the ADAM optimization algorithm.
momenta: Value of momenta used in the ADAM optimization algorithm.
velocities: Value of velocities used in the ADAM optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingAdadeltaParameters 

func LoadTPUEmbeddingAdadeltaParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, updates tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingAdadeltaParametersAttr) (o *tf.Operation)

Load Adadelta embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the Adadelta optimization algorithm.
accumulators: Value of accumulators used in the Adadelta optimization algorithm.
updates: Value of updates used in the Adadelta optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingAdagradMomentumParameters 

func LoadTPUEmbeddingAdagradMomentumParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, momenta tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingAdagradMomentumParametersAttr) (o *tf.Operation)

Load Adagrad Momentum embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the Adagrad Momentum optimization algorithm.
accumulators: Value of accumulators used in the Adagrad Momentum optimization algorithm.
momenta: Value of momenta used in the Adagrad Momentum optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingAdagradParameters 

func LoadTPUEmbeddingAdagradParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingAdagradParametersAttr) (o *tf.Operation)

Load Adagrad embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the Adagrad optimization algorithm.
accumulators: Value of accumulators used in the Adagrad optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingCenteredRMSPropParameters 

func LoadTPUEmbeddingCenteredRMSPropParameters(scope *Scope, parameters tf.Output, ms tf.Output, mom tf.Output, mg tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingCenteredRMSPropParametersAttr) (o *tf.Operation)

Load centered RMSProp embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the centered RMSProp optimization algorithm.
ms: Value of ms used in the centered RMSProp optimization algorithm.
mom: Value of mom used in the centered RMSProp optimization algorithm.
mg: Value of mg used in the centered RMSProp optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingFTRLParameters 

func LoadTPUEmbeddingFTRLParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, linears tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingFTRLParametersAttr) (o *tf.Operation)

Load FTRL embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the FTRL optimization algorithm.
accumulators: Value of accumulators used in the FTRL optimization algorithm.
linears: Value of linears used in the FTRL optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingFrequencyEstimatorParameters 

func LoadTPUEmbeddingFrequencyEstimatorParameters(scope *Scope, parameters tf.Output, last_hit_step tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingFrequencyEstimatorParametersAttr) (o *tf.Operation)

Load frequency estimator embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the frequency estimator optimization algorithm.
last_hit_step: Value of last_hit_step used in the frequency estimator optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingMDLAdagradLightParameters 

func LoadTPUEmbeddingMDLAdagradLightParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, weights tf.Output, benefits tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingMDLAdagradLightParametersAttr) (o *tf.Operation)

Load MDL Adagrad Light embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the MDL Adagrad Light optimization algorithm.
accumulators: Value of accumulators used in the MDL Adagrad Light optimization algorithm.
weights: Value of weights used in the MDL Adagrad Light optimization algorithm.
benefits: Value of benefits used in the MDL Adagrad Light optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingMomentumParameters 

func LoadTPUEmbeddingMomentumParameters(scope *Scope, parameters tf.Output, momenta tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingMomentumParametersAttr) (o *tf.Operation)

Load Momentum embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the Momentum optimization algorithm.
momenta: Value of momenta used in the Momentum optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingProximalAdagradParameters 

func LoadTPUEmbeddingProximalAdagradParameters(scope *Scope, parameters tf.Output, accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingProximalAdagradParametersAttr) (o *tf.Operation)

Load proximal Adagrad embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the proximal Adagrad optimization algorithm.
accumulators: Value of accumulators used in the proximal Adagrad optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingRMSPropParameters 

func LoadTPUEmbeddingRMSPropParameters(scope *Scope, parameters tf.Output, ms tf.Output, mom tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingRMSPropParametersAttr) (o *tf.Operation)

Load RMSProp embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the RMSProp optimization algorithm.
ms: Value of ms used in the RMSProp optimization algorithm.
mom: Value of mom used in the RMSProp optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingStochasticGradientDescentParameters 

func LoadTPUEmbeddingStochasticGradientDescentParameters(scope *Scope, parameters tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingStochasticGradientDescentParametersAttr) (o *tf.Operation)

Load SGD embedding parameters.

An op that loads optimization parameters into HBM for embedding. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to install parameters that are loaded from a checkpoint before a training loop is executed.

Arguments: 

parameters: Value of parameters used in the stochastic gradient descent optimization algorithm.

Returns the created operation.

func Log 

func Log(scope *Scope, x tf.Output) (y tf.Output)

Computes natural logarithm of x element-wise.

I.e., \\(y = \log_e x\\).

Example:

```python x = tf.constant([0, 0.5, 1, 5]) tf.math.log(x) ==> [-inf, -0.6931472, 0. , 1.609438] ```

func Log1p 

func Log1p(scope *Scope, x tf.Output) (y tf.Output)

Computes natural logarithm of (1 + x) element-wise.

I.e., \\(y = \log_e (1 + x)\\).

Example:

```python x = tf.constant([0, 0.5, 1, 5]) tf.math.log1p(x) ==> [0., 0.4054651, 0.6931472, 1.7917595] ```

func LogMatrixDeterminant 

func LogMatrixDeterminant(scope *Scope, input tf.Output) (sign tf.Output, log_abs_determinant tf.Output)

Computes the sign and the log of the absolute value of the determinant of

one or more square matrices.

The input is a tensor of shape `[N, M, M]` whose inner-most 2 dimensions form square matrices. The outputs are two tensors containing the signs and absolute values of the log determinants for all N input submatrices `[..., :, :]` such that `determinant = sign*exp(log_abs_determinant)`. The `log_abs_determinant` is computed as `det(P)*sum(log(diag(LU)))` where `LU` is the `LU` decomposition of the input and `P` is the corresponding permutation matrix.

Arguments: 

input: Shape is `[N, M, M]`.

Returns: 

sign: The signs of the log determinants of the inputs. Shape is `[N]`.
log_abs_determinant: The logs of the absolute values of the determinants

of the N input matrices. Shape is `[N]`.

func LogSoftmax 

func LogSoftmax(scope *Scope, logits tf.Output) (logsoftmax tf.Output)

Computes log softmax activations.

For each batch `i` and class `j` we have 

logsoftmax[i, j] = logits[i, j] - log(sum(exp(logits[i])))

Arguments: 

logits: 2-D with shape `[batch_size, num_classes]`.

Returns Same shape as `logits`.

func LogUniformCandidateSampler 

func LogUniformCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...LogUniformCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a log-uniform distribution.

See explanations of candidate sampling and the data formats at go/candidate-sampling.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to randomly sample.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities. 

range_max: The sampler will sample integers from the interval [0, range_max).

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func LogicalAnd 

func LogicalAnd(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of x AND y element-wise.

*NOTE*: `LogicalAnd` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func LogicalNot 

func LogicalNot(scope *Scope, x tf.Output) (y tf.Output)

Returns the truth value of `NOT x` element-wise.

Arguments: 

x: A `Tensor` of type `bool`.

Returns A `Tensor` of type `bool` with the same shape as `x`. The logical negation of `x`.

func LogicalOr 

func LogicalOr(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the truth value of x OR y element-wise.

*NOTE*: `LogicalOr` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func LookupTableExportV2 

func LookupTableExportV2(scope *Scope, table_handle tf.Output, Tkeys tf.DataType, Tvalues tf.DataType) (keys tf.Output, values tf.Output)

Outputs all keys and values in the table.

Arguments: 

table_handle: Handle to the table.

Returns: 

keys: Vector of all keys present in the table.
values: Tensor of all values in the table. Indexed in parallel with `keys`.

func LookupTableFindV2 

func LookupTableFindV2(scope *Scope, table_handle tf.Output, keys tf.Output, default_value tf.Output) (values tf.Output)

Looks up keys in a table, outputs the corresponding values.

The tensor `keys` must of the same type as the keys of the table. The output `values` is of the type of the table values.

The scalar `default_value` is the value output for keys not present in the table. It must also be of the same type as the table values.

Arguments: 

table_handle: Handle to the table.
keys: Any shape.  Keys to look up.

Returns Same shape as `keys`. Values found in the table, or `default_values` for missing keys.

func LookupTableImportV2 

func LookupTableImportV2(scope *Scope, table_handle tf.Output, keys tf.Output, values tf.Output) (o *tf.Operation)

Replaces the contents of the table with the specified keys and values.

The tensor `keys` must be of the same type as the keys of the table. The tensor `values` must be of the type of the table values.

Arguments: 

table_handle: Handle to the table.
keys: Any shape.  Keys to look up.
values: Values to associate with keys.

Returns the created operation.

func LookupTableInsertV2 

func LookupTableInsertV2(scope *Scope, table_handle tf.Output, keys tf.Output, values tf.Output) (o *tf.Operation)

Updates the table to associates keys with values.

The tensor `keys` must be of the same type as the keys of the table. The tensor `values` must be of the type of the table values.

Arguments: 

table_handle: Handle to the table.
keys: Any shape.  Keys to look up.
values: Values to associate with keys.

Returns the created operation.

func LookupTableRemoveV2 

func LookupTableRemoveV2(scope *Scope, table_handle tf.Output, keys tf.Output) (o *tf.Operation)

Removes keys and its associated values from a table.

The tensor `keys` must of the same type as the keys of the table. Keys not already in the table are silently ignored.

Arguments: 

table_handle: Handle to the table.
keys: Any shape.  Keys of the elements to remove.

Returns the created operation.

func LookupTableSizeV2 

func LookupTableSizeV2(scope *Scope, table_handle tf.Output) (size tf.Output)

Computes the number of elements in the given table.

Arguments: 

table_handle: Handle to the table.

Returns Scalar that contains number of elements in the table.

func LoopCond 

func LoopCond(scope *Scope, input tf.Output) (output tf.Output)

Forwards the input to the output.

This operator represents the loop termination condition used by the "pivot" switches of a loop.

Arguments: 

input: A boolean scalar, representing the branch predicate of the Switch op.

Returns The same tensor as `input`.

func LowerBound 

func LowerBound(scope *Scope, sorted_inputs tf.Output, values tf.Output, optional ...LowerBoundAttr) (output tf.Output)

Applies lower_bound(sorted_search_values, values) along each row.

Each set of rows with the same index in (sorted_inputs, values) is treated independently. The resulting row is the equivalent of calling `np.searchsorted(sorted_inputs, values, side='left')`.

The result is not a global index to the entire `Tensor`, but rather just the index in the last dimension.

A 2-D example: 

sorted_sequence = [[0, 3, 9, 9, 10],
                   [1, 2, 3, 4, 5]]
values = [[2, 4, 9],
          [0, 2, 6]]

result = LowerBound(sorted_sequence, values)

result == [[1, 2, 2],
           [0, 1, 5]]

Arguments: 

sorted_inputs: 2-D Tensor where each row is ordered.
values: 2-D Tensor with the same numbers of rows as `sorted_search_values`. Contains

the values that will be searched for in `sorted_search_values`.

Returns A `Tensor` with the same shape as `values`. It contains the first scalar index into the last dimension where values can be inserted without changing the ordered property.

func Lu 

func Lu(scope *Scope, input tf.Output, optional ...LuAttr) (lu tf.Output, p tf.Output)

Computes the LU decomposition of one or more square matrices.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices.

The input has to be invertible.

The output consists of two tensors LU and P containing the LU decomposition of all input submatrices `[..., :, :]`. LU encodes the lower triangular and upper triangular factors.

For each input submatrix of shape `[M, M]`, L is a lower triangular matrix of shape `[M, M]` with unit diagonal whose entries correspond to the strictly lower triangular part of LU. U is a upper triangular matrix of shape `[M, M]` whose entries correspond to the upper triangular part, including the diagonal, of LU.

P represents a permutation matrix encoded as a list of indices each between `0` and `M-1`, inclusive. If P_mat denotes the permutation matrix corresponding to P, then the L, U and P satisfies P_mat * input = L * U.

Arguments: 

input: A tensor of shape `[..., M, M]` whose inner-most 2 dimensions form matrices of

size `[M, M]`.

Returns: 

lu: A tensor of shape `[..., M, M]` whose strictly lower triangular part denotes the

lower triangular factor `L` with unit diagonal, and whose upper triangular part denotes the upper triangular factor `U`. 

p: Permutation of the rows encoded as a list of indices in `0..M-1`. Shape is

`[..., M]`. @compatibility(scipy) Similar to `scipy.linalg.lu`, except the triangular factors `L` and `U` are packed into a single tensor, the permutation is applied to `input` instead of the right hand side and the permutation `P` is returned as a list of indices instead of a permutation matrix. @end_compatibility

func MakeIterator 

func MakeIterator(scope *Scope, dataset tf.Output, iterator tf.Output) (o *tf.Operation)

Makes a new iterator from the given `dataset` and stores it in `iterator`.

This operation may be executed multiple times. Each execution will reset the iterator in `iterator` to the first element of `dataset`.

Returns the created operation.

func MakeUnique 

func MakeUnique(scope *Scope, input tf.Output) (output tf.Output)

Make all elements in the non-Batch dimension unique, but \"close\" to

their initial value. Never returns a sub-normal number. Never returns zero. The sign of each input element is always identical to the sign of the corresponding output element. Behavior for infinite elements is undefined. Behavior for subnormal elements is undefined.

func MapClear 

func MapClear(scope *Scope, dtypes []tf.DataType, optional ...MapClearAttr) (o *tf.Operation)

Op removes all elements in the underlying container.

Returns the created operation.

func MapIncompleteSize 

func MapIncompleteSize(scope *Scope, dtypes []tf.DataType, optional ...MapIncompleteSizeAttr) (size tf.Output)

Op returns the number of incomplete elements in the underlying container.

func MapPeek 

func MapPeek(scope *Scope, key tf.Output, indices tf.Output, dtypes []tf.DataType, optional ...MapPeekAttr) (values []tf.Output)

Op peeks at the values at the specified key. If the

underlying container does not contain this key this op will block until it does.

func MapSize 

func MapSize(scope *Scope, dtypes []tf.DataType, optional ...MapSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func MapStage 

func MapStage(scope *Scope, key tf.Output, indices tf.Output, values []tf.Output, dtypes []tf.DataType, optional ...MapStageAttr) (o *tf.Operation)

Stage (key, values) in the underlying container which behaves like a hashtable.

Arguments: 

key: int64

values: a list of tensors

dtypes A list of data types that inserted values should adhere to.

Returns the created operation.

func MapUnstage 

func MapUnstage(scope *Scope, key tf.Output, indices tf.Output, dtypes []tf.DataType, optional ...MapUnstageAttr) (values []tf.Output)

Op removes and returns the values associated with the key

from the underlying container. If the underlying container does not contain this key, the op will block until it does.

func MapUnstageNoKey 

func MapUnstageNoKey(scope *Scope, indices tf.Output, dtypes []tf.DataType, optional ...MapUnstageNoKeyAttr) (key tf.Output, values []tf.Output)

Op removes and returns a random (key, value)

from the underlying container. If the underlying container does not contain elements, the op will block until it does.

func MatMul 

func MatMul(scope *Scope, a tf.Output, b tf.Output, optional ...MatMulAttr) (product tf.Output)

Multiply the matrix "a" by the matrix "b".

The inputs must be two-dimensional matrices and the inner dimension of "a" (after being transposed if transpose_a is true) must match the outer dimension of "b" (after being transposed if transposed_b is true).

*Note*: The default kernel implementation for MatMul on GPUs uses cublas.

func MatchingFiles 

func MatchingFiles(scope *Scope, pattern tf.Output) (filenames tf.Output)

Returns the set of files matching one or more glob patterns.

Note that this routine only supports wildcard characters in the basename portion of the pattern, not in the directory portion. Note also that the order of filenames returned is deterministic.

Arguments: 

pattern: Shell wildcard pattern(s). Scalar or vector of type string.

Returns A vector of matching filenames.

func MatrixBandPart 

func MatrixBandPart(scope *Scope, input tf.Output, num_lower tf.Output, num_upper tf.Output) (band tf.Output)

Copy a tensor setting everything outside a central band in each innermost matrix to zero.

The `band` part is computed as follows: Assume `input` has `k` dimensions `[I, J, K, ..., M, N]`, then the output is a tensor with the same shape where

`band[i, j, k, ..., m, n] = in_band(m, n) * input[i, j, k, ..., m, n]`.

The indicator function

`in_band(m, n) = (num_lower < 0 || (m-n) <= num_lower)) && 

(num_upper < 0 || (n-m) <= num_upper)`.

For example:

``` # if 'input' is [[ 0, 1, 2, 3] # [-1, 0, 1, 2] # [-2, -1, 0, 1] # [-3, -2, -1, 0]],

tf.linalg.band_part(input, 1, -1) ==> [[ 0, 1, 2, 3] 

[-1,  0,  1, 2]
[ 0, -1,  0, 1]
[ 0,  0, -1, 0]],

tf.linalg.band_part(input, 2, 1) ==> [[ 0, 1, 0, 0] 

[-1,  0,  1, 0]
[-2, -1,  0, 1]
[ 0, -2, -1, 0]]

```

Useful special cases:

``` 

tf.linalg.band_part(input, 0, -1) ==> Upper triangular part.
tf.linalg.band_part(input, -1, 0) ==> Lower triangular part.
tf.linalg.band_part(input, 0, 0) ==> Diagonal.

```

Arguments: 

input: Rank `k` tensor.
num_lower: 0-D tensor. Number of subdiagonals to keep. If negative, keep entire

lower triangle. 

num_upper: 0-D tensor. Number of superdiagonals to keep. If negative, keep

entire upper triangle.

Returns Rank `k` tensor of the same shape as input. The extracted banded tensor.

func MatrixDeterminant 

func MatrixDeterminant(scope *Scope, input tf.Output) (output tf.Output)

Computes the determinant of one or more square matrices.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor containing the determinants for all input submatrices `[..., :, :]`.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[...]`.

func MatrixDiag 

func MatrixDiag(scope *Scope, diagonal tf.Output) (output tf.Output)

Returns a batched diagonal tensor with a given batched diagonal values.

Given a `diagonal`, this operation returns a tensor with the `diagonal` and everything else padded with zeros. The diagonal is computed as follows:

Assume `diagonal` has `k` dimensions `[I, J, K, ..., N]`, then the output is a tensor of rank `k+1` with dimensions [I, J, K, ..., N, N]` where:

`output[i, j, k, ..., m, n] = 1{m=n} * diagonal[i, j, k, ..., n]`.

For example:

``` # 'diagonal' is [[1, 2, 3, 4], [5, 6, 7, 8]]

and diagonal.shape = (2, 4)

tf.matrix_diag(diagonal) ==> [[[1, 0, 0, 0] 

 [0, 2, 0, 0]
 [0, 0, 3, 0]
 [0, 0, 0, 4]],
[[5, 0, 0, 0]
 [0, 6, 0, 0]
 [0, 0, 7, 0]
 [0, 0, 0, 8]]]

which has shape (2, 4, 4) ```

Arguments: 

diagonal: Rank `k`, where `k >= 1`.

Returns Rank `k+1`, with `output.shape = diagonal.shape + [diagonal.shape[-1]]`.

func MatrixDiagPart 

func MatrixDiagPart(scope *Scope, input tf.Output) (diagonal tf.Output)

Returns the batched diagonal part of a batched tensor.

This operation returns a tensor with the `diagonal` part of the batched `input`. The `diagonal` part is computed as follows:

Assume `input` has `k` dimensions `[I, J, K, ..., M, N]`, then the output is a tensor of rank `k - 1` with dimensions `[I, J, K, ..., min(M, N)]` where:

`diagonal[i, j, k, ..., n] = input[i, j, k, ..., n, n]`.

The input must be at least a matrix.

For example:

``` # 'input' is [[[1, 0, 0, 0] 

 [0, 2, 0, 0]
 [0, 0, 3, 0]
 [0, 0, 0, 4]],
[[5, 0, 0, 0]
 [0, 6, 0, 0]
 [0, 0, 7, 0]
 [0, 0, 0, 8]]]

and input.shape = (2, 4, 4)

tf.matrix_diag_part(input) ==> [[1, 2, 3, 4], [5, 6, 7, 8]]

which has shape (2, 4) ```

Arguments: 

input: Rank `k` tensor where `k >= 2`.

Returns The extracted diagonal(s) having shape `diagonal.shape = input.shape[:-2] + [min(input.shape[-2:])]`.

func MatrixDiagPartV2 

func MatrixDiagPartV2(scope *Scope, input tf.Output, k tf.Output, padding_value tf.Output) (diagonal tf.Output)

Returns the batched diagonal part of a batched tensor.

Returns a tensor with the `k[0]`-th to `k[1]`-th diagonals of the batched `input`.

Assume `input` has `r` dimensions `[I, J, ..., L, M, N]`. Let `max_diag_len` be the maximum length among all diagonals to be extracted, `max_diag_len = min(M + min(k[1], 0), N + min(-k[0], 0))` Let `num_diags` be the number of diagonals to extract, `num_diags = k[1] - k[0] + 1`.

If `num_diags == 1`, the output tensor is of rank `r - 1` with shape `[I, J, ..., L, max_diag_len]` and values:

``` diagonal[i, j, ..., l, n] 

= input[i, j, ..., l, n+y, n+x] ; if 0 <= n+y < M and 0 <= n+x < N,
  padding_value                 ; otherwise.

``` where `y = max(-k[1], 0)`, `x = max(k[1], 0)`.

Otherwise, the output tensor has rank `r` with dimensions `[I, J, ..., L, num_diags, max_diag_len]` with values:

``` diagonal[i, j, ..., l, m, n] 

= input[i, j, ..., l, n+y, n+x] ; if 0 <= n+y < M and 0 <= n+x < N,
  padding_value                 ; otherwise.

``` where `d = k[1] - m`, `y = max(-d, 0)`, and `x = max(d, 0)`.

The input must be at least a matrix.

For example:

``` input = np.array([[[1, 2, 3, 4], # Input shape: (2, 3, 4) 

 [5, 6, 7, 8],
 [9, 8, 7, 6]],
[[5, 4, 3, 2],
 [1, 2, 3, 4],
 [5, 6, 7, 8]]])

# A main diagonal from each batch. tf.matrix_diag_part(input) ==> [[1, 6, 7], # Output shape: (2, 3) 

[5, 2, 7]]

# A superdiagonal from each batch. tf.matrix_diag_part(input, k = 1) 

==> [[2, 7, 6],  # Output shape: (2, 3)
     [4, 3, 8]]

# A tridiagonal band from each batch. tf.matrix_diag_part(input, k = (-1, 1)) 

==> [[[2, 7, 6],  # Output shape: (2, 3, 3)
      [1, 6, 7],
      [5, 8, 0]],
     [[4, 3, 8],
      [5, 2, 7],
      [1, 6, 0]]]

# Padding value = 9 tf.matrix_diag_part(input, k = (1, 3), padding_value = 9) 

==> [[[4, 9, 9],  # Output shape: (2, 3, 3)
      [3, 8, 9],
      [2, 7, 6]],
     [[2, 9, 9],
      [3, 4, 9],
      [4, 3, 8]]]

```

Arguments: 

input: Rank `r` tensor where `r >= 2`.
k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`. 

padding_value: The value to fill the area outside the specified diagonal band with.

Default is 0.

Returns The extracted diagonal(s).

func MatrixDiagPartV3 

func MatrixDiagPartV3(scope *Scope, input tf.Output, k tf.Output, padding_value tf.Output, optional ...MatrixDiagPartV3Attr) (diagonal tf.Output)

Returns the batched diagonal part of a batched tensor.

Returns a tensor with the `k[0]`-th to `k[1]`-th diagonals of the batched `input`.

Assume `input` has `r` dimensions `[I, J, ..., L, M, N]`. Let `max_diag_len` be the maximum length among all diagonals to be extracted, `max_diag_len = min(M + min(k[1], 0), N + min(-k[0], 0))` Let `num_diags` be the number of diagonals to extract, `num_diags = k[1] - k[0] + 1`.

If `num_diags == 1`, the output tensor is of rank `r - 1` with shape `[I, J, ..., L, max_diag_len]` and values:

``` diagonal[i, j, ..., l, n] 

= input[i, j, ..., l, n+y, n+x] ; if 0 <= n+y < M and 0 <= n+x < N,
  padding_value                 ; otherwise.

``` where `y = max(-k[1], 0)`, `x = max(k[1], 0)`.

Otherwise, the output tensor has rank `r` with dimensions `[I, J, ..., L, num_diags, max_diag_len]` with values:

``` diagonal[i, j, ..., l, m, n] 

= input[i, j, ..., l, n+y, n+x] ; if 0 <= n+y < M and 0 <= n+x < N,
  padding_value                 ; otherwise.

``` where `d = k[1] - m`, `y = max(-d, 0) - offset`, and `x = max(d, 0) - offset`.

`offset` is zero except when the alignment of the diagonal is to the right. ``` offset = max_diag_len - diag_len(d) ; if (`align` in {RIGHT_LEFT, RIGHT_RIGHT} 

                                  and `d >= 0`) or
                                (`align` in {LEFT_RIGHT, RIGHT_RIGHT}
                                  and `d <= 0`)
0                          ; otherwise

``` where `diag_len(d) = min(cols - max(d, 0), rows + min(d, 0))`.

The input must be at least a matrix.

For example:

``` input = np.array([[[1, 2, 3, 4], # Input shape: (2, 3, 4) 

 [5, 6, 7, 8],
 [9, 8, 7, 6]],
[[5, 4, 3, 2],
 [1, 2, 3, 4],
 [5, 6, 7, 8]]])

# A main diagonal from each batch. tf.matrix_diag_part(input) ==> [[1, 6, 7], # Output shape: (2, 3) 

[5, 2, 7]]

# A superdiagonal from each batch. tf.matrix_diag_part(input, k = 1) 

==> [[2, 7, 6],  # Output shape: (2, 3)
     [4, 3, 8]]

# A band from each batch. tf.matrix_diag_part(input, k = (-1, 2)) 

==> [[[0, 3, 8],  # Output shape: (2, 4, 3)
      [2, 7, 6],
      [1, 6, 7],
      [5, 8, 0]],
     [[0, 3, 4],
      [4, 3, 8],
      [5, 2, 7],
      [1, 6, 0]]]

# LEFT_RIGHT alignment. tf.matrix_diag_part(input, k = (-1, 2), align="LEFT_RIGHT") 

==> [[[3, 8, 0],  # Output shape: (2, 4, 3)
      [2, 7, 6],
      [1, 6, 7],
      [0, 5, 8]],
     [[3, 4, 0],
      [4, 3, 8],
      [5, 2, 7],
      [0, 1, 6]]]

# max_diag_len can be shorter than the main diagonal. tf.matrix_diag_part(input, k = (-2, -1)) 

==> [[[5, 8],
      [9, 0]],
     [[1, 6],
      [5, 0]]]

# padding_value = 9 tf.matrix_diag_part(input, k = (1, 3), padding_value = 9) 

==> [[[9, 9, 4],  # Output shape: (2, 3, 3)
      [9, 3, 8],
      [2, 7, 6]],
     [[9, 9, 2],
      [9, 3, 4],
      [4, 3, 8]]]

```

Arguments: 

input: Rank `r` tensor where `r >= 2`.
k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`. 

padding_value: The value to fill the area outside the specified diagonal band with.

Default is 0.

Returns The extracted diagonal(s).

func MatrixDiagV2 

func MatrixDiagV2(scope *Scope, diagonal tf.Output, k tf.Output, num_rows tf.Output, num_cols tf.Output, padding_value tf.Output) (output tf.Output)

Returns a batched diagonal tensor with given batched diagonal values.

Returns a tensor with the contents in `diagonal` as `k[0]`-th to `k[1]`-th diagonals of a matrix, with everything else padded with `padding`. `num_rows` and `num_cols` specify the dimension of the innermost matrix of the output. If both are not specified, the op assumes the innermost matrix is square and infers its size from `k` and the innermost dimension of `diagonal`. If only one of them is specified, the op assumes the unspecified value is the smallest possible based on other criteria.

Let `diagonal` have `r` dimensions `[I, J, ..., L, M, N]`. The output tensor has rank `r+1` with shape `[I, J, ..., L, M, num_rows, num_cols]` when only one diagonal is given (`k` is an integer or `k[0] == k[1]`). Otherwise, it has rank `r` with shape `[I, J, ..., L, num_rows, num_cols]`.

The second innermost dimension of `diagonal` has double meaning. When `k` is scalar or `k[0] == k[1]`, `M` is part of the batch size [I, J, ..., M], and the output tensor is:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, n-max(d_upper, 0)] ; if n - m == d_upper
  padding_value                             ; otherwise

```

Otherwise, `M` is treated as the number of diagonals for the matrix in the same batch (`M = k[1]-k[0]+1`), and the output tensor is:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, diag_index, index_in_diag] ; if k[0] <= d <= k[1]
  padding_value                                     ; otherwise

``` where `d = n - m`, `diag_index = k[1] - d`, and `index_in_diag = n - max(d, 0)`.

For example:

``` # The main diagonal. diagonal = np.array([[1, 2, 3, 4], # Input shape: (2, 4) 

[5, 6, 7, 8]])

tf.matrix_diag(diagonal) ==> [[[1, 0, 0, 0], # Output shape: (2, 4, 4) 

 [0, 2, 0, 0],
 [0, 0, 3, 0],
 [0, 0, 0, 4]],
[[5, 0, 0, 0],
 [0, 6, 0, 0],
 [0, 0, 7, 0],
 [0, 0, 0, 8]]]

# A superdiagonal (per batch). diagonal = np.array([[1, 2, 3], # Input shape: (2, 3) 

[4, 5, 6]])

tf.matrix_diag(diagonal, k = 1) 

==> [[[0, 1, 0, 0],  # Output shape: (2, 4, 4)
      [0, 0, 2, 0],
      [0, 0, 0, 3],
      [0, 0, 0, 0]],
     [[0, 4, 0, 0],
      [0, 0, 5, 0],
      [0, 0, 0, 6],
      [0, 0, 0, 0]]]

# A band of diagonals. diagonals = np.array([[[1, 2, 3], # Input shape: (2, 2, 3) 

 [4, 5, 0]],
[[6, 7, 9],
 [9, 1, 0]]])

tf.matrix_diag(diagonals, k = (-1, 0)) 

==> [[[1, 0, 0],  # Output shape: (2, 3, 3)
      [4, 2, 0],
      [0, 5, 3]],
     [[6, 0, 0],
      [9, 7, 0],
      [0, 1, 9]]]

# Rectangular matrix. diagonal = np.array([1, 2]) # Input shape: (2) tf.matrix_diag(diagonal, k = -1, num_rows = 3, num_cols = 4) 

==> [[0, 0, 0, 0],  # Output shape: (3, 4)
     [1, 0, 0, 0],
     [0, 2, 0, 0]]

# Rectangular matrix with inferred num_cols and padding_value = 9. tf.matrix_diag(diagonal, k = -1, num_rows = 3, padding_value = 9) 

==> [[9, 9],  # Output shape: (3, 2)
     [1, 9],
     [9, 2]]

```

Arguments: 

diagonal: Rank `r`, where `r >= 1`
k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`. 

num_rows: The number of rows of the output matrix. If it is not provided, the op assumes

the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of `diagonal`. 

num_cols: The number of columns of the output matrix. If it is not provided, the op

assumes the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of `diagonal`. 

padding_value: The number to fill the area outside the specified diagonal band with.

Default is 0.

Returns Has rank `r+1` when `k` is an integer or `k[0] == k[1]`, rank `r` otherwise.

func MatrixDiagV3 

func MatrixDiagV3(scope *Scope, diagonal tf.Output, k tf.Output, num_rows tf.Output, num_cols tf.Output, padding_value tf.Output, optional ...MatrixDiagV3Attr) (output tf.Output)

Returns a batched diagonal tensor with given batched diagonal values.

Returns a tensor with the contents in `diagonal` as `k[0]`-th to `k[1]`-th diagonals of a matrix, with everything else padded with `padding`. `num_rows` and `num_cols` specify the dimension of the innermost matrix of the output. If both are not specified, the op assumes the innermost matrix is square and infers its size from `k` and the innermost dimension of `diagonal`. If only one of them is specified, the op assumes the unspecified value is the smallest possible based on other criteria.

Let `diagonal` have `r` dimensions `[I, J, ..., L, M, N]`. The output tensor has rank `r+1` with shape `[I, J, ..., L, M, num_rows, num_cols]` when only one diagonal is given (`k` is an integer or `k[0] == k[1]`). Otherwise, it has rank `r` with shape `[I, J, ..., L, num_rows, num_cols]`.

The second innermost dimension of `diagonal` has double meaning. When `k` is scalar or `k[0] == k[1]`, `M` is part of the batch size [I, J, ..., M], and the output tensor is:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, n-max(d_upper, 0)] ; if n - m == d_upper
  padding_value                             ; otherwise

```

Otherwise, `M` is treated as the number of diagonals for the matrix in the same batch (`M = k[1]-k[0]+1`), and the output tensor is:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, diag_index, index_in_diag] ; if k[0] <= d <= k[1]
  padding_value                                     ; otherwise

``` where `d = n - m`, `diag_index = [k] - d`, and `index_in_diag = n - max(d, 0) + offset`.

`offset` is zero except when the alignment of the diagonal is to the right. ``` offset = max_diag_len - diag_len(d) ; if (`align` in {RIGHT_LEFT, RIGHT_RIGHT} 

                                  and `d >= 0`) or
                                (`align` in {LEFT_RIGHT, RIGHT_RIGHT}
                                  and `d <= 0`)
0                          ; otherwise

``` where `diag_len(d) = min(cols - max(d, 0), rows + min(d, 0))`.

For example:

``` # The main diagonal. diagonal = np.array([[1, 2, 3, 4], # Input shape: (2, 4) 

[5, 6, 7, 8]])

tf.matrix_diag(diagonal) ==> [[[1, 0, 0, 0], # Output shape: (2, 4, 4) 

 [0, 2, 0, 0],
 [0, 0, 3, 0],
 [0, 0, 0, 4]],
[[5, 0, 0, 0],
 [0, 6, 0, 0],
 [0, 0, 7, 0],
 [0, 0, 0, 8]]]

# A superdiagonal (per batch). diagonal = np.array([[1, 2, 3], # Input shape: (2, 3) 

[4, 5, 6]])

tf.matrix_diag(diagonal, k = 1) 

==> [[[0, 1, 0, 0],  # Output shape: (2, 4, 4)
      [0, 0, 2, 0],
      [0, 0, 0, 3],
      [0, 0, 0, 0]],
     [[0, 4, 0, 0],
      [0, 0, 5, 0],
      [0, 0, 0, 6],
      [0, 0, 0, 0]]]

# A tridiagonal band (per batch). diagonals = np.array([[[0, 8, 9], # Input shape: (2, 2, 3) 

 [1, 2, 3],
 [4, 5, 0]],
[[0, 2, 3],
 [6, 7, 9],
 [9, 1, 0]]])

tf.matrix_diag(diagonals, k = (-1, 1)) 

==> [[[1, 8, 0],  # Output shape: (2, 3, 3)
      [4, 2, 9],
      [0, 5, 3]],
     [[6, 2, 0],
      [9, 7, 3],
      [0, 1, 9]]]

# LEFT_RIGHT alignment. diagonals = np.array([[[8, 9, 0], # Input shape: (2, 2, 3) 

 [1, 2, 3],
 [0, 4, 5]],
[[2, 3, 0],
 [6, 7, 9],
 [0, 9, 1]]])

tf.matrix_diag(diagonals, k = (-1, 1), align="LEFT_RIGHT") 

==> [[[1, 8, 0],  # Output shape: (2, 3, 3)
      [4, 2, 9],
      [0, 5, 3]],
     [[6, 2, 0],
      [9, 7, 3],
      [0, 1, 9]]]

# Rectangular matrix. diagonal = np.array([1, 2]) # Input shape: (2) tf.matrix_diag(diagonal, k = -1, num_rows = 3, num_cols = 4) 

==> [[0, 0, 0, 0],  # Output shape: (3, 4)
     [1, 0, 0, 0],
     [0, 2, 0, 0]]

# Rectangular matrix with inferred num_cols and padding_value = 9. tf.matrix_diag(diagonal, k = -1, num_rows = 3, padding_value = 9) 

==> [[9, 9],  # Output shape: (3, 2)
     [1, 9],
     [9, 2]]

```

Arguments: 

diagonal: Rank `r`, where `r >= 1`
k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`. 

num_rows: The number of rows of the output matrix. If it is not provided, the op assumes

the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of `diagonal`. 

num_cols: The number of columns of the output matrix. If it is not provided, the op

assumes the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of `diagonal`. 

padding_value: The number to fill the area outside the specified diagonal band with.

Default is 0.

Returns Has rank `r+1` when `k` is an integer or `k[0] == k[1]`, rank `r` otherwise.

func MatrixExponential 

func MatrixExponential(scope *Scope, input tf.Output) (output tf.Output)

Deprecated, use python implementation tf.linalg.matrix_exponential.

DEPRECATED at GraphDef version 27: Use Python implementation tf.linalg.matrix_exponential instead.

func MatrixInverse 

func MatrixInverse(scope *Scope, input tf.Output, optional ...MatrixInverseAttr) (output tf.Output)

Computes the inverse of one or more square invertible matrices or their adjoints (conjugate transposes).

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor of the same shape as the input containing the inverse for all input submatrices `[..., :, :]`.

The op uses LU decomposition with partial pivoting to compute the inverses.

If a matrix is not invertible there is no guarantee what the op does. It may detect the condition and raise an exception or it may simply return a garbage result.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[..., M, M]`.

@compatibility(numpy) Equivalent to np.linalg.inv @end_compatibility

func MatrixLogarithm 

func MatrixLogarithm(scope *Scope, input tf.Output) (output tf.Output)

Computes the matrix logarithm of one or more square matrices:

\\(log(exp(A)) = A\\)

This op is only defined for complex matrices. If A is positive-definite and real, then casting to a complex matrix, taking the logarithm and casting back to a real matrix will give the correct result.

This function computes the matrix logarithm using the Schur-Parlett algorithm. Details of the algorithm can be found in Section 11.6.2 of: Nicholas J. Higham, Functions of Matrices: Theory and Computation, SIAM 2008. ISBN 978-0-898716-46-7.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor of the same shape as the input containing the exponential for all input submatrices `[..., :, :]`.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[..., M, M]`.

@compatibility(scipy) Equivalent to scipy.linalg.logm @end_compatibility

func MatrixSetDiag 

func MatrixSetDiag(scope *Scope, input tf.Output, diagonal tf.Output) (output tf.Output)

Returns a batched matrix tensor with new batched diagonal values.

Given `input` and `diagonal`, this operation returns a tensor with the same shape and values as `input`, except for the main diagonal of the innermost matrices. These will be overwritten by the values in `diagonal`.

The output is computed as follows:

Assume `input` has `k+1` dimensions `[I, J, K, ..., M, N]` and `diagonal` has `k` dimensions `[I, J, K, ..., min(M, N)]`. Then the output is a tensor of rank `k+1` with dimensions `[I, J, K, ..., M, N]` where:

  • `output[i, j, k, ..., m, n] = diagonal[i, j, k, ..., n]` for `m == n`.
  • `output[i, j, k, ..., m, n] = input[i, j, k, ..., m, n]` for `m != n`.

Arguments: 

input: Rank `k+1`, where `k >= 1`.
diagonal: Rank `k`, where `k >= 1`.

Returns Rank `k+1`, with `output.shape = input.shape`.

func MatrixSetDiagV2 

func MatrixSetDiagV2(scope *Scope, input tf.Output, diagonal tf.Output, k tf.Output) (output tf.Output)

Returns a batched matrix tensor with new batched diagonal values.

Given `input` and `diagonal`, this operation returns a tensor with the same shape and values as `input`, except for the specified diagonals of the innermost matrices. These will be overwritten by the values in `diagonal`.

`input` has `r+1` dimensions `[I, J, ..., L, M, N]`. When `k` is scalar or `k[0] == k[1]`, `diagonal` has `r` dimensions `[I, J, ..., L, max_diag_len]`. Otherwise, it has `r+1` dimensions `[I, J, ..., L, num_diags, max_diag_len]`. `num_diags` is the number of diagonals, `num_diags = k[1] - k[0] + 1`. `max_diag_len` is the longest diagonal in the range `[k[0], k[1]]`, `max_diag_len = min(M + min(k[1], 0), N + min(-k[0], 0))`

The output is a tensor of rank `k+1` with dimensions `[I, J, ..., L, M, N]`. If `k` is scalar or `k[0] == k[1]`:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, n-max(k[1], 0)] ; if n - m == k[1]
  input[i, j, ..., l, m, n]              ; otherwise

```

Otherwise,

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, diag_index, index_in_diag] ; if k[0] <= d <= k[1]
  input[i, j, ..., l, m, n]                         ; otherwise

``` where `d = n - m`, `diag_index = k[1] - d`, and `index_in_diag = n - max(d, 0)`.

For example:

``` # The main diagonal. input = np.array([[[7, 7, 7, 7], # Input shape: (2, 3, 4) 

 [7, 7, 7, 7],
 [7, 7, 7, 7]],
[[7, 7, 7, 7],
 [7, 7, 7, 7],
 [7, 7, 7, 7]]])

diagonal = np.array([[1, 2, 3], # Diagonal shape: (2, 3) 

[4, 5, 6]])

tf.matrix_set_diag(diagonal) ==> [[[1, 7, 7, 7], # Output shape: (2, 3, 4) 

 [7, 2, 7, 7],
 [7, 7, 3, 7]],
[[4, 7, 7, 7],
 [7, 5, 7, 7],
 [7, 7, 6, 7]]]

# A superdiagonal (per batch). tf.matrix_set_diag(diagonal, k = 1) 

==> [[[7, 1, 7, 7],  # Output shape: (2, 3, 4)
      [7, 7, 2, 7],
      [7, 7, 7, 3]],
     [[7, 4, 7, 7],
      [7, 7, 5, 7],
      [7, 7, 7, 6]]]

# A band of diagonals. diagonals = np.array([[[1, 2, 3], # Diagonal shape: (2, 2, 3) 

 [4, 5, 0]],
[[6, 1, 2],
 [3, 4, 0]]])

tf.matrix_set_diag(diagonals, k = (-1, 0)) 

==> [[[1, 7, 7, 7],  # Output shape: (2, 3, 4)
      [4, 2, 7, 7],
      [0, 5, 3, 7]],
     [[6, 7, 7, 7],
      [3, 1, 7, 7],
      [7, 4, 2, 7]]]

```

Arguments: 

input: Rank `r+1`, where `r >= 1`.
diagonal: Rank `r` when `k` is an integer or `k[0] == k[1]`. Otherwise, it has rank `r+1`.

`k >= 1`. 

k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`.

Returns Rank `r+1`, with `output.shape = input.shape`.

func MatrixSetDiagV3 

func MatrixSetDiagV3(scope *Scope, input tf.Output, diagonal tf.Output, k tf.Output, optional ...MatrixSetDiagV3Attr) (output tf.Output)

Returns a batched matrix tensor with new batched diagonal values.

Given `input` and `diagonal`, this operation returns a tensor with the same shape and values as `input`, except for the specified diagonals of the innermost matrices. These will be overwritten by the values in `diagonal`.

`input` has `r+1` dimensions `[I, J, ..., L, M, N]`. When `k` is scalar or `k[0] == k[1]`, `diagonal` has `r` dimensions `[I, J, ..., L, max_diag_len]`. Otherwise, it has `r+1` dimensions `[I, J, ..., L, num_diags, max_diag_len]`. `num_diags` is the number of diagonals, `num_diags = k[1] - k[0] + 1`. `max_diag_len` is the longest diagonal in the range `[k[0], k[1]]`, `max_diag_len = min(M + min(k[1], 0), N + min(-k[0], 0))`

The output is a tensor of rank `k+1` with dimensions `[I, J, ..., L, M, N]`. If `k` is scalar or `k[0] == k[1]`:

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, n-max(k[1], 0)] ; if n - m == k[1]
  input[i, j, ..., l, m, n]              ; otherwise

```

Otherwise,

``` output[i, j, ..., l, m, n] 

= diagonal[i, j, ..., l, diag_index, index_in_diag] ; if k[0] <= d <= k[1]
  input[i, j, ..., l, m, n]                         ; otherwise

``` where `d = n - m`, `diag_index = k[1] - d`, and `index_in_diag = n - max(d, 0) + offset`.

`offset` is zero except when the alignment of the diagonal is to the right. ``` offset = max_diag_len - diag_len(d) ; if (`align` in {RIGHT_LEFT, RIGHT_RIGHT} 

                                  and `d >= 0`) or
                                (`align` in {LEFT_RIGHT, RIGHT_RIGHT}
                                  and `d <= 0`)
0                          ; otherwise

``` where `diag_len(d) = min(cols - max(d, 0), rows + min(d, 0))`.

For example:

``` # The main diagonal. input = np.array([[[7, 7, 7, 7], # Input shape: (2, 3, 4) 

 [7, 7, 7, 7],
 [7, 7, 7, 7]],
[[7, 7, 7, 7],
 [7, 7, 7, 7],
 [7, 7, 7, 7]]])

diagonal = np.array([[1, 2, 3], # Diagonal shape: (2, 3) 

[4, 5, 6]])

tf.matrix_set_diag(input, diagonal) 

==> [[[1, 7, 7, 7],  # Output shape: (2, 3, 4)
      [7, 2, 7, 7],
      [7, 7, 3, 7]],
     [[4, 7, 7, 7],
      [7, 5, 7, 7],
      [7, 7, 6, 7]]]

# A superdiagonal (per batch). tf.matrix_set_diag(input, diagonal, k = 1) 

==> [[[7, 1, 7, 7],  # Output shape: (2, 3, 4)
      [7, 7, 2, 7],
      [7, 7, 7, 3]],
     [[7, 4, 7, 7],
      [7, 7, 5, 7],
      [7, 7, 7, 6]]]

# A band of diagonals. diagonals = np.array([[[0, 9, 1], # Diagonal shape: (2, 4, 3) 

 [6, 5, 8],
 [1, 2, 3],
 [4, 5, 0]],
[[0, 1, 2],
 [5, 6, 4],
 [6, 1, 2],
 [3, 4, 0]]])

tf.matrix_set_diag(input, diagonals, k = (-1, 2)) 

==> [[[1, 6, 9, 7],  # Output shape: (2, 3, 4)
      [4, 2, 5, 1],
      [7, 5, 3, 8]],
     [[6, 5, 1, 7],
      [3, 1, 6, 2],
      [7, 4, 2, 4]]]

# LEFT_RIGHT alignment. diagonals = np.array([[[9, 1, 0], # Diagonal shape: (2, 4, 3) 

 [6, 5, 8],
 [1, 2, 3],
 [0, 4, 5]],
[[1, 2, 0],
 [5, 6, 4],
 [6, 1, 2],
 [0, 3, 4]]])

tf.matrix_set_diag(input, diagonals, k = (-1, 2), align="LEFT_RIGHT") 

==> [[[1, 6, 9, 7],  # Output shape: (2, 3, 4)
      [4, 2, 5, 1],
      [7, 5, 3, 8]],
     [[6, 5, 1, 7],
      [3, 1, 6, 2],
      [7, 4, 2, 4]]]

```

Arguments: 

input: Rank `r+1`, where `r >= 1`.
diagonal: Rank `r` when `k` is an integer or `k[0] == k[1]`. Otherwise, it has rank `r+1`.

`k >= 1`. 

k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main

diagonal, and negative value means subdiagonals. `k` can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. `k[0]` must not be larger than `k[1]`.

Returns Rank `r+1`, with `output.shape = input.shape`.

func MatrixSolve 

func MatrixSolve(scope *Scope, matrix tf.Output, rhs tf.Output, optional ...MatrixSolveAttr) (output tf.Output)

Solves systems of linear equations.

`Matrix` is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. `Rhs` is a tensor of shape `[..., M, K]`. The `output` is a tensor shape `[..., M, K]`. If `adjoint` is `False` then each output matrix satisfies `matrix[..., :, :] * output[..., :, :] = rhs[..., :, :]`. If `adjoint` is `True` then each output matrix satisfies `adjoint(matrix[..., :, :]) * output[..., :, :] = rhs[..., :, :]`.

Arguments: 

matrix: Shape is `[..., M, M]`.
rhs: Shape is `[..., M, K]`.

Returns Shape is `[..., M, K]`.

func MatrixSolveLs 

func MatrixSolveLs(scope *Scope, matrix tf.Output, rhs tf.Output, l2_regularizer tf.Output, optional ...MatrixSolveLsAttr) (output tf.Output)

Solves one or more linear least-squares problems.

`matrix` is a tensor of shape `[..., M, N]` whose inner-most 2 dimensions form real or complex matrices of size `[M, N]`. `Rhs` is a tensor of the same type as `matrix` and shape `[..., M, K]`. The output is a tensor shape `[..., N, K]` where each output matrix solves each of the equations `matrix[..., :, :]` * `output[..., :, :]` = `rhs[..., :, :]` in the least squares sense.

We use the following notation for (complex) matrix and right-hand sides in the batch:

`matrix`=\\(A \in \mathbb{C}^{m \times n}\\), `rhs`=\\(B \in \mathbb{C}^{m \times k}\\), `output`=\\(X \in \mathbb{C}^{n \times k}\\), `l2_regularizer`=\\(\lambda \in \mathbb{R}\\).

If `fast` is `True`, then the solution is computed by solving the normal equations using Cholesky decomposition. Specifically, if \\(m \ge n\\) then \\(X = (A^H A + \lambda I)^{-1} A^H B\\), which solves the least-squares problem \\(X = \mathrm{argmin}_{Z \in \Re^{n \times k} } ||A Z - B||_F^2 + \lambda ||Z||_F^2\\). If \\(m \lt n\\) then `output` is computed as \\(X = A^H (A A^H + \lambda I)^{-1} B\\), which (for \\(\lambda = 0\\)) is the minimum-norm solution to the under-determined linear system, i.e. \\(X = \mathrm{argmin}_{Z \in \mathbb{C}^{n \times k} } ||Z||_F^2 \\), subject to \\(A Z = B\\). Notice that the fast path is only numerically stable when \\(A\\) is numerically full rank and has a condition number \\(\mathrm{cond}(A) \lt \frac{1}{\sqrt{\epsilon_{mach} } }\\) or \\(\lambda\\) is sufficiently large.

If `fast` is `False` an algorithm based on the numerically robust complete orthogonal decomposition is used. This computes the minimum-norm least-squares solution, even when \\(A\\) is rank deficient. This path is typically 6-7 times slower than the fast path. If `fast` is `False` then `l2_regularizer` is ignored.

Arguments: 

matrix: Shape is `[..., M, N]`.
rhs: Shape is `[..., M, K]`.
l2_regularizer: Scalar tensor.

@compatibility(numpy) Equivalent to np.linalg.lstsq @end_compatibility

Returns Shape is `[..., N, K]`.

func MatrixSquareRoot 

func MatrixSquareRoot(scope *Scope, input tf.Output) (output tf.Output)

Computes the matrix square root of one or more square matrices:

matmul(sqrtm(A), sqrtm(A)) = A

The input matrix should be invertible. If the input matrix is real, it should have no eigenvalues which are real and negative (pairs of complex conjugate eigenvalues are allowed).

The matrix square root is computed by first reducing the matrix to quasi-triangular form with the real Schur decomposition. The square root of the quasi-triangular matrix is then computed directly. Details of the algorithm can be found in: Nicholas J. Higham, "Computing real square roots of a real matrix", Linear Algebra Appl., 1987.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor of the same shape as the input containing the matrix square root for all input submatrices `[..., :, :]`.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[..., M, M]`.

@compatibility(scipy) Equivalent to scipy.linalg.sqrtm @end_compatibility

func MatrixTriangularSolve 

func MatrixTriangularSolve(scope *Scope, matrix tf.Output, rhs tf.Output, optional ...MatrixTriangularSolveAttr) (output tf.Output)

Solves systems of linear equations with upper or lower triangular matrices by backsubstitution.

`matrix` is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. If `lower` is `True` then the strictly upper triangular part of each inner-most matrix is assumed to be zero and not accessed. If `lower` is False then the strictly lower triangular part of each inner-most matrix is assumed to be zero and not accessed. `rhs` is a tensor of shape `[..., M, N]`.

The output is a tensor of shape `[..., M, N]`. If `adjoint` is `True` then the innermost matrices in `output` satisfy matrix equations `matrix[..., :, :] * output[..., :, :] = rhs[..., :, :]`. If `adjoint` is `False` then the strictly then the innermost matrices in `output` satisfy matrix equations `adjoint(matrix[..., i, k]) * output[..., k, j] = rhs[..., i, j]`.

Note, the batch shapes for the inputs only need to broadcast.

Example: ```python

a = tf.constant([[3, 0, 0, 0], 

[2,  1,  0,  0],
[1,  0,  1,  0],
[1,  1,  1,  1]], dtype=tf.float32)

b = tf.constant([[4], 

[2],
[4],
[2]], dtype=tf.float32)

x = tf.linalg.triangular_solve(a, b, lower=True) x # <tf.Tensor: shape=(4, 1), dtype=float32, numpy= # array([[ 1.3333334 ], # [-0.66666675], # [ 2.6666665 ], # [-1.3333331 ]], dtype=float32)>

# in python3 one can use `a@x` tf.matmul(a, x) # <tf.Tensor: shape=(4, 1), dtype=float32, numpy= # array([[4. ], # [2. ], # [4. ], # [1.9999999]], dtype=float32)> ```

Arguments: 

matrix: Shape is `[..., M, M]`.
rhs: Shape is `[..., M, K]`.

Returns Shape is `[..., M, K]`.

func Max 

func Max(scope *Scope, input tf.Output, axis tf.Output, optional ...MaxAttr) (output tf.Output)

Computes the maximum of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func MaxIntraOpParallelismDataset 

func MaxIntraOpParallelismDataset(scope *Scope, input_dataset tf.Output, max_intra_op_parallelism tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that overrides the maximum intra-op parallelism.

Arguments: 

max_intra_op_parallelism: Identifies the maximum intra-op parallelism to use.

func MaxPool 

func MaxPool(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolAttr) (output tf.Output)

Performs max pooling on the input.

Arguments: 

input: 4-D input to pool over.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns The max pooled output tensor.

func MaxPool3D 

func MaxPool3D(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPool3DAttr) (output tf.Output)

Performs 3D max pooling on the input.

Arguments: 

input: Shape `[batch, depth, rows, cols, channels]` tensor to pool over.
ksize: 1-D tensor of length 5. The size of the window for each dimension of

the input tensor. Must have `ksize[0] = ksize[4] = 1`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

Returns The max pooled output tensor.

func MaxPool3DGrad 

func MaxPool3DGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPool3DGradAttr) (output tf.Output)

Computes gradients of 3D max pooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: Output backprop of shape `[batch, depth, rows, cols, channels]`.
ksize: 1-D tensor of length 5. The size of the window for each dimension of

the input tensor. Must have `ksize[0] = ksize[4] = 1`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

func MaxPool3DGradGrad 

func MaxPool3DGradGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPool3DGradGradAttr) (output tf.Output)

Computes second-order gradients of the maxpooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: Output backprop of shape `[batch, depth, rows, cols, channels]`.
ksize: 1-D tensor of length 5. The size of the window for each dimension of

the input tensor. Must have `ksize[0] = ksize[4] = 1`. 

strides: 1-D tensor of length 5. The stride of the sliding window for each

dimension of `input`. Must have `strides[0] = strides[4] = 1`. 

padding: The type of padding algorithm to use.

Returns Gradients of gradients w.r.t. the input to `max_pool`.

func MaxPoolGrad 

func MaxPoolGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolGradAttr) (output tf.Output)

Computes gradients of the maxpooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: 4-D.  Gradients w.r.t. the output of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients w.r.t. the input to `max_pool`.

func MaxPoolGradGrad 

func MaxPoolGradGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolGradGradAttr) (output tf.Output)

Computes second-order gradients of the maxpooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: 4-D.  Gradients of gradients w.r.t. the input of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients of gradients w.r.t. the input to `max_pool`.

func MaxPoolGradGradV2 

func MaxPoolGradGradV2(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize tf.Output, strides tf.Output, padding string, optional ...MaxPoolGradGradV2Attr) (output tf.Output)

Computes second-order gradients of the maxpooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: 4-D.  Gradients of gradients w.r.t. the input of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients of gradients w.r.t. the input to `max_pool`.

func MaxPoolGradGradWithArgmax 

func MaxPoolGradGradWithArgmax(scope *Scope, input tf.Output, grad tf.Output, argmax tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolGradGradWithArgmaxAttr) (output tf.Output)

Computes second-order gradients of the maxpooling function.

Arguments: 

input: The original input.
grad: 4-D with shape `[batch, height, width, channels]`.  Gradients w.r.t. the

input of `max_pool`. 

argmax: The indices of the maximum values chosen for each output of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients of gradients w.r.t. the input of `max_pool`.

func MaxPoolGradV2 

func MaxPoolGradV2(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize tf.Output, strides tf.Output, padding string, optional ...MaxPoolGradV2Attr) (output tf.Output)

Computes gradients of the maxpooling function.

Arguments: 

orig_input: The original input tensor.
orig_output: The original output tensor.
grad: 4-D.  Gradients w.r.t. the output of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients w.r.t. the input to `max_pool`.

func MaxPoolGradWithArgmax 

func MaxPoolGradWithArgmax(scope *Scope, input tf.Output, grad tf.Output, argmax tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolGradWithArgmaxAttr) (output tf.Output)

Computes gradients of the maxpooling function.

Arguments: 

input: The original input.
grad: 4-D with shape `[batch, height, width, channels]`.  Gradients w.r.t. the

output of `max_pool`. 

argmax: The indices of the maximum values chosen for each output of `max_pool`.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns Gradients w.r.t. the input of `max_pool`.

func MaxPoolV2 

func MaxPoolV2(scope *Scope, input tf.Output, ksize tf.Output, strides tf.Output, padding string, optional ...MaxPoolV2Attr) (output tf.Output)

Performs max pooling on the input.

Arguments: 

input: 4-D input to pool over.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns The max pooled output tensor.

func MaxPoolWithArgmax 

func MaxPoolWithArgmax(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolWithArgmaxAttr) (output tf.Output, argmax tf.Output)

Performs max pooling on the input and outputs both max values and indices.

The indices in `argmax` are flattened, so that a maximum value at position `[b, y, x, c]` becomes flattened index: `(y * width + x) * channels + c` if `include_batch_in_index` is False; `((b * height + y) * width + x) * channels + c` if `include_batch_in_index` is True.

The indices returned are always in `[0, height) x [0, width)` before flattening, even if padding is involved and the mathematically correct answer is outside (either negative or too large). This is a bug, but fixing it is difficult to do in a safe backwards compatible way, especially due to flattening.

Arguments: 

input: 4-D with shape `[batch, height, width, channels]`.  Input to pool over.
ksize: The size of the window for each dimension of the input tensor.
strides: The stride of the sliding window for each dimension of the

input tensor. 

padding: The type of padding algorithm to use.

Returns: 

output: The max pooled output tensor.
argmax: 4-D.  The flattened indices of the max values chosen for each output.

func Maximum 

func Maximum(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the max of x and y (i.e. x > y ? x : y) element-wise.

*NOTE*: `Maximum` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func Mean 

func Mean(scope *Scope, input tf.Output, axis tf.Output, optional ...MeanAttr) (output tf.Output)

Computes the mean of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func Merge 

func Merge(scope *Scope, inputs []tf.Output) (output tf.Output, value_index tf.Output)

Forwards the value of an available tensor from `inputs` to `output`.

`Merge` waits for at least one of the tensors in `inputs` to become available. It is usually combined with `Switch` to implement branching.

`Merge` forwards the first tensor to become available to `output`, and sets `value_index` to its index in `inputs`.

Arguments: 

inputs: The input tensors, exactly one of which will become available.

Returns: 

output: Will be set to the available input tensor.
value_index: The index of the chosen input tensor in `inputs`.

func MergeDedupData added in v0.4.0 

func MergeDedupData(scope *Scope, integer_tensor tf.Output, float_tensor tf.Output, tuple_mask string, optional ...MergeDedupDataAttr) (output tf.Output)

An op merges elements of integer and float tensors into deduplication data as XLA tuple.

This op merges outputs of SplitDedupDataOp, which gives two 1-D tensors, integer and floating point. With respect to tuple_mask, this op merges values of these two tensors into an XLA tuple, which should be as same as input to SplitDedupDataOp.

Arguments: 

integer_tensor: A 1-D integer tensor, includes integer elements of deduplication data tuple.
float_tensor: A 1-D float tensor, includes float elements of deduplication data tuple.
tuple_mask: A serialized TensorProto string of output tuple mask. This mask is a 2-D tensor,

with first column as tuple element type, and second column as span of this type. For example, an output tuple of (1, 2, 0.1, 3), its mask is [[0, 2], [1, 1], [0, 1]]. We expect only two types of elements: integer(0) and float(1).

Returns An XLA tuple merging integer and float elements as deduplication data tuple.

func MergeSummary 

func MergeSummary(scope *Scope, inputs []tf.Output) (summary tf.Output)

Merges summaries.

This op creates a [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) protocol buffer that contains the union of all the values in the input summaries.

When the Op is run, it reports an `InvalidArgument` error if multiple values in the summaries to merge use the same tag.

Arguments: 

inputs: Can be of any shape.  Each must contain serialized `Summary` protocol

buffers.

Returns Scalar. Serialized `Summary` protocol buffer.

func MergeV2Checkpoints 

func MergeV2Checkpoints(scope *Scope, checkpoint_prefixes tf.Output, destination_prefix tf.Output, optional ...MergeV2CheckpointsAttr) (o *tf.Operation)

V2 format specific: merges the metadata files of sharded checkpoints. The

result is one logical checkpoint, with one physical metadata file and renamed data files.

Intended for "grouping" multiple checkpoints in a sharded checkpoint setup.

If delete_old_dirs is true, attempts to delete recursively the dirname of each path in the input checkpoint_prefixes. This is useful when those paths are non user-facing temporary locations.

If allow_missing_files is true, merges the checkpoint prefixes as long as at least one file exists. Otherwise, if no files exist, an error will be thrown. The default value for allow_missing_files is false.

Arguments: 

checkpoint_prefixes: prefixes of V2 checkpoints to merge.
destination_prefix: scalar.  The desired final prefix.  Allowed to be the same

as one of the checkpoint_prefixes.

Returns the created operation.

func Mfcc 

func Mfcc(scope *Scope, spectrogram tf.Output, sample_rate tf.Output, optional ...MfccAttr) (output tf.Output)

Transforms a spectrogram into a form that's useful for speech recognition.

Mel Frequency Cepstral Coefficients are a way of representing audio data that's been effective as an input feature for machine learning. They are created by taking the spectrum of a spectrogram (a 'cepstrum'), and discarding some of the higher frequencies that are less significant to the human ear. They have a long history in the speech recognition world, and https://en.wikipedia.org/wiki/Mel-frequency_cepstrum is a good resource to learn more.

Arguments: 

spectrogram: Typically produced by the Spectrogram op, with magnitude_squared

set to true. 

sample_rate: How many samples per second the source audio used.

func Min 

func Min(scope *Scope, input tf.Output, axis tf.Output, optional ...MinAttr) (output tf.Output)

Computes the minimum of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func Minimum 

func Minimum(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns the min of x and y (i.e. x < y ? x : y) element-wise.

*NOTE*: `Minimum` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func MirrorPad 

func MirrorPad(scope *Scope, input tf.Output, paddings tf.Output, mode string) (output tf.Output)

Pads a tensor with mirrored values.

This operation pads a `input` with mirrored values according to the `paddings` you specify. `paddings` is an integer tensor with shape `[n, 2]`, where n is the rank of `input`. For each dimension D of `input`, `paddings[D, 0]` indicates how many values to add before the contents of `input` in that dimension, and `paddings[D, 1]` indicates how many values to add after the contents of `input` in that dimension. Both `paddings[D, 0]` and `paddings[D, 1]` must be no greater than `input.dim_size(D)` (or `input.dim_size(D) - 1`) if `copy_border` is true (if false, respectively).

The padded size of each dimension D of the output is:

`paddings(D, 0) + input.dim_size(D) + paddings(D, 1)`

For example:

``` # 't' is [[1, 2, 3], [4, 5, 6]]. # 'paddings' is [[1, 1]], [2, 2]]. # 'mode' is SYMMETRIC. # rank of 't' is 2. pad(t, paddings) ==> [[2, 1, 1, 2, 3, 3, 2] 

[2, 1, 1, 2, 3, 3, 2]
[5, 4, 4, 5, 6, 6, 5]
[5, 4, 4, 5, 6, 6, 5]]

```

Arguments: 

input: The input tensor to be padded.
paddings: A two-column matrix specifying the padding sizes. The number of

rows must be the same as the rank of `input`. 

mode: Either `REFLECT` or `SYMMETRIC`. In reflect mode the padded regions

do not include the borders, while in symmetric mode the padded regions do include the borders. For example, if `input` is `[1, 2, 3]` and `paddings` is `[0, 2]`, then the output is `[1, 2, 3, 2, 1]` in reflect mode, and it is `[1, 2, 3, 3, 2]` in symmetric mode.

Returns The padded tensor.

func MirrorPadGrad 

func MirrorPadGrad(scope *Scope, input tf.Output, paddings tf.Output, mode string) (output tf.Output)

Gradient op for `MirrorPad` op. This op folds a mirror-padded tensor.

This operation folds the padded areas of `input` by `MirrorPad` according to the `paddings` you specify. `paddings` must be the same as `paddings` argument given to the corresponding `MirrorPad` op.

The folded size of each dimension D of the output is:

`input.dim_size(D) - paddings(D, 0) - paddings(D, 1)`

For example:

``` # 't' is [[1, 2, 3], [4, 5, 6], [7, 8, 9]]. # 'paddings' is [[0, 1]], [0, 1]]. # 'mode' is SYMMETRIC. # rank of 't' is 2. pad(t, paddings) ==> [[ 1, 5] 

[11, 28]]

```

Arguments: 

input: The input tensor to be folded.
paddings: A two-column matrix specifying the padding sizes. The number of

rows must be the same as the rank of `input`. 

mode: The mode used in the `MirrorPad` op.

Returns The folded tensor.

func MlirPassthroughOp 

func MlirPassthroughOp(scope *Scope, inputs []tf.Output, mlir_module string, Toutputs []tf.DataType) (outputs []tf.Output)

Wraps an arbitrary MLIR computation expressed as a module with a main() function.

This operation does not have an associated kernel and is not intended to be executed in a regular TensorFlow session. Instead it is intended to be used for testing or for special case where a user intends to pass custom MLIR computation through a TensorFlow graph with the intent of having custom tooling processing it downstream (when targeting a different environment, like TensorFlow lite for example). The MLIR module is expected to have a main() function that will be used as an entry point. The inputs to the operations will be passed as argument to the main() function and the returned values of the main function mapped to the outputs. Example usage:

``` import tensorflow as tf from tensorflow.compiler.mlir.tensorflow.gen_mlir_passthrough_op import mlir_passthrough_op

mlir_module = ”'python 

func @main(%arg0 : tensor<10xf32>, %arg1 : tensor<10xf32>) -> tensor<10x10xf32> {
   %add = "magic.op"(%arg0, %arg1) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10x10xf32>
   return %ret : tensor<10x10xf32>
}

”'

@tf.function def foo(x, y): 

return mlir_passthrough_op([x, y], mlir_module, Toutputs=[tf.float32])

graph_def = foo.get_concrete_function(tf.TensorSpec([10], tf.float32), tf.TensorSpec([10], tf.float32)).graph.as_graph_def() ```

func Mod 

func Mod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns element-wise remainder of division. This emulates C semantics in that

the result here is consistent with a truncating divide. E.g. `tf.truncatediv(x, y) * y + truncate_mod(x, y) = x`.

*NOTE*: `Mod` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func ModelDataset 

func ModelDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ModelDatasetAttr) (handle tf.Output)

Identity transformation that models performance.

Identity transformation that models performance.

Arguments: 

input_dataset: A variant tensor representing the input dataset.

func Mul 

func Mul(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x * y element-wise.

*NOTE*: `Multiply` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func MulNoNan 

func MulNoNan(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x * y element-wise. Returns zero if y is zero, even if x if infinite or NaN.

*NOTE*: `MulNoNan` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func MultiDeviceIterator 

func MultiDeviceIterator(scope *Scope, devices []string, shared_name string, container string, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a MultiDeviceIterator resource.

Arguments: 

devices: A list of devices the iterator works across.
shared_name: If non-empty, this resource will be shared under the given name

across multiple sessions. 

container: If non-empty, this resource is placed in the given container.

Otherwise, a default container is used. 

output_types: The type list for the return values.
output_shapes: The list of shapes being produced.

Returns Handle to the resource created.

func MultiDeviceIteratorFromStringHandle 

func MultiDeviceIteratorFromStringHandle(scope *Scope, string_handle tf.Output, optional ...MultiDeviceIteratorFromStringHandleAttr) (multi_device_iterator tf.Output)

Generates a MultiDeviceIterator resource from its provided string handle.

Arguments: 

string_handle: String representing the resource.

Returns A MultiDeviceIterator resource.

func MultiDeviceIteratorGetNextFromShard 

func MultiDeviceIteratorGetNextFromShard(scope *Scope, multi_device_iterator tf.Output, shard_num tf.Output, incarnation_id tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Gets next element for the provided shard number.

Arguments: 

multi_device_iterator: A MultiDeviceIterator resource.
shard_num: Integer representing which shard to fetch data for.
incarnation_id: Which incarnation of the MultiDeviceIterator is running.
output_types: The type list for the return values.
output_shapes: The list of shapes being produced.

Returns Result of the get_next on the dataset.

func MultiDeviceIteratorInit 

func MultiDeviceIteratorInit(scope *Scope, dataset tf.Output, multi_device_iterator tf.Output, max_buffer_size tf.Output) (incarnation_id tf.Output)

Initializes the multi device iterator with the given dataset.

Arguments: 

dataset: Dataset to be iterated upon.
multi_device_iterator: A MultiDeviceIteratorResource.
max_buffer_size: The maximum size of the host side per device buffer to keep.

Returns An int64 indicating which incarnation of the MultiDeviceIterator is running.

func MultiDeviceIteratorToStringHandle 

func MultiDeviceIteratorToStringHandle(scope *Scope, multi_device_iterator tf.Output) (string_handle tf.Output)

Produces a string handle for the given MultiDeviceIterator.

Arguments: 

multi_device_iterator: A MultiDeviceIterator resource.

Returns A string representing the resource.

func Multinomial 

func Multinomial(scope *Scope, logits tf.Output, num_samples tf.Output, optional ...MultinomialAttr) (output tf.Output)

Draws samples from a multinomial distribution.

Arguments: 

logits: 2-D Tensor with shape `[batch_size, num_classes]`.  Each slice `[i, :]`

represents the unnormalized log probabilities for all classes. 

num_samples: 0-D.  Number of independent samples to draw for each row slice.

Returns 2-D Tensor with shape `[batch_size, num_samples]`. Each slice `[i, :]` contains the drawn class labels with range `[0, num_classes)`.

func MutableDenseHashTableV2 

func MutableDenseHashTableV2(scope *Scope, empty_key tf.Output, deleted_key tf.Output, value_dtype tf.DataType, optional ...MutableDenseHashTableV2Attr) (table_handle tf.Output)

Creates an empty hash table that uses tensors as the backing store.

It uses "open addressing" with quadratic reprobing to resolve collisions.

This op creates a mutable hash table, specifying the type of its keys and values. Each value must be a scalar. Data can be inserted into the table using the insert operations. It does not support the initialization operation.

Arguments: 

empty_key: The key used to represent empty key buckets internally. Must not

be used in insert or lookup operations. 

value_dtype: Type of the table values.

Returns Handle to a table.

func MutableHashTableOfTensorsV2 

func MutableHashTableOfTensorsV2(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType, optional ...MutableHashTableOfTensorsV2Attr) (table_handle tf.Output)

Creates an empty hash table.

This op creates a mutable hash table, specifying the type of its keys and values. Each value must be a vector. Data can be inserted into the table using the insert operations. It does not support the initialization operation.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns Handle to a table.

func MutableHashTableV2 

func MutableHashTableV2(scope *Scope, key_dtype tf.DataType, value_dtype tf.DataType, optional ...MutableHashTableV2Attr) (table_handle tf.Output)

Creates an empty hash table.

This op creates a mutable hash table, specifying the type of its keys and values. Each value must be a scalar. Data can be inserted into the table using the insert operations. It does not support the initialization operation.

Arguments: 

key_dtype: Type of the table keys.
value_dtype: Type of the table values.

Returns Handle to a table.

func MutexLock 

func MutexLock(scope *Scope, mutex tf.Output) (mutex_lock tf.Output)

Locks a mutex resource. The output is the lock. So long as the lock tensor

is alive, any other request to use `MutexLock` with this mutex will wait.

This is particularly useful for creating a critical section when used in conjunction with `MutexLockIdentity`:

```python

mutex = mutex_v2( 

shared_name=handle_name, container=container, name=name)

def execute_in_critical_section(fn, *args, **kwargs): 

lock = gen_resource_variable_ops.mutex_lock(mutex)

with ops.control_dependencies([lock]):
  r = fn(*args, **kwargs)

with ops.control_dependencies(nest.flatten(r)):
  with ops.colocate_with(mutex):
    ensure_lock_exists = mutex_lock_identity(lock)

  # Make sure that if any element of r is accessed, all of
  # them are executed together.
  r = nest.map_structure(tf.identity, r)

with ops.control_dependencies([ensure_lock_exists]):
  return nest.map_structure(tf.identity, r)

```

While `fn` is running in the critical section, no other functions which wish to use this critical section may run.

Often the use case is that two executions of the same graph, in parallel, wish to run `fn`; and we wish to ensure that only one of them executes at a time. This is especially important if `fn` modifies one or more variables at a time.

It is also useful if two separate functions must share a resource, but we wish to ensure the usage is exclusive.

Arguments: 

mutex: The mutex resource to lock.

Returns A tensor that keeps a shared pointer to a lock on the mutex; when the Tensor is destroyed, the use count on the shared pointer is decreased by 1. When it reaches 0, the lock is released.

func MutexV2 

func MutexV2(scope *Scope, optional ...MutexV2Attr) (resource tf.Output)

Creates a Mutex resource that can be locked by `MutexLock`.

Returns The mutex resource.

func NcclAllReduce 

func NcclAllReduce(scope *Scope, input tf.Output, reduction string, num_devices int64, shared_name string) (data tf.Output)

Outputs a tensor containing the reduction across all input tensors.

Outputs a tensor containing the reduction across all input tensors passed to ops within the same `shared_name.

The graph should be constructed so if one op runs with shared_name value `c`, then `num_devices` ops will run with shared_name value `c`. Failure to do so will cause the graph execution to fail to complete.

input: the input to the reduction data: the value of the reduction across all `num_devices` devices. reduction: the reduction operation to perform. num_devices: The number of devices participating in this reduction. shared_name: Identifier that shared between ops of the same reduction.

func NcclBroadcast 

func NcclBroadcast(scope *Scope, input tf.Output, shape tf.Shape) (output tf.Output)

Sends `input` to all devices that are connected to the output.

Sends `input` to all devices that are connected to the output.

The graph should be constructed so that all ops connected to the output have a valid device assignment, and the op itself is assigned one of these devices.

input: The input to the broadcast. output: The same as input. shape: The shape of the input tensor.

func NcclReduce 

func NcclReduce(scope *Scope, input []tf.Output, reduction string) (data tf.Output)

Reduces `input` from `num_devices` using `reduction` to a single device.

Reduces `input` from `num_devices` using `reduction` to a single device.

The graph should be constructed so that all inputs have a valid device assignment, and the op itself is assigned one of these devices.

input: The input to the reduction. data: the value of the reduction across all `num_devices` devices. reduction: the reduction operation to perform.

func NearestNeighbors 

func NearestNeighbors(scope *Scope, points tf.Output, centers tf.Output, k tf.Output) (nearest_center_indices tf.Output, nearest_center_distances tf.Output)

Selects the k nearest centers for each point.

Rows of points are assumed to be input points. Rows of centers are assumed to be the list of candidate centers. For each point, the k centers that have least L2 distance to it are computed.

Arguments: 

points: Matrix of shape (n, d). Rows are assumed to be input points.
centers: Matrix of shape (m, d). Rows are assumed to be centers.
k: Number of nearest centers to return for each point. If k is larger than m, then

only m centers are returned.

Returns: 

nearest_center_indices: Matrix of shape (n, min(m, k)). Each row contains the indices of the centers

closest to the corresponding point, ordered by increasing distance. 

nearest_center_distances: Matrix of shape (n, min(m, k)). Each row contains the squared L2 distance to the

corresponding center in nearest_center_indices.

func Neg 

func Neg(scope *Scope, x tf.Output) (y tf.Output)

Computes numerical negative value element-wise.

I.e., \\(y = -x\\).

func NextAfter 

func NextAfter(scope *Scope, x1 tf.Output, x2 tf.Output) (output tf.Output)

Returns the next representable value of `x1` in the direction of `x2`, element-wise.

This operation returns the same result as the C++ std::nextafter function.

It can also return a subnormal number.

@compatibility(cpp) Equivalent to C++ std::nextafter function. @end_compatibility

func NextIteration 

func NextIteration(scope *Scope, data tf.Output) (output tf.Output)

Makes its input available to the next iteration.

Arguments: 

data: The tensor to be made available to the next iteration.

Returns The same tensor as `data`.

func NoOp 

func NoOp(scope *Scope) (o *tf.Operation)

Does nothing. Only useful as a placeholder for control edges.

Returns the created operation.

func NonDeterministicInts 

func NonDeterministicInts(scope *Scope, shape tf.Output, optional ...NonDeterministicIntsAttr) (output tf.Output)

Non-deterministically generates some integers.

This op may use some OS-provided source of non-determinism (e.g. an RNG), so each execution will give different results.

Arguments: 

shape: The shape of the output tensor.

Returns Non-deterministic integer values with specified shape.

func NonMaxSuppression 

func NonMaxSuppression(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, optional ...NonMaxSuppressionAttr) (selected_indices tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system. Note that this algorithm is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm. The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression(
    boxes, scores, max_output_size, iou_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

Arguments: 

boxes: A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression.

Returns A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.

func NonMaxSuppressionV2 

func NonMaxSuppressionV2(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, iou_threshold tf.Output) (selected_indices tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system. Note that this algorithm is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm.

The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression_v2(
    boxes, scores, max_output_size, iou_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

Arguments: 

boxes: A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression. 

iou_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too much with respect to IOU.

Returns A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.

func NonMaxSuppressionV3 

func NonMaxSuppressionV3(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, iou_threshold tf.Output, score_threshold tf.Output) (selected_indices tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes with score less than `score_threshold` are removed. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system and more generally is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm. The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression_v2(
    boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

Arguments: 

boxes: A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression. 

iou_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too much with respect to IOU. 

score_threshold: A 0-D float tensor representing the threshold for deciding when to remove

boxes based on score.

Returns A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.

func NonMaxSuppressionV4 

func NonMaxSuppressionV4(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, iou_threshold tf.Output, score_threshold tf.Output, optional ...NonMaxSuppressionV4Attr) (selected_indices tf.Output, valid_outputs tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes with score less than `score_threshold` are removed. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system and more generally is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm. The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression_v2(
    boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

Arguments: 

boxes: A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression. 

iou_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too much with respect to IOU. 

score_threshold: A 0-D float tensor representing the threshold for deciding when to remove

boxes based on score.

Returns: 

selected_indices: A 1-D integer tensor of shape `[M]` representing the selected

indices from the boxes tensor, where `M <= max_output_size`. 

valid_outputs: A 0-D integer tensor representing the number of valid elements in

`selected_indices`, with the valid elements appearing first.

func NonMaxSuppressionV5 

func NonMaxSuppressionV5(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, iou_threshold tf.Output, score_threshold tf.Output, soft_nms_sigma tf.Output, optional ...NonMaxSuppressionV5Attr) (selected_indices tf.Output, selected_scores tf.Output, valid_outputs tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes with score less than `score_threshold` are removed. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system and more generally is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm. The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression_v2(
    boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

This op also supports a Soft-NMS (with Gaussian weighting) mode (c.f. Bodla et al, https://arxiv.org/abs/1704.04503) where boxes reduce the score of other overlapping boxes instead of directly causing them to be pruned. To enable this Soft-NMS mode, set the `soft_nms_sigma` parameter to be larger than 0.

Arguments: 

boxes: A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression. 

iou_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too much with respect to IOU. 

score_threshold: A 0-D float tensor representing the threshold for deciding when to remove

boxes based on score. 

soft_nms_sigma: A 0-D float tensor representing the sigma parameter for Soft NMS; see Bodla et

al (c.f. https://arxiv.org/abs/1704.04503). When `soft_nms_sigma=0.0` (which is default), we fall back to standard (hard) NMS.

Returns: 

selected_indices: A 1-D integer tensor of shape `[M]` representing the selected

indices from the boxes tensor, where `M <= max_output_size`. 

selected_scores: A 1-D float tensor of shape `[M]` representing the corresponding

scores for each selected box, where `M <= max_output_size`. Scores only differ from corresponding input scores when using Soft NMS (i.e. when `soft_nms_sigma>0`) 

valid_outputs: A 0-D integer tensor representing the number of valid elements in

`selected_indices`, with the valid elements appearing first.

func NonMaxSuppressionWithOverlaps 

func NonMaxSuppressionWithOverlaps(scope *Scope, overlaps tf.Output, scores tf.Output, max_output_size tf.Output, overlap_threshold tf.Output, score_threshold tf.Output) (selected_indices tf.Output)

Greedily selects a subset of bounding boxes in descending order of score,

pruning away boxes that have high overlaps with previously selected boxes. Bounding boxes with score less than `score_threshold` are removed. N-by-n overlap values are supplied as square matrix, which allows for defining a custom overlap criterium (eg. intersection over union, intersection over area, etc.).

The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example: 

selected_indices = tf.image.non_max_suppression_with_overlaps(
    overlaps, scores, max_output_size, overlap_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)

Arguments: 

overlaps: A 2-D float tensor of shape `[num_boxes, num_boxes]` representing

the n-by-n box overlap values. 

scores: A 1-D float tensor of shape `[num_boxes]` representing a single

score corresponding to each box (each row of boxes). 

max_output_size: A scalar integer tensor representing the maximum number of

boxes to be selected by non max suppression. 

overlap_threshold: A 0-D float tensor representing the threshold for deciding whether

boxes overlap too. 

score_threshold: A 0-D float tensor representing the threshold for deciding when to remove

boxes based on score.

Returns A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.

func NotEqual 

func NotEqual(scope *Scope, x tf.Output, y tf.Output, optional ...NotEqualAttr) (z tf.Output)

Returns the truth value of (x != y) element-wise.

*NOTE*: `NotEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func NthElement 

func NthElement(scope *Scope, input tf.Output, n tf.Output, optional ...NthElementAttr) (values tf.Output)

Finds values of the `n`-th order statistic for the last dimension.

If the input is a vector (rank-1), finds the entries which is the nth-smallest value in the vector and outputs their values as scalar tensor.

For matrices (resp. higher rank input), computes the entries which is the nth-smallest value in each row (resp. vector along the last dimension). Thus, 

values.shape = input.shape[:-1]

Arguments: 

input: 1-D or higher with last dimension at least `n+1`.
n: 0-D. Position of sorted vector to select along the last dimension (along

each row for matrices). Valid range of n is `[0, input.shape[:-1])`

Returns The `n`-th order statistic along each last dimensional slice.

func OneHot 

func OneHot(scope *Scope, indices tf.Output, depth tf.Output, on_value tf.Output, off_value tf.Output, optional ...OneHotAttr) (output tf.Output)

Returns a one-hot tensor.

The locations represented by indices in `indices` take value `on_value`, while all other locations take value `off_value`.

If the input `indices` is rank `N`, the output will have rank `N+1`, The new axis is created at dimension `axis` (default: the new axis is appended at the end).

If `indices` is a scalar the output shape will be a vector of length `depth`.

If `indices` is a vector of length `features`, the output shape will be: ``` 

features x depth if axis == -1
depth x features if axis == 0

```

If `indices` is a matrix (batch) with shape `[batch, features]`, the output shape will be: ``` 

batch x features x depth if axis == -1
batch x depth x features if axis == 1
depth x batch x features if axis == 0

```

Examples =========

Suppose that ``` 

indices = [0, 2, -1, 1]
depth = 3
on_value = 5.0
off_value = 0.0
axis = -1

```

Then output is `[4 x 3]`: ``` output = 

[5.0 0.0 0.0]  // one_hot(0)
[0.0 0.0 5.0]  // one_hot(2)
[0.0 0.0 0.0]  // one_hot(-1)
[0.0 5.0 0.0]  // one_hot(1)

```

Suppose that ``` 

indices = [0, 2, -1, 1]
depth = 3
on_value = 0.0
off_value = 3.0
axis = 0

```

Then output is `[3 x 4]`: ``` output = 

[0.0 3.0 3.0 3.0]
[3.0 3.0 3.0 0.0]
[3.0 3.0 3.0 3.0]
[3.0 0.0 3.0 3.0]

// ^ one_hot(0) // ^ one_hot(2) // ^ one_hot(-1) // ^ one_hot(1) ```

Suppose that ``` 

indices = [[0, 2], [1, -1]]
depth = 3
on_value = 1.0
off_value = 0.0
axis = -1

```

Then output is `[2 x 2 x 3]`: ``` output = 

[
  [1.0, 0.0, 0.0]  // one_hot(0)
  [0.0, 0.0, 1.0]  // one_hot(2)
][
  [0.0, 1.0, 0.0]  // one_hot(1)
  [0.0, 0.0, 0.0]  // one_hot(-1)
]

```

Arguments: 

indices: A tensor of indices.
depth: A scalar defining the depth of the one hot dimension.
on_value: A scalar defining the value to fill in output when `indices[j] = i`.
off_value: A scalar defining the value to fill in output when `indices[j] != i`.

Returns The one-hot tensor.

func OnesLike 

func OnesLike(scope *Scope, x tf.Output) (y tf.Output)

Returns a tensor of ones with the same shape and type as x.

Arguments: 

x: a tensor of type T.

Returns a tensor of the same shape and type as x but filled with ones.

func OptimizeDataset 

func OptimizeDataset(scope *Scope, input_dataset tf.Output, optimizations tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...OptimizeDatasetAttr) (handle tf.Output)

Creates a dataset by applying optimizations to `input_dataset`.

Creates a dataset by applying optimizations to `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
optimizations: A `tf.string` vector `tf.Tensor` identifying optimizations to use.

func OptimizeDatasetV2 

func OptimizeDatasetV2(scope *Scope, input_dataset tf.Output, optimizations_enabled tf.Output, optimizations_disabled tf.Output, optimizations_default tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...OptimizeDatasetV2Attr) (handle tf.Output)

Creates a dataset by applying related optimizations to `input_dataset`.

Creates a dataset by applying related optimizations to `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
optimizations_enabled: A `tf.string` vector `tf.Tensor` identifying user enabled optimizations.
optimizations_disabled: A `tf.string` vector `tf.Tensor` identifying user disabled optimizations.
optimizations_default: A `tf.string` vector `tf.Tensor` identifying optimizations by default.

func OptionalFromValue 

func OptionalFromValue(scope *Scope, components []tf.Output) (optional tf.Output)

Constructs an Optional variant from a tuple of tensors.

func OptionalGetValue 

func OptionalGetValue(scope *Scope, optional tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Returns the value stored in an Optional variant or raises an error if none exists.

func OptionalHasValue 

func OptionalHasValue(scope *Scope, optional tf.Output) (has_value tf.Output)

Returns true if and only if the given Optional variant has a value.

func OptionalNone 

func OptionalNone(scope *Scope) (optional tf.Output)

Creates an Optional variant with no value.

func OptionsDataset 

func OptionsDataset(scope *Scope, input_dataset tf.Output, serialized_options string, output_types []tf.DataType, output_shapes []tf.Shape, optional ...OptionsDatasetAttr) (handle tf.Output)

Creates a dataset by attaching tf.data.Options to `input_dataset`.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
serialized_options: A `tf.string` scalar `tf.Tensor` of serialized `tf.data.Options` protocol buffer.

func OrderedMapClear 

func OrderedMapClear(scope *Scope, dtypes []tf.DataType, optional ...OrderedMapClearAttr) (o *tf.Operation)

Op removes all elements in the underlying container.

Returns the created operation.

func OrderedMapIncompleteSize 

func OrderedMapIncompleteSize(scope *Scope, dtypes []tf.DataType, optional ...OrderedMapIncompleteSizeAttr) (size tf.Output)

Op returns the number of incomplete elements in the underlying container.

func OrderedMapPeek 

func OrderedMapPeek(scope *Scope, key tf.Output, indices tf.Output, dtypes []tf.DataType, optional ...OrderedMapPeekAttr) (values []tf.Output)

Op peeks at the values at the specified key. If the

underlying container does not contain this key this op will block until it does. This Op is optimized for performance.

func OrderedMapSize 

func OrderedMapSize(scope *Scope, dtypes []tf.DataType, optional ...OrderedMapSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func OrderedMapStage 

func OrderedMapStage(scope *Scope, key tf.Output, indices tf.Output, values []tf.Output, dtypes []tf.DataType, optional ...OrderedMapStageAttr) (o *tf.Operation)

Stage (key, values) in the underlying container which behaves like a ordered

associative container. Elements are ordered by key.

Arguments: 

key: int64

values: a list of tensors

dtypes A list of data types that inserted values should adhere to.

Returns the created operation.

func OrderedMapUnstage 

func OrderedMapUnstage(scope *Scope, key tf.Output, indices tf.Output, dtypes []tf.DataType, optional ...OrderedMapUnstageAttr) (values []tf.Output)

Op removes and returns the values associated with the key

from the underlying container. If the underlying container does not contain this key, the op will block until it does.

func OrderedMapUnstageNoKey 

func OrderedMapUnstageNoKey(scope *Scope, indices tf.Output, dtypes []tf.DataType, optional ...OrderedMapUnstageNoKeyAttr) (key tf.Output, values []tf.Output)

Op removes and returns the (key, value) element with the smallest

key from the underlying container. If the underlying container does not contain elements, the op will block until it does.

func OutfeedDequeue 

func OutfeedDequeue(scope *Scope, dtype tf.DataType, shape tf.Shape, optional ...OutfeedDequeueAttr) (output tf.Output)

Retrieves a single tensor from the computation outfeed.

This operation will block indefinitely until data is available.

Arguments: 

dtype: The type of elements in the tensor.
shape: The shape of the tensor.

Returns A tensor that will be read from the device outfeed.

func OutfeedDequeueTuple 

func OutfeedDequeueTuple(scope *Scope, dtypes []tf.DataType, shapes []tf.Shape, optional ...OutfeedDequeueTupleAttr) (outputs []tf.Output)

Retrieve multiple values from the computation outfeed.

This operation will block indefinitely until data is available. Output `i` corresponds to XLA tuple element `i`.

Arguments: 

dtypes: The element types of each element in `outputs`.
shapes: The shapes of each tensor in `outputs`.

Returns A list of tensors that will be read from the outfeed.

func OutfeedDequeueTupleV2 

func OutfeedDequeueTupleV2(scope *Scope, device_ordinal tf.Output, dtypes []tf.DataType, shapes []tf.Shape) (outputs []tf.Output)

Retrieve multiple values from the computation outfeed. Device ordinal is a tensor allowing dynamic outfeed.

This operation will block indefinitely until data is available. Output `i` corresponds to XLA tuple element `i`.

Arguments: 

device_ordinal: An int scalar tensor, representing the TPU device to use. This should be -1 when

the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. 

dtypes: The element types of each element in `outputs`.
shapes: The shapes of each tensor in `outputs`.

Returns A list of tensors that will be read from the outfeed.

func OutfeedDequeueV2 

func OutfeedDequeueV2(scope *Scope, device_ordinal tf.Output, dtype tf.DataType, shape tf.Shape) (output tf.Output)

Retrieves a single tensor from the computation outfeed. Device ordinal is a tensor allowing dynamic outfeed.

This operation will block indefinitely until data is available.

Arguments: 

device_ordinal: An int scalar tensor, representing the TPU device to use. This should be -1 when

the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. 

dtype: The type of elements in the tensor.
shape: The shape of the tensor.

Returns A tensor that will be read from the device outfeed.

func OutfeedEnqueue 

func OutfeedEnqueue(scope *Scope, input tf.Output) (o *tf.Operation)

Enqueue a Tensor on the computation outfeed.

Arguments: 

input: A tensor that will be inserted into the outfeed queue.

Returns the created operation.

func OutfeedEnqueueTuple 

func OutfeedEnqueueTuple(scope *Scope, inputs []tf.Output) (o *tf.Operation)

Enqueue multiple Tensor values on the computation outfeed.

Arguments: 

inputs: A list of tensors that will be inserted into the outfeed queue as an

XLA tuple.

Returns the created operation.

func Pack 

func Pack(scope *Scope, values []tf.Output, optional ...PackAttr) (output tf.Output)

Packs a list of `N` rank-`R` tensors into one rank-`(R+1)` tensor.

Packs the `N` tensors in `values` into a tensor with rank one higher than each tensor in `values`, by packing them along the `axis` dimension. Given a list of tensors of shape `(A, B, C)`;

if `axis == 0` then the `output` tensor will have the shape `(N, A, B, C)`. if `axis == 1` then the `output` tensor will have the shape `(A, N, B, C)`. Etc.

For example:

``` # 'x' is [1, 4] # 'y' is [2, 5] # 'z' is [3, 6] pack([x, y, z]) => [[1, 4], [2, 5], [3, 6]] # Pack along first dim. pack([x, y, z], axis=1) => [[1, 2, 3], [4, 5, 6]] ```

This is the opposite of `unpack`.

Arguments: 

values: Must be of same shape and type.

Returns The packed tensor.

func Pad 

func Pad(scope *Scope, input tf.Output, paddings tf.Output) (output tf.Output)

Pads a tensor with zeros.

This operation pads a `input` with zeros according to the `paddings` you specify. `paddings` is an integer tensor with shape `[Dn, 2]`, where n is the rank of `input`. For each dimension D of `input`, `paddings[D, 0]` indicates how many zeros to add before the contents of `input` in that dimension, and `paddings[D, 1]` indicates how many zeros to add after the contents of `input` in that dimension.

The padded size of each dimension D of the output is:

`paddings(D, 0) + input.dim_size(D) + paddings(D, 1)`

For example:

``` # 't' is [[1, 1], [2, 2]] # 'paddings' is [[1, 1], [2, 2]] # rank of 't' is 2 pad(t, paddings) ==> [[0, 0, 0, 0, 0, 0] 

[0, 0, 1, 1, 0, 0]
[0, 0, 2, 2, 0, 0]
[0, 0, 0, 0, 0, 0]]

```

func PadV2 

func PadV2(scope *Scope, input tf.Output, paddings tf.Output, constant_values tf.Output) (output tf.Output)

Pads a tensor.

This operation pads `input` according to the `paddings` and `constant_values` you specify. `paddings` is an integer tensor with shape `[Dn, 2]`, where n is the rank of `input`. For each dimension D of `input`, `paddings[D, 0]` indicates how many padding values to add before the contents of `input` in that dimension, and `paddings[D, 1]` indicates how many padding values to add after the contents of `input` in that dimension. `constant_values` is a scalar tensor of the same type as `input` that indicates the value to use for padding `input`.

The padded size of each dimension D of the output is:

`paddings(D, 0) + input.dim_size(D) + paddings(D, 1)`

For example:

``` # 't' is [[1, 1], [2, 2]] # 'paddings' is [[1, 1], [2, 2]] # 'constant_values' is 0 # rank of 't' is 2 pad(t, paddings) ==> [[0, 0, 0, 0, 0, 0] 

[0, 0, 1, 1, 0, 0]
[0, 0, 2, 2, 0, 0]
[0, 0, 0, 0, 0, 0]]

```

func PaddedBatchDataset 

func PaddedBatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, padded_shapes []tf.Output, padding_values []tf.Output, output_shapes []tf.Shape, optional ...PaddedBatchDatasetAttr) (handle tf.Output)

Creates a dataset that batches and pads `batch_size` elements from the input.

Arguments: 

batch_size: A scalar representing the number of elements to accumulate in a

batch. 

padded_shapes: A list of int64 tensors representing the desired padded shapes

of the corresponding output components. These shapes may be partially specified, using `-1` to indicate that a particular dimension should be padded to the maximum size of all batch elements. 

padding_values: A list of scalars containing the padding value to use for

each of the outputs.

func PaddedBatchDatasetV2 

func PaddedBatchDatasetV2(scope *Scope, input_dataset tf.Output, batch_size tf.Output, padded_shapes []tf.Output, padding_values []tf.Output, drop_remainder tf.Output, output_shapes []tf.Shape, optional ...PaddedBatchDatasetV2Attr) (handle tf.Output)

Creates a dataset that batches and pads `batch_size` elements from the input.

Arguments: 

batch_size: A scalar representing the number of elements to accumulate in a

batch. 

padded_shapes: A list of int64 tensors representing the desired padded shapes

of the corresponding output components. These shapes may be partially specified, using `-1` to indicate that a particular dimension should be padded to the maximum size of all batch elements. 

padding_values: A list of scalars containing the padding value to use for

each of the outputs. 

drop_remainder: A scalar representing whether the last batch should be dropped in case its size

is smaller than desired.

func PaddingFIFOQueueV2 

func PaddingFIFOQueueV2(scope *Scope, component_types []tf.DataType, optional ...PaddingFIFOQueueV2Attr) (handle tf.Output)

A queue that produces elements in first-in first-out order.

Variable-size shapes are allowed by setting the corresponding shape dimensions to 0 in the shape attr. In this case DequeueMany will pad up to the maximum size of any given element in the minibatch. See below for details.

Arguments: 

component_types: The type of each component in a value.

Returns The handle to the queue.

func ParallelConcat 

func ParallelConcat(scope *Scope, values []tf.Output, shape tf.Shape) (output tf.Output)

Concatenates a list of `N` tensors along the first dimension.

The input tensors are all required to have size 1 in the first dimension.

For example:

``` # 'x' is [[1, 4]] # 'y' is [[2, 5]] # 'z' is [[3, 6]] parallel_concat([x, y, z]) => [[1, 4], [2, 5], [3, 6]] # Pack along first dim. ```

The difference between concat and parallel_concat is that concat requires all of the inputs be computed before the operation will begin but doesn't require that the input shapes be known during graph construction. Parallel concat will copy pieces of the input into the output as they become available, in some situations this can provide a performance benefit.

Arguments: 

values: Tensors to be concatenated. All must have size 1 in the first dimension

and same shape. 

shape: the final shape of the result; should be equal to the shapes of any input

but with the number of input values in the first dimension.

Returns The concatenated tensor.

func ParallelDynamicStitch 

func ParallelDynamicStitch(scope *Scope, indices []tf.Output, data []tf.Output) (merged tf.Output)

Interleave the values from the `data` tensors into a single tensor.

Builds a merged tensor such that

```python 

merged[indices[m][i, ..., j], ...] = data[m][i, ..., j, ...]

```

For example, if each `indices[m]` is scalar or vector, we have

```python 

# Scalar indices:
merged[indices[m], ...] = data[m][...]

# Vector indices:
merged[indices[m][i], ...] = data[m][i, ...]

```

Each `data[i].shape` must start with the corresponding `indices[i].shape`, and the rest of `data[i].shape` must be constant w.r.t. `i`. That is, we must have `data[i].shape = indices[i].shape + constant`. In terms of this `constant`, the output shape is 

merged.shape = [max(indices)] + constant

Values may be merged in parallel, so if an index appears in both `indices[m][i]` and `indices[n][j]`, the result may be invalid. This differs from the normal DynamicStitch operator that defines the behavior in that case.

For example:

```python 

indices[0] = 6
indices[1] = [4, 1]
indices[2] = [[5, 2], [0, 3]]
data[0] = [61, 62]
data[1] = [[41, 42], [11, 12]]
data[2] = [[[51, 52], [21, 22]], [[1, 2], [31, 32]]]
merged = [[1, 2], [11, 12], [21, 22], [31, 32], [41, 42],
          [51, 52], [61, 62]]

```

This method can be used to merge partitions created by `dynamic_partition` as illustrated on the following example:

```python 

# Apply function (increments x_i) on elements for which a certain condition
# apply (x_i != -1 in this example).
x=tf.constant([0.1, -1., 5.2, 4.3, -1., 7.4])
condition_mask=tf.not_equal(x,tf.constant(-1.))
partitioned_data = tf.dynamic_partition(
    x, tf.cast(condition_mask, tf.int32) , 2)
partitioned_data[1] = partitioned_data[1] + 1.0
condition_indices = tf.dynamic_partition(
    tf.range(tf.shape(x)[0]), tf.cast(condition_mask, tf.int32) , 2)
x = tf.dynamic_stitch(condition_indices, partitioned_data)
# Here x=[1.1, -1., 6.2, 5.3, -1, 8.4], the -1. values remain
# unchanged.

```

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/DynamicStitch.png" alt> </div>

func ParameterizedTruncatedNormal 

func ParameterizedTruncatedNormal(scope *Scope, shape tf.Output, means tf.Output, stdevs tf.Output, minvals tf.Output, maxvals tf.Output, optional ...ParameterizedTruncatedNormalAttr) (output tf.Output)

Outputs random values from a normal distribution. The parameters may each be a

scalar which applies to the entire output, or a vector of length shape[0] which stores the parameters for each batch.

Arguments: 

shape: The shape of the output tensor. Batches are indexed by the 0th dimension.
means: The mean parameter of each batch.
stdevs: The standard deviation parameter of each batch. Must be greater than 0.
minvals: The minimum cutoff. May be -infinity.
maxvals: The maximum cutoff. May be +infinity, and must be more than the minval

for each batch.

Returns A matrix of shape num_batches x samples_per_batch, filled with random truncated normal values using the parameters for each row.

func ParseExample 

func ParseExample(scope *Scope, serialized tf.Output, names tf.Output, sparse_keys []tf.Output, dense_keys []tf.Output, dense_defaults []tf.Output, sparse_types []tf.DataType, dense_shapes []tf.Shape) (sparse_indices []tf.Output, sparse_values []tf.Output, sparse_shapes []tf.Output, dense_values []tf.Output)

Transforms a vector of brain.Example protos (as strings) into typed tensors.

Arguments: 

serialized: A vector containing a batch of binary serialized Example protos.
names: A vector containing the names of the serialized protos.

May contain, for example, table key (descriptive) names for the corresponding serialized protos. These are purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty vector if no names are available. If non-empty, this vector must be the same length as "serialized". 

sparse_keys: A list of Nsparse string Tensors (scalars).

The keys expected in the Examples' features associated with sparse values. 

dense_keys: A list of Ndense string Tensors (scalars).

The keys expected in the Examples' features associated with dense values. 

dense_defaults: A list of Ndense Tensors (some may be empty).

dense_defaults[j] provides default values when the example's feature_map lacks dense_key[j]. If an empty Tensor is provided for dense_defaults[j], then the Feature dense_keys[j] is required. The input type is inferred from dense_defaults[j], even when it's empty. If dense_defaults[j] is not empty, and dense_shapes[j] is fully defined, then the shape of dense_defaults[j] must match that of dense_shapes[j]. If dense_shapes[j] has an undefined major dimension (variable strides dense feature), dense_defaults[j] must contain a single element: the padding element. 

sparse_types: A list of Nsparse types; the data types of data in each Feature

given in sparse_keys. Currently the ParseExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). 

dense_shapes: A list of Ndense shapes; the shapes of data in each Feature

given in dense_keys. The number of elements in the Feature corresponding to dense_key[j] must always equal dense_shapes[j].NumEntries(). If dense_shapes[j] == (D0, D1, ..., DN) then the shape of output Tensor dense_values[j] will be (|serialized|, D0, D1, ..., DN): The dense outputs are just the inputs row-stacked by batch. This works for dense_shapes[j] = (-1, D1, ..., DN). In this case the shape of the output Tensor dense_values[j] will be (|serialized|, M, D1, .., DN), where M is the maximum number of blocks of elements of length D1 * .... * DN, across all minibatch entries in the input. Any minibatch entry with less than M blocks of elements of length D1 * ... * DN will be padded with the corresponding default_value scalar element along the second dimension.

func ParseExampleDataset 

func ParseExampleDataset(scope *Scope, input_dataset tf.Output, num_parallel_calls tf.Output, dense_defaults []tf.Output, sparse_keys []string, dense_keys []string, sparse_types []tf.DataType, dense_shapes []tf.Shape, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ParseExampleDatasetAttr) (handle tf.Output)

Transforms `input_dataset` containing `Example` protos as vectors of DT_STRING into a dataset of `Tensor` or `SparseTensor` objects representing the parsed features.

Arguments: 

dense_defaults: A dict mapping string keys to `Tensor`s.

The keys of the dict must match the dense_keys of the feature. 

sparse_keys: A list of string keys in the examples features.

The results for these keys will be returned as `SparseTensor` objects. 

dense_keys: A list of Ndense string Tensors (scalars).

The keys expected in the Examples features associated with dense values. 

sparse_types: A list of `DTypes` of the same length as `sparse_keys`.

Only `tf.float32` (`FloatList`), `tf.int64` (`Int64List`), and `tf.string` (`BytesList`) are supported. 

dense_shapes: List of tuples with the same length as `dense_keys`.

The shape of the data for each dense feature referenced by `dense_keys`. Required for any input tensors identified by `dense_keys`. Must be either fully defined, or may contain an unknown first dimension. An unknown first dimension means the feature is treated as having a variable number of blocks, and the output shape along this dimension is considered unknown at graph build time. Padding is applied for minibatch elements smaller than the maximum number of blocks for the given feature along this dimension. 

output_types: The type list for the return values.
output_shapes: The list of shapes being produced.

func ParseExampleDatasetV2 

func ParseExampleDatasetV2(scope *Scope, input_dataset tf.Output, num_parallel_calls tf.Output, dense_defaults []tf.Output, sparse_keys []string, dense_keys []string, sparse_types []tf.DataType, dense_shapes []tf.Shape, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ParseExampleDatasetV2Attr) (handle tf.Output)

Transforms `input_dataset` containing `Example` protos as vectors of DT_STRING into a dataset of `Tensor` or `SparseTensor` objects representing the parsed features.

Arguments: 

dense_defaults: A dict mapping string keys to `Tensor`s.

The keys of the dict must match the dense_keys of the feature. 

sparse_keys: A list of string keys in the examples features.

The results for these keys will be returned as `SparseTensor` objects. 

dense_keys: A list of Ndense string Tensors (scalars).

The keys expected in the Examples features associated with dense values. 

sparse_types: A list of `DTypes` of the same length as `sparse_keys`.

Only `tf.float32` (`FloatList`), `tf.int64` (`Int64List`), and `tf.string` (`BytesList`) are supported. 

dense_shapes: List of tuples with the same length as `dense_keys`.

The shape of the data for each dense feature referenced by `dense_keys`. Required for any input tensors identified by `dense_keys`. Must be either fully defined, or may contain an unknown first dimension. An unknown first dimension means the feature is treated as having a variable number of blocks, and the output shape along this dimension is considered unknown at graph build time. Padding is applied for minibatch elements smaller than the maximum number of blocks for the given feature along this dimension. 

output_types: The type list for the return values.
output_shapes: The list of shapes being produced.

func ParseExampleV2 

func ParseExampleV2(scope *Scope, serialized tf.Output, names tf.Output, sparse_keys tf.Output, dense_keys tf.Output, ragged_keys tf.Output, dense_defaults []tf.Output, num_sparse int64, sparse_types []tf.DataType, ragged_value_types []tf.DataType, ragged_split_types []tf.DataType, dense_shapes []tf.Shape) (sparse_indices []tf.Output, sparse_values []tf.Output, sparse_shapes []tf.Output, dense_values []tf.Output, ragged_values []tf.Output, ragged_row_splits []tf.Output)

Transforms a vector of tf.Example protos (as strings) into typed tensors.

Arguments: 

serialized: A scalar or vector containing binary serialized Example protos.
names: A tensor containing the names of the serialized protos.

Corresponds 1:1 with the `serialized` tensor. May contain, for example, table key (descriptive) names for the corresponding serialized protos. These are purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty vector if no names are available. If non-empty, this tensor must have the same shape as "serialized". 

sparse_keys: Vector of strings.

The keys expected in the Examples' features associated with sparse values. 

dense_keys: Vector of strings.

The keys expected in the Examples' features associated with dense values. 

ragged_keys: Vector of strings.

The keys expected in the Examples' features associated with ragged values. 

dense_defaults: A list of Tensors (some may be empty).  Corresponds 1:1 with `dense_keys`.

dense_defaults[j] provides default values when the example's feature_map lacks dense_key[j]. If an empty Tensor is provided for dense_defaults[j], then the Feature dense_keys[j] is required. The input type is inferred from dense_defaults[j], even when it's empty. If dense_defaults[j] is not empty, and dense_shapes[j] is fully defined, then the shape of dense_defaults[j] must match that of dense_shapes[j]. If dense_shapes[j] has an undefined major dimension (variable strides dense feature), dense_defaults[j] must contain a single element: the padding element. 

num_sparse: The number of sparse keys.
sparse_types: A list of `num_sparse` types; the data types of data in each Feature

given in sparse_keys. Currently the ParseExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). 

ragged_value_types: A list of `num_ragged` types; the data types of data in each Feature

given in ragged_keys (where `num_ragged = sparse_keys.size()`). Currently the ParseExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). 

ragged_split_types: A list of `num_ragged` types; the data types of row_splits in each Feature

given in ragged_keys (where `num_ragged = sparse_keys.size()`). May be DT_INT32 or DT_INT64. 

dense_shapes: A list of `num_dense` shapes; the shapes of data in each Feature

given in dense_keys (where `num_dense = dense_keys.size()`). The number of elements in the Feature corresponding to dense_key[j] must always equal dense_shapes[j].NumEntries(). If dense_shapes[j] == (D0, D1, ..., DN) then the shape of output Tensor dense_values[j] will be (|serialized|, D0, D1, ..., DN): The dense outputs are just the inputs row-stacked by batch. This works for dense_shapes[j] = (-1, D1, ..., DN). In this case the shape of the output Tensor dense_values[j] will be (|serialized|, M, D1, .., DN), where M is the maximum number of blocks of elements of length D1 * .... * DN, across all minibatch entries in the input. Any minibatch entry with less than M blocks of elements of length D1 * ... * DN will be padded with the corresponding default_value scalar element along the second dimension.

func ParseSequenceExample 

func ParseSequenceExample(scope *Scope, serialized tf.Output, debug_name tf.Output, context_dense_defaults []tf.Output, feature_list_dense_missing_assumed_empty []string, context_sparse_keys []string, context_dense_keys []string, feature_list_sparse_keys []string, feature_list_dense_keys []string, optional ...ParseSequenceExampleAttr) (context_sparse_indices []tf.Output, context_sparse_values []tf.Output, context_sparse_shapes []tf.Output, context_dense_values []tf.Output, feature_list_sparse_indices []tf.Output, feature_list_sparse_values []tf.Output, feature_list_sparse_shapes []tf.Output, feature_list_dense_values []tf.Output, feature_list_dense_lengths []tf.Output)

Transforms a vector of brain.SequenceExample protos (as strings) into typed tensors.

Arguments: 

serialized: A vector containing binary serialized SequenceExample protos.
debug_name: A vector containing the names of the serialized protos.

May contain, for example, table key (descriptive) name for the corresponding serialized proto. This is purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty vector if no name is available. 

context_dense_defaults: A list of Ncontext_dense Tensors (some may be empty).

context_dense_defaults[j] provides default values when the SequenceExample's context map lacks context_dense_key[j]. If an empty Tensor is provided for context_dense_defaults[j], then the Feature context_dense_keys[j] is required. The input type is inferred from context_dense_defaults[j], even when it's empty. If context_dense_defaults[j] is not empty, its shape must match context_dense_shapes[j]. 

feature_list_dense_missing_assumed_empty: A vector listing the

FeatureList keys which may be missing from the SequenceExamples. If the associated FeatureList is missing, it is treated as empty. By default, any FeatureList not listed in this vector must exist in the SequenceExamples. 

context_sparse_keys: A list of Ncontext_sparse string Tensors (scalars).

The keys expected in the Examples' features associated with context_sparse values. 

context_dense_keys: A list of Ncontext_dense string Tensors (scalars).

The keys expected in the SequenceExamples' context features associated with dense values. 

feature_list_sparse_keys: A list of Nfeature_list_sparse string Tensors

(scalars). The keys expected in the FeatureLists associated with sparse values. 

feature_list_dense_keys: A list of Nfeature_list_dense string Tensors (scalars).

The keys expected in the SequenceExamples' feature_lists associated with lists of dense values.

func ParseSequenceExampleV2 

func ParseSequenceExampleV2(scope *Scope, serialized tf.Output, debug_name tf.Output, context_sparse_keys tf.Output, context_dense_keys tf.Output, context_ragged_keys tf.Output, feature_list_sparse_keys tf.Output, feature_list_dense_keys tf.Output, feature_list_ragged_keys tf.Output, feature_list_dense_missing_assumed_empty tf.Output, context_dense_defaults []tf.Output, optional ...ParseSequenceExampleV2Attr) (context_sparse_indices []tf.Output, context_sparse_values []tf.Output, context_sparse_shapes []tf.Output, context_dense_values []tf.Output, context_ragged_values []tf.Output, context_ragged_row_splits []tf.Output, feature_list_sparse_indices []tf.Output, feature_list_sparse_values []tf.Output, feature_list_sparse_shapes []tf.Output, feature_list_dense_values []tf.Output, feature_list_dense_lengths []tf.Output, feature_list_ragged_values []tf.Output, feature_list_ragged_outer_splits []tf.Output, feature_list_ragged_inner_splits []tf.Output)

Transforms a vector of tf.io.SequenceExample protos (as strings) into typed tensors.

Arguments: 

serialized: A scalar or vector containing binary serialized SequenceExample protos.
debug_name: A scalar or vector containing the names of the serialized protos.

May contain, for example, table key (descriptive) name for the corresponding serialized proto. This is purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty vector if no name is available. 

context_sparse_keys: The keys expected in the Examples' features associated with context_sparse

values. 

context_dense_keys: The keys expected in the SequenceExamples' context features associated with

dense values. 

context_ragged_keys: The keys expected in the Examples' features associated with context_ragged

values. 

feature_list_sparse_keys: The keys expected in the FeatureLists associated with sparse values.
feature_list_dense_keys: The keys expected in the SequenceExamples' feature_lists associated

with lists of dense values. 

feature_list_ragged_keys: The keys expected in the FeatureLists associated with ragged values.
feature_list_dense_missing_assumed_empty: A vector corresponding 1:1 with feature_list_dense_keys, indicating which

features may be missing from the SequenceExamples. If the associated FeatureList is missing, it is treated as empty. 

context_dense_defaults: A list of Ncontext_dense Tensors (some may be empty).

context_dense_defaults[j] provides default values when the SequenceExample's context map lacks context_dense_key[j]. If an empty Tensor is provided for context_dense_defaults[j], then the Feature context_dense_keys[j] is required. The input type is inferred from context_dense_defaults[j], even when it's empty. If context_dense_defaults[j] is not empty, its shape must match context_dense_shapes[j].

func ParseSingleExample 

func ParseSingleExample(scope *Scope, serialized tf.Output, dense_defaults []tf.Output, num_sparse int64, sparse_keys []string, dense_keys []string, sparse_types []tf.DataType, dense_shapes []tf.Shape) (sparse_indices []tf.Output, sparse_values []tf.Output, sparse_shapes []tf.Output, dense_values []tf.Output)

Transforms a tf.Example proto (as a string) into typed tensors.

Arguments: 

serialized: A vector containing a batch of binary serialized Example protos.
dense_defaults: A list of Tensors (some may be empty), whose length matches

the length of `dense_keys`. dense_defaults[j] provides default values when the example's feature_map lacks dense_key[j]. If an empty Tensor is provided for dense_defaults[j], then the Feature dense_keys[j] is required. The input type is inferred from dense_defaults[j], even when it's empty. If dense_defaults[j] is not empty, and dense_shapes[j] is fully defined, then the shape of dense_defaults[j] must match that of dense_shapes[j]. If dense_shapes[j] has an undefined major dimension (variable strides dense feature), dense_defaults[j] must contain a single element: the padding element. 

num_sparse: The number of sparse features to be parsed from the example. This

must match the lengths of `sparse_keys` and `sparse_types`. 

sparse_keys: A list of `num_sparse` strings.

The keys expected in the Examples' features associated with sparse values. 

dense_keys: The keys expected in the Examples' features associated with dense

values. 

sparse_types: A list of `num_sparse` types; the data types of data in each

Feature given in sparse_keys. Currently the ParseSingleExample op supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). 

dense_shapes: The shapes of data in each Feature given in dense_keys.

The length of this list must match the length of `dense_keys`. The number of elements in the Feature corresponding to dense_key[j] must always equal dense_shapes[j].NumEntries(). If dense_shapes[j] == (D0, D1, ..., DN) then the shape of output Tensor dense_values[j] will be (D0, D1, ..., DN): In the case dense_shapes[j] = (-1, D1, ..., DN), the shape of the output Tensor dense_values[j] will be (M, D1, .., DN), where M is the number of blocks of elements of length D1 * .... * DN, in the input.

func ParseSingleSequenceExample 

func ParseSingleSequenceExample(scope *Scope, serialized tf.Output, feature_list_dense_missing_assumed_empty tf.Output, context_sparse_keys []tf.Output, context_dense_keys []tf.Output, feature_list_sparse_keys []tf.Output, feature_list_dense_keys []tf.Output, context_dense_defaults []tf.Output, debug_name tf.Output, optional ...ParseSingleSequenceExampleAttr) (context_sparse_indices []tf.Output, context_sparse_values []tf.Output, context_sparse_shapes []tf.Output, context_dense_values []tf.Output, feature_list_sparse_indices []tf.Output, feature_list_sparse_values []tf.Output, feature_list_sparse_shapes []tf.Output, feature_list_dense_values []tf.Output)

Transforms a scalar brain.SequenceExample proto (as strings) into typed tensors.

Arguments: 

serialized: A scalar containing a binary serialized SequenceExample proto.
feature_list_dense_missing_assumed_empty: A vector listing the

FeatureList keys which may be missing from the SequenceExample. If the associated FeatureList is missing, it is treated as empty. By default, any FeatureList not listed in this vector must exist in the SequenceExample. 

context_sparse_keys: A list of Ncontext_sparse string Tensors (scalars).

The keys expected in the Examples' features associated with context_sparse values. 

context_dense_keys: A list of Ncontext_dense string Tensors (scalars).

The keys expected in the SequenceExamples' context features associated with dense values. 

feature_list_sparse_keys: A list of Nfeature_list_sparse string Tensors

(scalars). The keys expected in the FeatureLists associated with sparse values. 

feature_list_dense_keys: A list of Nfeature_list_dense string Tensors (scalars).

The keys expected in the SequenceExamples' feature_lists associated with lists of dense values. 

context_dense_defaults: A list of Ncontext_dense Tensors (some may be empty).

context_dense_defaults[j] provides default values when the SequenceExample's context map lacks context_dense_key[j]. If an empty Tensor is provided for context_dense_defaults[j], then the Feature context_dense_keys[j] is required. The input type is inferred from context_dense_defaults[j], even when it's empty. If context_dense_defaults[j] is not empty, its shape must match context_dense_shapes[j]. 

debug_name: A scalar containing the name of the serialized proto.

May contain, for example, table key (descriptive) name for the corresponding serialized proto. This is purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty scalar if no name is available.

func ParseTensor 

func ParseTensor(scope *Scope, serialized tf.Output, out_type tf.DataType) (output tf.Output)

Transforms a serialized tensorflow.TensorProto proto into a Tensor.

Arguments: 

serialized: A scalar string containing a serialized TensorProto proto.
out_type: The type of the serialized tensor.  The provided type must match the

type of the serialized tensor and no implicit conversion will take place.

Returns A Tensor of type `out_type`.

func Placeholder 

func Placeholder(scope *Scope, dtype tf.DataType, optional ...PlaceholderAttr) (output tf.Output)

A placeholder op for a value that will be fed into the computation.

N.B. This operation will fail with an error if it is executed. It is intended as a way to represent a value that will always be fed, and to provide attrs that enable the fed value to be checked at runtime.

Arguments: 

dtype: The type of elements in the tensor.

Returns A placeholder tensor that must be replaced using the feed mechanism.

func PlaceholderV2 

func PlaceholderV2(scope *Scope, dtype tf.DataType, shape tf.Shape) (output tf.Output)

A placeholder op for a value that will be fed into the computation.

DEPRECATED at GraphDef version 23: Placeholder now behaves the same as PlaceholderV2.

N.B. This operation will fail with an error if it is executed. It is intended as a way to represent a value that will always be fed, and to provide attrs that enable the fed value to be checked at runtime.

Arguments: 

dtype: The type of elements in the tensor.
shape: The shape of the tensor. The shape can be any partially-specified

shape. To be unconstrained, pass in a shape with unknown rank.

Returns A placeholder tensor that must be replaced using the feed mechanism.

func PlaceholderWithDefault 

func PlaceholderWithDefault(scope *Scope, input tf.Output, shape tf.Shape) (output tf.Output)

A placeholder op that passes through `input` when its output is not fed.

Arguments: 

input: The default value to produce when `output` is not fed.
shape: The (possibly partial) shape of the tensor.

Returns A placeholder tensor that defaults to `input` if it is not fed.

func Polygamma 

func Polygamma(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)

Compute the polygamma function \\(\psi^{(n)}(x)\\).

The polygamma function is defined as:

\\(\psi^{(a)}(x) = \frac{d^a}{dx^a} \psi(x)\\)

where \\(\psi(x)\\) is the digamma function. The polygamma function is defined only for non-negative integer orders \\a\\.

func PopulationCount 

func PopulationCount(scope *Scope, x tf.Output) (y tf.Output)

Computes element-wise population count (a.k.a. popcount, bitsum, bitcount).

For each entry in `x`, calculates the number of `1` (on) bits in the binary representation of that entry.

**NOTE**: It is more efficient to first `tf.bitcast` your tensors into `int32` or `int64` and perform the bitcount on the result, than to feed in 8- or 16-bit inputs and then aggregate the resulting counts.

func Pow 

func Pow(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Computes the power of one value to another.

Given a tensor `x` and a tensor `y`, this operation computes \\(x^y\\) for corresponding elements in `x` and `y`. For example:

``` # tensor 'x' is [[2, 2]], [3, 3]] # tensor 'y' is [[8, 16], [2, 3]] tf.pow(x, y) ==> [[256, 65536], [9, 27]] ```

func PrefetchDataset 

func PrefetchDataset(scope *Scope, input_dataset tf.Output, buffer_size tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...PrefetchDatasetAttr) (handle tf.Output)

Creates a dataset that asynchronously prefetches elements from `input_dataset`.

Arguments: 

buffer_size: The maximum number of elements to buffer in an iterator over

this dataset.

func Prelinearize 

func Prelinearize(scope *Scope, input tf.Output, optional ...PrelinearizeAttr) (output tf.Output)

An op which linearizes one Tensor value to an opaque variant tensor.

Arguments: 

input: A tensor that will be linearized.

func PrelinearizeTuple 

func PrelinearizeTuple(scope *Scope, inputs []tf.Output, shapes []tf.Shape, optional ...PrelinearizeTupleAttr) (output tf.Output)

An op which linearizes multiple Tensor values to an opaque variant tensor.

Arguments: 

inputs: A list of tensors that will be provided using the infeed mechanism.
shapes: The shapes of each tensor in `inputs`.

func PreventGradient 

func PreventGradient(scope *Scope, input tf.Output, optional ...PreventGradientAttr) (output tf.Output)

An identity op that triggers an error if a gradient is requested.

When executed in a graph, this op outputs its input tensor as-is.

When building ops to compute gradients, the TensorFlow gradient system will return an error when trying to lookup the gradient of this op, because no gradient must ever be registered for this function. This op exists to prevent subtle bugs from silently returning unimplemented gradients in some corner cases.

Arguments: 

input: any tensor.

Returns the same input tensor.

func Print 

func Print(scope *Scope, input tf.Output, data []tf.Output, optional ...PrintAttr) (output tf.Output)

Prints a list of tensors.

Passes `input` through to `output` and prints `data` when evaluating.

Arguments: 

input: The tensor passed to `output`
data: A list of tensors to print out when op is evaluated.

Returns The unmodified `input` tensor

func PrintV2 

func PrintV2(scope *Scope, input tf.Output, optional ...PrintV2Attr) (o *tf.Operation)

Prints a string scalar.

Prints a string scalar to the desired output_stream.

Arguments: 

input: The string scalar to print.

Returns the created operation.

func PriorityQueueV2 

func PriorityQueueV2(scope *Scope, shapes []tf.Shape, optional ...PriorityQueueV2Attr) (handle tf.Output)

A queue that produces elements sorted by the first component value.

Note that the PriorityQueue requires the first component of any element to be a scalar int64, in addition to the other elements declared by component_types. Therefore calls to Enqueue and EnqueueMany (resp. Dequeue and DequeueMany) on a PriorityQueue will all require (resp. output) one extra entry in their input (resp. output) lists.

Arguments: 

shapes: The shape of each component in a value. The length of this attr must

be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time.

Returns The handle to the queue.

func PrivateThreadPoolDataset 

func PrivateThreadPoolDataset(scope *Scope, input_dataset tf.Output, num_threads tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

num_threads: Identifies the number of threads to use for the private threadpool.

func Prod 

func Prod(scope *Scope, input tf.Output, axis tf.Output, optional ...ProdAttr) (output tf.Output)

Computes the product of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func Qr 

func Qr(scope *Scope, input tf.Output, optional ...QrAttr) (q tf.Output, r tf.Output)

Computes the QR decompositions of one or more matrices.

Computes the QR decomposition of each inner matrix in `tensor` such that `tensor[..., :, :] = q[..., :, :] * r[..., :,:])`

Currently, the gradient for the QR decomposition is well-defined only when the first `P` columns of the inner matrix are linearly independent, where `P` is the minimum of `M` and `N`, the 2 inner-most dimmensions of `tensor`.

```python # a is a tensor. # q is a tensor of orthonormal matrices. # r is a tensor of upper triangular matrices. q, r = qr(a) q_full, r_full = qr(a, full_matrices=True) ```

Arguments: 

input: A tensor of shape `[..., M, N]` whose inner-most 2 dimensions

form matrices of size `[M, N]`. Let `P` be the minimum of `M` and `N`.

Returns: 

q: Orthonormal basis for range of `a`. If `full_matrices` is `False` then

shape is `[..., M, P]`; if `full_matrices` is `True` then shape is `[..., M, M]`. 

r: Triangular factor. If `full_matrices` is `False` then shape is

`[..., P, N]`. If `full_matrices` is `True` then shape is `[..., M, N]`.

func QuantizeAndDequantize 

func QuantizeAndDequantize(scope *Scope, input tf.Output, optional ...QuantizeAndDequantizeAttr) (output tf.Output)

Use QuantizeAndDequantizeV2 instead.

DEPRECATED at GraphDef version 22: Replaced by QuantizeAndDequantizeV2

func QuantizeAndDequantizeV2 

func QuantizeAndDequantizeV2(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, optional ...QuantizeAndDequantizeV2Attr) (output tf.Output)

Quantizes then dequantizes a tensor.

This op simulates the precision loss from the quantized forward pass by:

  1. Quantizing the tensor to fixed point numbers, which should match the target quantization method when it is used in inference.
  2. Dequantizing it back to floating point numbers for the following ops, most likely matmul.

There are different ways to quantize. This version uses only scaling, so 0.0 maps to 0.

From the specified 'num_bits' in the quantized output type, it determines minimum and maximum representable quantized values.

e.g.

* [-128, 127] for signed, num_bits = 8, or * [0, 255] for unsigned, num_bits = 8.

If range_given == False, the initial input_min, input_max will be determined automatically as the minimum and maximum values in the input tensor, otherwise the specified values of input_min, input_max are used.

Note: If the input_min, input_max are specified, they do not need to equal the actual minimum and maximum values in the tensor. e.g. in some cases it may be beneficial to specify these values such that the low probability extremes of the input distribution are clipped.

This op determines the maximum scale_factor that would map the initial [input_min, input_max] range to a range that lies within the representable quantized range.

It determines the scale from one of input_min and input_max, then updates the other one to maximize the representable range.

e.g.

  • if the output is signed, num_bits = 8, [input_min, input_max] = [-10.0, 5.0]: it would use a scale_factor of -128 / -10.0 = 12.8 In this case, it would update input_max to be 127 / 12.8 = 9.921875
  • if the output is signed, num_bits = 8, [input_min, input_max] = [-10.0, 10.0]: it would use a scale_factor of 127 / 10.0 = 12.7 In this case, it would update input_min to be 128.0 / 12.7 = -10.07874
  • if the output is unsigned, input_min is forced to be 0, and only the specified input_max is used.

After determining the scale_factor and updating the input range, it applies the following to each value in the 'input' tensor.

output = round(clamp(value, input_min, input_max) * scale_factor) / scale_factor.

The above round function rounds the value based on the given round_mode.

Arguments: 

input: Tensor to quantize and then dequantize.
input_min: If `range_given == True`, this specifies the minimum input value that needs to

be represented, otherwise it is determined from the min value of the `input` tensor. 

input_max: If `range_given == True`, this specifies the maximum input value that needs to

be represented, otherwise it is determined from the max value of the `input` tensor.

func QuantizeAndDequantizeV3 

func QuantizeAndDequantizeV3(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, num_bits tf.Output, optional ...QuantizeAndDequantizeV3Attr) (output tf.Output)

Quantizes then dequantizes a tensor.

This is almost identical to QuantizeAndDequantizeV2, except that num_bits is a tensor, so its value can change during training.

func QuantizeAndDequantizeV4 

func QuantizeAndDequantizeV4(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, optional ...QuantizeAndDequantizeV4Attr) (output tf.Output)

Quantizes then dequantizes a tensor.

This is almost identical to QuantizeAndDequantizeV2, except that it returns a gradient of 1 for inputs that are within the quantization range, or 0 otherwise.

Arguments: 

input: Tensor to quantize and then dequantize.
input_min: If `range_given == True`, this specifies the minimum input value that needs to

be represented, otherwise it is determined from the min value of the `input` tensor. 

input_max: If `range_given == True`, this specifies the maximum input value that needs to

be represented, otherwise it is determined from the max value of the `input` tensor.

func QuantizeAndDequantizeV4Grad 

func QuantizeAndDequantizeV4Grad(scope *Scope, gradients tf.Output, input tf.Output, input_min tf.Output, input_max tf.Output, optional ...QuantizeAndDequantizeV4GradAttr) (input_backprop tf.Output, input_min_backprop tf.Output, input_max_backprop tf.Output)

Returns the gradient of `QuantizeAndDequantizeV4`.

Returns a gradient of 1 for inputs that are within the quantization range, or 0 otherwise.

func QuantizeDownAndShrinkRange 

func QuantizeDownAndShrinkRange(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, out_type tf.DataType) (output tf.Output, output_min tf.Output, output_max tf.Output)

Convert the quantized 'input' tensor into a lower-precision 'output', using the

actual distribution of the values to maximize the usage of the lower bit depth and adjusting the output min and max ranges accordingly.

[input_min, input_max] are scalar floats that specify the range for the float interpretation of the 'input' data. For example, if input_min is -1.0f and input_max is 1.0f, and we are dealing with quint16 quantized data, then a 0 value in the 16-bit data should be interpreted as -1.0f, and a 65535 means 1.0f.

This operator tries to squeeze as much precision as possible into an output with a lower bit depth by calculating the actual min and max values found in the data. For example, maybe that quint16 input has no values lower than 16,384 and none higher than 49,152. That means only half the range is actually needed, all the float interpretations are between -0.5f and 0.5f, so if we want to compress the data into a quint8 output, we can use that range rather than the theoretical -1.0f to 1.0f that is suggested by the input min and max.

In practice, this is most useful for taking output from operations like QuantizedMatMul that can produce higher bit-depth outputs than their inputs and may have large potential output ranges, but in practice have a distribution of input values that only uses a small fraction of the possible range. By feeding that output into this operator, we can reduce it from 32 bits down to 8 with minimal loss of accuracy.

Arguments: 

input_min: The float value that the minimum quantized input value represents.
input_max: The float value that the maximum quantized input value represents.
out_type: The type of the output. Should be a lower bit depth than Tinput.

Returns: 

output
output_min: The float value that the minimum quantized output value represents.
output_max: The float value that the maximum quantized output value represents.

func QuantizeV2 

func QuantizeV2(scope *Scope, input tf.Output, min_range tf.Output, max_range tf.Output, T tf.DataType, optional ...QuantizeV2Attr) (output tf.Output, output_min tf.Output, output_max tf.Output)

Quantize the 'input' tensor of type float to 'output' tensor of type 'T'.

[min_range, max_range] are scalar floats that specify the range for the 'input' data. The 'mode' attribute controls exactly which calculations are used to convert the float values to their quantized equivalents. The 'round_mode' attribute controls which rounding tie-breaking algorithm is used when rounding float values to their quantized equivalents.

In 'MIN_COMBINED' mode, each value of the tensor will undergo the following:

``` out[i] = (in[i] - min_range) * range(T) / (max_range - min_range) if T == qint8: out[i] -= (range(T) + 1) / 2.0 ```

here `range(T) = numeric_limits<T>::max() - numeric_limits<T>::min()`

*MIN_COMBINED Mode Example*

Assume the input is type float and has a possible range of [0.0, 6.0] and the output type is quint8 ([0, 255]). The min_range and max_range values should be specified as 0.0 and 6.0. Quantizing from float to quint8 will multiply each value of the input by 255/6 and cast to quint8.

If the output type was qint8 ([-128, 127]), the operation will additionally subtract each value by 128 prior to casting, so that the range of values aligns with the range of qint8.

If the mode is 'MIN_FIRST', then this approach is used:

``` num_discrete_values = 1 << (# of bits in T) range_adjust = num_discrete_values / (num_discrete_values - 1) range = (range_max - range_min) * range_adjust range_scale = num_discrete_values / range quantized = round(input * range_scale) - round(range_min * range_scale) + 

numeric_limits<T>::min()

quantized = max(quantized, numeric_limits<T>::min()) quantized = min(quantized, numeric_limits<T>::max()) ```

The biggest difference between this and MIN_COMBINED is that the minimum range is rounded first, before it's subtracted from the rounded value. With MIN_COMBINED, a small bias is introduced where repeated iterations of quantizing and dequantizing will introduce a larger and larger error.

*SCALED mode Example*

`SCALED` mode matches the quantization approach used in `QuantizeAndDequantize{V2|V3}`.

If the mode is `SCALED`, the quantization is performed by multiplying each input value by a scaling_factor. The scaling_factor is determined from `min_range` and `max_range` to be as large as possible such that the range from `min_range` to `max_range` is representable within values of type T.

```c++ 

const int min_T = std::numeric_limits<T>::min();
const int max_T = std::numeric_limits<T>::max();
const float max_float = std::numeric_limits<float>::max();

const float scale_factor_from_min_side =
    (min_T * min_range > 0) ? min_T / min_range : max_float;
const float scale_factor_from_max_side =
    (max_T * max_range > 0) ? max_T / max_range : max_float;

const float scale_factor = std::min(scale_factor_from_min_side,
                                    scale_factor_from_max_side);

```

We next use the scale_factor to adjust min_range and max_range as follows:

```c++ 

min_range = min_T / scale_factor;
max_range = max_T / scale_factor;

```

e.g. if T = qint8, and initially min_range = -10, and max_range = 9, we would compare -128/-10.0 = 12.8 to 127/9.0 = 14.11, and set scaling_factor = 12.8 In this case, min_range would remain -10, but max_range would be adjusted to 127 / 12.8 = 9.921875

So we will quantize input values in the range (-10, 9.921875) to (-128, 127).

The input tensor can now be quantized by clipping values to the range `min_range` to `max_range`, then multiplying by scale_factor as follows:

```c++ result = round(min(max_range, max(min_range, input)) * scale_factor) ```

The adjusted `min_range` and `max_range` are returned as outputs 2 and 3 of this operation. These outputs should be used as the range for any further calculations.

*narrow_range (bool) attribute*

If true, we do not use the minimum quantized value. i.e. for int8 the quantized output, it would be restricted to the range -127..127 instead of the full -128..127 range. This is provided for compatibility with certain inference backends. (Only applies to SCALED mode)

*axis (int) attribute*

An optional `axis` attribute can specify a dimension index of the input tensor, such that quantization ranges will be calculated and applied separately for each slice of the tensor along that dimension. This is useful for per-channel quantization.

If axis is specified, min_range and max_range

if `axis`=None, per-tensor quantization is performed as normal.

*ensure_minimum_range (float) attribute*

Ensures the minimum quantization range is at least this value. The legacy default value for this is 0.01, but it is strongly suggested to set it to 0 for new uses.

Arguments: 

min_range: The minimum value of the quantization range. This value may be adjusted by the

op depending on other parameters. The adjusted value is written to `output_min`. If the `axis` attribute is specified, this must be a 1-D tensor whose size matches the `axis` dimension of the input and output tensors. 

max_range: The maximum value of the quantization range. This value may be adjusted by the

op depending on other parameters. The adjusted value is written to `output_max`. If the `axis` attribute is specified, this must be a 1-D tensor whose size matches the `axis` dimension of the input and output tensors.

Returns: 

output: The quantized data produced from the float input.
output_min: The final quantization range minimum, used to clip input values before scaling

and rounding them to quantized values. If the `axis` attribute is specified, this will be a 1-D tensor whose size matches the `axis` dimension of the input and output tensors. 

output_max: The final quantization range maximum, used to clip input values before scaling

and rounding them to quantized values. If the `axis` attribute is specified, this will be a 1-D tensor whose size matches the `axis` dimension of the input and output tensors.

func QuantizedAdd 

func QuantizedAdd(scope *Scope, x tf.Output, y tf.Output, min_x tf.Output, max_x tf.Output, min_y tf.Output, max_y tf.Output, optional ...QuantizedAddAttr) (z tf.Output, min_z tf.Output, max_z tf.Output)

Returns x + y element-wise, working on quantized buffers.

Arguments: 

min_x: The float value that the lowest quantized `x` value represents.
max_x: The float value that the highest quantized `x` value represents.
min_y: The float value that the lowest quantized `y` value represents.
max_y: The float value that the highest quantized `y` value represents.

Returns: 

z
min_z: The float value that the lowest quantized output value represents.
max_z: The float value that the highest quantized output value represents.

*NOTE*: `QuantizedAdd` supports limited forms of broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func QuantizedAvgPool 

func QuantizedAvgPool(scope *Scope, input tf.Output, min_input tf.Output, max_input tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output, min_output tf.Output, max_output tf.Output)

Produces the average pool of the input tensor for quantized types.

Arguments: 

input: 4-D with shape `[batch, height, width, channels]`.
min_input: The float value that the lowest quantized input value represents.
max_input: The float value that the highest quantized input value represents.
ksize: The size of the window for each dimension of the input tensor.

The length must be 4 to match the number of dimensions of the input. 

strides: The stride of the sliding window for each dimension of the input

tensor. The length must be 4 to match the number of dimensions of the input. 

padding: The type of padding algorithm to use.

Returns: 

output
min_output: The float value that the lowest quantized output value represents.
max_output: The float value that the highest quantized output value represents.

func QuantizedBatchNormWithGlobalNormalization 

func QuantizedBatchNormWithGlobalNormalization(scope *Scope, t tf.Output, t_min tf.Output, t_max tf.Output, m tf.Output, m_min tf.Output, m_max tf.Output, v tf.Output, v_min tf.Output, v_max tf.Output, beta tf.Output, beta_min tf.Output, beta_max tf.Output, gamma tf.Output, gamma_min tf.Output, gamma_max tf.Output, out_type tf.DataType, variance_epsilon float32, scale_after_normalization bool) (result tf.Output, result_min tf.Output, result_max tf.Output)

Quantized Batch normalization.

This op is deprecated and will be removed in the future. Prefer `tf.nn.batch_normalization`.

Arguments: 

t: A 4D input Tensor.
t_min: The value represented by the lowest quantized input.
t_max: The value represented by the highest quantized input.
m: A 1D mean Tensor with size matching the last dimension of t.

This is the first output from tf.nn.moments, or a saved moving average thereof. 

m_min: The value represented by the lowest quantized mean.
m_max: The value represented by the highest quantized mean.
v: A 1D variance Tensor with size matching the last dimension of t.

This is the second output from tf.nn.moments, or a saved moving average thereof. 

v_min: The value represented by the lowest quantized variance.
v_max: The value represented by the highest quantized variance.
beta: A 1D beta Tensor with size matching the last dimension of t.

An offset to be added to the normalized tensor. 

beta_min: The value represented by the lowest quantized offset.
beta_max: The value represented by the highest quantized offset.
gamma: A 1D gamma Tensor with size matching the last dimension of t.

If "scale_after_normalization" is true, this tensor will be multiplied with the normalized tensor. 

gamma_min: The value represented by the lowest quantized gamma.
gamma_max: The value represented by the highest quantized gamma.

variance_epsilon: A small float number to avoid dividing by 0.
scale_after_normalization: A bool indicating whether the resulted tensor

needs to be multiplied with gamma.

func QuantizedBiasAdd 

func QuantizedBiasAdd(scope *Scope, input tf.Output, bias tf.Output, min_input tf.Output, max_input tf.Output, min_bias tf.Output, max_bias tf.Output, out_type tf.DataType) (output tf.Output, min_out tf.Output, max_out tf.Output)

Adds Tensor 'bias' to Tensor 'input' for Quantized types.

Broadcasts the values of bias on dimensions 0..N-2 of 'input'.

Arguments: 

bias: A 1D bias Tensor with size matching the last dimension of 'input'.
min_input: The float value that the lowest quantized input value represents.
max_input: The float value that the highest quantized input value represents.
min_bias: The float value that the lowest quantized bias value represents.
max_bias: The float value that the highest quantized bias value represents.

Returns: 

output
min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents.

func QuantizedConcat 

func QuantizedConcat(scope *Scope, concat_dim tf.Output, values []tf.Output, input_mins []tf.Output, input_maxes []tf.Output) (output tf.Output, output_min tf.Output, output_max tf.Output)

Concatenates quantized tensors along one dimension.

Arguments: 

concat_dim: 0-D.  The dimension along which to concatenate.  Must be in the

range [0, rank(values)). 

values: The `N` Tensors to concatenate. Their ranks and types must match,

and their sizes must match in all dimensions except `concat_dim`. 

input_mins: The minimum scalar values for each of the input tensors.
input_maxes: The maximum scalar values for each of the input tensors.

Returns: 

output: A `Tensor` with the concatenation of values stacked along the

`concat_dim` dimension. This tensor's shape matches that of `values` except in `concat_dim` where it has the sum of the sizes. 

output_min: The float value that the minimum quantized output value represents.
output_max: The float value that the maximum quantized output value represents.

func QuantizedConv2D 

func QuantizedConv2D(scope *Scope, input tf.Output, filter tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, strides []int64, padding string, optional ...QuantizedConv2DAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes a 2D convolution given quantized 4D input and filter tensors.

The inputs are quantized tensors where the lowest value represents the real number of the associated minimum, and the highest represents the maximum. This means that you can only interpret the quantized output in the same way, by taking the returned minimum and maximum values into account.

Arguments: 

filter: filter's input_depth dimension must match input's depth dimensions.
min_input: The float value that the lowest quantized input value represents.
max_input: The float value that the highest quantized input value represents.
min_filter: The float value that the lowest quantized filter value represents.
max_filter: The float value that the highest quantized filter value represents.
strides: The stride of the sliding window for each dimension of the input

tensor. 

padding: The type of padding algorithm to use.

Returns: 

output
min_output: The float value that the lowest quantized output value represents.
max_output: The float value that the highest quantized output value represents.

func QuantizedConv2DPerChannel 

func QuantizedConv2DPerChannel(scope *Scope, input tf.Output, filter tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, strides []int64, padding string, optional ...QuantizedConv2DPerChannelAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes QuantizedConv2D per channel.

Arguments: 

input: The original input tensor.
filter: The original filter tensor.
min_input: The minimum value of the input tensor
max_input: The maximum value of the input tensor.
min_filter: The minimum value of the filter tensor.
max_filter: The maximum value of the filter tensor.
strides: list of stride values.

Returns: 

output: The output tensor.
min_output: The minimum value of the final output tensor.
max_output: The maximum value of the final output tensor.

func QuantizedDepthwiseConv2D 

func QuantizedDepthwiseConv2D(scope *Scope, input tf.Output, filter tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, strides []int64, padding string, optional ...QuantizedDepthwiseConv2DAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes quantized depthwise Conv2D.

Arguments: 

input: The original input tensor.
filter: The original filter tensor.
min_input: The float value that the minimum quantized input value represents.
max_input: The float value that the maximum quantized input value represents.
min_filter: The float value that the minimum quantized filter value represents.
max_filter: The float value that the maximum quantized filter value represents.
strides: List of stride values.

Returns: 

output: The output tensor.
min_output: The float value that the minimum quantized output value represents.
max_output: The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBias 

func QuantizedDepthwiseConv2DWithBias(scope *Scope, input tf.Output, filter tf.Output, bias tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, strides []int64, padding string, optional ...QuantizedDepthwiseConv2DWithBiasAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes quantized depthwise Conv2D with Bias.

Arguments: 

input: The original input tensor.
filter: The original filter tensor.
bias: The original bias tensor.
min_input: The float value that the minimum quantized input value represents.
max_input: The float value that the maximum quantized input value represents.
min_filter: The float value that the minimum quantized filter value represents.
max_filter: The float value that the maximum quantized filter value represents.
strides: List of stride values.

Returns: 

output: The output tensor.
min_output: The float value that the minimum quantized output value represents.
max_output: The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBiasAndRelu 

func QuantizedDepthwiseConv2DWithBiasAndRelu(scope *Scope, input tf.Output, filter tf.Output, bias tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, strides []int64, padding string, optional ...QuantizedDepthwiseConv2DWithBiasAndReluAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes quantized depthwise Conv2D with Bias and Relu.

Arguments: 

input: The original input tensor.
filter: The original filter tensor.
bias: The original bias tensor.
min_input: The float value that the minimum quantized input value represents.
max_input: The float value that the maximum quantized input value represents.
min_filter: The float value that the minimum quantized filter value represents.
max_filter: The float value that the maximum quantized filter value represents.
strides: List of stride values.

Returns: 

output: The output tensor.
min_output: The float value that the minimum quantized output value represents.
max_output: The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize 

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize(scope *Scope, input tf.Output, filter tf.Output, bias tf.Output, min_input tf.Output, max_input tf.Output, min_filter tf.Output, max_filter tf.Output, min_freezed_output tf.Output, max_freezed_output tf.Output, strides []int64, padding string, optional ...QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr) (output tf.Output, min_output tf.Output, max_output tf.Output)

Computes quantized depthwise Conv2D with Bias, Relu and Requantize.

Arguments: 

input: The original input tensor.
filter: The original filter tensor.
bias: The original bias tensor.
min_input: The float value that the minimum quantized input value represents.
max_input: The float value that the maximum quantized input value represents.
min_filter: The float value that the minimum quantized filter value represents.
max_filter: The float value that the maximum quantized filter value represents.
min_freezed_output: The minimum float value of the output tensor.
max_freezed_output: The maximum float value of the output tensor.
strides: List of stride values.

Returns: 

output: The output tensor.
min_output: The float value that the minimum quantized output value represents.
max_output: The float value that the maximum quantized output value represents.

func QuantizedInstanceNorm 

func QuantizedInstanceNorm(scope *Scope, x tf.Output, x_min tf.Output, x_max tf.Output, optional ...QuantizedInstanceNormAttr) (y tf.Output, y_min tf.Output, y_max tf.Output)

Quantized Instance normalization.

Arguments: 

x: A 4D input Tensor.
x_min: The value represented by the lowest quantized input.
x_max: The value represented by the highest quantized input.

Returns: 

y: A 4D Tensor.
y_min: The value represented by the lowest quantized output.
y_max: The value represented by the highest quantized output.

func QuantizedMatMul 

func QuantizedMatMul(scope *Scope, a tf.Output, b tf.Output, min_a tf.Output, max_a tf.Output, min_b tf.Output, max_b tf.Output, optional ...QuantizedMatMulAttr) (out tf.Output, min_out tf.Output, max_out tf.Output)

Perform a quantized matrix multiplication of `a` by the matrix `b`.

The inputs must be two-dimensional matrices and the inner dimension of `a` (after being transposed if `transpose_a` is non-zero) must match the outer dimension of `b` (after being transposed if `transposed_b` is non-zero).

Arguments: 

a: Must be a two-dimensional tensor.
b: Must be a two-dimensional tensor.
min_a: The float value that the lowest quantized `a` value represents.
max_a: The float value that the highest quantized `a` value represents.
min_b: The float value that the lowest quantized `b` value represents.
max_b: The float value that the highest quantized `b` value represents.

Returns: 

out
min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents.

func QuantizedMatMulWithBias 

func QuantizedMatMulWithBias(scope *Scope, a tf.Output, b tf.Output, bias tf.Output, min_a tf.Output, max_a tf.Output, min_b tf.Output, max_b tf.Output, optional ...QuantizedMatMulWithBiasAttr) (out tf.Output, min_out tf.Output, max_out tf.Output)

Performs a quantized matrix multiplication of `a` by the matrix `b` with bias add.

The inputs must be two-dimensional matrices and 1D bias vector. And the inner dimension of `a` (after being transposed if `transpose_a` is non-zero) must match the outer dimension of `b` (after being transposed if `transposed_b` is non-zero). Then do broadcast add operation with bias values on the matrix multiplication result. The bias size must match inner dimension of `b`.

Arguments: 

a: A matrix to be multiplied. Must be a two-dimensional tensor of type `quint8`.
b: A matrix to be multiplied and must be a two-dimensional tensor of type `qint8`.
bias: A 1D bias tensor with size matching inner dimension of `b` (after being

transposed if `transposed_b` is non-zero). 

min_a: The float value that the lowest quantized `a` value represents.
max_a: The float value that the highest quantized `a` value represents.
min_b: The float value that the lowest quantized `b` value represents.
max_b: The float value that the highest quantized `b` value represents.

Returns: 

out
min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents.

func QuantizedMatMulWithBiasAndRelu 

func QuantizedMatMulWithBiasAndRelu(scope *Scope, a tf.Output, b tf.Output, bias tf.Output, min_a tf.Output, max_a tf.Output, min_b tf.Output, max_b tf.Output, optional ...QuantizedMatMulWithBiasAndReluAttr) (out tf.Output, min_out tf.Output, max_out tf.Output)

Perform a quantized matrix multiplication of `a` by the matrix `b` with bias add and relu fusion.

The inputs must be two-dimensional matrices and 1D bias vector. And the inner dimension of `a` (after being transposed if `transpose_a` is non-zero) must match the outer dimension of `b` (after being transposed if `transposed_b` is non-zero). Then do broadcast add operation with bias values on the matrix multiplication result. The bias size must match inner dimension of `b`. Then do relu activation to get non-negative result.

Arguments: 

a: A matrix to be multiplied. Must be a two-dimensional tensor of type `quint8`.
b: A matrix to be multiplied and must be a two-dimensional tensor of type `qint8`.
bias: A 1D bias tensor with size matching with inner dimension of `b` (after being

transposed if `transposed_b` is non-zero). 

min_a: The float value that the lowest quantized `a` value represents.
max_a: The float value that the highest quantized `a` value represents.
min_b: The float value that the lowest quantized `b` value represents.
max_b: The float value that the highest quantized `b` value represents.

Returns: 

out
min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents.

func QuantizedMatMulWithBiasAndReluAndRequantize 

func QuantizedMatMulWithBiasAndReluAndRequantize(scope *Scope, a tf.Output, b tf.Output, bias tf.Output, min_a tf.Output, max_a tf.Output, min_b tf.Output, max_b tf.Output, min_freezed_output tf.Output, max_freezed_output tf.Output, optional ...QuantizedMatMulWithBiasAndReluAndRequantizeAttr) (out tf.Output, min_out tf.Output, max_out tf.Output)

Perform a quantized matrix multiplication of `a` by the matrix `b` with bias add and relu and requantize fusion.

The inputs must be two-dimensional matrices and 1D bias vector. And the inner dimension of `a` (after being transposed if `transpose_a` is non-zero) must match the outer dimension of `b` (after being transposed if `transposed_b` is non-zero). Then do broadcast add operation with bias values on the matrix multiplication result. The bias size must match inner dimension of `b`. Then do relu activation to get non-negative result. Then do requantize operation to get final uint8 result.

Arguments: 

a: A matrix to be multiplied. Must be a two-dimensional tensor of type `quint8`.
b: A matrix to be multiplied and must be a two-dimensional tensor of type `qint8`.
bias: A 1D bias tensor with size matching with inner dimension of `b` (after being

transposed if `transposed_b` is non-zero). 

min_a: The float value that the lowest quantized `a` value represents.
max_a: The float value that the highest quantized `a` value represents.
min_b: The float value that the lowest quantized `b` value represents.
max_b: The float value that the highest quantized `b` value represents.
min_freezed_output: The float value that the highest quantized output value after requantize.

Returns: 

out
min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents.

func QuantizedMaxPool 

func QuantizedMaxPool(scope *Scope, input tf.Output, min_input tf.Output, max_input tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output, min_output tf.Output, max_output tf.Output)

Produces the max pool of the input tensor for quantized types.

Arguments: 

input: The 4D (batch x rows x cols x depth) Tensor to MaxReduce over.
min_input: The float value that the lowest quantized input value represents.
max_input: The float value that the highest quantized input value represents.
ksize: The size of the window for each dimension of the input tensor.

The length must be 4 to match the number of dimensions of the input. 

strides: The stride of the sliding window for each dimension of the input

tensor. The length must be 4 to match the number of dimensions of the input. 

padding: The type of padding algorithm to use.

Returns: 

output
min_output: The float value that the lowest quantized output value represents.
max_output: The float value that the highest quantized output value represents.

func QuantizedMul 

func QuantizedMul(scope *Scope, x tf.Output, y tf.Output, min_x tf.Output, max_x tf.Output, min_y tf.Output, max_y tf.Output, optional ...QuantizedMulAttr) (z tf.Output, min_z tf.Output, max_z tf.Output)

Returns x * y element-wise, working on quantized buffers.

Arguments: 

min_x: The float value that the lowest quantized `x` value represents.
max_x: The float value that the highest quantized `x` value represents.
min_y: The float value that the lowest quantized `y` value represents.
max_y: The float value that the highest quantized `y` value represents.

Returns: 

z
min_z: The float value that the lowest quantized output value represents.
max_z: The float value that the highest quantized output value represents.

*NOTE*: `QuantizedMul` supports limited forms of broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func QuantizedRelu 

func QuantizedRelu(scope *Scope, features tf.Output, min_features tf.Output, max_features tf.Output, optional ...QuantizedReluAttr) (activations tf.Output, min_activations tf.Output, max_activations tf.Output)

Computes Quantized Rectified Linear: `max(features, 0)`

Arguments: 

min_features: The float value that the lowest quantized value represents.
max_features: The float value that the highest quantized value represents.

Returns: 

activations: Has the same output shape as "features".
min_activations: The float value that the lowest quantized value represents.
max_activations: The float value that the highest quantized value represents.

func QuantizedRelu6 

func QuantizedRelu6(scope *Scope, features tf.Output, min_features tf.Output, max_features tf.Output, optional ...QuantizedRelu6Attr) (activations tf.Output, min_activations tf.Output, max_activations tf.Output)

Computes Quantized Rectified Linear 6: `min(max(features, 0), 6)`

Arguments: 

min_features: The float value that the lowest quantized value represents.
max_features: The float value that the highest quantized value represents.

Returns: 

activations: Has the same output shape as "features".
min_activations: The float value that the lowest quantized value represents.
max_activations: The float value that the highest quantized value represents.

func QuantizedReluX 

func QuantizedReluX(scope *Scope, features tf.Output, max_value tf.Output, min_features tf.Output, max_features tf.Output, optional ...QuantizedReluXAttr) (activations tf.Output, min_activations tf.Output, max_activations tf.Output)

Computes Quantized Rectified Linear X: `min(max(features, 0), max_value)`

Arguments: 

min_features: The float value that the lowest quantized value represents.
max_features: The float value that the highest quantized value represents.

Returns: 

activations: Has the same output shape as "features".
min_activations: The float value that the lowest quantized value represents.
max_activations: The float value that the highest quantized value represents.

func QuantizedReshape 

func QuantizedReshape(scope *Scope, tensor tf.Output, shape tf.Output, input_min tf.Output, input_max tf.Output) (output tf.Output, output_min tf.Output, output_max tf.Output)

Reshapes a quantized tensor as per the Reshape op.

```

Arguments: 

shape: Defines the shape of the output tensor.
input_min: The minimum value of the input.
input_max: The maximum value of the input.

Returns: 

output
output_min: This value is copied from input_min.
output_max: This value is copied from input_max.

func QuantizedResizeBilinear 

func QuantizedResizeBilinear(scope *Scope, images tf.Output, size tf.Output, min tf.Output, max tf.Output, optional ...QuantizedResizeBilinearAttr) (resized_images tf.Output, out_min tf.Output, out_max tf.Output)

Resize quantized `images` to `size` using quantized bilinear interpolation.

Input images and output images must be quantized types.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images.

Returns: 

resized_images: 4-D with shape

`[batch, new_height, new_width, channels]`. 

out_min
out_max

func QueueCloseV2 

func QueueCloseV2(scope *Scope, handle tf.Output, optional ...QueueCloseV2Attr) (o *tf.Operation)

Closes the given queue.

This operation signals that no more elements will be enqueued in the given queue. Subsequent Enqueue(Many) operations will fail. Subsequent Dequeue(Many) operations will continue to succeed if sufficient elements remain in the queue. Subsequent Dequeue(Many) operations that would block will fail immediately.

Arguments: 

handle: The handle to a queue.

Returns the created operation.

func QueueDequeueManyV2 

func QueueDequeueManyV2(scope *Scope, handle tf.Output, n tf.Output, component_types []tf.DataType, optional ...QueueDequeueManyV2Attr) (components []tf.Output)

Dequeues `n` tuples of one or more tensors from the given queue.

If the queue is closed and there are fewer than `n` elements, then an OutOfRange error is returned.

This operation concatenates queue-element component tensors along the 0th dimension to make a single component tensor. All of the components in the dequeued tuple will have size `n` in the 0th dimension.

This operation has `k` outputs, where `k` is the number of components in the tuples stored in the given queue, and output `i` is the ith component of the dequeued tuple.

N.B. If the queue is empty, this operation will block until `n` elements have been dequeued (or 'timeout_ms' elapses, if specified).

Arguments: 

handle: The handle to a queue.
n: The number of tuples to dequeue.
component_types: The type of each component in a tuple.

Returns One or more tensors that were dequeued as a tuple.

func QueueDequeueUpToV2 

func QueueDequeueUpToV2(scope *Scope, handle tf.Output, n tf.Output, component_types []tf.DataType, optional ...QueueDequeueUpToV2Attr) (components []tf.Output)

Dequeues `n` tuples of one or more tensors from the given queue.

This operation is not supported by all queues. If a queue does not support DequeueUpTo, then an Unimplemented error is returned.

If the queue is closed and there are more than 0 but less than `n` elements remaining, then instead of returning an OutOfRange error like QueueDequeueMany, less than `n` elements are returned immediately. If the queue is closed and there are 0 elements left in the queue, then an OutOfRange error is returned just like in QueueDequeueMany. Otherwise the behavior is identical to QueueDequeueMany:

This operation concatenates queue-element component tensors along the 0th dimension to make a single component tensor. All of the components in the dequeued tuple will have size n in the 0th dimension.

This operation has `k` outputs, where `k` is the number of components in the tuples stored in the given queue, and output `i` is the ith component of the dequeued tuple.

Arguments: 

handle: The handle to a queue.
n: The number of tuples to dequeue.
component_types: The type of each component in a tuple.

Returns One or more tensors that were dequeued as a tuple.

func QueueDequeueV2 

func QueueDequeueV2(scope *Scope, handle tf.Output, component_types []tf.DataType, optional ...QueueDequeueV2Attr) (components []tf.Output)

Dequeues a tuple of one or more tensors from the given queue.

This operation has k outputs, where k is the number of components in the tuples stored in the given queue, and output i is the ith component of the dequeued tuple.

N.B. If the queue is empty, this operation will block until an element has been dequeued (or 'timeout_ms' elapses, if specified).

Arguments: 

handle: The handle to a queue.
component_types: The type of each component in a tuple.

Returns One or more tensors that were dequeued as a tuple.

func QueueEnqueueManyV2 

func QueueEnqueueManyV2(scope *Scope, handle tf.Output, components []tf.Output, optional ...QueueEnqueueManyV2Attr) (o *tf.Operation)

Enqueues zero or more tuples of one or more tensors in the given queue.

This operation slices each component tensor along the 0th dimension to make multiple queue elements. All of the tuple components must have the same size in the 0th dimension.

The components input has k elements, which correspond to the components of tuples stored in the given queue.

N.B. If the queue is full, this operation will block until the given elements have been enqueued (or 'timeout_ms' elapses, if specified).

Arguments: 

handle: The handle to a queue.
components: One or more tensors from which the enqueued tensors should

be taken.

Returns the created operation.

func QueueEnqueueV2 

func QueueEnqueueV2(scope *Scope, handle tf.Output, components []tf.Output, optional ...QueueEnqueueV2Attr) (o *tf.Operation)

Enqueues a tuple of one or more tensors in the given queue.

The components input has k elements, which correspond to the components of tuples stored in the given queue.

N.B. If the queue is full, this operation will block until the given element has been enqueued (or 'timeout_ms' elapses, if specified).

Arguments: 

handle: The handle to a queue.
components: One or more tensors from which the enqueued tensors should be taken.

Returns the created operation.

func QueueIsClosedV2 

func QueueIsClosedV2(scope *Scope, handle tf.Output) (is_closed tf.Output)

Returns true if queue is closed.

This operation returns true if the queue is closed and false if the queue is open.

Arguments: 

handle: The handle to a queue.

func QueueSizeV2 

func QueueSizeV2(scope *Scope, handle tf.Output) (size tf.Output)

Computes the number of elements in the given queue.

Arguments: 

handle: The handle to a queue.

Returns The number of elements in the given queue.

func RFFT 

func RFFT(scope *Scope, input tf.Output, fft_length tf.Output, optional ...RFFTAttr) (output tf.Output)

Real-valued fast Fourier transform.

Computes the 1-dimensional discrete Fourier transform of a real-valued signal over the inner-most dimension of `input`.

Since the DFT of a real signal is Hermitian-symmetric, `RFFT` only returns the `fft_length / 2 + 1` unique components of the FFT: the zero-frequency term, followed by the `fft_length / 2` positive-frequency terms.

Along the axis `RFFT` is computed on, if `fft_length` is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A float32 tensor.
fft_length: An int32 tensor of shape [1]. The FFT length.

Returns A complex64 tensor of the same rank as `input`. The inner-most 

dimension of `input` is replaced with the `fft_length / 2 + 1` unique
frequency components of its 1D Fourier transform.

@compatibility(numpy) Equivalent to np.fft.rfft @end_compatibility

func RFFT2D 

func RFFT2D(scope *Scope, input tf.Output, fft_length tf.Output, optional ...RFFT2DAttr) (output tf.Output)

2D real-valued fast Fourier transform.

Computes the 2-dimensional discrete Fourier transform of a real-valued signal over the inner-most 2 dimensions of `input`.

Since the DFT of a real signal is Hermitian-symmetric, `RFFT2D` only returns the `fft_length / 2 + 1` unique components of the FFT for the inner-most dimension of `output`: the zero-frequency term, followed by the `fft_length / 2` positive-frequency terms.

Along each axis `RFFT2D` is computed on, if `fft_length` is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A float32 tensor.
fft_length: An int32 tensor of shape [2]. The FFT length for each dimension.

Returns A complex64 tensor of the same rank as `input`. The inner-most 2 

dimensions of `input` are replaced with their 2D Fourier transform. The
inner-most dimension contains `fft_length / 2 + 1` unique frequency
components.

@compatibility(numpy) Equivalent to np.fft.rfft2 @end_compatibility

func RFFT3D 

func RFFT3D(scope *Scope, input tf.Output, fft_length tf.Output, optional ...RFFT3DAttr) (output tf.Output)

3D real-valued fast Fourier transform.

Computes the 3-dimensional discrete Fourier transform of a real-valued signal over the inner-most 3 dimensions of `input`.

Since the DFT of a real signal is Hermitian-symmetric, `RFFT3D` only returns the `fft_length / 2 + 1` unique components of the FFT for the inner-most dimension of `output`: the zero-frequency term, followed by the `fft_length / 2` positive-frequency terms.

Along each axis `RFFT3D` is computed on, if `fft_length` is smaller than the corresponding dimension of `input`, the dimension is cropped. If it is larger, the dimension is padded with zeros.

Arguments: 

input: A float32 tensor.
fft_length: An int32 tensor of shape [3]. The FFT length for each dimension.

Returns A complex64 tensor of the same rank as `input`. The inner-most 3 

dimensions of `input` are replaced with the their 3D Fourier transform. The
inner-most dimension contains `fft_length / 2 + 1` unique frequency
components.

@compatibility(numpy) Equivalent to np.fft.rfftn with 3 dimensions. @end_compatibility

func RFFTND added in v0.7.0 

func RFFTND(scope *Scope, input tf.Output, fft_length tf.Output, axes tf.Output, optional ...RFFTNDAttr) (output tf.Output)

ND fast real Fourier transform.

Computes the n-dimensional real discrete Fourier transform over designated dimensions of `input`. The designated dimensions of `input` are assumed to be the result of `RFFTND`. The length of the last axis transformed will be fft_length[-1]//2+1.

If fft_length[i]<shape(input)[i], the input is cropped. If fft_length[i]>shape(input)[i], the input is padded with zeros. If fft_length is not given, the default shape(input) is used.

Axes mean the dimensions to perform the transform on. Default is to perform on all axes.

Arguments: 

input: A complex tensor.
fft_length: An int32 tensor. The FFT length for each dimension.
axes: An int32 tensor with a same shape as fft_length. Axes to perform the transform.

Returns A complex tensor of the same shape as `input`. The designated dimensions of `input` are replaced with their real Fourier transforms.

@compatibility(numpy) Equivalent to np.fft.rfftn. @end_compatibility

func RGBToHSV 

func RGBToHSV(scope *Scope, images tf.Output) (output tf.Output)

Converts one or more images from RGB to HSV.

Outputs a tensor of the same shape as the `images` tensor, containing the HSV value of the pixels. The output is only well defined if the value in `images` are in `[0,1]`.

`output[..., 0]` contains hue, `output[..., 1]` contains saturation, and `output[..., 2]` contains value. All HSV values are in `[0,1]`. A hue of 0 corresponds to pure red, hue 1/3 is pure green, and 2/3 is pure blue.

Usage Example:

>>> blue_image = tf.stack([ ... tf.zeros([5,5]), ... tf.zeros([5,5]), ... tf.ones([5,5])], ... axis=-1) >>> blue_hsv_image = tf.image.rgb_to_hsv(blue_image) >>> blue_hsv_image[0,0].numpy() array([0.6666667, 1. , 1. ], dtype=float32)

Arguments: 

images: 1-D or higher rank. RGB data to convert. Last dimension must be size 3.

Returns `images` converted to HSV.

func RaggedBincount 

func RaggedBincount(scope *Scope, splits tf.Output, values tf.Output, size tf.Output, weights tf.Output, optional ...RaggedBincountAttr) (output tf.Output)

Counts the number of occurrences of each value in an integer array.

Outputs a vector with length `size` and the same dtype as `weights`. If `weights` are empty, then index `i` stores the number of times the value `i` is counted in `arr`. If `weights` are non-empty, then index `i` stores the sum of the value in `weights` at each index where the corresponding value in `arr` is `i`.

Values in `arr` outside of the range [0, size) are ignored.

Arguments: 

splits: 1D int64 `Tensor`.
values: 2D int `Tensor`.
size: non-negative int scalar `Tensor`.
weights: is an int32, int64, float32, or float64 `Tensor` with the same

shape as `input`, or a length-0 `Tensor`, in which case it acts as all weights equal to 1.

Returns 1D `Tensor` with length equal to `size` or 2D `Tensor` with [batch_size, `size`]. The counts or summed weights for each value in the range [0, size).

func RaggedCountSparseOutput 

func RaggedCountSparseOutput(scope *Scope, splits tf.Output, values tf.Output, weights tf.Output, binary_output bool, optional ...RaggedCountSparseOutputAttr) (output_indices tf.Output, output_values tf.Output, output_dense_shape tf.Output)

Performs sparse-output bin counting for a ragged tensor input. 

Counts the number of times each value occurs in the input.

Arguments: 

splits: Tensor containing the row splits of the ragged tensor to count.
values: Tensor containing values of the sparse tensor to count.
weights: A Tensor of the same shape as indices containing per-index weight values.

May also be the empty tensor if no weights are used. 

binary_output: Whether to output the number of occurrences of each value or 1.

Returns: 

	output_indices: Indices tensor for the resulting sparse tensor object.
	output_values: Values tensor for the resulting sparse tensor object.
	output_dense_shape: Shape tensor for the resulting sparse tensor object.
  END
  }
  attr {
    name: "T"
    description: <<END

Dtype of the input values tensor.

func RaggedCross 

func RaggedCross(scope *Scope, ragged_values []tf.Output, ragged_row_splits []tf.Output, sparse_indices []tf.Output, sparse_values []tf.Output, sparse_shape []tf.Output, dense_inputs []tf.Output, input_order string, hashed_output bool, num_buckets int64, hash_key int64, out_values_type tf.DataType, out_row_splits_type tf.DataType) (output_values tf.Output, output_row_splits tf.Output)

Generates a feature cross from a list of tensors, and returns it as a RaggedTensor. See `tf.ragged.cross` for more details.

Arguments: 

ragged_values: The values tensor for each RaggedTensor input.
ragged_row_splits: The row_splits tensor for each RaggedTensor input.
sparse_indices: The indices tensor for each SparseTensor input.
sparse_values: The values tensor for each SparseTensor input.
sparse_shape: The dense_shape tensor for each SparseTensor input.
dense_inputs: The tf.Tensor inputs.
input_order: String specifying the tensor type for each input.  The `i`th character in

this string specifies the type of the `i`th input, and is one of: 'R' (ragged), 'D' (dense), or 'S' (sparse). This attr is used to ensure that the crossed values are combined in the order of the inputs from the call to tf.ragged.cross.

Returns: 

output_values: The `values` for the returned `RaggedTensor`.
output_row_splits: The `row_splits` for the returned `RaggedTensor`.

func RaggedGather 

func RaggedGather(scope *Scope, params_nested_splits []tf.Output, params_dense_values tf.Output, indices tf.Output, OUTPUT_RAGGED_RANK int64) (output_nested_splits []tf.Output, output_dense_values tf.Output)

Gather ragged slices from `params` axis `0` according to `indices`.

Outputs a `RaggedTensor` output composed from `output_dense_values` and `output_nested_splits`, such that:

```python output.shape = indices.shape + params.shape[1:] output.ragged_rank = indices.shape.ndims + params.ragged_rank output[i...j, d0...dn] = params[indices[i...j], d0...dn] ```

where

  • `params = ragged.from_nested_row_splits(params_dense_values, params_nested_splits)` provides the values that should be gathered.
  • `indices` ia a dense tensor with dtype `int32` or `int64`, indicating which values should be gathered.
  • `output = ragged.from_nested_row_splits(output_dense_values, output_nested_splits)` is the output tensor.

(Note: This c++ op is used to implement the higher-level python `tf.ragged.gather` op, which also supports ragged indices.)

Arguments: 

params_nested_splits: The `nested_row_splits` tensors that define the row-partitioning for the

`params` RaggedTensor input. 

params_dense_values: The `flat_values` for the `params` RaggedTensor. There was a terminology change

at the python level from dense_values to flat_values, so dense_values is the deprecated name. 

indices: Indices in the outermost dimension of `params` of the values that should be

gathered. 

OUTPUT_RAGGED_RANK: The ragged rank of the output RaggedTensor. `output_nested_splits` will contain

this number of `row_splits` tensors. This value should equal `indices.shape.ndims + params.ragged_rank - 1`.

Returns: 

output_nested_splits: The `nested_row_splits` tensors that define the row-partitioning for the

returned RaggedTensor. 

output_dense_values: The `flat_values` for the returned RaggedTensor.

func RaggedRange 

func RaggedRange(scope *Scope, starts tf.Output, limits tf.Output, deltas tf.Output, optional ...RaggedRangeAttr) (rt_nested_splits tf.Output, rt_dense_values tf.Output)

Returns a `RaggedTensor` containing the specified sequences of numbers.

Returns a `RaggedTensor` `result` composed from `rt_dense_values` and `rt_nested_splits`, such that `result[i] = range(starts[i], limits[i], deltas[i])`.

```python (rt_nested_splits, rt_dense_values) = ragged_range( 

starts=[2, 5, 8], limits=[3, 5, 12], deltas=1)

result = tf.ragged.from_row_splits(rt_dense_values, rt_nested_splits) print(result) <tf.RaggedTensor [[2], [], [8, 9, 10, 11]] > ```

The input tensors `starts`, `limits`, and `deltas` may be scalars or vectors. The vector inputs must all have the same size. Scalar inputs are broadcast to match the size of the vector inputs.

Arguments: 

starts: The starts of each range.
limits: The limits of each range.
deltas: The deltas of each range.

Returns: 

rt_nested_splits: The `row_splits` for the returned `RaggedTensor`.
rt_dense_values: The `flat_values` for the returned `RaggedTensor`.

func RaggedTensorFromVariant 

func RaggedTensorFromVariant(scope *Scope, encoded_ragged tf.Output, input_ragged_rank int64, output_ragged_rank int64, Tvalues tf.DataType, optional ...RaggedTensorFromVariantAttr) (output_nested_splits []tf.Output, output_dense_values tf.Output)

Decodes a `variant` Tensor into a `RaggedTensor`.

Decodes the given `variant` Tensor and returns a `RaggedTensor`. The input could be a scalar, meaning it encodes a single `RaggedTensor` with ragged_rank `output_ragged_rank`. It could also have an arbitrary rank, in which case each element is decoded into a `RaggedTensor` with ragged_rank `input_ragged_rank` and these are then stacked according to the input shape to output a single `RaggedTensor` with ragged_rank `output_ragged_rank`. Each `variant` element in the input Tensor is decoded by retrieving from the element a 1-D `variant` Tensor with `input_ragged_rank + 1` Tensors, corresponding to the splits and values of the decoded `RaggedTensor`. If `input_ragged_rank` is -1, then it is inferred as `output_ragged_rank` - `rank(encoded_ragged)`. See `RaggedTensorToVariant` for the corresponding encoding logic.

Arguments: 

encoded_ragged: A `variant` Tensor containing encoded `RaggedTensor`s.
input_ragged_rank: The ragged rank of each encoded `RaggedTensor` component in the input. If set to

-1, this is inferred as `output_ragged_rank` - `rank(encoded_ragged)` 

output_ragged_rank: The expected ragged rank of the output `RaggedTensor`. The following must hold:

`output_ragged_rank = rank(encoded_ragged) + input_ragged_rank`.

Returns: 

output_nested_splits: A list of one or more Tensors representing the splits of the output

`RaggedTensor`. 

output_dense_values: A Tensor representing the values of the output `RaggedTensor`.

func RaggedTensorToSparse 

func RaggedTensorToSparse(scope *Scope, rt_nested_splits []tf.Output, rt_dense_values tf.Output) (sparse_indices tf.Output, sparse_values tf.Output, sparse_dense_shape tf.Output)

Converts a `RaggedTensor` into a `SparseTensor` with the same values.

input=ragged.from_nested_row_splits(rt_dense_values, rt_nested_splits) output=SparseTensor(indices=sparse_indices, values=sparse_values, 

dense_shape=sparse_dense_shape)

Arguments: 

rt_nested_splits: The `row_splits` for the `RaggedTensor`.
rt_dense_values: The `flat_values` for the `RaggedTensor`.

Returns: 

sparse_indices: The indices for the `SparseTensor`.
sparse_values: The values of the `SparseTensor`.
sparse_dense_shape: `sparse_dense_shape` is a tight bounding box of the input `RaggedTensor`.

func RaggedTensorToTensor 

func RaggedTensorToTensor(scope *Scope, shape tf.Output, values tf.Output, default_value tf.Output, row_partition_tensors []tf.Output, row_partition_types []string) (result tf.Output)

Create a dense tensor from a ragged tensor, possibly altering its shape.

The `ragged_to_dense` op creates a dense tensor from a list of row partition tensors, a value vector, and default values. If the shape is unspecified, the minimal shape required to contain all the elements in the ragged tensor (the natural shape) will be used. If some dimensions are left unspecified, then the size of the natural shape is used in that dimension.

The default_value will be broadcast to the output shape. After that, the values from the ragged tensor overwrite the default values. Note that the default_value must have less dimensions than the value.

The row partition tensors are in the order of the dimensions. At present, the types can be:

  • "ROW_SPLITS": the row_splits tensor from the ragged tensor.
  • "VALUE_ROWIDS": the value_rowids tensor from the ragged tensor.
  • "FIRST_DIM_SIZE": if value_rowids is used for the first dimension, then it is preceded by "FIRST_DIM_SIZE".

Arguments: 

shape: The desired shape of the output tensor. If left unspecified (empty),

the minimal shape required to contain all the elements in the ragged tensor (the natural shape) will be used. If some dimensions are left unspecified, then the size of the natural shape is used in that dimension.

Note that dense dimensions cannot be modified by the shape argument. Trying to change the size of a dense dimension will cause the op to fail. Examples: natural shape: [4, 5, 6] shape: -1 output shape: [4, 5, 6]

natural shape: [4, 5, 6] shape: [3, -1, 2] output shape: [3, 5, 2]

natural shape: [4, 5, 6] shape: [3, 7, 2] output shape: [3, 7, 2] 

values: A 1D tensor representing the values of the ragged tensor.
default_value: The default_value when the shape is larger than the ragged tensor. The

default_value is broadcast until it is the shape of the output tensor, and then overwritten by values in the ragged tensor. The default value must be compatible with this broadcast operation, and must have fewer dimensions than the value tensor. 

	row_partition_types: The types of the row partition tensors. At present, these can be:
  - "ROW_SPLITS": the row_splits tensor from the ragged tensor.
  - "VALUE_ROWIDS": the value_rowids tensor from the ragged tensor.
  - "FIRST_DIM_SIZE": if value_rowids is used for the first dimension, then it
    is preceeded by "FIRST_DIM_SIZE".

The tensors are in the order of the dimensions.

Returns The resulting dense tensor.

func RaggedTensorToVariant 

func RaggedTensorToVariant(scope *Scope, rt_nested_splits []tf.Output, rt_dense_values tf.Output, batched_input bool) (encoded_ragged tf.Output)

Encodes a `RaggedTensor` into a `variant` Tensor.

Encodes the given `RaggedTensor` and returns a `variant` Tensor. If `batched_input` is True, then input `RaggedTensor` is unbatched along the zero-th dimension, each component `RaggedTensor` is encoded into a scalar `variant` Tensor, and these are stacked to return a 1-D `variant` Tensor. If `batched_input` is False, then the input `RaggedTensor` is encoded as is and a scalar `variant` Tensor is returned. A `RaggedTensor` is encoded by first creating a 1-D `variant` Tensor with `ragged_rank + 1` elements, containing the splits and values Tensors of the `RaggedTensor`. Then the 1-D `variant` Tensor is wrapped in a scalar `variant` Tensor. See `RaggedTensorFromVariant` for the corresponding decoding logic.

Arguments: 

rt_nested_splits: A list of one or more Tensors representing the splits of the input

`RaggedTensor`. 

rt_dense_values: A Tensor representing the values of the input `RaggedTensor`.
batched_input: A `bool` denoting whether the input is a batched `RaggedTensor`.

Returns A `variant` Tensor that containing encoded `RaggedTensor`.

func RaggedTensorToVariantGradient 

func RaggedTensorToVariantGradient(scope *Scope, encoded_ragged_grad tf.Output, row_splits tf.Output, dense_values_shape tf.Output, Tvalues tf.DataType) (dense_values_grad tf.Output)

Helper used to compute the gradient for `RaggedTensorToVariant`.

Computes the gradient for the dense_values input to the RaggedTensorToVariant op, given the variant-encoded ragged gradients of the outputs, along with the outer row-splits and the shape of the dense-values that were provided as inputs to the RaggedTensorToVariant op.

Arguments: 

encoded_ragged_grad: A `variant` Tensor containing encoded `RaggedTensor` gradients.
row_splits: Outermost row-splits that were used as input to the RaggedTensorToVariant op.
dense_values_shape: Shape of the dense_values that was used as an input to the

RaggedTensorToVariant op.

Returns Gradient for the dense_values of the RaggedTensorToVariant op.

func RandomCrop 

func RandomCrop(scope *Scope, image tf.Output, size tf.Output, optional ...RandomCropAttr) (output tf.Output)

Randomly crop `image`.

DEPRECATED at GraphDef version 8: Random crop is now pure Python

`size` is a 1-D int64 tensor with 2 elements representing the crop height and width. The values must be non negative.

This Op picks a random location in `image` and crops a `height` by `width` rectangle from that location. The random location is picked so the cropped area will fit inside the original image.

Arguments: 

image: 3-D of shape `[height, width, channels]`.
size: 1-D of length 2 containing: `crop_height`, `crop_width`..

Returns 3-D of shape `[crop_height, crop_width, channels].`

func RandomDataset 

func RandomDataset(scope *Scope, seed tf.Output, seed2 tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...RandomDatasetAttr) (handle tf.Output)

Creates a Dataset that returns pseudorandom numbers.

Creates a Dataset that returns a stream of uniformly distributed pseudorandom 64-bit signed integers.

In the TensorFlow Python API, you can instantiate this dataset via the class `tf.data.experimental.RandomDataset`.

Instances of this dataset are also created as a result of the `hoist_random_uniform` static optimization. Whether this optimization is performed is determined by the `experimental_optimization.hoist_random_uniform` option of `tf.data.Options`.

Arguments: 

seed: A scalar seed for the random number generator. If either seed or

seed2 is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. 

seed2: A second scalar seed to avoid seed collision.

func RandomDatasetV2 added in v0.4.0 

func RandomDatasetV2(scope *Scope, seed tf.Output, seed2 tf.Output, seed_generator tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...RandomDatasetV2Attr) (handle tf.Output)

Creates a Dataset that returns pseudorandom numbers.

Creates a Dataset that returns a stream of uniformly distributed pseudorandom 64-bit signed integers. It accepts a boolean attribute that determines if the random number generators are re-applied at each epoch. The default value is True which means that the seeds are applied and the same sequence of random numbers are generated at each epoch. If set to False, the seeds are not re-applied and a different sequence of random numbers are generated at each epoch.

In the TensorFlow Python API, you can instantiate this dataset via the class `tf.data.experimental.RandomDatasetV2`.

Arguments: 

seed: A scalar seed for the random number generator. If either seed or

seed2 is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. 

seed2: A second scalar seed to avoid seed collision.
seed_generator: A resource for the random number seed generator.

func RandomGamma 

func RandomGamma(scope *Scope, shape tf.Output, alpha tf.Output, optional ...RandomGammaAttr) (output tf.Output)

Outputs random values from the Gamma distribution(s) described by alpha.

This op uses the algorithm by Marsaglia et al. to acquire samples via transformation-rejection from pairs of uniform and normal random variables. See http://dl.acm.org/citation.cfm?id=358414

Arguments: 

shape: 1-D integer tensor. Shape of independent samples to draw from each

distribution described by the shape parameters given in alpha. 

alpha: A tensor in which each scalar is a "shape" parameter describing the

associated gamma distribution.

Returns A tensor with shape `shape + shape(alpha)`. Each slice `[:, ..., :, i0, i1, ...iN]` contains the samples drawn for `alpha[i0, i1, ...iN]`. The dtype of the output matches the dtype of alpha.

func RandomGammaGrad 

func RandomGammaGrad(scope *Scope, alpha tf.Output, sample tf.Output) (output tf.Output)

Computes the derivative of a Gamma random sample w.r.t. `alpha`.

func RandomIndexShuffle 

func RandomIndexShuffle(scope *Scope, index tf.Output, seed tf.Output, max_index tf.Output, optional ...RandomIndexShuffleAttr) (output tf.Output)

Outputs the position of `value` in a permutation of [0, ..., max_index].

Output values are a bijection of the `index` for any combination and `seed` and `max_index`.

If multiple inputs are vectors (matrix in case of seed) then the size of the first dimension must match.

The outputs are deterministic.

Arguments: 

index: A scalar tensor or a vector of dtype `dtype`. The index (or indices) to be shuffled. Must be within [0, max_index].
seed: A tensor of dtype `Tseed` and shape [3] or [n, 3]. The random seed.
max_index: A scalar tensor or vector of dtype `dtype`. The upper bound(s) of the interval (inclusive).

Returns A scalar tensor of dtype `dtype`, within [0, max_index]. The randomly shuffled index.

func RandomPoisson 

func RandomPoisson(scope *Scope, shape tf.Output, rate tf.Output, optional ...RandomPoissonAttr) (output tf.Output)

Use RandomPoissonV2 instead.

DEPRECATED at GraphDef version 25: Replaced by RandomPoissonV2

func RandomPoissonV2 

func RandomPoissonV2(scope *Scope, shape tf.Output, rate tf.Output, optional ...RandomPoissonV2Attr) (output tf.Output)

Outputs random values from the Poisson distribution(s) described by rate.

This op uses two algorithms, depending on rate. If rate >= 10, then the algorithm by Hormann is used to acquire samples via transformation-rejection. See http://www.sciencedirect.com/science/article/pii/0167668793909974.

Otherwise, Knuth's algorithm is used to acquire samples via multiplying uniform random variables. See Donald E. Knuth (1969). Seminumerical Algorithms. The Art of Computer Programming, Volume 2. Addison Wesley

Arguments: 

shape: 1-D integer tensor. Shape of independent samples to draw from each

distribution described by the shape parameters given in rate. 

rate: A tensor in which each scalar is a "rate" parameter describing the

associated poisson distribution.

Returns A tensor with shape `shape + shape(rate)`. Each slice `[:, ..., :, i0, i1, ...iN]` contains the samples drawn for `rate[i0, i1, ...iN]`.

func RandomShuffle 

func RandomShuffle(scope *Scope, value tf.Output, optional ...RandomShuffleAttr) (output tf.Output)

Randomly shuffles a tensor along its first dimension. 

The tensor is shuffled along dimension 0, such that each `value[j]` is mapped
to one and only one `output[i]`. For example, a mapping that might occur for a
3x2 tensor is:

``` [[1, 2], [[5, 6], 

[3, 4],  ==>   [1, 2],
[5, 6]]        [3, 4]]

```

Arguments: 

value: The tensor to be shuffled.

Returns A tensor of same shape and type as `value`, shuffled along its first dimension.

func RandomShuffleQueueV2 

func RandomShuffleQueueV2(scope *Scope, component_types []tf.DataType, optional ...RandomShuffleQueueV2Attr) (handle tf.Output)

A queue that randomizes the order of elements.

Arguments: 

component_types: The type of each component in a value.

Returns The handle to the queue.

func RandomStandardNormal 

func RandomStandardNormal(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...RandomStandardNormalAttr) (output tf.Output)

Outputs random values from a normal distribution.

The generated values will have mean 0 and standard deviation 1.

Arguments: 

shape: The shape of the output tensor.
dtype: The type of the output.

Returns A tensor of the specified shape filled with random normal values.

func RandomUniform 

func RandomUniform(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...RandomUniformAttr) (output tf.Output)

Outputs random values from a uniform distribution.

The generated values follow a uniform distribution in the range `[0, 1)`. The lower bound 0 is included in the range, while the upper bound 1 is excluded.

Arguments: 

shape: The shape of the output tensor.
dtype: The type of the output.

Returns A tensor of the specified shape filled with uniform random values.

func RandomUniformInt 

func RandomUniformInt(scope *Scope, shape tf.Output, minval tf.Output, maxval tf.Output, optional ...RandomUniformIntAttr) (output tf.Output)

Outputs random integers from a uniform distribution.

The generated values are uniform integers in the range `[minval, maxval)`. The lower bound `minval` is included in the range, while the upper bound `maxval` is excluded.

The random integers are slightly biased unless `maxval - minval` is an exact power of two. The bias is small for values of `maxval - minval` significantly smaller than the range of the output (either `2^32` or `2^64`).

Arguments: 

shape: The shape of the output tensor.
minval: 0-D.  Inclusive lower bound on the generated integers.
maxval: 0-D.  Exclusive upper bound on the generated integers.

Returns A tensor of the specified shape filled with uniform random integers.

func Range 

func Range(scope *Scope, start tf.Output, limit tf.Output, delta tf.Output) (output tf.Output)

Creates a sequence of numbers.

This operation creates a sequence of numbers that begins at `start` and extends by increments of `delta` up to but not including `limit`.

For example:

``` # 'start' is 3 # 'limit' is 18 # 'delta' is 3 tf.range(start, limit, delta) ==> [3, 6, 9, 12, 15] ```

Arguments: 

start: 0-D (scalar). First entry in the sequence.
limit: 0-D (scalar). Upper limit of sequence, exclusive.
delta: 0-D (scalar). Optional. Default is 1. Number that increments `start`.

Returns 1-D.

func RangeDataset 

func RangeDataset(scope *Scope, start tf.Output, stop tf.Output, step tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...RangeDatasetAttr) (handle tf.Output)

Creates a dataset with a range of values. Corresponds to python's xrange.

Arguments: 

start: corresponds to start in python's xrange().
stop: corresponds to stop in python's xrange().
step: corresponds to step in python's xrange().

func Rank 

func Rank(scope *Scope, input tf.Output) (output tf.Output)

Returns the rank of a tensor.

This operation returns an integer representing the rank of `input`.

For example:

``` # 't' is [[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]] # shape of tensor 't' is [2, 2, 3] rank(t) ==> 3 ```

**Note**: The rank of a tensor is not the same as the rank of a matrix. The rank of a tensor is the number of indices required to uniquely select each element of the tensor. Rank is also known as "order", "degree", or "ndims."

func ReadFile 

func ReadFile(scope *Scope, filename tf.Output) (contents tf.Output)

Reads and outputs the entire contents of the input filename.

func ReadVariableOp 

func ReadVariableOp(scope *Scope, resource tf.Output, dtype tf.DataType) (value tf.Output)

Reads the value of a variable.

The tensor returned by this operation is immutable.

The value returned by this operation is guaranteed to be influenced by all the writes on which this operation depends directly or indirectly, and to not be influenced by any of the writes which depend directly or indirectly on this operation.

Arguments: 

resource: handle to the resource in which to store the variable.
dtype: the dtype of the value.

func ReadVariableXlaSplitND 

func ReadVariableXlaSplitND(scope *Scope, resource tf.Output, T tf.DataType, N int64, num_splits []int64, optional ...ReadVariableXlaSplitNDAttr) (outputs []tf.Output)

Splits resource variable input tensor across all dimensions.

An op which splits the resource variable input tensor based on the given num_splits attribute, pads slices optionally, and returned the slices. Slices are returned in row-major order.

This op may be generated via the TPU bridge.

For example, with `input` tensor: ``` [[0, 1, 2], 

[3, 4, 5],
[6, 7, 8]]

``` `num_splits`: ``` [2, 2] ``` and `paddings`: ``` [1, 1] ``` the expected `outputs` is: ``` [[0, 1], 

[3, 4]]

[[2, 0], 

[5, 0]]

[[6, 7], 

[0, 0]]

[[8, 0], 

[0, 0]]

```

Arguments: 

resource: Resource variable of input tensor to split across all dimensions.

num_splits: Number of ways to split per dimension. Shape dimensions must be evenly

divisible.

Returns Output slices based on input and num_splits defined, in row-major order.

func ReaderNumRecordsProducedV2 

func ReaderNumRecordsProducedV2(scope *Scope, reader_handle tf.Output) (records_produced tf.Output)

Returns the number of records this Reader has produced.

This is the same as the number of ReaderRead executions that have succeeded.

Arguments: 

reader_handle: Handle to a Reader.

func ReaderNumWorkUnitsCompletedV2 

func ReaderNumWorkUnitsCompletedV2(scope *Scope, reader_handle tf.Output) (units_completed tf.Output)

Returns the number of work units this Reader has finished processing.

Arguments: 

reader_handle: Handle to a Reader.

func ReaderReadUpToV2 

func ReaderReadUpToV2(scope *Scope, reader_handle tf.Output, queue_handle tf.Output, num_records tf.Output) (keys tf.Output, values tf.Output)

Returns up to `num_records` (key, value) pairs produced by a Reader.

Will dequeue from the input queue if necessary (e.g. when the Reader needs to start reading from a new file since it has finished with the previous file). It may return less than `num_records` even before the last batch.

Arguments: 

reader_handle: Handle to a `Reader`.
queue_handle: Handle to a `Queue`, with string work items.
num_records: number of records to read from `Reader`.

Returns: 

keys: A 1-D tensor.
values: A 1-D tensor.

func ReaderReadV2 

func ReaderReadV2(scope *Scope, reader_handle tf.Output, queue_handle tf.Output) (key tf.Output, value tf.Output)

Returns the next record (key, value pair) produced by a Reader.

Will dequeue from the input queue if necessary (e.g. when the Reader needs to start reading from a new file since it has finished with the previous file).

Arguments: 

reader_handle: Handle to a Reader.
queue_handle: Handle to a Queue, with string work items.

Returns: 

key: A scalar.
value: A scalar.

func ReaderResetV2 

func ReaderResetV2(scope *Scope, reader_handle tf.Output) (o *tf.Operation)

Restore a Reader to its initial clean state.

Arguments: 

reader_handle: Handle to a Reader.

Returns the created operation.

func ReaderRestoreStateV2 

func ReaderRestoreStateV2(scope *Scope, reader_handle tf.Output, state tf.Output) (o *tf.Operation)

Restore a reader to a previously saved state.

Not all Readers support being restored, so this can produce an Unimplemented error.

Arguments: 

reader_handle: Handle to a Reader.
state: Result of a ReaderSerializeState of a Reader with type

matching reader_handle.

Returns the created operation.

func ReaderSerializeStateV2 

func ReaderSerializeStateV2(scope *Scope, reader_handle tf.Output) (state tf.Output)

Produce a string tensor that encodes the state of a Reader.

Not all Readers support being serialized, so this can produce an Unimplemented error.

Arguments: 

reader_handle: Handle to a Reader.

func Real 

func Real(scope *Scope, input tf.Output, optional ...RealAttr) (output tf.Output)

Returns the real part of a complex number.

Given a tensor `input` of complex numbers, this operation returns a tensor of type `float` that is the real part of each element in `input`. All elements in `input` must be complex numbers of the form \\(a + bj\\), where *a* is the real 

part returned by this operation and *b* is the imaginary part.

For example:

``` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.real(input) ==> [-2.25, 3.25] ```

func RealDiv 

func RealDiv(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x / y element-wise for real types.

If `x` and `y` are reals, this will return the floating-point division.

*NOTE*: `Div` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func RebatchDataset 

func RebatchDataset(scope *Scope, input_dataset tf.Output, num_replicas tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...RebatchDatasetAttr) (handle tf.Output)

Creates a dataset that changes the batch size.

Creates a dataset that changes the batch size of the dataset to current batch size // num_workers.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
num_replicas: A scalar representing the number of replicas to distribute this batch across. As

a result of this transformation the current batch size would end up being divided by this parameter.

func RebatchDatasetV2 

func RebatchDatasetV2(scope *Scope, input_dataset tf.Output, batch_sizes tf.Output, drop_remainder tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that changes the batch size.

Creates a dataset that rebatches elements from `input_dataset` into new batch sizes.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
batch_sizes: A vector of integers representing the size of batches to produce. These values

are cycled through in order.

func Reciprocal 

func Reciprocal(scope *Scope, x tf.Output) (y tf.Output)

Computes the reciprocal of x element-wise.

I.e., \\(y = 1 / x\\).

func ReciprocalGrad 

func ReciprocalGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient for the inverse of `x` wrt its input.

Specifically, `grad = -dy * y*y`, where `y = 1/x`, and `dy` is the corresponding input gradient.

func RecordInput 

func RecordInput(scope *Scope, file_pattern string, optional ...RecordInputAttr) (records tf.Output)

Emits randomized records.

Arguments: 

file_pattern: Glob pattern for the data files.

Returns A tensor of shape [batch_size].

func Recv 

func Recv(scope *Scope, tensor_type tf.DataType, tensor_name string, send_device string, send_device_incarnation int64, recv_device string, optional ...RecvAttr) (tensor tf.Output)

Receives the named tensor from send_device on recv_device.

Arguments: 

tensor_name: The name of the tensor to receive.
send_device: The name of the device sending the tensor.
send_device_incarnation: The current incarnation of send_device.
recv_device: The name of the device receiving the tensor.

Returns The tensor to receive.

func RecvTPUEmbeddingActivations 

func RecvTPUEmbeddingActivations(scope *Scope, num_outputs int64, config string) (outputs []tf.Output)

An op that receives embedding activations on the TPU.

The TPU system performs the embedding lookups and aggregations specified by the arguments to TPUEmbeddingEnqueue(Integer/Sparse/SparseTensor)Batch. The results of these aggregations are visible to the Tensorflow Graph as the outputs of a RecvTPUEmbeddingActivations op. This op returns a list containing one Tensor of activations per table specified in the model. There can be at most one RecvTPUEmbeddingActivations op in the TPU graph.

Arguments: 

num_outputs: The number of output activation tensors, equal to the number of

embedding tables in the model. 

config: Serialized TPUEmbeddingConfiguration proto.

Returns A TensorList of embedding activations containing one Tensor per embedding table in the model.

func ReduceJoin 

func ReduceJoin(scope *Scope, inputs tf.Output, reduction_indices tf.Output, optional ...ReduceJoinAttr) (output tf.Output)

Joins a string Tensor across the given dimensions.

Computes the string join across dimensions in the given string Tensor of shape `[\\(d_0, d_1, ..., d_{n-1}\\)]`. Returns a new Tensor created by joining the input strings with the given separator (default: empty string). Negative indices are counted backwards from the end, with `-1` being equivalent to `n - 1`. If indices are not specified, joins across all dimensions beginning from `n - 1` through `0`.

For example:

```python # tensor `a` is [["a", "b"], ["c", "d"]] tf.reduce_join(a, 0) ==> ["ac", "bd"] tf.reduce_join(a, 1) ==> ["ab", "cd"] tf.reduce_join(a, -2) = tf.reduce_join(a, 0) ==> ["ac", "bd"] tf.reduce_join(a, -1) = tf.reduce_join(a, 1) ==> ["ab", "cd"] tf.reduce_join(a, 0, keep_dims=True) ==> [["ac", "bd"]] tf.reduce_join(a, 1, keep_dims=True) ==> [["ab"], ["cd"]] tf.reduce_join(a, 0, separator=".") ==> ["a.c", "b.d"] tf.reduce_join(a, [0, 1]) ==> "acbd" tf.reduce_join(a, [1, 0]) ==> "abcd" tf.reduce_join(a, []) ==> [["a", "b"], ["c", "d"]] tf.reduce_join(a) = tf.reduce_join(a, [1, 0]) ==> "abcd" ```

Arguments: 

inputs: The input to be joined.  All reduced indices must have non-zero size.
reduction_indices: The dimensions to reduce over.  Dimensions are reduced in the

order specified. Omitting `reduction_indices` is equivalent to passing `[n-1, n-2, ..., 0]`. Negative indices from `-n` to `-1` are supported.

Returns Has shape equal to that of the input with reduced dimensions removed or set to `1` depending on `keep_dims`.

func RegexFullMatch 

func RegexFullMatch(scope *Scope, input tf.Output, pattern tf.Output) (output tf.Output)

Check if the input matches the regex pattern.

The input is a string tensor of any shape. The pattern is a scalar string tensor which is applied to every element of the input tensor. The boolean values (True or False) of the output tensor indicate if the input matches the regex pattern provided.

The pattern follows the re2 syntax (https://github.com/google/re2/wiki/Syntax)

Examples:

>>> tf.strings.regex_full_match(["TF lib", "lib TF"], ".*lib$") <tf.Tensor: shape=(2,), dtype=bool, numpy=array([ True, False])> >>> tf.strings.regex_full_match(["TF lib", "lib TF"], ".*TF$") <tf.Tensor: shape=(2,), dtype=bool, numpy=array([False, True])>

Arguments: 

input: A string tensor of the text to be processed.
pattern: A scalar string tensor containing the regular expression to match the input.

Returns A bool tensor with the same shape as `input`.

func RegexReplace 

func RegexReplace(scope *Scope, input tf.Output, pattern tf.Output, rewrite tf.Output, optional ...RegexReplaceAttr) (output tf.Output)

Replaces matches of the `pattern` regular expression in `input` with the replacement string provided in `rewrite`.

It follows the re2 syntax (https://github.com/google/re2/wiki/Syntax)

Arguments: 

input: The text to be processed.
pattern: The regular expression to be matched in the `input` strings.
rewrite: The rewrite string to be substituted for the `pattern` expression where it is

matched in the `input` strings.

Returns The text after applying pattern match and rewrite substitution.

func RegisterDataset 

func RegisterDataset(scope *Scope, dataset tf.Output, address tf.Output, protocol tf.Output, external_state_policy int64, optional ...RegisterDatasetAttr) (dataset_id tf.Output)

Registers a dataset with the tf.data service.

func RegisterDatasetV2 added in v0.2.0 

func RegisterDatasetV2(scope *Scope, dataset tf.Output, address tf.Output, protocol tf.Output, external_state_policy int64, optional ...RegisterDatasetV2Attr) (dataset_id tf.Output)

Registers a dataset with the tf.data service.

func Relu 

func Relu(scope *Scope, features tf.Output) (activations tf.Output)

Computes rectified linear: `max(features, 0)`.

See: https://en.wikipedia.org/wiki/Rectifier_(neural_networks) Example usage: >>> tf.nn.relu([-2., 0., 3.]).numpy() array([0., 0., 3.], dtype=float32)

func Relu6 

func Relu6(scope *Scope, features tf.Output) (activations tf.Output)

Computes rectified linear 6: `min(max(features, 0), 6)`.

func Relu6Grad 

func Relu6Grad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)

Computes rectified linear 6 gradients for a Relu6 operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding Relu6 operation.
features: The features passed as input to the corresponding Relu6 operation, or

its output; using either one produces the same result.

Returns The gradients: `gradients * (features > 0) * (features < 6)`.

func ReluGrad 

func ReluGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)

Computes rectified linear gradients for a Relu operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding Relu operation.
features: The features passed as input to the corresponding Relu operation, OR

the outputs of that operation (both work equivalently).

Returns `gradients * (features > 0)`.

func RepeatDataset 

func RepeatDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...RepeatDatasetAttr) (handle tf.Output)

Creates a dataset that emits the outputs of `input_dataset` `count` times.

Arguments: 

count: A scalar representing the number of times that `input_dataset` should

be repeated. A value of `-1` indicates that it should be repeated infinitely.

func RequantizationRange 

func RequantizationRange(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output) (output_min tf.Output, output_max tf.Output)

Computes a range that covers the actual values present in a quantized tensor.

Given a quantized tensor described by `(input, input_min, input_max)`, outputs a range that covers the actual values present in that tensor. This op is typically used to produce the `requested_output_min` and `requested_output_max` for `Requantize`.

Arguments: 

input_min: The float value that the minimum quantized input value represents.
input_max: The float value that the maximum quantized input value represents.

Returns: 

output_min: The computed min output.
output_max: the computed max output.

func RequantizationRangePerChannel 

func RequantizationRangePerChannel(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, clip_value_max float32) (output_min tf.Output, output_max tf.Output)

Computes requantization range per channel.

Arguments: 

input: The original input tensor.
input_min: The minimum value of the input tensor
input_max: The maximum value of the input tensor.
clip_value_max: The maximum value of the output that needs to be clipped.

Example: set this to 6 for Relu6.

Returns: 

output_min: The minimum value of the final output tensor
output_max: The maximum value of the final output tensor.

func Requantize 

func Requantize(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, requested_output_min tf.Output, requested_output_max tf.Output, out_type tf.DataType) (output tf.Output, output_min tf.Output, output_max tf.Output)

Converts the quantized `input` tensor into a lower-precision `output`.

Converts the quantized `input` tensor into a lower-precision `output`, using the output range specified with `requested_output_min` and `requested_output_max`.

`[input_min, input_max]` are scalar floats that specify the range for the float interpretation of the `input` data. For example, if `input_min` is -1.0f and `input_max` is 1.0f, and we are dealing with `quint16` quantized data, then a 0 value in the 16-bit data should be interpreted as -1.0f, and a 65535 means 1.0f.

Arguments: 

input_min: The float value that the minimum quantized input value represents.
input_max: The float value that the maximum quantized input value represents.
requested_output_min: The float value that the minimum quantized output value represents.
requested_output_max: The float value that the maximum quantized output value represents.
out_type: The type of the output. Should be a lower bit depth than Tinput.

Returns: 

output
output_min: The requested_output_min value is copied into this output.
output_max: The requested_output_max value is copied into this output.

func RequantizePerChannel 

func RequantizePerChannel(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, requested_output_min tf.Output, requested_output_max tf.Output, optional ...RequantizePerChannelAttr) (output tf.Output, output_min tf.Output, output_max tf.Output)

Requantizes input with min and max values known per channel.

Arguments: 

input: The original input tensor.
input_min: The minimum value of the input tensor
input_max: The maximum value of the input tensor.
requested_output_min: The minimum value of the output tensor requested.
requested_output_max: The maximum value of the output tensor requested.

Returns: 

output: Output tensor.
output_min: The minimum value of the final output tensor
output_max: The maximum value of the final output tensor.

func Reshape 

func Reshape(scope *Scope, tensor tf.Output, shape tf.Output) (output tf.Output)

Reshapes a tensor.

Given `tensor`, this operation returns a tensor that has the same values as `tensor` with shape `shape`.

If one component of 1-D tensor `shape` is the special value -1, the size of that dimension is computed so that the total size remains constant. In particular, a `shape` of `[-1]` flattens into 1-D. At most one component of `shape` may be unknown.

The `shape` must be 1-D and the operation returns a tensor with shape `shape` filled with the values of `tensor`. In this case, the number of elements implied by `shape` must be the same as the number of elements in `tensor`.

It is an error if `shape` is not 1-D.

For example:

``` # tensor 't' is [1, 2, 3, 4, 5, 6, 7, 8, 9] # tensor 't' has shape [9] reshape(t, [3, 3]) ==> [[1, 2, 3], 

[4, 5, 6],
[7, 8, 9]]

# tensor 't' is [[[1, 1], [2, 2]], # [[3, 3], [4, 4]]] # tensor 't' has shape [2, 2, 2] reshape(t, [2, 4]) ==> [[1, 1, 2, 2], 

[3, 3, 4, 4]]

# tensor 't' is [[[1, 1, 1], # [2, 2, 2]], # [[3, 3, 3], # [4, 4, 4]], # [[5, 5, 5], # [6, 6, 6]]] # tensor 't' has shape [3, 2, 3] # pass '[-1]' to flatten 't' reshape(t, [-1]) ==> [1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6]

-1 can also be used to infer the shape

# -1 is inferred to be 9: reshape(t, [2, -1]) ==> [[1, 1, 1, 2, 2, 2, 3, 3, 3], 

[4, 4, 4, 5, 5, 5, 6, 6, 6]]

# -1 is inferred to be 2: reshape(t, [-1, 9]) ==> [[1, 1, 1, 2, 2, 2, 3, 3, 3], 

[4, 4, 4, 5, 5, 5, 6, 6, 6]]

# -1 is inferred to be 3: reshape(t, [ 2, -1, 3]) ==> [[[1, 1, 1], 

 [2, 2, 2],
 [3, 3, 3]],
[[4, 4, 4],
 [5, 5, 5],
 [6, 6, 6]]]

# tensor 't' is [7] # shape `[]` reshapes to a scalar reshape(t, []) ==> 7 ```

Arguments: 

shape: Defines the shape of the output tensor.

func ResizeArea 

func ResizeArea(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeAreaAttr) (resized_images tf.Output)

Resize `images` to `size` using area interpolation.

Input images can be of different types but output images are always float.

The range of pixel values for the output image might be slightly different from the range for the input image because of limited numerical precision. To guarantee an output range, for example `[0.0, 1.0]`, apply `tf.clip_by_value` to the output.

Each output pixel is computed by first transforming the pixel's footprint into the input tensor and then averaging the pixels that intersect the footprint. An input pixel's contribution to the average is weighted by the fraction of its area that intersects the footprint. This is the same as OpenCV's INTER_AREA.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images.

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ResizeBicubic 

func ResizeBicubic(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeBicubicAttr) (resized_images tf.Output)

Resize `images` to `size` using bicubic interpolation.

Input images can be of different types but output images are always float.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images.

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ResizeBicubicGrad 

func ResizeBicubicGrad(scope *Scope, grads tf.Output, original_image tf.Output, optional ...ResizeBicubicGradAttr) (output tf.Output)

Computes the gradient of bicubic interpolation.

Arguments: 

grads: 4-D with shape `[batch, height, width, channels]`.
original_image: 4-D with shape `[batch, orig_height, orig_width, channels]`,

The image tensor that was resized.

Returns 4-D with shape `[batch, orig_height, orig_width, channels]`. Gradients with respect to the input image. Input image must have been float or double.

func ResizeBilinear 

func ResizeBilinear(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeBilinearAttr) (resized_images tf.Output)

Resize `images` to `size` using bilinear interpolation.

Input images can be of different types but output images are always float.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images.

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ResizeBilinearGrad 

func ResizeBilinearGrad(scope *Scope, grads tf.Output, original_image tf.Output, optional ...ResizeBilinearGradAttr) (output tf.Output)

Computes the gradient of bilinear interpolation.

Arguments: 

grads: 4-D with shape `[batch, height, width, channels]`.
original_image: 4-D with shape `[batch, orig_height, orig_width, channels]`,

The image tensor that was resized.

Returns 4-D with shape `[batch, orig_height, orig_width, channels]`. Gradients with respect to the input image. Input image must have been float or double.

func ResizeNearestNeighbor 

func ResizeNearestNeighbor(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeNearestNeighborAttr) (resized_images tf.Output)

Resize `images` to `size` using nearest neighbor interpolation.

Arguments: 

images: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`.  The

new size for the images.

Returns 4-D with shape `[batch, new_height, new_width, channels]`.

func ResizeNearestNeighborGrad 

func ResizeNearestNeighborGrad(scope *Scope, grads tf.Output, size tf.Output, optional ...ResizeNearestNeighborGradAttr) (output tf.Output)

Computes the gradient of nearest neighbor interpolation.

Arguments: 

grads: 4-D with shape `[batch, height, width, channels]`.
size: = A 1-D int32 Tensor of 2 elements: `orig_height, orig_width`. The

original input size.

Returns 4-D with shape `[batch, orig_height, orig_width, channels]`. Gradients with respect to the input image.

func ResourceAccumulatorApplyGradient 

func ResourceAccumulatorApplyGradient(scope *Scope, handle tf.Output, local_step tf.Output, gradient tf.Output) (o *tf.Operation)

Applies a gradient to a given accumulator.

Does not add if local_step is lesser than the accumulator's global_step.

Arguments: 

handle: The handle to a accumulator.
local_step: The local_step value at which the gradient was computed.
gradient: A tensor of the gradient to be accumulated.

Returns the created operation.

func ResourceAccumulatorNumAccumulated 

func ResourceAccumulatorNumAccumulated(scope *Scope, handle tf.Output) (num_accumulated tf.Output)

Returns the number of gradients aggregated in the given accumulators.

Arguments: 

handle: The handle to an accumulator.

Returns The number of gradients aggregated in the given accumulator.

func ResourceAccumulatorSetGlobalStep 

func ResourceAccumulatorSetGlobalStep(scope *Scope, handle tf.Output, new_global_step tf.Output) (o *tf.Operation)

Updates the accumulator with a new value for global_step.

Logs warning if the accumulator's value is already higher than new_global_step.

Arguments: 

handle: The handle to an accumulator.
new_global_step: The new global_step value to set.

Returns the created operation.

func ResourceAccumulatorTakeGradient 

func ResourceAccumulatorTakeGradient(scope *Scope, handle tf.Output, num_required tf.Output, dtype tf.DataType) (average tf.Output)

Extracts the average gradient in the given ConditionalAccumulator.

The op blocks until sufficient (i.e., more than num_required) gradients have been accumulated. If the accumulator has already aggregated more than num_required gradients, it returns the average of the accumulated gradients. Also automatically increments the recorded global_step in the accumulator by 1, and resets the aggregate to 0.

Arguments: 

handle: The handle to an accumulator.
num_required: Number of gradients required before we return an aggregate.
dtype: The data type of accumulated gradients. Needs to correspond to the type

of the accumulator.

Returns The average of the accumulated gradients.

func ResourceApplyAdaMax 

func ResourceApplyAdaMax(scope *Scope, var_ tf.Output, m tf.Output, v tf.Output, beta1_power tf.Output, lr tf.Output, beta1 tf.Output, beta2 tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdaMaxAttr) (o *tf.Operation)

Update '*var' according to the AdaMax algorithm.

m_t <- beta1 * m_{t-1} + (1 - beta1) * g v_t <- max(beta2 * v_{t-1}, abs(g)) variable <- variable - learning_rate / (1 - beta1^t) * m_t / (v_t + epsilon)

Arguments: 

var_: Should be from a Variable().
m: Should be from a Variable().
v: Should be from a Variable().
beta1_power: Must be a scalar.
lr: Scaling factor. Must be a scalar.
beta1: Momentum factor. Must be a scalar.
beta2: Momentum factor. Must be a scalar.
epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAdadelta 

func ResourceApplyAdadelta(scope *Scope, var_ tf.Output, accum tf.Output, accum_update tf.Output, lr tf.Output, rho tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdadeltaAttr) (o *tf.Operation)

Update '*var' according to the adadelta scheme.

accum = rho() * accum + (1 - rho()) * grad.square(); update = (update_accum + epsilon).sqrt() * (accum + epsilon()).rsqrt() * grad; update_accum = rho() * update_accum + (1 - rho()) * update.square(); var -= update;

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
accum_update: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
rho: Decay factor. Must be a scalar.
epsilon: Constant factor. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAdagrad 

func ResourceApplyAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, optional ...ResourceApplyAdagradAttr) (o *tf.Operation)

Update '*var' according to the adagrad scheme.

accum += grad * grad var -= lr * grad * (1 / sqrt(accum))

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAdagradDA 

func ResourceApplyAdagradDA(scope *Scope, var_ tf.Output, gradient_accumulator tf.Output, gradient_squared_accumulator tf.Output, grad tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, global_step tf.Output, optional ...ResourceApplyAdagradDAAttr) (o *tf.Operation)

Update '*var' according to the proximal adagrad scheme.

Arguments: 

var_: Should be from a Variable().
gradient_accumulator: Should be from a Variable().
gradient_squared_accumulator: Should be from a Variable().
grad: The gradient.
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
global_step: Training step number. Must be a scalar.

Returns the created operation.

func ResourceApplyAdagradV2 

func ResourceApplyAdagradV2(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdagradV2Attr) (o *tf.Operation)

Update '*var' according to the adagrad scheme.

accum += grad * grad var -= lr * grad * (1 / (sqrt(accum) + epsilon))

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
epsilon: Constant factor. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAdam 

func ResourceApplyAdam(scope *Scope, var_ tf.Output, m tf.Output, v tf.Output, beta1_power tf.Output, beta2_power tf.Output, lr tf.Output, beta1 tf.Output, beta2 tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdamAttr) (o *tf.Operation)

Update '*var' according to the Adam algorithm.

$$\text{lr}_t := \mathrm{lr} \cdot \frac{\sqrt{1 - \beta_2^t}}{1 - \beta_1^t}$$ $$m_t := \beta_1 \cdot m_{t-1} + (1 - \beta_1) \cdot g$$ $$v_t := \beta_2 \cdot v_{t-1} + (1 - \beta_2) \cdot g^2$$ $$\text{var} := \begin{cases} \text{var} - (m_t \beta_1 + g \cdot (1 - \beta_1))\cdot\text{lr}_t/(\sqrt{v_t} + \epsilon), &\text{if use_nesterov}\\\\ \text{var} - m_t \cdot \text{lr}_t /(\sqrt{v_t} + \epsilon), &\text{otherwise} \end{cases}$$

Arguments: 

var_: Should be from a Variable().
m: Should be from a Variable().
v: Should be from a Variable().
beta1_power: Must be a scalar.
beta2_power: Must be a scalar.
lr: Scaling factor. Must be a scalar.
beta1: Momentum factor. Must be a scalar.
beta2: Momentum factor. Must be a scalar.
epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAdamWithAmsgrad 

func ResourceApplyAdamWithAmsgrad(scope *Scope, var_ tf.Output, m tf.Output, v tf.Output, vhat tf.Output, beta1_power tf.Output, beta2_power tf.Output, lr tf.Output, beta1 tf.Output, beta2 tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdamWithAmsgradAttr) (o *tf.Operation)

Update '*var' according to the Adam algorithm.

$$\text{lr}_t := \mathrm{learning_rate} * \sqrt{1 - \beta_2^t} / (1 - \beta_1^t)$$ $$m_t := \beta_1 * m_{t-1} + (1 - \beta_1) * g$$ $$v_t := \beta_2 * v_{t-1} + (1 - \beta_2) * g * g$$ $$\hat{v}_t := max{\hat{v}_{t-1}, v_t}$$ $$\text{variable} := \text{variable} - \text{lr}_t * m_t / (\sqrt{\hat{v}_t} + \epsilon)$$

Arguments: 

var_: Should be from a Variable().
m: Should be from a Variable().
v: Should be from a Variable().
vhat: Should be from a Variable().
beta1_power: Must be a scalar.
beta2_power: Must be a scalar.
lr: Scaling factor. Must be a scalar.
beta1: Momentum factor. Must be a scalar.
beta2: Momentum factor. Must be a scalar.
epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyAddSign 

func ResourceApplyAddSign(scope *Scope, var_ tf.Output, m tf.Output, lr tf.Output, alpha tf.Output, sign_decay tf.Output, beta tf.Output, grad tf.Output, optional ...ResourceApplyAddSignAttr) (o *tf.Operation)

Update '*var' according to the AddSign update.

m_t <- beta1 * m_{t-1} + (1 - beta1) * g update <- (alpha + sign_decay * sign(g) *sign(m)) * g variable <- variable - lr_t * update

Arguments: 

var_: Should be from a Variable().
m: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
alpha: Must be a scalar.
sign_decay: Must be a scalar.
beta: Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyCenteredRMSProp 

func ResourceApplyCenteredRMSProp(scope *Scope, var_ tf.Output, mg tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyCenteredRMSPropAttr) (o *tf.Operation)

Update '*var' according to the centered RMSProp algorithm.

The centered RMSProp algorithm uses an estimate of the centered second moment (i.e., the variance) for normalization, as opposed to regular RMSProp, which uses the (uncentered) second moment. This often helps with training, but is slightly more expensive in terms of computation and memory.

Note that in dense implementation of this algorithm, mg, ms, and mom will update even if the grad is zero, but in this sparse implementation, mg, ms, and mom will not update in iterations during which the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2 mean_grad = decay * mean_grad + (1-decay) * gradient

Delta = learning_rate * gradient / sqrt(mean_square + epsilon - mean_grad ** 2)

mg <- rho * mg_{t-1} + (1-rho) * grad ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms - mg * mg + epsilon) var <- var - mom

Arguments: 

var_: Should be from a Variable().
mg: Should be from a Variable().
ms: Should be from a Variable().
mom: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
rho: Decay rate. Must be a scalar.
momentum: Momentum Scale. Must be a scalar.
epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyFtrl 

func ResourceApplyFtrl(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, lr_power tf.Output, optional ...ResourceApplyFtrlAttr) (o *tf.Operation)

Update '*var' according to the Ftrl-proximal scheme.

accum_new = accum + grad * grad linear += grad - (accum_new^(-lr_power) - accum^(-lr_power)) / lr * var quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
linear: Should be from a Variable().
grad: The gradient.
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceApplyFtrlV2 

func ResourceApplyFtrlV2(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, l2_shrinkage tf.Output, lr_power tf.Output, optional ...ResourceApplyFtrlV2Attr) (o *tf.Operation)

Update '*var' according to the Ftrl-proximal scheme.

accum_new = accum + grad * grad grad_with_shrinkage = grad + 2 * l2_shrinkage * var linear += grad_with_shrinkage + 

(accum_new^(-lr_power) - accum^(-lr_power)) / lr * var

quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
linear: Should be from a Variable().
grad: The gradient.
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 shrinkage regularization. Must be a scalar.

lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceApplyGradientDescent 

func ResourceApplyGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, delta tf.Output, optional ...ResourceApplyGradientDescentAttr) (o *tf.Operation)

Update '*var' by subtracting 'alpha' * 'delta' from it.

Arguments: 

var_: Should be from a Variable().
alpha: Scaling factor. Must be a scalar.
delta: The change.

Returns the created operation.

func ResourceApplyKerasMomentum 

func ResourceApplyKerasMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, momentum tf.Output, optional ...ResourceApplyKerasMomentumAttr) (o *tf.Operation)

Update '*var' according to the momentum scheme.

Set use_nesterov = True if you want to use Nesterov momentum.

accum = accum * momentum - lr * grad var += accum

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
grad: The gradient.
momentum: Momentum. Must be a scalar.

Returns the created operation.

func ResourceApplyMomentum 

func ResourceApplyMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, momentum tf.Output, optional ...ResourceApplyMomentumAttr) (o *tf.Operation)

Update '*var' according to the momentum scheme.

Set use_nesterov = True if you want to use Nesterov momentum.

accum = accum * momentum + grad var -= lr * accum

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
grad: The gradient.
momentum: Momentum. Must be a scalar.

Returns the created operation.

func ResourceApplyPowerSign 

func ResourceApplyPowerSign(scope *Scope, var_ tf.Output, m tf.Output, lr tf.Output, logbase tf.Output, sign_decay tf.Output, beta tf.Output, grad tf.Output, optional ...ResourceApplyPowerSignAttr) (o *tf.Operation)

Update '*var' according to the AddSign update.

m_t <- beta1 * m_{t-1} + (1 - beta1) * g update <- exp(logbase * sign_decay * sign(g) * sign(m_t)) * g variable <- variable - lr_t * update

Arguments: 

var_: Should be from a Variable().
m: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
logbase: Must be a scalar.
sign_decay: Must be a scalar.
beta: Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyProximalAdagrad 

func ResourceApplyProximalAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, optional ...ResourceApplyProximalAdagradAttr) (o *tf.Operation)

Update '*var' and '*accum' according to FOBOS with Adagrad learning rate.

accum += grad * grad prox_v = var - lr * grad * (1 / sqrt(accum)) var = sign(prox_v)/(1+lr*l2) * max{|prox_v|-lr*l1,0}

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyProximalGradientDescent 

func ResourceApplyProximalGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, l1 tf.Output, l2 tf.Output, delta tf.Output, optional ...ResourceApplyProximalGradientDescentAttr) (o *tf.Operation)

Update '*var' as FOBOS algorithm with fixed learning rate.

prox_v = var - alpha * delta var = sign(prox_v)/(1+alpha*l2) * max{|prox_v|-alpha*l1,0}

Arguments: 

var_: Should be from a Variable().
alpha: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
delta: The change.

Returns the created operation.

func ResourceApplyRMSProp 

func ResourceApplyRMSProp(scope *Scope, var_ tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyRMSPropAttr) (o *tf.Operation)

Update '*var' according to the RMSProp algorithm.

Note that in dense implementation of this algorithm, ms and mom will update even if the grad is zero, but in this sparse implementation, ms and mom will not update in iterations during which the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2 Delta = learning_rate * gradient / sqrt(mean_square + epsilon)

ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom

Arguments: 

var_: Should be from a Variable().
ms: Should be from a Variable().
mom: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
rho: Decay rate. Must be a scalar.

epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceConditionalAccumulator 

func ResourceConditionalAccumulator(scope *Scope, dtype tf.DataType, shape tf.Shape, optional ...ResourceConditionalAccumulatorAttr) (handle tf.Output)

A conditional accumulator for aggregating gradients.

The accumulator accepts gradients marked with local_step greater or equal to the most recent global_step known to the accumulator. The average can be extracted from the accumulator, provided sufficient gradients have been accumulated. Extracting the average automatically resets the aggregate to 0, and increments the global_step recorded by the accumulator. This is a resource version of ConditionalAccumulator that will work in TF2.0 with tf.cond version 2.

Arguments: 

dtype: The type of the value being accumulated.
shape: The shape of the values, can be [], in which case shape is unknown.

Returns The handle to the accumulator.

func ResourceCountUpTo 

func ResourceCountUpTo(scope *Scope, resource tf.Output, limit int64, T tf.DataType) (output tf.Output)

Increments variable pointed to by 'resource' until it reaches 'limit'.

Arguments: 

resource: Should be from a scalar `Variable` node.
limit: If incrementing ref would bring it above limit, instead generates an

'OutOfRange' error.

Returns A copy of the input before increment. If nothing else modifies the input, the values produced will all be distinct.

func ResourceGather 

func ResourceGather(scope *Scope, resource tf.Output, indices tf.Output, dtype tf.DataType, optional ...ResourceGatherAttr) (output tf.Output)

Gather slices from the variable pointed to by `resource` according to `indices`.

`indices` must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape `indices.shape + params.shape[1:]` where:

```python 

# Scalar indices
output[:, ..., :] = params[indices, :, ... :]

# Vector indices
output[i, :, ..., :] = params[indices[i], :, ... :]

# Higher rank indices
output[i, ..., j, :, ... :] = params[indices[i, ..., j], :, ..., :]

```

func ResourceScatterAdd 

func ResourceScatterAdd(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Adds sparse updates to the variable referenced by `resource`.

This operation computes 

# Scalar indices
ref[indices, ...] += updates[...]

# Vector indices (for each i)
ref[indices[i], ...] += updates[i, ...]

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] += updates[i, ..., j, ...]

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions add.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterDiv 

func ResourceScatterDiv(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Divides sparse updates into the variable referenced by `resource`.

This operation computes 

# Scalar indices
ref[indices, ...] /= updates[...]

# Vector indices (for each i)
ref[indices[i], ...] /= updates[i, ...]

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] /= updates[i, ..., j, ...]

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions multiply.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterMax 

func ResourceScatterMax(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Reduces sparse updates into the variable referenced by `resource` using the `max` operation.

This operation computes 

# Scalar indices
ref[indices, ...] = max(ref[indices, ...], updates[...])

# Vector indices (for each i)
ref[indices[i], ...] = max(ref[indices[i], ...], updates[i, ...])

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] = max(ref[indices[i, ..., j], ...], updates[i, ..., j, ...])

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions are combined.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterMin 

func ResourceScatterMin(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Reduces sparse updates into the variable referenced by `resource` using the `min` operation.

This operation computes 

# Scalar indices
ref[indices, ...] = min(ref[indices, ...], updates[...])

# Vector indices (for each i)
ref[indices[i], ...] = min(ref[indices[i], ...], updates[i, ...])

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] = min(ref[indices[i, ..., j], ...], updates[i, ..., j, ...])

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions are combined.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterMul 

func ResourceScatterMul(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Multiplies sparse updates into the variable referenced by `resource`.

This operation computes 

# Scalar indices
ref[indices, ...] *= updates[...]

# Vector indices (for each i)
ref[indices[i], ...] *= updates[i, ...]

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] *= updates[i, ..., j, ...]

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions multiply.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterNdAdd 

func ResourceScatterNdAdd(scope *Scope, ref tf.Output, indices tf.Output, updates tf.Output, optional ...ResourceScatterNdAddAttr) (o *tf.Operation)

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

`ref` is a `Tensor` with rank `P` and `indices` is a `Tensor` of rank `Q`.

`indices` must be integer tensor, containing indices into `ref`. 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 `K`th dimension of `ref`.

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

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

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

```python ref = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8], use_resource=True) indices = tf.constant([[4], [3], [1], [7]]) updates = tf.constant([9, 10, 11, 12]) add = tf.scatter_nd_add(ref, indices, updates) with tf.Session() as sess: 

print sess.run(add)

```

The resulting update to ref 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.

Arguments: 

ref: A resource handle. Must be from a VarHandleOp.
indices: A Tensor. Must be one of the following types: int32, int64.

A tensor of indices into ref. 

updates: A Tensor. Must have the same type as ref. A tensor of

values to add to ref.

Returns the created operation.

func ResourceScatterNdSub 

func ResourceScatterNdSub(scope *Scope, ref tf.Output, indices tf.Output, updates tf.Output, optional ...ResourceScatterNdSubAttr) (o *tf.Operation)

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

`ref` is a `Tensor` with rank `P` and `indices` is a `Tensor` of rank `Q`.

`indices` must be integer tensor, containing indices into `ref`. 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 `K`th dimension of `ref`.

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

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

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

```python ref = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8], use_resource=True) indices = tf.constant([[4], [3], [1], [7]]) updates = tf.constant([9, 10, 11, 12]) sub = tf.scatter_nd_sub(ref, indices, updates) with tf.Session() as sess: 

print sess.run(sub)

```

The resulting update to ref would look like this: 

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

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

Arguments: 

ref: A resource handle. Must be from a VarHandleOp.
indices: A Tensor. Must be one of the following types: int32, int64.

A tensor of indices into ref. 

updates: A Tensor. Must have the same type as ref. A tensor of

values to add to ref.

Returns the created operation.

func ResourceScatterNdUpdate 

func ResourceScatterNdUpdate(scope *Scope, ref tf.Output, indices tf.Output, updates tf.Output, optional ...ResourceScatterNdUpdateAttr) (o *tf.Operation)

Applies sparse `updates` to individual values or slices within a given

variable according to `indices`.

`ref` is a `Tensor` with rank `P` and `indices` is a `Tensor` of rank `Q`.

`indices` must be integer tensor, containing indices into `ref`. 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 `K`th dimension of `ref`.

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

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

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

```python 

ref = tf.Variable([1, 2, 3, 4, 5, 6, 7, 8])
indices = tf.constant([[4], [3], [1] ,[7]])
updates = tf.constant([9, 10, 11, 12])
update = tf.scatter_nd_update(ref, indices, updates)
with tf.Session() as sess:
  print sess.run(update)

```

The resulting update to ref 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.

Arguments: 

ref: A resource handle. Must be from a VarHandleOp.
indices: A Tensor. Must be one of the following types: int32, int64.

A tensor of indices into ref. 

updates: A Tensor. Must have the same type as ref. A tensor of updated

values to add to ref.

Returns the created operation.

func ResourceScatterSub 

func ResourceScatterSub(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Subtracts sparse updates from the variable referenced by `resource`.

This operation computes 

# Scalar indices
ref[indices, ...] -= updates[...]

# Vector indices (for each i)
ref[indices[i], ...] -= updates[i, ...]

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] -= updates[i, ..., j, ...]

Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions add.

Requires `updates.shape = indices.shape + ref.shape[1:]` or `updates.shape = []`.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src='https://www.tensorflow.org/images/ScatterAdd.png' alt> </div>

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceScatterUpdate 

func ResourceScatterUpdate(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)

Assigns sparse updates to the variable referenced by `resource`.

This operation computes 

# Scalar indices
ref[indices, ...] = updates[...]

# Vector indices (for each i)
ref[indices[i], ...] = updates[i, ...]

# High rank indices (for each i, ..., j)
ref[indices[i, ..., j], ...] = updates[i, ..., j, ...]

Arguments: 

resource: Should be from a `Variable` node.
indices: A tensor of indices into the first dimension of `ref`.
updates: A tensor of updated values to add to `ref`.

Returns the created operation.

func ResourceSparseApplyAdadelta 

func ResourceSparseApplyAdadelta(scope *Scope, var_ tf.Output, accum tf.Output, accum_update tf.Output, lr tf.Output, rho tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyAdadeltaAttr) (o *tf.Operation)

var: Should be from a Variable().

Arguments: 

accum: Should be from a Variable().
accum_update: : Should be from a Variable().
lr: Learning rate. Must be a scalar.
rho: Decay factor. Must be a scalar.
epsilon: Constant factor. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.

Returns the created operation.

func ResourceSparseApplyAdagrad 

func ResourceSparseApplyAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyAdagradAttr) (o *tf.Operation)

Update relevant entries in '*var' and '*accum' according to the adagrad scheme.

That is for rows we have grad for, we update var and accum as follows: accum += grad * grad var -= lr * grad * (1 / sqrt(accum))

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Learning rate. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.

Returns the created operation.

func ResourceSparseApplyAdagradDA 

func ResourceSparseApplyAdagradDA(scope *Scope, var_ tf.Output, gradient_accumulator tf.Output, gradient_squared_accumulator tf.Output, grad tf.Output, indices tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, global_step tf.Output, optional ...ResourceSparseApplyAdagradDAAttr) (o *tf.Operation)

Update entries in '*var' and '*accum' according to the proximal adagrad scheme.

Arguments: 

var_: Should be from a Variable().
gradient_accumulator: Should be from a Variable().
gradient_squared_accumulator: Should be from a Variable().
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.
lr: Learning rate. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
global_step: Training step number. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyAdagradV2 

func ResourceSparseApplyAdagradV2(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyAdagradV2Attr) (o *tf.Operation)

Update relevant entries in '*var' and '*accum' according to the adagrad scheme.

That is for rows we have grad for, we update var and accum as follows: accum += grad * grad var -= lr * grad * (1 / sqrt(accum))

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Learning rate. Must be a scalar.
epsilon: Constant factor. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.

Returns the created operation.

func ResourceSparseApplyCenteredRMSProp 

func ResourceSparseApplyCenteredRMSProp(scope *Scope, var_ tf.Output, mg tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyCenteredRMSPropAttr) (o *tf.Operation)

Update '*var' according to the centered RMSProp algorithm.

The centered RMSProp algorithm uses an estimate of the centered second moment (i.e., the variance) for normalization, as opposed to regular RMSProp, which uses the (uncentered) second moment. This often helps with training, but is slightly more expensive in terms of computation and memory.

Note that in dense implementation of this algorithm, mg, ms, and mom will update even if the grad is zero, but in this sparse implementation, mg, ms, and mom will not update in iterations during which the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2 mean_grad = decay * mean_grad + (1-decay) * gradient Delta = learning_rate * gradient / sqrt(mean_square + epsilon - mean_grad ** 2)

ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom

Arguments: 

var_: Should be from a Variable().
mg: Should be from a Variable().
ms: Should be from a Variable().
mom: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
rho: Decay rate. Must be a scalar.

epsilon: Ridge term. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var, ms and mom.

Returns the created operation.

func ResourceSparseApplyFtrl 

func ResourceSparseApplyFtrl(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, indices tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, lr_power tf.Output, optional ...ResourceSparseApplyFtrlAttr) (o *tf.Operation)

Update relevant entries in '*var' according to the Ftrl-proximal scheme.

That is for rows we have grad for, we update var, accum and linear as follows: accum_new = accum + grad * grad linear += grad - (accum_new^(-lr_power) - accum^(-lr_power)) / lr * var quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
linear: Should be from a Variable().
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyFtrlV2 

func ResourceSparseApplyFtrlV2(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, indices tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, l2_shrinkage tf.Output, lr_power tf.Output, optional ...ResourceSparseApplyFtrlV2Attr) (o *tf.Operation)

Update relevant entries in '*var' according to the Ftrl-proximal scheme.

That is for rows we have grad for, we update var, accum and linear as follows: grad_with_shrinkage = grad + 2 * l2_shrinkage * var accum_new = accum + grad_with_shrinkage * grad_with_shrinkage linear += grad_with_shrinkage + 

(accum_new^(-lr_power) - accum^(-lr_power)) / lr * var

quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
linear: Should be from a Variable().
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.
lr: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 shrinkage regularization. Must be a scalar.

lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyKerasMomentum 

func ResourceSparseApplyKerasMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, indices tf.Output, momentum tf.Output, optional ...ResourceSparseApplyKerasMomentumAttr) (o *tf.Operation)

Update relevant entries in '*var' and '*accum' according to the momentum scheme.

Set use_nesterov = True if you want to use Nesterov momentum.

That is for rows we have grad for, we update var and accum as follows:

accum = accum * momentum - lr * grad var += accum

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Learning rate. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.
momentum: Momentum. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyMomentum 

func ResourceSparseApplyMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, indices tf.Output, momentum tf.Output, optional ...ResourceSparseApplyMomentumAttr) (o *tf.Operation)

Update relevant entries in '*var' and '*accum' according to the momentum scheme.

Set use_nesterov = True if you want to use Nesterov momentum.

That is for rows we have grad for, we update var and accum as follows:

accum = accum * momentum + grad var -= lr * accum

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Learning rate. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.
momentum: Momentum. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyProximalAdagrad 

func ResourceSparseApplyProximalAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyProximalAdagradAttr) (o *tf.Operation)

Sparse update entries in '*var' and '*accum' according to FOBOS algorithm.

That is for rows we have grad for, we update var and accum as follows: accum += grad * grad prox_v = var prox_v -= lr * grad * (1 / sqrt(accum)) var = sign(prox_v)/(1+lr*l2) * max{|prox_v|-lr*l1,0}

Arguments: 

var_: Should be from a Variable().
accum: Should be from a Variable().
lr: Learning rate. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.

Returns the created operation.

func ResourceSparseApplyProximalGradientDescent 

func ResourceSparseApplyProximalGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyProximalGradientDescentAttr) (o *tf.Operation)

Sparse update '*var' as FOBOS algorithm with fixed learning rate.

That is for rows we have grad for, we update var as follows: prox_v = var - alpha * grad var = sign(prox_v)/(1+alpha*l2) * max{|prox_v|-alpha*l1,0}

Arguments: 

var_: Should be from a Variable().
alpha: Scaling factor. Must be a scalar.
l1: L1 regularization. Must be a scalar.
l2: L2 regularization. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var and accum.

Returns the created operation.

func ResourceSparseApplyRMSProp 

func ResourceSparseApplyRMSProp(scope *Scope, var_ tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyRMSPropAttr) (o *tf.Operation)

Update '*var' according to the RMSProp algorithm.

Note that in dense implementation of this algorithm, ms and mom will update even if the grad is zero, but in this sparse implementation, ms and mom will not update in iterations during which the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2 Delta = learning_rate * gradient / sqrt(mean_square + epsilon)

ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom

Arguments: 

var_: Should be from a Variable().
ms: Should be from a Variable().
mom: Should be from a Variable().
lr: Scaling factor. Must be a scalar.
rho: Decay rate. Must be a scalar.

epsilon: Ridge term. Must be a scalar.
grad: The gradient.
indices: A vector of indices into the first dimension of var, ms and mom.

Returns the created operation.

func ResourceStridedSliceAssign 

func ResourceStridedSliceAssign(scope *Scope, ref tf.Output, begin tf.Output, end tf.Output, strides tf.Output, value tf.Output, optional ...ResourceStridedSliceAssignAttr) (o *tf.Operation)

Assign `value` to the sliced l-value reference of `ref`.

The values of `value` are assigned to the positions in the variable `ref` that are selected by the slice parameters. The slice parameters `begin, `end`, `strides`, etc. work exactly as in `StridedSlice`.

NOTE this op currently does not support broadcasting and so `value`'s shape must be exactly the shape produced by the slice of `ref`.

Returns the created operation.

func Restore 

func Restore(scope *Scope, file_pattern tf.Output, tensor_name tf.Output, dt tf.DataType, optional ...RestoreAttr) (tensor tf.Output)

Restores a tensor from checkpoint files.

Reads a tensor stored in one or several files. If there are several files (for instance because a tensor was saved as slices), `file_pattern` may contain wildcard symbols (`*` and `?`) in the filename portion only, not in the directory portion.

If a `file_pattern` matches several files, `preferred_shard` can be used to hint in which file the requested tensor is likely to be found. This op will first open the file at index `preferred_shard` in the list of matching files and try to restore tensors from that file. Only if some tensors or tensor slices are not found in that first file, then the Op opens all the files. Setting `preferred_shard` to match the value passed as the `shard` input of a matching `Save` Op may speed up Restore. This attribute only affects performance, not correctness. The default value -1 means files are processed in order.

See also `RestoreSlice`.

Arguments: 

file_pattern: Must have a single element. The pattern of the files from

which we read the tensor. 

tensor_name: Must have a single element. The name of the tensor to be

restored. 

dt: The type of the tensor to be restored.

Returns The restored tensor.

func RestoreSlice 

func RestoreSlice(scope *Scope, file_pattern tf.Output, tensor_name tf.Output, shape_and_slice tf.Output, dt tf.DataType, optional ...RestoreSliceAttr) (tensor tf.Output)

Restores a tensor from checkpoint files.

This is like `Restore` except that restored tensor can be listed as filling only a slice of a larger tensor. `shape_and_slice` specifies the shape of the larger tensor and the slice that the restored tensor covers.

The `shape_and_slice` input has the same format as the elements of the `shapes_and_slices` input of the `SaveSlices` op.

Arguments: 

file_pattern: Must have a single element. The pattern of the files from

which we read the tensor. 

tensor_name: Must have a single element. The name of the tensor to be

restored. 

shape_and_slice: Scalar. The shapes and slice specifications to use when

restoring a tensors. 

dt: The type of the tensor to be restored.

Returns The restored tensor.

func RestoreV2 

func RestoreV2(scope *Scope, prefix tf.Output, tensor_names tf.Output, shape_and_slices tf.Output, dtypes []tf.DataType) (tensors []tf.Output)

Restores tensors from a V2 checkpoint.

For backward compatibility with the V1 format, this Op currently allows restoring from a V1 checkpoint as well:

  • This Op first attempts to find the V2 index file pointed to by "prefix", and if found proceed to read it as a V2 checkpoint;
  • Otherwise the V1 read path is invoked.

Relying on this behavior is not recommended, as the ability to fall back to read V1 might be deprecated and eventually removed.

By default, restores the named tensors in full. If the caller wishes to restore specific slices of stored tensors, "shape_and_slices" should be non-empty strings and correspondingly well-formed.

Callers must ensure all the named tensors are indeed stored in the checkpoint.

Arguments: 

prefix: Must have a single element.  The prefix of a V2 checkpoint.
tensor_names: shape {N}.  The names of the tensors to be restored.
shape_and_slices: shape {N}.  The slice specs of the tensors to be restored.

Empty strings indicate that they are non-partitioned tensors. 

dtypes: shape {N}.  The list of expected dtype for the tensors.  Must match

those stored in the checkpoint.

Returns shape {N}. The restored tensors, whose shapes are read from the checkpoint directly.

func RetrieveAllTPUEmbeddingParameters 

func RetrieveAllTPUEmbeddingParameters(scope *Scope, NumTables int64, config string, num_shards int64, shard_id int64) (parameters []tf.Output, auxiliary1 []tf.Output, auxiliary2 []tf.Output, auxiliary3 []tf.Output, auxiliary4 []tf.Output, auxiliary5 []tf.Output, auxiliary6 []tf.Output, auxiliary7 []tf.Output)

An op that retrieves optimization parameters from embedding to host memory.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint. For Adagrad, auxiliary1 will contain the accumulators after running this op. For SGD, all of the auxiliary* values will be empty (0x0 tensors for that table). For FTRL, auxiliary1 will contain the accumulators and auxiliary2 will contain the linear terms. For ADAM, auxiliary1 will contain the momenta and auxiliary2 will contain the velocities.

Arguments: 

NumTables: The number of embedding tables.
config: An TPUEmbeddingConfiguration proto describing the

table parameters being loaded, serialized to a string. 

num_shards: Number of shards into which the embedding tables are divided.
shard_id: Identifier of shard for this operation.

Returns: 

parameters:  A list of tensors, one for each embedding table, containing the

stored embedding table parameters. 

auxiliary1: A list of tensors, one for each embedding table, containing the

first auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary2: A list of tensors, one for each embedding table, containing the

second auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary3: A list of tensors, one for each embedding table, containing the

third auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary4: A list of tensors, one for each embedding table, containing the

fourth auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary5: A list of tensors, one for each embedding table, containing the

fifth auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary6: A list of tensors, one for each embedding table, containing the

six auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table). 

auxiliary7: A list of tensors, one for each embedding table, containing the

seventh auxiliary optimization parameter stored. Elements are present in the list, but have zero size, for unused optimization parameters (based on the algorithm in use for each table).

func RetrieveTPUEmbeddingADAMParameters 

func RetrieveTPUEmbeddingADAMParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingADAMParametersAttr) (parameters tf.Output, momenta tf.Output, velocities tf.Output)

Retrieve ADAM embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the ADAM optimization algorithm.
momenta: Parameter momenta updated by the ADAM optimization algorithm.
velocities: Parameter velocities updated by the ADAM optimization algorithm.

func RetrieveTPUEmbeddingAdadeltaParameters 

func RetrieveTPUEmbeddingAdadeltaParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingAdadeltaParametersAttr) (parameters tf.Output, accumulators tf.Output, updates tf.Output)

Retrieve Adadelta embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the Adadelta optimization algorithm.
accumulators: Parameter accumulators updated by the Adadelta optimization algorithm.
updates: Parameter updates updated by the Adadelta optimization algorithm.

func RetrieveTPUEmbeddingAdagradMomentumParameters 

func RetrieveTPUEmbeddingAdagradMomentumParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingAdagradMomentumParametersAttr) (parameters tf.Output, accumulators tf.Output, momenta tf.Output)

Retrieve Adagrad Momentum embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the Adagrad Momentum optimization algorithm.
accumulators: Parameter accumulators updated by the Adagrad Momentum optimization algorithm.
momenta: Parameter momenta updated by the Adagrad Momentum optimization algorithm.

func RetrieveTPUEmbeddingAdagradParameters 

func RetrieveTPUEmbeddingAdagradParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingAdagradParametersAttr) (parameters tf.Output, accumulators tf.Output)

Retrieve Adagrad embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the Adagrad optimization algorithm.
accumulators: Parameter accumulators updated by the Adagrad optimization algorithm.

func RetrieveTPUEmbeddingCenteredRMSPropParameters 

func RetrieveTPUEmbeddingCenteredRMSPropParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingCenteredRMSPropParametersAttr) (parameters tf.Output, ms tf.Output, mom tf.Output, mg tf.Output)

Retrieve centered RMSProp embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the centered RMSProp optimization algorithm.
ms: Parameter ms updated by the centered RMSProp optimization algorithm.
mom: Parameter mom updated by the centered RMSProp optimization algorithm.
mg: Parameter mg updated by the centered RMSProp optimization algorithm.

func RetrieveTPUEmbeddingFTRLParameters 

func RetrieveTPUEmbeddingFTRLParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingFTRLParametersAttr) (parameters tf.Output, accumulators tf.Output, linears tf.Output)

Retrieve FTRL embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the FTRL optimization algorithm.
accumulators: Parameter accumulators updated by the FTRL optimization algorithm.
linears: Parameter linears updated by the FTRL optimization algorithm.

func RetrieveTPUEmbeddingFrequencyEstimatorParameters 

func RetrieveTPUEmbeddingFrequencyEstimatorParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr) (parameters tf.Output, last_hit_step tf.Output)

Retrieve frequency estimator embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the frequency estimator optimization algorithm.
last_hit_step: Parameter last_hit_step updated by the frequency estimator optimization

algorithm.

func RetrieveTPUEmbeddingMDLAdagradLightParameters 

func RetrieveTPUEmbeddingMDLAdagradLightParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingMDLAdagradLightParametersAttr) (parameters tf.Output, accumulators tf.Output, weights tf.Output, benefits tf.Output)

Retrieve MDL Adagrad Light embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the MDL Adagrad Light optimization algorithm.
accumulators: Parameter accumulators updated by the MDL Adagrad Light optimization algorithm.
weights: Parameter weights updated by the MDL Adagrad Light optimization algorithm.
benefits: Parameter benefits updated by the MDL Adagrad Light optimization algorithm.

func RetrieveTPUEmbeddingMomentumParameters 

func RetrieveTPUEmbeddingMomentumParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingMomentumParametersAttr) (parameters tf.Output, momenta tf.Output)

Retrieve Momentum embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the Momentum optimization algorithm.
momenta: Parameter momenta updated by the Momentum optimization algorithm.

func RetrieveTPUEmbeddingProximalAdagradParameters 

func RetrieveTPUEmbeddingProximalAdagradParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingProximalAdagradParametersAttr) (parameters tf.Output, accumulators tf.Output)

Retrieve proximal Adagrad embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the proximal Adagrad optimization algorithm.
accumulators: Parameter accumulators updated by the proximal Adagrad optimization algorithm.

func RetrieveTPUEmbeddingRMSPropParameters 

func RetrieveTPUEmbeddingRMSPropParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingRMSPropParametersAttr) (parameters tf.Output, ms tf.Output, mom tf.Output)

Retrieve RMSProp embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns: 

parameters: Parameter parameters updated by the RMSProp optimization algorithm.
ms: Parameter ms updated by the RMSProp optimization algorithm.
mom: Parameter mom updated by the RMSProp optimization algorithm.

func RetrieveTPUEmbeddingStochasticGradientDescentParameters 

func RetrieveTPUEmbeddingStochasticGradientDescentParameters(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr) (parameters tf.Output)

Retrieve SGD embedding parameters.

An op that retrieves optimization parameters from embedding to host memory. Must be preceded by a ConfigureTPUEmbeddingHost op that sets up the correct embedding table configuration. For example, this op is used to retrieve updated parameters before saving a checkpoint.

Returns Parameter parameters updated by the stochastic gradient descent optimization algorithm.

func Reverse 

func Reverse(scope *Scope, tensor tf.Output, dims tf.Output) (output tf.Output)

Reverses specific dimensions of a tensor.

Given a `tensor`, and a `bool` tensor `dims` representing the dimensions of `tensor`, this operation reverses each dimension i of `tensor` where `dims[i]` is `True`.

`tensor` can have up to 8 dimensions. The number of dimensions of `tensor` must equal the number of elements in `dims`. In other words:

`rank(tensor) = size(dims)`

For example:

``` # tensor 't' is [[[[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]], # [[12, 13, 14, 15], # [16, 17, 18, 19], # [20, 21, 22, 23]]]] # tensor 't' shape is [1, 2, 3, 4]

# 'dims' is [False, False, False, True] reverse(t, dims) ==> [[[[ 3, 2, 1, 0], 

 [ 7,  6,  5,  4],
 [ 11, 10, 9, 8]],
[[15, 14, 13, 12],
 [19, 18, 17, 16],
 [23, 22, 21, 20]]]]

# 'dims' is [False, True, False, False] reverse(t, dims) ==> [[[[12, 13, 14, 15], 

 [16, 17, 18, 19],
 [20, 21, 22, 23]
[[ 0,  1,  2,  3],
 [ 4,  5,  6,  7],
 [ 8,  9, 10, 11]]]]

# 'dims' is [False, False, True, False] reverse(t, dims) ==> [[[[8, 9, 10, 11], 

 [4, 5, 6, 7],
 [0, 1, 2, 3]]
[[20, 21, 22, 23],
 [16, 17, 18, 19],
 [12, 13, 14, 15]]]]

```

Arguments: 

tensor: Up to 8-D.
dims: 1-D. The dimensions to reverse.

Returns The same shape as `tensor`.

func ReverseSequence 

func ReverseSequence(scope *Scope, input tf.Output, seq_lengths tf.Output, seq_dim int64, optional ...ReverseSequenceAttr) (output tf.Output)

Reverses variable length slices.

This op first slices `input` along the dimension `batch_dim`, and for each slice `i`, reverses the first `seq_lengths[i]` elements along the dimension `seq_dim`.

The elements of `seq_lengths` must obey `seq_lengths[i] <= input.dims[seq_dim]`, and `seq_lengths` must be a vector of length `input.dims[batch_dim]`.

The output slice `i` along dimension `batch_dim` is then given by input slice `i`, with the first `seq_lengths[i]` slices along dimension `seq_dim` reversed.

For example:

``` # Given this: batch_dim = 0 seq_dim = 1 input.dims = (4, 8, ...) seq_lengths = [7, 2, 3, 5]

# then slices of input are reversed on seq_dim, but only up to seq_lengths: output[0, 0:7, :, ...] = input[0, 7:0:-1, :, ...] output[1, 0:2, :, ...] = input[1, 2:0:-1, :, ...] output[2, 0:3, :, ...] = input[2, 3:0:-1, :, ...] output[3, 0:5, :, ...] = input[3, 5:0:-1, :, ...]

# while entries past seq_lens are copied through: output[0, 7:, :, ...] = input[0, 7:, :, ...] output[1, 2:, :, ...] = input[1, 2:, :, ...] output[2, 3:, :, ...] = input[2, 3:, :, ...] output[3, 2:, :, ...] = input[3, 2:, :, ...] ```

In contrast, if:

``` # Given this: batch_dim = 2 seq_dim = 0 input.dims = (8, ?, 4, ...) seq_lengths = [7, 2, 3, 5]

# then slices of input are reversed on seq_dim, but only up to seq_lengths: output[0:7, :, 0, :, ...] = input[7:0:-1, :, 0, :, ...] output[0:2, :, 1, :, ...] = input[2:0:-1, :, 1, :, ...] output[0:3, :, 2, :, ...] = input[3:0:-1, :, 2, :, ...] output[0:5, :, 3, :, ...] = input[5:0:-1, :, 3, :, ...]

# while entries past seq_lens are copied through: output[7:, :, 0, :, ...] = input[7:, :, 0, :, ...] output[2:, :, 1, :, ...] = input[2:, :, 1, :, ...] output[3:, :, 2, :, ...] = input[3:, :, 2, :, ...] output[2:, :, 3, :, ...] = input[2:, :, 3, :, ...] ```

Arguments: 

input: The input to reverse.
seq_lengths: 1-D with length `input.dims(batch_dim)` and

`max(seq_lengths) <= input.dims(seq_dim)` 

seq_dim: The dimension which is partially reversed.

Returns The partially reversed input. It has the same shape as `input`.

func ReverseV2 

func ReverseV2(scope *Scope, tensor tf.Output, axis tf.Output) (output tf.Output)

Reverses specific dimensions of a tensor.

Given a `tensor`, and a `int32` tensor `axis` representing the set of dimensions of `tensor` to reverse. This operation reverses each dimension `i` for which there exists `j` s.t. `axis[j] == i`.

`tensor` can have up to 8 dimensions. The number of dimensions specified in `axis` may be 0 or more entries. If an index is specified more than once, a InvalidArgument error is raised.

For example:

``` # tensor 't' is [[[[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]], # [[12, 13, 14, 15], # [16, 17, 18, 19], # [20, 21, 22, 23]]]] # tensor 't' shape is [1, 2, 3, 4]

# 'dims' is [3] or 'dims' is [-1] reverse(t, dims) ==> [[[[ 3, 2, 1, 0], 

 [ 7,  6,  5,  4],
 [ 11, 10, 9, 8]],
[[15, 14, 13, 12],
 [19, 18, 17, 16],
 [23, 22, 21, 20]]]]

# 'dims' is '[1]' (or 'dims' is '[-3]') reverse(t, dims) ==> [[[[12, 13, 14, 15], 

 [16, 17, 18, 19],
 [20, 21, 22, 23]
[[ 0,  1,  2,  3],
 [ 4,  5,  6,  7],
 [ 8,  9, 10, 11]]]]

# 'dims' is '[2]' (or 'dims' is '[-2]') reverse(t, dims) ==> [[[[8, 9, 10, 11], 

 [4, 5, 6, 7],
 [0, 1, 2, 3]]
[[20, 21, 22, 23],
 [16, 17, 18, 19],
 [12, 13, 14, 15]]]]

```

Arguments: 

tensor: Up to 8-D.
axis: 1-D. The indices of the dimensions to reverse. Must be in the range

`[-rank(tensor), rank(tensor))`.

Returns The same shape as `tensor`.

func RightShift 

func RightShift(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Elementwise computes the bitwise right-shift of `x` and `y`.

Performs a logical shift for unsigned integer types, and an arithmetic shift for signed integer types.

If `y` is negative, or greater than or equal to than the width of `x` in bits the result is implementation defined.

Example:

```python import tensorflow as tf from tensorflow.python.ops import bitwise_ops import numpy as np dtype_list = [tf.int8, tf.int16, tf.int32, tf.int64]

for dtype in dtype_list: 

lhs = tf.constant([-1, -5, -3, -14], dtype=dtype)
rhs = tf.constant([5, 0, 7, 11], dtype=dtype)

right_shift_result = bitwise_ops.right_shift(lhs, rhs)

print(right_shift_result)

# This will print: # tf.Tensor([-1 -5 -1 -1], shape=(4,), dtype=int8) # tf.Tensor([-1 -5 -1 -1], shape=(4,), dtype=int16) # tf.Tensor([-1 -5 -1 -1], shape=(4,), dtype=int32) # tf.Tensor([-1 -5 -1 -1], shape=(4,), dtype=int64)

lhs = np.array([-2, 64, 101, 32], dtype=np.int8) rhs = np.array([-1, -5, -3, -14], dtype=np.int8) bitwise_ops.right_shift(lhs, rhs) # <tf.Tensor: shape=(4,), dtype=int8, numpy=array([ -2, 64, 101, 32], dtype=int8)> ```

func Rint 

func Rint(scope *Scope, x tf.Output) (y tf.Output)

Returns element-wise integer closest to x.

If the result is midway between two representable values, the even representable is chosen. For example:

``` rint(-1.5) ==> -2.0 rint(0.5000001) ==> 1.0 rint([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0]) ==> [-2., -2., -0., 0., 2., 2., 2.] ```

func RiscAdd 

func RiscAdd(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x + y element-wise.

*NOTE*: `RiscAdd` does not supports broadcasting.

Given two input tensors, the `tf.risc_add` operation computes the sum for every element in the tensor.

Both input and output have a range `(-inf, inf)`.

func RiscMax 

func RiscMax(scope *Scope, x tf.Output, y tf.Output) (max tf.Output)

Returns max(x, y) element-wise.

*NOTE*: `RiscMax` does not supports broadcasting.

Given two input tensors, the `tf.risc_max` operation computes the maximum for every element in the tensor.

func RngReadAndSkip 

func RngReadAndSkip(scope *Scope, resource tf.Output, alg tf.Output, delta tf.Output) (value tf.Output)

Advance the counter of a counter-based RNG.

The state of the RNG after `rng_read_and_skip(n)` will be the same as that after `uniform([n])` (or any other distribution). The actual increment added to the counter is an unspecified implementation choice.

In the case that the input algorithm is RNG_ALG_AUTO_SELECT, the counter in the state needs to be of size int64[2], the current maximal counter size among algorithms. In this case, this op will manage the counter as if it is an 128-bit integer with layout [lower_64bits, higher_64bits]. If an algorithm needs less than 128 bits for the counter, it should use the left portion of the int64[2]. In this way, the int64[2] is compatible with all current RNG algorithms (Philox, ThreeFry and xla::RandomAlgorithm::RNG_DEFAULT). Downstream RNG ops can thus use this counter with any RNG algorithm.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG. The state consists of the counter followed by the key.
alg: The RNG algorithm.
delta: The amount of advancement.

Returns The old value of the resource variable, before incrementing. Since state size is algorithm-dependent, this output will be right-padded with zeros to reach shape int64[3] (the current maximal state size among algorithms).

func RngSkip 

func RngSkip(scope *Scope, resource tf.Output, algorithm tf.Output, delta tf.Output) (o *tf.Operation)

Advance the counter of a counter-based RNG.

The state of the RNG after `rng_skip(n)` will be the same as that after `stateful_uniform([n])` (or any other distribution). The actual increment added to the counter is an unspecified implementation detail.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
delta: The amount of advancement.

Returns the created operation.

func Roll 

func Roll(scope *Scope, input tf.Output, shift tf.Output, axis tf.Output) (output tf.Output)

Rolls the elements of a tensor along an axis.

The elements are shifted positively (towards larger indices) by the offset of `shift` along the dimension of `axis`. Negative `shift` values will shift elements in the opposite direction. Elements that roll passed the last position will wrap around to the first and vice versa. Multiple shifts along multiple axes may be specified.

For example:

``` # 't' is [0, 1, 2, 3, 4] roll(t, shift=2, axis=0) ==> [3, 4, 0, 1, 2]

# shifting along multiple dimensions # 't' is [[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]] roll(t, shift=[1, -2], axis=[0, 1]) ==> [[7, 8, 9, 5, 6], [2, 3, 4, 0, 1]]

# shifting along the same axis multiple times # 't' is [[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]] roll(t, shift=[2, -3], axis=[1, 1]) ==> [[1, 2, 3, 4, 0], [6, 7, 8, 9, 5]] ```

Arguments: 

shift: Dimension must be 0-D or 1-D. `shift[i]` specifies the number of places by which

elements are shifted positively (towards larger indices) along the dimension specified by `axis[i]`. Negative shifts will roll the elements in the opposite direction. 

axis: Dimension must be 0-D or 1-D. `axis[i]` specifies the dimension that the shift

`shift[i]` should occur. If the same axis is referenced more than once, the total shift for that axis will be the sum of all the shifts that belong to that axis.

Returns Has the same shape and size as the input. The elements are shifted positively (towards larger indices) by the offsets of `shift` along the dimensions of `axis`.

func Round 

func Round(scope *Scope, x tf.Output) (y tf.Output)

Rounds the values of a tensor to the nearest integer, element-wise.

Rounds half to even. Also known as bankers rounding. If you want to round according to the current system rounding mode use std::cint.

func Rsqrt 

func Rsqrt(scope *Scope, x tf.Output) (y tf.Output)

Computes reciprocal of square root of x element-wise.

I.e., \\(y = 1 / \sqrt{x}\\).

func RsqrtGrad 

func RsqrtGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient for the rsqrt of `x` wrt its input.

Specifically, `grad = dy * -0.5 * y^3`, where `y = rsqrt(x)`, and `dy` is the corresponding input gradient.

func SampleDistortedBoundingBox 

func SampleDistortedBoundingBox(scope *Scope, image_size tf.Output, bounding_boxes tf.Output, optional ...SampleDistortedBoundingBoxAttr) (begin tf.Output, size tf.Output, bboxes tf.Output)

Generate a single randomly distorted bounding box for an image.

Bounding box annotations are often supplied in addition to ground-truth labels in image recognition or object localization tasks. A common technique for training such a system is to randomly distort an image while preserving its content, i.e. *data augmentation*. This Op outputs a randomly distorted localization of an object, i.e. bounding box, given an `image_size`, `bounding_boxes` and a series of constraints.

The output of this Op is a single bounding box that may be used to crop the original image. The output is returned as 3 tensors: `begin`, `size` and `bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize what the bounding box looks like.

Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.

For example,

```python 

# Generate a single distorted bounding box.
begin, size, bbox_for_draw = tf.image.sample_distorted_bounding_box(
    tf.shape(image),
    bounding_boxes=bounding_boxes)

# Draw the bounding box in an image summary.
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
                                              bbox_for_draw)
tf.summary.image('images_with_box', image_with_box)

# Employ the bounding box to distort the image.
distorted_image = tf.slice(image, begin, size)

```

Note that if no bounding box information is available, setting `use_image_if_no_bounding_boxes = true` will assume there is a single implicit bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is false and no bounding boxes are supplied, an error is raised.

Arguments: 

image_size: 1-D, containing `[height, width, channels]`.
bounding_boxes: 3-D with shape `[batch, N, 4]` describing the N bounding boxes

associated with the image.

Returns: 

begin: 1-D, containing `[offset_height, offset_width, 0]`. Provide as input to

`tf.slice`. 

size: 1-D, containing `[target_height, target_width, -1]`. Provide as input to

`tf.slice`. 

bboxes: 3-D with shape `[1, 1, 4]` containing the distorted bounding box.

Provide as input to `tf.image.draw_bounding_boxes`.

func SampleDistortedBoundingBoxV2 

func SampleDistortedBoundingBoxV2(scope *Scope, image_size tf.Output, bounding_boxes tf.Output, min_object_covered tf.Output, optional ...SampleDistortedBoundingBoxV2Attr) (begin tf.Output, size tf.Output, bboxes tf.Output)

Generate a single randomly distorted bounding box for an image.

Bounding box annotations are often supplied in addition to ground-truth labels in image recognition or object localization tasks. A common technique for training such a system is to randomly distort an image while preserving its content, i.e. *data augmentation*. This Op outputs a randomly distorted localization of an object, i.e. bounding box, given an `image_size`, `bounding_boxes` and a series of constraints.

The output of this Op is a single bounding box that may be used to crop the original image. The output is returned as 3 tensors: `begin`, `size` and `bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize what the bounding box looks like.

Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.

For example,

```python 

# Generate a single distorted bounding box.
begin, size, bbox_for_draw = tf.image.sample_distorted_bounding_box(
    tf.shape(image),
    bounding_boxes=bounding_boxes)

# Draw the bounding box in an image summary.
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
                                              bbox_for_draw)
tf.summary.image('images_with_box', image_with_box)

# Employ the bounding box to distort the image.
distorted_image = tf.slice(image, begin, size)

```

Note that if no bounding box information is available, setting `use_image_if_no_bounding_boxes = true` will assume there is a single implicit bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is false and no bounding boxes are supplied, an error is raised.

Arguments: 

image_size: 1-D, containing `[height, width, channels]`.
bounding_boxes: 3-D with shape `[batch, N, 4]` describing the N bounding boxes

associated with the image. 

min_object_covered: The cropped area of the image must contain at least this

fraction of any bounding box supplied. The value of this parameter should be non-negative. In the case of 0, the cropped area does not need to overlap any of the bounding boxes supplied.

Returns: 

begin: 1-D, containing `[offset_height, offset_width, 0]`. Provide as input to

`tf.slice`. 

size: 1-D, containing `[target_height, target_width, -1]`. Provide as input to

`tf.slice`. 

bboxes: 3-D with shape `[1, 1, 4]` containing the distorted bounding box.

Provide as input to `tf.image.draw_bounding_boxes`.

func SamplingDataset 

func SamplingDataset(scope *Scope, input_dataset tf.Output, rate tf.Output, seed tf.Output, seed2 tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that takes a Bernoulli sample of the contents of another dataset.

There is no transformation in the `tf.data` Python API for creating this dataset. Instead, it is created as a result of the `filter_with_random_uniform_fusion` static optimization. Whether this optimization is performed is determined by the `experimental_optimization.filter_with_random_uniform_fusion` option of `tf.data.Options`.

Arguments: 

rate: A scalar representing the sample rate. Each element of `input_dataset` is

retained with this probability, independent of all other elements. 

seed: A scalar representing seed of random number generator.
seed2: A scalar representing seed2 of random number generator.

func Save 

func Save(scope *Scope, filename tf.Output, tensor_names tf.Output, data []tf.Output) (o *tf.Operation)

Saves the input tensors to disk.

The size of `tensor_names` must match the number of tensors in `data`. `data[i]` is written to `filename` with name `tensor_names[i]`.

See also `SaveSlices`.

Arguments: 

filename: Must have a single element. The name of the file to which we write

the tensor. 

tensor_names: Shape `[N]`. The names of the tensors to be saved.
data: `N` tensors to save.

Returns the created operation.

func SaveSlices 

func SaveSlices(scope *Scope, filename tf.Output, tensor_names tf.Output, shapes_and_slices tf.Output, data []tf.Output) (o *tf.Operation)

Saves input tensors slices to disk.

This is like `Save` except that tensors can be listed in the saved file as being a slice of a larger tensor. `shapes_and_slices` specifies the shape of the larger tensor and the slice that this tensor covers. `shapes_and_slices` must have as many elements as `tensor_names`.

Elements of the `shapes_and_slices` input must either be:

  • The empty string, in which case the corresponding tensor is saved normally.
  • A string of the form `dim0 dim1 ... dimN-1 slice-spec` where the `dimI` are the dimensions of the larger tensor and `slice-spec` specifies what part is covered by the tensor to save.

`slice-spec` itself is a `:`-separated list: `slice0:slice1:...:sliceN-1` where each `sliceI` is either:

  • The string `-` meaning that the slice covers all indices of this dimension
  • `start,length` where `start` and `length` are integers. In that case the slice covers `length` indices starting at `start`.

See also `Save`.

Arguments: 

filename: Must have a single element. The name of the file to which we write the

tensor. 

tensor_names: Shape `[N]`. The names of the tensors to be saved.
shapes_and_slices: Shape `[N]`.  The shapes and slice specifications to use when

saving the tensors. 

data: `N` tensors to save.

Returns the created operation.

func SaveV2 

func SaveV2(scope *Scope, prefix tf.Output, tensor_names tf.Output, shape_and_slices tf.Output, tensors []tf.Output) (o *tf.Operation)

Saves tensors in V2 checkpoint format.

By default, saves the named tensors in full. If the caller wishes to save specific slices of full tensors, "shape_and_slices" should be non-empty strings and correspondingly well-formed.

Arguments: 

prefix: Must have a single element. The prefix of the V2 checkpoint to which we

write the tensors. 

tensor_names: shape {N}. The names of the tensors to be saved.
shape_and_slices: shape {N}.  The slice specs of the tensors to be saved.

Empty strings indicate that they are non-partitioned tensors. 

tensors: `N` tensors to save.

Returns the created operation.

func ScalarSummary 

func ScalarSummary(scope *Scope, tags tf.Output, values tf.Output) (summary tf.Output)

Outputs a `Summary` protocol buffer with scalar values.

The input `tags` and `values` must have the same shape. The generated summary has a summary value for each tag-value pair in `tags` and `values`.

Arguments: 

tags: Tags for the summary.
values: Same shape as `tags.  Values for the summary.

Returns Scalar. Serialized `Summary` protocol buffer.

func ScatterNd 

func ScatterNd(scope *Scope, indices tf.Output, updates tf.Output, shape tf.Output, optional ...ScatterNdAttr) (output tf.Output)

Scatters `updates` into a tensor of shape `shape` according to `indices`.

Scatter sparse `updates` according to individual values at the specified `indices`. This op returns an output tensor with the `shape` you specify. This op is the inverse of the `tf.gather_nd` operator which extracts values or slices from a given tensor.

This operation is similar to `tf.tensor_scatter_nd_add`, except that the tensor is zero-initialized. Calling `tf.scatter_nd(indices, updates, shape)` is identical to calling `tf.tensor_scatter_nd_add(tf.zeros(shape, updates.dtype), indices, updates)`

If `indices` contains duplicates, the associated `updates` are accumulated (summed) into the output tensor.

**WARNING**: For floating-point data types, the output may be nondeterministic. This is because the order in which the updates are applied is nondeterministic and when floating-point numbers are added in different orders the resulting numerical approximation error can be slightly different. However, the output will be deterministic if op determinism is enabled via `tf.config.experimental.enable_op_determinism`.

`indices` is an integer tensor containing indices into the output tensor. The last dimension of `indices` can be at most the rank of `shape`: 

indices.shape[-1] <= shape.rank

The last dimension of `indices` corresponds to indices of elements (if `indices.shape[-1] = shape.rank`) or slices (if `indices.shape[-1] < shape.rank`) along dimension `indices.shape[-1]` of `shape`.

`updates` is a tensor with shape: 

indices.shape[:-1] + shape[indices.shape[-1]:]

The simplest form of the scatter op is to insert individual elements in a tensor by index. Consider an example where you want to insert 4 scattered elements in a rank-1 tensor with 8 elements.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/ScatterNd1.png" alt> </div>

In Python, this scatter operation would look like this:

```python 

indices = tf.constant([[4], [3], [1], [7]])
updates = tf.constant([9, 10, 11, 12])
shape = tf.constant([8])
scatter = tf.scatter_nd(indices, updates, shape)
print(scatter)

```

The resulting tensor would look like this: 

[0, 11, 0, 10, 9, 0, 0, 12]

You can also insert entire slices of a higher rank tensor all at once. For example, you can insert two slices in the first dimension of a rank-3 tensor with two matrices of new values.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/ScatterNd2.png" alt> </div>

In Python, this scatter operation would look like this:

```python 

indices = tf.constant([[1], [3]])
updates = tf.constant([[[5, 5, 5, 5], [6, 6, 6, 6],
                        [7, 7, 7, 7], [8, 8, 8, 8]],
                       [[5, 5, 5, 5], [6, 6, 6, 6],
                        [7, 7, 7, 7], [8, 8, 8, 8]]])
shape = tf.constant([4, 4, 4])
scatter = tf.scatter_nd(indices, updates, shape)
print(scatter)

```

The resulting tensor would look like this: 

[[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]],
 [[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]],
 [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]],
 [[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]]]

If `indices` contains any out-of-bound indices, depending on `bad_indices_policy`, the op will either return an error or ignore the out-of-bound indices. `bad_indices_policy` can be one of the following values:

  1. "" or "DEFAULT": raises on CPU and ignore on GPU. This is because historically on CPU and GPU we handle errors in different ways, and for backward compatibility we keep the default behavior.
  2. "ERROR": raises error; GPU does not support this value.
  3. "IGNORE": ignore the bad indices; supported on both CPU and GPU.

Arguments: 

indices: Tensor of indices.
updates: Values to scatter into the output tensor.
shape: 1-D. The shape of the output tensor.

Returns A new tensor with the given shape and updates applied according to the indices.

func ScatterNdNonAliasingAdd 

func ScatterNdNonAliasingAdd(scope *Scope, input tf.Output, indices tf.Output, updates tf.Output, optional ...ScatterNdNonAliasingAddAttr) (output tf.Output)

Applies sparse addition to `input` using individual values or slices

from `updates` according to indices `indices`. The updates are non-aliasing: `input` is only modified in-place if no other operations will use it. Otherwise, a copy of `input` is made. This operation has a gradient with respect to both `input` and `updates`.

`input` is a `Tensor` with rank `P` and `indices` is a `Tensor` of rank `Q`.

`indices` must be integer tensor, containing indices into `input`. 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 `(P-K)`-dimensional slices (if `K < P`) along the `K`th dimension of `input`.

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

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

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

input = tf.constant([1, 2, 3, 4, 5, 6, 7, 8])
indices = tf.constant([[4], [3], [1], [7]])
updates = tf.constant([9, 10, 11, 12])
output = tf.scatter_nd_non_aliasing_add(input, indices, updates)
with tf.Session() as sess:
  print(sess.run(output))

The resulting value `output` 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.

Arguments: 

input: A Tensor.
indices: A Tensor. Must be one of the following types: `int32`, `int64`.

A tensor of indices into `input`. 

updates: A Tensor. Must have the same type as ref. A tensor of updated values

to add to `input`.

Returns A `Tensor` with the same shape as `input`, containing values of `input` updated with `updates`.

func SdcaFprint 

func SdcaFprint(scope *Scope, input tf.Output) (output tf.Output)

Computes fingerprints of the input strings.

Arguments: 

input: vector of strings to compute fingerprints on.

Returns a (N,2) shaped matrix where N is the number of elements in the input vector. Each row contains the low and high parts of the fingerprint.

func SdcaOptimizer 

func SdcaOptimizer(scope *Scope, sparse_example_indices []tf.Output, sparse_feature_indices []tf.Output, sparse_feature_values []tf.Output, dense_features []tf.Output, example_weights tf.Output, example_labels tf.Output, sparse_indices []tf.Output, sparse_weights []tf.Output, dense_weights []tf.Output, example_state_data tf.Output, loss_type string, l1 float32, l2 float32, num_loss_partitions int64, num_inner_iterations int64, optional ...SdcaOptimizerAttr) (out_example_state_data tf.Output, out_delta_sparse_weights []tf.Output, out_delta_dense_weights []tf.Output)

Distributed version of Stochastic Dual Coordinate Ascent (SDCA) optimizer for

linear models with L1 + L2 regularization. As global optimization objective is strongly-convex, the optimizer optimizes the dual objective at each step. The optimizer applies each update one example at a time. Examples are sampled uniformly, and the optimizer is learning rate free and enjoys linear convergence rate.

[Proximal Stochastic Dual Coordinate Ascent](http://arxiv.org/pdf/1211.2717v1.pdf).<br> Shai Shalev-Shwartz, Tong Zhang. 2012

$$Loss Objective = \sum f_{i} (wx_{i}) + (l2 / 2) * |w|^2 + l1 * |w|$$

[Adding vs. Averaging in Distributed Primal-Dual Optimization](http://arxiv.org/abs/1502.03508).<br> Chenxin Ma, Virginia Smith, Martin Jaggi, Michael I. Jordan, Peter Richtarik, Martin Takac. 2015

[Stochastic Dual Coordinate Ascent with Adaptive Probabilities](https://arxiv.org/abs/1502.08053).<br> Dominik Csiba, Zheng Qu, Peter Richtarik. 2015

Arguments: 

sparse_example_indices: a list of vectors which contain example indices.
sparse_feature_indices: a list of vectors which contain feature indices.
sparse_feature_values: a list of vectors which contains feature value

associated with each feature group. 

dense_features: a list of matrices which contains the dense feature values.
example_weights: a vector which contains the weight associated with each

example. 

example_labels: a vector which contains the label/target associated with each

example. 

sparse_indices: a list of vectors where each value is the indices which has

corresponding weights in sparse_weights. This field maybe omitted for the dense approach. 

sparse_weights: a list of vectors where each value is the weight associated with

a sparse feature group. 

dense_weights: a list of vectors where the values are the weights associated

with a dense feature group. 

example_state_data: a list of vectors containing the example state data.
loss_type: Type of the primal loss. Currently SdcaSolver supports logistic,

squared and hinge losses. 

l1: Symmetric l1 regularization strength.
l2: Symmetric l2 regularization strength.
num_loss_partitions: Number of partitions of the global loss function.
num_inner_iterations: Number of iterations per mini-batch.

Returns: 

out_example_state_data: a list of vectors containing the updated example state

data. 

out_delta_sparse_weights: a list of vectors where each value is the delta

weights associated with a sparse feature group. 

out_delta_dense_weights: a list of vectors where the values are the delta

weights associated with a dense feature group.

func SdcaOptimizerV2 

func SdcaOptimizerV2(scope *Scope, sparse_example_indices []tf.Output, sparse_feature_indices []tf.Output, sparse_feature_values []tf.Output, dense_features []tf.Output, example_weights tf.Output, example_labels tf.Output, sparse_indices []tf.Output, sparse_weights []tf.Output, dense_weights []tf.Output, example_state_data tf.Output, loss_type string, l1 float32, l2 float32, num_loss_partitions int64, num_inner_iterations int64, optional ...SdcaOptimizerV2Attr) (out_example_state_data tf.Output, out_delta_sparse_weights []tf.Output, out_delta_dense_weights []tf.Output)

Distributed version of Stochastic Dual Coordinate Ascent (SDCA) optimizer for

linear models with L1 + L2 regularization. As global optimization objective is strongly-convex, the optimizer optimizes the dual objective at each step. The optimizer applies each update one example at a time. Examples are sampled uniformly, and the optimizer is learning rate free and enjoys linear convergence rate.

[Proximal Stochastic Dual Coordinate Ascent](http://arxiv.org/pdf/1211.2717v1.pdf).<br> Shai Shalev-Shwartz, Tong Zhang. 2012

$$Loss Objective = \sum f_{i} (wx_{i}) + (l2 / 2) * |w|^2 + l1 * |w|$$

[Adding vs. Averaging in Distributed Primal-Dual Optimization](http://arxiv.org/abs/1502.03508).<br> Chenxin Ma, Virginia Smith, Martin Jaggi, Michael I. Jordan, Peter Richtarik, Martin Takac. 2015

[Stochastic Dual Coordinate Ascent with Adaptive Probabilities](https://arxiv.org/abs/1502.08053).<br> Dominik Csiba, Zheng Qu, Peter Richtarik. 2015

Arguments: 

sparse_example_indices: a list of vectors which contain example indices.
sparse_feature_indices: a list of vectors which contain feature indices.
sparse_feature_values: a list of vectors which contains feature value

associated with each feature group. 

dense_features: a list of matrices which contains the dense feature values.
example_weights: a vector which contains the weight associated with each

example. 

example_labels: a vector which contains the label/target associated with each

example. 

sparse_indices: a list of vectors where each value is the indices which has

corresponding weights in sparse_weights. This field maybe omitted for the dense approach. 

sparse_weights: a list of vectors where each value is the weight associated with

a sparse feature group. 

dense_weights: a list of vectors where the values are the weights associated

with a dense feature group. 

example_state_data: a list of vectors containing the example state data.
loss_type: Type of the primal loss. Currently SdcaSolver supports logistic,

squared and hinge losses. 

l1: Symmetric l1 regularization strength.
l2: Symmetric l2 regularization strength.
num_loss_partitions: Number of partitions of the global loss function.
num_inner_iterations: Number of iterations per mini-batch.

Returns: 

out_example_state_data: a list of vectors containing the updated example state

data. 

out_delta_sparse_weights: a list of vectors where each value is the delta

weights associated with a sparse feature group. 

out_delta_dense_weights: a list of vectors where the values are the delta

weights associated with a dense feature group.

func SegmentMax 

func SegmentMax(scope *Scope, data tf.Output, segment_ids tf.Output) (output tf.Output)

Computes the maximum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \max_j(data_j)\\) where `max` is over `j` such that `segment_ids[j] == i`.

If the max is empty for a given segment ID `i`, `output[i] = 0`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/SegmentMax.png" alt> </div>

For example:

>>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> tf.math.segment_max(c, tf.constant([0, 0, 1])).numpy() array([[4, 3, 3, 4], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentMaxV2 added in v0.5.0 

func SegmentMaxV2(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the maximum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \max_j(data_j)\\) where `max` is over `j` such that `segment_ids[j] == i`.

If the maximum is empty for a given segment ID `i`, it outputs the smallest possible value for the specific numeric type, `output[i] = numeric_limits<T>::lowest()`.

Note: That this op is currently only supported with jit_compile=True.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

The only difference with SegmentMax is the additional input `num_segments`. This helps in evaluating the output shape in compile time. `num_segments` should be consistent with segment_ids. e.g. Max(segment_ids) should be equal to `num_segments` - 1 for a 1-d segment_ids With inconsistent num_segments, the op still runs. only difference is, the output takes the size of num_segments irrespective of size of segment_ids and data. for num_segments less than expected output size, the last elements are ignored for num_segments more than the expected output size, last elements are assigned smallest possible value for the specific numeric type.

For example:

>>> @tf.function(jit_compile=True) ... def test(c): ... return tf.raw_ops.SegmentMaxV2(data=c, segment_ids=tf.constant([0, 0, 1]), num_segments=2) >>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> test(c).numpy() array([[4, 3, 3, 4], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated. The values must be less than `num_segments`.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimensionw which has size `num_segments`.

func SegmentMean 

func SegmentMean(scope *Scope, data tf.Output, segment_ids tf.Output) (output tf.Output)

Computes the mean along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \frac{\sum_j data_j}{N}\\) where `mean` is over `j` such that `segment_ids[j] == i` and `N` is the total number of values summed.

If the mean is empty for a given segment ID `i`, `output[i] = 0`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as a smaller following index when computing the numerator of the mean.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/SegmentMean.png" alt> </div>

For example:

>>> c = tf.constant([[1.0,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> tf.math.segment_mean(c, tf.constant([0, 0, 1])).numpy() array([[2.5, 2.5, 2.5, 2.5], 

[5., 6., 7., 8.]], dtype=float32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentMin 

func SegmentMin(scope *Scope, data tf.Output, segment_ids tf.Output) (output tf.Output)

Computes the minimum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \min_j(data_j)\\) where `min` is over `j` such that `segment_ids[j] == i`.

If the min is empty for a given segment ID `i`, `output[i] = 0`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/SegmentMin.png" alt> </div>

For example:

>>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> tf.math.segment_min(c, tf.constant([0, 0, 1])).numpy() array([[1, 2, 2, 1], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentMinV2 added in v0.5.0 

func SegmentMinV2(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the minimum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \min_j(data_j)\\) where `min` is over `j` such that `segment_ids[j] == i`.

If the minimum is empty for a given segment ID `i`, it outputs the largest possible value for the specific numeric type, `output[i] = numeric_limits<T>::max()`.

Note: That this op is currently only supported with jit_compile=True.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

The only difference with SegmentMin is the additional input `num_segments`. This helps in evaluating the output shape in compile time. `num_segments` should be consistent with segment_ids. e.g. Max(segment_ids) should be equal to `num_segments` - 1 for a 1-d segment_ids With inconsistent num_segments, the op still runs. only difference is, the output takes the size of num_segments irrespective of size of segment_ids and data. for num_segments less than expected output size, the last elements are ignored for num_segments more than the expected output size, last elements are assigned the largest possible value for the specific numeric type.

For example:

>>> @tf.function(jit_compile=True) ... def test(c): ... return tf.raw_ops.SegmentMinV2(data=c, segment_ids=tf.constant([0, 0, 1]), num_segments=2) >>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> test(c).numpy() array([[1, 2, 2, 1], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated. The values must be less than `num_segments`.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimensionw which has size `num_segments`.

func SegmentProd 

func SegmentProd(scope *Scope, data tf.Output, segment_ids tf.Output) (output tf.Output)

Computes the product along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \prod_j data_j\\) where the product is over `j` such that `segment_ids[j] == i`.

If the product is empty for a given segment ID `i`, `output[i] = 1`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/SegmentProd.png" alt> </div>

For example:

>>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> tf.math.segment_prod(c, tf.constant([0, 0, 1])).numpy() array([[4, 6, 6, 4], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentProdV2 added in v0.4.0 

func SegmentProdV2(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the product along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \prod_j data_j\\) where the product is over `j` such that `segment_ids[j] == i`.

If the product is empty for a given segment ID `i`, `output[i] = 1`.

Note: That this op is currently only supported with jit_compile=True.

The only difference with SegmentProd is the additional input `num_segments`. This helps in evaluating the output shape in compile time. `num_segments` should be consistent with segment_ids. e.g. Max(segment_ids) - 1 should be equal to `num_segments` for a 1-d segment_ids With inconsistent num_segments, the op still runs. only difference is, the output takes the size of num_segments irrespective of size of segment_ids and data. for num_segments less than expected output size, the last elements are ignored for num_segments more than the expected output size, last elements are assigned 1.

For example:

>>> @tf.function(jit_compile=True) ... def test(c): ... return tf.raw_ops.SegmentProdV2(data=c, segment_ids=tf.constant([0, 0, 1]), num_segments=2) >>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> test(c).numpy() array([[4, 6, 6, 4], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated. The values must be less than `num_segments`.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimensionw which has size `num_segments`.

func SegmentSum 

func SegmentSum(scope *Scope, data tf.Output, segment_ids tf.Output) (output tf.Output)

Computes the sum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \sum_j data_j\\) where sum is over `j` such that `segment_ids[j] == i`.

If the sum is empty for a given segment ID `i`, `output[i] = 0`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted, and an error is thrown for indices that are not increasing. On GPU, this does not throw an error for unsorted indices. On GPU, out-of-order indices result in safe but unspecified behavior, which may include treating out-of-order indices as the same as a smaller following index.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/SegmentSum.png" alt> </div>

For example:

>>> c = tf.constant([[1,2,3,4], [4, 3, 2, 1], [5,6,7,8]]) >>> tf.math.segment_sum(c, tf.constant([0, 0, 1])).numpy() array([[5, 5, 5, 5], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentSumV2 added in v0.4.0 

func SegmentSumV2(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the sum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output_i = \sum_j data_j\\) where sum is over `j` such that `segment_ids[j] == i`.

If the sum is empty for a given segment ID `i`, `output[i] = 0`.

Note that this op is currently only supported with jit_compile=True. </div>

Arguments: 

segment_ids: A 1-D tensor whose size is equal to the size of `data`'s

first dimension. Values should be sorted and can be repeated. The values must be less than `num_segments`.

Caution: The values are always validated to be sorted on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimension which has size `num_segments`.

func Select 

func Select(scope *Scope, condition tf.Output, x tf.Output, y tf.Output) (output tf.Output)

Selects elements from `x` or `y`, depending on `condition`.

The `x`, and `y` tensors must all have the same shape, and the output will also have that shape.

The `condition` tensor must be a scalar if `x` and `y` are scalars. If `x` and `y` are vectors or higher rank, then `condition` must be either a scalar, a vector with size matching the first dimension of `x`, or must have the same shape as `x`.

The `condition` tensor acts as a mask that chooses, based on the value at each element, whether the corresponding element / row in the output should be taken from `x` (if true) or `y` (if false).

If `condition` is a vector and `x` and `y` are higher rank matrices, then it chooses which row (outer dimension) to copy from `x` and `y`. If `condition` has the same shape as `x` and `y`, then it chooses which element to copy from `x` and `y`.

For example:

```python # 'condition' tensor is [[True, False] # [False, True]] # 't' is [[1, 2], # [3, 4]] # 'e' is [[5, 6], # [7, 8]] select(condition, t, e) # => [[1, 6], [7, 4]]

# 'condition' tensor is [True, False] # 't' is [[1, 2], # [3, 4]] # 'e' is [[5, 6], # [7, 8]] select(condition, t, e) ==> [[1, 2], 

[7, 8]]

```

Arguments: 

x: = A `Tensor` which may have the same shape as `condition`.

If `condition` is rank 1, `x` may have higher rank, but its first dimension must match the size of `condition`. 

y: = A `Tensor` with the same type and shape as `x`.

Returns = A `Tensor` with the same type and shape as `x` and `y`.

func SelfAdjointEig 

func SelfAdjointEig(scope *Scope, input tf.Output) (output tf.Output)

Computes the Eigen Decomposition of a batch of square self-adjoint matrices.

DEPRECATED at GraphDef version 11: Use SelfAdjointEigV2 instead.

The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices, with the same constraints as the single matrix SelfAdjointEig.

The result is a [..., M+1, M] matrix with [..., 0,:] containing the eigenvalues, and subsequent [...,1:, :] containing the eigenvectors. The eigenvalues are sorted in non-decreasing order.

Arguments: 

input: Shape is `[..., M, M]`.

Returns Shape is `[..., M+1, M]`.

func SelfAdjointEigV2 

func SelfAdjointEigV2(scope *Scope, input tf.Output, optional ...SelfAdjointEigV2Attr) (e tf.Output, v tf.Output)

Computes the eigen decomposition of one or more square self-adjoint matrices.

Computes the eigenvalues and (optionally) eigenvectors of each inner matrix in `input` such that `input[..., :, :] = v[..., :, :] * diag(e[..., :])`. The eigenvalues are sorted in non-decreasing order.

```python # a is a tensor. # e is a tensor of eigenvalues. # v is a tensor of eigenvectors. e, v = self_adjoint_eig(a) e = self_adjoint_eig(a, compute_v=False) ```

Arguments: 

input: `Tensor` input of shape `[N, N]`.

Returns: 

e: Eigenvalues. Shape is `[N]`.
v: Eigenvectors. Shape is `[N, N]`.

func Selu 

func Selu(scope *Scope, features tf.Output) (activations tf.Output)

Computes scaled exponential linear: `scale * alpha * (exp(features) - 1)`

if < 0, `scale * features` otherwise.

To be used together with `initializer = tf.variance_scaling_initializer(factor=1.0, mode='FAN_IN')`. For correct dropout, use `tf.contrib.nn.alpha_dropout`.

See [Self-Normalizing Neural Networks](https://arxiv.org/abs/1706.02515)

func SeluGrad 

func SeluGrad(scope *Scope, gradients tf.Output, outputs tf.Output) (backprops tf.Output)

Computes gradients for the scaled exponential linear (Selu) operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding Selu operation.
outputs: The outputs of the corresponding Selu operation.

Returns The gradients: `gradients * (outputs + scale * alpha)` if outputs < 0, `scale * gradients` otherwise.

func Send 

func Send(scope *Scope, tensor tf.Output, tensor_name string, send_device string, send_device_incarnation int64, recv_device string, optional ...SendAttr) (o *tf.Operation)

Sends the named tensor from send_device to recv_device.

Arguments: 

tensor: The tensor to send.
tensor_name: The name of the tensor to send.
send_device: The name of the device sending the tensor.
send_device_incarnation: The current incarnation of send_device.
recv_device: The name of the device receiving the tensor.

Returns the created operation.

func SendTPUEmbeddingGradients 

func SendTPUEmbeddingGradients(scope *Scope, inputs []tf.Output, learning_rates []tf.Output, config string) (o *tf.Operation)

Performs gradient updates of embedding tables.

Arguments: 

inputs: A TensorList of gradients with which to update embedding tables.

This argument has the same length and shapes as the return value of RecvTPUEmbeddingActivations, but contains gradients of the model's loss with respect to the embedding activations. The embedding tables are updated from these gradients via the optimizer specified in the TPU embedding configuration given to tpu.initialize_system. 

learning_rates: A TensorList of float32 scalars, one for each dynamic learning

rate tag: see the comments in //third_party/tensorflow/core/protobuf/tpu/optimization_parameters.proto. Multiple tables can share the same dynamic learning rate tag as specified in the configuration. If the learning rates for all tables are constant, this list should be empty. 

config: Serialized TPUEmbeddingConfiguration proto.

Returns the created operation.

func SerializeIterator 

func SerializeIterator(scope *Scope, resource_handle tf.Output, optional ...SerializeIteratorAttr) (serialized tf.Output)

Converts the given `resource_handle` representing an iterator to a variant tensor.

Arguments: 

resource_handle: A handle to an iterator resource.

Returns A variant tensor storing the state of the iterator contained in the resource.

func SerializeManySparse 

func SerializeManySparse(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...SerializeManySparseAttr) (serialized_sparse tf.Output)

Serialize an `N`-minibatch `SparseTensor` into an `[N, 3]` `Tensor` object.

The `SparseTensor` must have rank `R` greater than 1, and the first dimension is treated as the minibatch dimension. Elements of the `SparseTensor` must be sorted in increasing order of this first dimension. The serialized `SparseTensor` objects going into each row of `serialized_sparse` will have rank `R-1`.

The minibatch size `N` is extracted from `sparse_shape[0]`.

Arguments: 

sparse_indices: 2-D.  The `indices` of the minibatch `SparseTensor`.
sparse_values: 1-D.  The `values` of the minibatch `SparseTensor`.
sparse_shape: 1-D.  The `shape` of the minibatch `SparseTensor`.

func SerializeSparse 

func SerializeSparse(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...SerializeSparseAttr) (serialized_sparse tf.Output)

Serialize a `SparseTensor` into a `[3]` `Tensor` object.

Arguments: 

sparse_indices: 2-D.  The `indices` of the `SparseTensor`.
sparse_values: 1-D.  The `values` of the `SparseTensor`.
sparse_shape: 1-D.  The `shape` of the `SparseTensor`.

func SerializeTensor 

func SerializeTensor(scope *Scope, tensor tf.Output) (serialized tf.Output)

Transforms a Tensor into a serialized TensorProto proto.

Arguments: 

tensor: A Tensor of type `T`.

Returns A serialized TensorProto proto of the input tensor.

func SetSize 

func SetSize(scope *Scope, set_indices tf.Output, set_values tf.Output, set_shape tf.Output, optional ...SetSizeAttr) (size tf.Output)

Number of unique elements along last dimension of input `set`.

Input `set` is a `SparseTensor` represented by `set_indices`, `set_values`, and `set_shape`. The last dimension contains values in a set, duplicates are allowed but ignored.

If `validate_indices` is `True`, this op validates the order and range of `set` indices. Setting is to `False` while passing invalid arguments results in undefined behavior.

Arguments: 

set_indices: 2D `Tensor`, indices of a `SparseTensor`.
set_values: 1D `Tensor`, values of a `SparseTensor`.
set_shape: 1D `Tensor`, shape of a `SparseTensor`.

Returns For `set` ranked `n`, this is a `Tensor` with rank `n-1`, and the same 1st `n-1` dimensions as `set`. Each value is the number of unique elements in the corresponding `[0...n-1]` dimension of `set`.

func Shape 

func Shape(scope *Scope, input tf.Output, optional ...ShapeAttr) (output tf.Output)

Returns the shape of a tensor.

This operation returns a 1-D integer tensor representing the shape of `input`.

For example:

``` # 't' is [[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]] shape(t) ==> [2, 2, 3] ```

func ShapeN 

func ShapeN(scope *Scope, input []tf.Output, optional ...ShapeNAttr) (output []tf.Output)

Returns shape of tensors.

This operation returns N 1-D integer tensors representing shape of `input[i]s`.

func ShardDataset 

func ShardDataset(scope *Scope, input_dataset tf.Output, num_shards tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ShardDatasetAttr) (handle tf.Output)

Creates a `Dataset` that includes only 1/`num_shards` of this dataset.

Arguments: 

num_shards: An integer representing the number of shards operating in parallel.
index: An integer representing the current worker index.

func ShardedFilename 

func ShardedFilename(scope *Scope, basename tf.Output, shard tf.Output, num_shards tf.Output) (filename tf.Output)

Generate a sharded filename. The filename is printf formatted as 

%s-%05d-of-%05d, basename, shard, num_shards.

func ShardedFilespec 

func ShardedFilespec(scope *Scope, basename tf.Output, num_shards tf.Output) (filename tf.Output)

Generate a glob pattern matching all sharded file names.

func ShuffleAndRepeatDataset 

func ShuffleAndRepeatDataset(scope *Scope, input_dataset tf.Output, buffer_size tf.Output, seed tf.Output, seed2 tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ShuffleAndRepeatDatasetAttr) (handle tf.Output)

Creates a dataset that shuffles and repeats elements from `input_dataset`

pseudorandomly.

Arguments: 

buffer_size: The number of output elements to buffer in an iterator over

this dataset. Compare with the `min_after_dequeue` attr when creating a `RandomShuffleQueue`. 

seed: A scalar seed for the random number generator. If either `seed` or

`seed2` is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. 

seed2: A second scalar seed to avoid seed collision.
count: A scalar representing the number of times the underlying dataset

should be repeated. The default is `-1`, which results in infinite repetition.

func ShuffleDataset 

func ShuffleDataset(scope *Scope, input_dataset tf.Output, buffer_size tf.Output, seed tf.Output, seed2 tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ShuffleDatasetAttr) (handle tf.Output)

Creates a dataset that shuffles elements from `input_dataset` pseudorandomly.

Arguments: 

buffer_size: The number of output elements to buffer in an iterator over

this dataset. Compare with the `min_after_dequeue` attr when creating a `RandomShuffleQueue`. 

seed: A scalar seed for the random number generator. If either `seed` or

`seed2` is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. 

seed2: A second scalar seed to avoid seed collision.

func ShutdownDistributedTPU 

func ShutdownDistributedTPU(scope *Scope) (o *tf.Operation)

Shuts down a running distributed TPU system.

The op returns an error if no system is running.

Returns the created operation.

func ShutdownTPUSystem added in v0.2.0 

func ShutdownTPUSystem(scope *Scope) (success tf.Output)

An op that shuts down the TPU system.

Returns A boolean that indicates if the shut down process succeeds.

func Sigmoid 

func Sigmoid(scope *Scope, x tf.Output) (y tf.Output)

Computes sigmoid of `x` element-wise.

Specifically, `y = 1 / (1 + exp(-x))`.

func SigmoidGrad 

func SigmoidGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient of the sigmoid of `x` wrt its input.

Specifically, `grad = dy * y * (1 - y)`, where `y = sigmoid(x)`, and `dy` is the corresponding input gradient.

func Sign 

func Sign(scope *Scope, x tf.Output) (y tf.Output)

Returns an element-wise indication of the sign of a number.

`y = sign(x) = -1` if `x < 0`; 0 if `x == 0`; 1 if `x > 0`.

For complex numbers, `y = sign(x) = x / |x|` if `x != 0`, otherwise `y = 0`.

Example usage: >>> tf.math.sign([0., 2., -3.]) <tf.Tensor: shape=(3,), dtype=float32, numpy=array([ 0., 1., -1.], dtype=float32)>

func Sin 

func Sin(scope *Scope, x tf.Output) (y tf.Output)

Computes sine of x element-wise. 

Given an input tensor, this function computes sine of every
element in the tensor. Input range is `(-inf, inf)` and
output range is `[-1,1]`.

```python
x = tf.constant([-float("inf"), -9, -0.5, 1, 1.2, 200, 10, float("inf")])
tf.math.sin(x) ==> [nan -0.4121185 -0.47942555 0.84147096 0.9320391 -0.87329733 -0.54402107 nan]
```

func Sinh 

func Sinh(scope *Scope, x tf.Output) (y tf.Output)

Computes hyperbolic sine of x element-wise. 

Given an input tensor, this function computes hyperbolic sine of every
element in the tensor. Input range is `[-inf,inf]` and output range
is `[-inf,inf]`.

```python
x = tf.constant([-float("inf"), -9, -0.5, 1, 1.2, 2, 10, float("inf")])
tf.math.sinh(x) ==> [-inf -4.0515420e+03 -5.2109528e-01 1.1752012e+00 1.5094614e+00 3.6268604e+00 1.1013232e+04 inf]
```

func Size 

func Size(scope *Scope, input tf.Output, optional ...SizeAttr) (output tf.Output)

Returns the size of a tensor.

This operation returns an integer representing the number of elements in `input`.

For example:

``` # 't' is [[[1, 1,, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]]] size(t) ==> 12 ```

func SkipDataset 

func SkipDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...SkipDatasetAttr) (handle tf.Output)

Creates a dataset that skips `count` elements from the `input_dataset`.

Arguments: 

count: A scalar representing the number of elements from the `input_dataset`

that should be skipped. If count is -1, skips everything.

func Skipgram 

func Skipgram(scope *Scope, filename string, batch_size int64, optional ...SkipgramAttr) (vocab_word tf.Output, vocab_freq tf.Output, words_per_epoch tf.Output, current_epoch tf.Output, total_words_processed tf.Output, examples tf.Output, labels tf.Output)

Parses a text file and creates a batch of examples.

DEPRECATED at GraphDef version 19: Moving word2vec into tensorflow_models/tutorials and deprecating its ops here as a result

Arguments: 

filename: The corpus's text file name.
batch_size: The size of produced batch.

Returns: 

vocab_word: A vector of words in the corpus.
vocab_freq: Frequencies of words. Sorted in the non-ascending order.
words_per_epoch: Number of words per epoch in the data file.
current_epoch: The current epoch number.
total_words_processed: The total number of words processed so far.
examples: A vector of word ids.
labels: A vector of word ids.

func Slice 

func Slice(scope *Scope, input tf.Output, begin tf.Output, size tf.Output) (output tf.Output)

Return a slice from 'input'.

The output tensor is a tensor with dimensions described by 'size' whose values are extracted from 'input' starting at the offsets in 'begin'.

*Requirements*: 

0 <= begin[i] <= begin[i] + size[i] <= Di  for i in [0, n)

Arguments: 

begin: begin[i] specifies the offset into the 'i'th dimension of

'input' to slice from. 

size: size[i] specifies the number of elements of the 'i'th dimension

of 'input' to slice. If size[i] is -1, all remaining elements in dimension i are included in the slice (i.e. this is equivalent to setting size[i] = input.dim_size(i) - begin[i]).

func SlidingWindowDataset 

func SlidingWindowDataset(scope *Scope, input_dataset tf.Output, window_size tf.Output, window_shift tf.Output, window_stride tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...SlidingWindowDatasetAttr) (handle tf.Output)

Creates a dataset that passes a sliding window over `input_dataset`.

Arguments: 

window_size: A scalar representing the number of elements in the

sliding window. 

window_shift: A scalar representing the steps moving the sliding window

forward in one iteration. It must be positive. 

window_stride: A scalar representing the stride of the input elements of the sliding window.

It must be positive.

func Snapshot 

func Snapshot(scope *Scope, input tf.Output) (output tf.Output)

Returns a copy of the input tensor.

func SnapshotDataset 

func SnapshotDataset(scope *Scope, input_dataset tf.Output, path tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...SnapshotDatasetAttr) (handle tf.Output)

Creates a dataset that will write to / read from a snapshot.

This dataset attempts to determine whether a valid snapshot exists at the `snapshot_path`, and reads from the snapshot in lieu of using `input_dataset`. If not, it will run the preprocessing pipeline as usual, and write out a snapshot of the data processed for future use.

Arguments: 

input_dataset: A variant tensor representing the input dataset.
path: The path we should write snapshots to / read snapshots from.

func SobolSample 

func SobolSample(scope *Scope, dim tf.Output, num_results tf.Output, skip tf.Output, optional ...SobolSampleAttr) (samples tf.Output)

Generates points from the Sobol sequence.

Creates a Sobol sequence with `num_results` samples. Each sample has dimension `dim`. Skips the first `skip` samples.

Arguments: 

dim: Positive scalar `Tensor` representing each sample's dimension.
num_results: Positive scalar `Tensor` of dtype int32. The number of Sobol points to return

in the output. 

skip: Positive scalar `Tensor` of dtype int32. The number of initial points of the

Sobol sequence to skip.

Returns `Tensor` of samples from Sobol sequence with `shape` [num_results, dim].

func Softmax 

func Softmax(scope *Scope, logits tf.Output) (softmax tf.Output)

Computes softmax activations.

For each batch `i` and class `j` we have 

$$softmax[i, j] = exp(logits[i, j]) / sum_j(exp(logits[i, j]))$$

Arguments: 

logits: 2-D with shape `[batch_size, num_classes]`.

Returns Same shape as `logits`.

func SoftmaxCrossEntropyWithLogits 

func SoftmaxCrossEntropyWithLogits(scope *Scope, features tf.Output, labels tf.Output) (loss tf.Output, backprop tf.Output)

Computes softmax cross entropy cost and gradients to backpropagate.

Inputs are the logits, not probabilities.

Arguments: 

features: batch_size x num_classes matrix
labels: batch_size x num_classes matrix

The caller must ensure that each batch of labels represents a valid probability distribution.

Returns: 

loss: Per example loss (batch_size vector).
backprop: backpropagated gradients (batch_size x num_classes matrix).

func SoftplusGrad 

func SoftplusGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)

Computes softplus gradients for a softplus operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding softplus operation.
features: The features passed as input to the corresponding softplus operation.

Returns The gradients: `gradients / (1 + exp(-features))`.

func Softsign 

func Softsign(scope *Scope, features tf.Output) (activations tf.Output)

Computes softsign: `features / (abs(features) + 1)`.

func SoftsignGrad 

func SoftsignGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)

Computes softsign gradients for a softsign operation.

Arguments: 

gradients: The backpropagated gradients to the corresponding softsign operation.
features: The features passed as input to the corresponding softsign operation.

Returns The gradients: `gradients / (1 + abs(features)) ** 2`.

func SpaceToBatch 

func SpaceToBatch(scope *Scope, input tf.Output, paddings tf.Output, block_size int64) (output tf.Output)

SpaceToBatch for 4-D tensors of type T.

This is a legacy version of the more general SpaceToBatchND.

Zero-pads and then rearranges (permutes) blocks of spatial data into batch. More specifically, this op outputs a copy of the input tensor where values from the `height` and `width` dimensions are moved to the `batch` dimension. After the zero-padding, both `height` and `width` of the input must be divisible by the block size.

The attr `block_size` must be greater than one. It indicates the block size.

  • Non-overlapping blocks of size `block_size x block size` in the height and width dimensions are rearranged into the batch dimension at each location.
  • The batch of the output tensor is `batch * block_size * block_size`.
  • Both height_pad and width_pad must be divisible by block_size.

The shape of the output will be: 

[batch*block_size*block_size, height_pad/block_size, width_pad/block_size,
 depth]

Some examples:

(1) For the following input of shape `[1, 2, 2, 1]` and block_size of 2:

``` x = [[[[1], [2]], [[3], [4]]]] ```

The output tensor has shape `[4, 1, 1, 1]` and value:

``` [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] ```

(2) For the following input of shape `[1, 2, 2, 3]` and block_size of 2:

``` x = [[[[1, 2, 3], [4, 5, 6]], 

[[7, 8, 9], [10, 11, 12]]]]

```

The output tensor has shape `[4, 1, 1, 3]` and value:

``` [[[[1, 2, 3]]], [[[4, 5, 6]]], [[[7, 8, 9]]], [[[10, 11, 12]]]] ```

(3) For the following input of shape `[1, 4, 4, 1]` and block_size of 2:

``` x = [[[[1], [2], [3], [4]], 

[[5],   [6],  [7],  [8]],
[[9],  [10], [11],  [12]],
[[13], [14], [15],  [16]]]]

```

The output tensor has shape `[4, 2, 2, 1]` and value:

``` x = [[[[1], [3]], [[9], [11]]], 

[[[2], [4]], [[10], [12]]],
[[[5], [7]], [[13], [15]]],
[[[6], [8]], [[14], [16]]]]

```

(4) For the following input of shape `[2, 2, 4, 1]` and block_size of 2:

``` x = [[[[1], [2], [3], [4]], 

 [[5],   [6],  [7],  [8]]],
[[[9],  [10], [11],  [12]],
 [[13], [14], [15],  [16]]]]

```

The output tensor has shape `[8, 1, 2, 1]` and value:

``` x = [[[[1], [3]]], [[[9], [11]]], [[[2], [4]]], [[[10], [12]]], 

[[[5], [7]]], [[[13], [15]]], [[[6], [8]]], [[[14], [16]]]]

```

Among others, this operation is useful for reducing atrous convolution into regular convolution.

Arguments: 

	input: 4-D with shape `[batch, height, width, depth]`.
	paddings: 2-D tensor of non-negative integers with shape `[2, 2]`. It specifies
  the padding of the input with zeros across the spatial dimensions as follows:

      paddings = [[pad_top, pad_bottom], [pad_left, pad_right]]

  The effective spatial dimensions of the zero-padded input tensor will be:

      height_pad = pad_top + height + pad_bottom
      width_pad = pad_left + width + pad_right

func SpaceToBatchND 

func SpaceToBatchND(scope *Scope, input tf.Output, block_shape tf.Output, paddings tf.Output) (output tf.Output)

SpaceToBatch for N-D tensors of type T.

This operation divides "spatial" dimensions `[1, ..., M]` of the input into a grid of blocks of shape `block_shape`, and interleaves these blocks with the "batch" dimension (0) such that in the output, the spatial dimensions `[1, ..., M]` correspond to the position within the grid, and the batch dimension combines both the position within a spatial block and the original batch position. Prior to division into blocks, the spatial dimensions of the input are optionally zero padded according to `paddings`. See below for a precise description.

This operation is equivalent to the following steps:

  1. Zero-pad the start and end of dimensions `[1, ..., M]` of the input according to `paddings` to produce `padded` of shape `padded_shape`.

2. Reshape `padded` to `reshaped_padded` of shape: 

	[batch] +
	[padded_shape[1] / block_shape[0],
	  block_shape[0],
	 ...,
	 padded_shape[M] / block_shape[M-1],
	 block_shape[M-1]] +
	remaining_shape

 3. Permute dimensions of `reshaped_padded` to produce
    `permuted_reshaped_padded` of shape:

    block_shape +
    [batch] +
    [padded_shape[1] / block_shape[0],
    ...,
    padded_shape[M] / block_shape[M-1]] +
    remaining_shape

 4. Reshape `permuted_reshaped_padded` to flatten `block_shape` into the batch
    dimension, producing an output tensor of shape:

    [batch * prod(block_shape)] +
    [padded_shape[1] / block_shape[0],
    ...,
    padded_shape[M] / block_shape[M-1]] +
    remaining_shape

Some examples:

(1) For the following input of shape `[1, 2, 2, 1]`, `block_shape = [2, 2]`, and 

`paddings = [[0, 0], [0, 0]]`:

``` x = [[[[1], [2]], [[3], [4]]]] ```

The output tensor has shape `[4, 1, 1, 1]` and value:

``` [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] ```

(2) For the following input of shape `[1, 2, 2, 3]`, `block_shape = [2, 2]`, and 

`paddings = [[0, 0], [0, 0]]`:

``` x = [[[[1, 2, 3], [4, 5, 6]], 

[[7, 8, 9], [10, 11, 12]]]]

```

The output tensor has shape `[4, 1, 1, 3]` and value:

``` [[[[1, 2, 3]]], [[[4, 5, 6]]], [[[7, 8, 9]]], [[[10, 11, 12]]]] ```

(3) For the following input of shape `[1, 4, 4, 1]`, `block_shape = [2, 2]`, and 

`paddings = [[0, 0], [0, 0]]`:

``` x = [[[[1], [2], [3], [4]], 

[[5],   [6],  [7],  [8]],
[[9],  [10], [11],  [12]],
[[13], [14], [15],  [16]]]]

```

The output tensor has shape `[4, 2, 2, 1]` and value:

``` x = [[[[1], [3]], [[9], [11]]], 

[[[2], [4]], [[10], [12]]],
[[[5], [7]], [[13], [15]]],
[[[6], [8]], [[14], [16]]]]

```

(4) For the following input of shape `[2, 2, 4, 1]`, block_shape = `[2, 2]`, and 

paddings = `[[0, 0], [2, 0]]`:

``` x = [[[[1], [2], [3], [4]], 

 [[5],   [6],  [7],  [8]]],
[[[9],  [10], [11],  [12]],
 [[13], [14], [15],  [16]]]]

```

The output tensor has shape `[8, 1, 3, 1]` and value:

``` x = [[[[0], [1], [3]]], [[[0], [9], [11]]], 

[[[0], [2], [4]]], [[[0], [10], [12]]],
[[[0], [5], [7]]], [[[0], [13], [15]]],
[[[0], [6], [8]]], [[[0], [14], [16]]]]

```

Among others, this operation is useful for reducing atrous convolution into regular convolution.

Arguments: 

input: N-D with shape `input_shape = [batch] + spatial_shape + remaining_shape`,

where spatial_shape has `M` dimensions. 

	block_shape: 1-D with shape `[M]`, all values must be >= 1.
	paddings: 2-D with shape `[M, 2]`, all values must be >= 0.
  `paddings[i] = [pad_start, pad_end]` specifies the padding for input dimension
  `i + 1`, which corresponds to spatial dimension `i`.  It is required that
  `block_shape[i]` divides `input_shape[i + 1] + pad_start + pad_end`.

func SpaceToDepth 

func SpaceToDepth(scope *Scope, input tf.Output, block_size int64, optional ...SpaceToDepthAttr) (output tf.Output)

SpaceToDepth for tensors of type T.

Rearranges blocks of spatial data, into depth. More specifically, this op outputs a copy of the input tensor where values from the `height` and `width` dimensions are moved to the `depth` dimension. The attr `block_size` indicates the input block size.

  • Non-overlapping blocks of size `block_size x block size` are rearranged into depth at each location.
  • The depth of the output tensor is `block_size * block_size * input_depth`.
  • The Y, X coordinates within each block of the input become the high order component of the output channel index.
  • The input tensor's height and width must be divisible by block_size.

The `data_format` attr specifies the layout of the input and output tensors with the following options: 

"NHWC": `[ batch, height, width, channels ]`
"NCHW": `[ batch, channels, height, width ]`
"NCHW_VECT_C":
    `qint8 [ batch, channels / 4, height, width, 4 ]`

It is useful to consider the operation as transforming a 6-D Tensor. e.g. for data_format = NHWC, 

Each element in the input tensor can be specified via 6 coordinates,
ordered by decreasing memory layout significance as:
n,oY,bY,oX,bX,iC  (where n=batch index, oX, oY means X or Y coordinates
                   within the output image, bX, bY means coordinates
                   within the input block, iC means input channels).
The output would be a transpose to the following layout:
n,oY,oX,bY,bX,iC

This operation is useful for resizing the activations between convolutions (but keeping all data), e.g. instead of pooling. It is also useful for training purely convolutional models.

For example, given an input of shape `[1, 2, 2, 1]`, data_format = "NHWC" and block_size = 2:

``` x = [[[[1], [2]], 

[[3], [4]]]]

```

This operation will output a tensor of shape `[1, 1, 1, 4]`:

``` [[[[1, 2, 3, 4]]]] ```

Here, the input has a batch of 1 and each batch element has shape `[2, 2, 1]`, the corresponding output will have a single element (i.e. width and height are both 1) and will have a depth of 4 channels (1 * block_size * block_size). The output element shape is `[1, 1, 4]`.

For an input tensor with larger depth, here of shape `[1, 2, 2, 3]`, e.g.

``` x = [[[[1, 2, 3], [4, 5, 6]], 

[[7, 8, 9], [10, 11, 12]]]]

```

This operation, for block_size of 2, will return the following tensor of shape `[1, 1, 1, 12]`

``` [[[[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]]]] ```

Similarly, for the following input of shape `[1 4 4 1]`, and a block size of 2:

``` x = [[[[1], [2], [5], [6]], 

[[3],   [4],  [7],  [8]],
[[9],  [10], [13],  [14]],
[[11], [12], [15],  [16]]]]

```

the operator will return the following tensor of shape `[1 2 2 4]`:

``` x = [[[[1, 2, 3, 4], 

 [5, 6, 7, 8]],
[[9, 10, 11, 12],
 [13, 14, 15, 16]]]]

```

Arguments: 

block_size: The size of the spatial block.

func SparseAdd 

func SparseAdd(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output, thresh tf.Output) (sum_indices tf.Output, sum_values tf.Output, sum_shape tf.Output)

Adds two `SparseTensor` objects to produce another `SparseTensor`.

The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, before this step run `SparseReorder` to restore index ordering.

By default, if two values sum to zero at some index, the output `SparseTensor` would still include that particular location in its index, storing a zero in the corresponding value slot. To override this, callers can specify `thresh`, indicating that if the sum has a magnitude strictly smaller than `thresh`, its corresponding value and index would then not be included. In particular, `thresh == 0` (default) means everything is kept and actual thresholding happens only for a positive value.

In the following shapes, `nnz` is the count after taking `thresh` into account.

Arguments: 

a_indices: 2-D.  The `indices` of the first `SparseTensor`, size `[nnz, ndims]` Matrix.
a_values: 1-D.  The `values` of the first `SparseTensor`, size `[nnz]` Vector.
a_shape: 1-D.  The `shape` of the first `SparseTensor`, size `[ndims]` Vector.
b_indices: 2-D.  The `indices` of the second `SparseTensor`, size `[nnz, ndims]` Matrix.
b_values: 1-D.  The `values` of the second `SparseTensor`, size `[nnz]` Vector.
b_shape: 1-D.  The `shape` of the second `SparseTensor`, size `[ndims]` Vector.
thresh: 0-D.  The magnitude threshold that determines if an output value/index

pair takes space.

func SparseAddGrad 

func SparseAddGrad(scope *Scope, backprop_val_grad tf.Output, a_indices tf.Output, b_indices tf.Output, sum_indices tf.Output) (a_val_grad tf.Output, b_val_grad tf.Output)

The gradient operator for the SparseAdd op.

The SparseAdd op calculates A + B, where A, B, and the sum are all represented as `SparseTensor` objects. This op takes in the upstream gradient w.r.t. non-empty values of the sum, and outputs the gradients w.r.t. the non-empty values of A and B.

Arguments: 

backprop_val_grad: 1-D with shape `[nnz(sum)]`.  The gradient with respect to

the non-empty values of the sum. 

a_indices: 2-D.  The `indices` of the `SparseTensor` A, size `[nnz(A), ndims]`.
b_indices: 2-D.  The `indices` of the `SparseTensor` B, size `[nnz(B), ndims]`.
sum_indices: 2-D.  The `indices` of the sum `SparseTensor`, size

`[nnz(sum), ndims]`.

Returns: 

a_val_grad: 1-D with shape `[nnz(A)]`. The gradient with respect to the

non-empty values of A. 

b_val_grad: 1-D with shape `[nnz(B)]`. The gradient with respect to the

non-empty values of B.

func SparseBincount 

func SparseBincount(scope *Scope, indices tf.Output, values tf.Output, dense_shape tf.Output, size tf.Output, weights tf.Output, optional ...SparseBincountAttr) (output tf.Output)

Counts the number of occurrences of each value in an integer array.

Outputs a vector with length `size` and the same dtype as `weights`. If `weights` are empty, then index `i` stores the number of times the value `i` is counted in `arr`. If `weights` are non-empty, then index `i` stores the sum of the value in `weights` at each index where the corresponding value in `arr` is `i`.

Values in `arr` outside of the range [0, size) are ignored.

Arguments: 

indices: 2D int64 `Tensor`.
values: 1D int `Tensor`.
dense_shape: 1D int64 `Tensor`.
size: non-negative int scalar `Tensor`.
weights: is an int32, int64, float32, or float64 `Tensor` with the same

shape as `input`, or a length-0 `Tensor`, in which case it acts as all weights equal to 1.

Returns 1D `Tensor` with length equal to `size` or 2D `Tensor` with [batch_size, `size`]. The counts or summed weights for each value in the range [0, size).

func SparseConcat 

func SparseConcat(scope *Scope, indices []tf.Output, values []tf.Output, shapes []tf.Output, concat_dim int64) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Concatenates a list of `SparseTensor` along the specified dimension.

Concatenation is with respect to the dense versions of these sparse tensors. It is assumed that each input is a `SparseTensor` whose elements are ordered along increasing dimension number.

All inputs' shapes must match, except for the concat dimension. The `indices`, `values`, and `shapes` lists must have the same length.

The output shape is identical to the inputs', except along the concat dimension, where it is the sum of the inputs' sizes along that dimension.

The output elements will be resorted to preserve the sort order along increasing dimension number.

This op runs in `O(M log M)` time, where `M` is the total number of non-empty values across all inputs. This is due to the need for an internal sort in order to concatenate efficiently across an arbitrary dimension.

For example, if `concat_dim = 1` and the inputs are 

sp_inputs[0]: shape = [2, 3]
[0, 2]: "a"
[1, 0]: "b"
[1, 1]: "c"

sp_inputs[1]: shape = [2, 4]
[0, 1]: "d"
[0, 2]: "e"

then the output will be 

shape = [2, 7]
[0, 2]: "a"
[0, 4]: "d"
[0, 5]: "e"
[1, 0]: "b"
[1, 1]: "c"

Graphically this is equivalent to doing 

[    a] concat [  d e  ] = [    a   d e  ]
[b c  ]        [       ]   [b c          ]

Arguments: 

indices: 2-D.  Indices of each input `SparseTensor`.
values: 1-D.  Non-empty values of each `SparseTensor`.
shapes: 1-D.  Shapes of each `SparseTensor`.
concat_dim: Dimension to concatenate along. Must be in range [-rank, rank),

where rank is the number of dimensions in each input `SparseTensor`.

Returns: 

output_indices: 2-D.  Indices of the concatenated `SparseTensor`.
output_values: 1-D.  Non-empty values of the concatenated `SparseTensor`.
output_shape: 1-D.  Shape of the concatenated `SparseTensor`.

func SparseCountSparseOutput 

func SparseCountSparseOutput(scope *Scope, indices tf.Output, values tf.Output, dense_shape tf.Output, weights tf.Output, binary_output bool, optional ...SparseCountSparseOutputAttr) (output_indices tf.Output, output_values tf.Output, output_dense_shape tf.Output)

Performs sparse-output bin counting for a sparse tensor input. 

Counts the number of times each value occurs in the input.

Arguments: 

indices: Tensor containing the indices of the sparse tensor to count.
values: Tensor containing values of the sparse tensor to count.
dense_shape: Tensor containing the dense shape of the sparse tensor to count.
weights: A Tensor of the same shape as indices containing per-index weight values.

May also be the empty tensor if no weights are used. 

binary_output: Whether to output the number of occurrences of each value or 1.

Returns: 

output_indices: Indices tensor for the resulting sparse tensor object.
output_values: Values tensor for the resulting sparse tensor object.
output_dense_shape: Shape tensor for the resulting sparse tensor object.

func SparseCross 

func SparseCross(scope *Scope, indices []tf.Output, values []tf.Output, shapes []tf.Output, dense_inputs []tf.Output, hashed_output bool, num_buckets int64, hash_key int64, out_type tf.DataType, internal_type tf.DataType) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Generates sparse cross from a list of sparse and dense tensors.

The op takes two lists, one of 2D `SparseTensor` and one of 2D `Tensor`, each representing features of one feature column. It outputs a 2D `SparseTensor` with the batchwise crosses of these features.

For example, if the inputs are 

inputs[0]: SparseTensor with shape = [2, 2]
[0, 0]: "a"
[1, 0]: "b"
[1, 1]: "c"

inputs[1]: SparseTensor with shape = [2, 1]
[0, 0]: "d"
[1, 0]: "e"

inputs[2]: Tensor [["f"], ["g"]]

then the output will be 

shape = [2, 2]
[0, 0]: "a_X_d_X_f"
[1, 0]: "b_X_e_X_g"
[1, 1]: "c_X_e_X_g"

if hashed_output=true then the output will be 

shape = [2, 2]
[0, 0]: FingerprintCat64(
            Fingerprint64("f"), FingerprintCat64(
                Fingerprint64("d"), Fingerprint64("a")))
[1, 0]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("b")))
[1, 1]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("c")))

Arguments: 

indices: 2-D.  Indices of each input `SparseTensor`.
values: 1-D.   values of each `SparseTensor`.
shapes: 1-D.   Shapes of each `SparseTensor`.
dense_inputs: 2-D.    Columns represented by dense `Tensor`.
hashed_output: If true, returns the hash of the cross instead of the string.

This will allow us avoiding string manipulations. 

num_buckets: It is used if hashed_output is true.

output = hashed_value%num_buckets if num_buckets > 0 else hashed_value. 

hash_key: Specify the hash_key that will be used by the `FingerprintCat64`

function to combine the crosses fingerprints.

Returns: 

output_indices: 2-D.  Indices of the concatenated `SparseTensor`.
output_values: 1-D.  Non-empty values of the concatenated or hashed

`SparseTensor`. 

output_shape: 1-D.  Shape of the concatenated `SparseTensor`.

func SparseCrossHashed 

func SparseCrossHashed(scope *Scope, indices []tf.Output, values []tf.Output, shapes []tf.Output, dense_inputs []tf.Output, num_buckets tf.Output, strong_hash tf.Output, salt tf.Output) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Generates sparse cross from a list of sparse and dense tensors.

The op takes two lists, one of 2D `SparseTensor` and one of 2D `Tensor`, each representing features of one feature column. It outputs a 2D `SparseTensor` with the batchwise crosses of these features.

For example, if the inputs are 

inputs[0]: SparseTensor with shape = [2, 2]
[0, 0]: "a"
[1, 0]: "b"
[1, 1]: "c"

inputs[1]: SparseTensor with shape = [2, 1]
[0, 0]: "d"
[1, 0]: "e"

inputs[2]: Tensor [["f"], ["g"]]

then the output will be 

shape = [2, 2]
[0, 0]: "a_X_d_X_f"
[1, 0]: "b_X_e_X_g"
[1, 1]: "c_X_e_X_g"

if hashed_output=true then the output will be 

shape = [2, 2]
[0, 0]: FingerprintCat64(
            Fingerprint64("f"), FingerprintCat64(
                Fingerprint64("d"), Fingerprint64("a")))
[1, 0]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("b")))
[1, 1]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("c")))

Arguments: 

indices: 2-D.  Indices of each input `SparseTensor`.
values: 1-D.   values of each `SparseTensor`.
shapes: 1-D.   Shapes of each `SparseTensor`.
dense_inputs: 2-D.    Columns represented by dense `Tensor`.
num_buckets: It is used if hashed_output is true.

output = hashed_value%num_buckets if num_buckets > 0 else hashed_value. 

strong_hash: boolean, if true, siphash with salt will be used instead of farmhash.
salt: Specify the salt that will be used by the siphash function.

Returns: 

output_indices: 2-D.  Indices of the concatenated `SparseTensor`.
output_values: 1-D.  Non-empty values of the concatenated or hashed

`SparseTensor`. 

output_shape: 1-D.  Shape of the concatenated `SparseTensor`.

func SparseCrossV2 

func SparseCrossV2(scope *Scope, indices []tf.Output, values []tf.Output, shapes []tf.Output, dense_inputs []tf.Output, sep tf.Output) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Generates sparse cross from a list of sparse and dense tensors.

The op takes two lists, one of 2D `SparseTensor` and one of 2D `Tensor`, each representing features of one feature column. It outputs a 2D `SparseTensor` with the batchwise crosses of these features.

For example, if the inputs are 

inputs[0]: SparseTensor with shape = [2, 2]
[0, 0]: "a"
[1, 0]: "b"
[1, 1]: "c"

inputs[1]: SparseTensor with shape = [2, 1]
[0, 0]: "d"
[1, 0]: "e"

inputs[2]: Tensor [["f"], ["g"]]

then the output will be 

shape = [2, 2]
[0, 0]: "a_X_d_X_f"
[1, 0]: "b_X_e_X_g"
[1, 1]: "c_X_e_X_g"

if hashed_output=true then the output will be 

shape = [2, 2]
[0, 0]: FingerprintCat64(
            Fingerprint64("f"), FingerprintCat64(
                Fingerprint64("d"), Fingerprint64("a")))
[1, 0]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("b")))
[1, 1]: FingerprintCat64(
            Fingerprint64("g"), FingerprintCat64(
                Fingerprint64("e"), Fingerprint64("c")))

Arguments: 

indices: 2-D.  Indices of each input `SparseTensor`.
values: 1-D.   values of each `SparseTensor`.
shapes: 1-D.   Shapes of each `SparseTensor`.
dense_inputs: 2-D.    Columns represented by dense `Tensor`.
sep: string used when joining a list of string inputs, can be used as separator later.

Returns: 

output_indices: 2-D.  Indices of the concatenated `SparseTensor`.
output_values: 1-D.  Non-empty values of the concatenated or hashed

`SparseTensor`. 

output_shape: 1-D.  Shape of the concatenated `SparseTensor`.

func SparseDenseCwiseAdd 

func SparseDenseCwiseAdd(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)

Adds up a SparseTensor and a dense Tensor, using these special rules:

(1) Broadcasts the dense side to have the same shape as the sparse side, if 

eligible;

(2) Then, only the dense values pointed to by the indices of the SparseTensor 

participate in the cwise addition.

By these rules, the result is a logical SparseTensor with exactly the same indices and shape, but possibly with different non-zero values. The output of this Op is the resultant non-zero values.

Arguments: 

sp_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

sp_values: 1-D.  `N` non-empty values corresponding to `sp_indices`.
sp_shape: 1-D.  Shape of the input SparseTensor.
dense: `R`-D.  The dense Tensor operand.

Returns 1-D. The `N` values that are operated on.

func SparseDenseCwiseDiv 

func SparseDenseCwiseDiv(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)

Component-wise divides a SparseTensor by a dense Tensor.

*Limitation*: this Op only broadcasts the dense side to the sparse side, but not the other direction.

Arguments: 

sp_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

sp_values: 1-D.  `N` non-empty values corresponding to `sp_indices`.
sp_shape: 1-D.  Shape of the input SparseTensor.
dense: `R`-D.  The dense Tensor operand.

Returns 1-D. The `N` values that are operated on.

func SparseDenseCwiseMul 

func SparseDenseCwiseMul(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)

Component-wise multiplies a SparseTensor by a dense Tensor.

The output locations corresponding to the implicitly zero elements in the sparse tensor will be zero (i.e., will not take up storage space), regardless of the contents of the dense tensor (even if it's +/-INF and that INF*0 == NaN).

*Limitation*: this Op only broadcasts the dense side to the sparse side, but not the other direction.

Arguments: 

sp_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

sp_values: 1-D.  `N` non-empty values corresponding to `sp_indices`.
sp_shape: 1-D.  Shape of the input SparseTensor.
dense: `R`-D.  The dense Tensor operand.

Returns 1-D. The `N` values that are operated on.

func SparseFillEmptyRows 

func SparseFillEmptyRows(scope *Scope, indices tf.Output, values tf.Output, dense_shape tf.Output, default_value tf.Output) (output_indices tf.Output, output_values tf.Output, empty_row_indicator tf.Output, reverse_index_map tf.Output)

Fills empty rows in the input 2-D `SparseTensor` with a default value.

The input `SparseTensor` is represented via the tuple of inputs (`indices`, `values`, `dense_shape`). The output `SparseTensor` has the same `dense_shape` but with indices `output_indices` and values `output_values`.

This op inserts a single entry for every row that doesn't have any values. The index is created as `[row, 0, ..., 0]` and the inserted value is `default_value`.

For example, suppose `sp_input` has shape `[5, 6]` and non-empty values: 

[0, 1]: a
[0, 3]: b
[2, 0]: c
[3, 1]: d

Rows 1 and 4 are empty, so the output will be of shape `[5, 6]` with values: 

[0, 1]: a
[0, 3]: b
[1, 0]: default_value
[2, 0]: c
[3, 1]: d
[4, 0]: default_value

The output `SparseTensor` will be in row-major order and will have the same shape as the input.

This op also returns an indicator vector shaped `[dense_shape[0]]` such that 

empty_row_indicator[i] = True iff row i was an empty row.

And a reverse index map vector shaped `[indices.shape[0]]` that is used during backpropagation, 

reverse_index_map[j] = out_j s.t. indices[j, :] == output_indices[out_j, :]

Arguments: 

	indices: 2-D. the indices of the sparse tensor.
	values: 1-D. the values of the sparse tensor.
	dense_shape: 1-D. the shape of the sparse tensor.
	default_value: 0-D. default value to insert into location `[row, 0, ..., 0]`
  for rows missing from the input sparse tensor.

output indices: 2-D. the indices of the filled sparse tensor.

Returns: 

output_indices
output_values: 1-D. the values of the filled sparse tensor.
empty_row_indicator: 1-D. whether the dense row was missing in the

input sparse tensor. 

reverse_index_map: 1-D. a map from the input indices to the output indices.

func SparseFillEmptyRowsGrad 

func SparseFillEmptyRowsGrad(scope *Scope, reverse_index_map tf.Output, grad_values tf.Output) (d_values tf.Output, d_default_value tf.Output)

The gradient of SparseFillEmptyRows.

Takes vectors reverse_index_map, shaped `[N]`, and grad_values, shaped `[N_full]`, where `N_full >= N` and copies data into either `d_values` or `d_default_value`. Here `d_values` is shaped `[N]` and `d_default_value` is a scalar. 

d_values[j] = grad_values[reverse_index_map[j]]
d_default_value = sum_{k : 0 .. N_full - 1} (
   grad_values[k] * 1{k not in reverse_index_map})

Arguments: 

reverse_index_map: 1-D.  The reverse index map from SparseFillEmptyRows.
grad_values: 1-D.  The gradients from backprop.

Returns: 

d_values: 1-D.  The backprop into values.
d_default_value: 0-D.  The backprop into default_value.

func SparseMatMul 

func SparseMatMul(scope *Scope, a tf.Output, b tf.Output, optional ...SparseMatMulAttr) (product tf.Output)

Multiply matrix "a" by matrix "b".

The inputs must be two-dimensional matrices and the inner dimension of "a" must match the outer dimension of "b". Both "a" and "b" must be `Tensor`s not `SparseTensor`s. This op is optimized for the case where at least one of "a" or "b" is sparse, in the sense that they have a large proportion of zero values. The breakeven for using this versus a dense matrix multiply on one platform was 30% zero values in the sparse matrix.

The gradient computation of this operation will only take advantage of sparsity in the input gradient when that gradient comes from a Relu.

func SparseMatrixAdd 

func SparseMatrixAdd(scope *Scope, a tf.Output, b tf.Output, alpha tf.Output, beta tf.Output) (c tf.Output)

Sparse addition of two CSR matrices, C = alpha * A + beta * B.

The gradients of SparseMatrixAdd outputs with respect to alpha and beta are not currently defined (TensorFlow will return zeros for these entries).

Arguments: 

a: A CSRSparseMatrix.
b: A CSRSparseMatrix.
alpha: A constant scalar.
beta: A constant scalar.

Returns A CSRSparseMatrix.

func SparseMatrixMatMul 

func SparseMatrixMatMul(scope *Scope, a tf.Output, b tf.Output, optional ...SparseMatrixMatMulAttr) (output tf.Output)

Matrix-multiplies a sparse matrix with a dense matrix.

Returns a dense matrix. For inputs A and B, where A is CSR and B is dense; this op returns a dense C;

If transpose_output is false, returns: ``` 

C = A . B

```

If transpose_output is `true`, returns: ``` 

C = transpose(A . B) = transpose(B) . transpose(A)

``` where the transposition is performed along the two innermost (matrix) dimensions.

If conjugate_output is `true`, returns: ``` 

C = conjugate(A . B) = conjugate(A) . conjugate(B)

```

If both conjugate_output and transpose_output are `true`, returns: ``` 

C = conjugate(transpose(A . B)) = conjugate(transpose(B)) .
                                  conjugate(transpose(A))

```

Arguments: 

a: A CSRSparseMatrix.
b: A dense tensor.

Returns A dense output tensor.

func SparseMatrixMul 

func SparseMatrixMul(scope *Scope, a tf.Output, b tf.Output) (output tf.Output)

Element-wise multiplication of a sparse matrix with a dense tensor.

Returns a sparse matrix.

The dense tensor `b` may be either a scalar; otherwise `a` must be a rank-3 `SparseMatrix`; in this case `b` must be shaped `[batch_size, 1, 1]` and the multiply operation broadcasts.

**NOTE** even if `b` is zero, the sparsity structure of the output does not change.

Arguments: 

a: A CSRSparseMatrix.
b: A dense tensor.

Returns A dense output tensor.

func SparseMatrixNNZ 

func SparseMatrixNNZ(scope *Scope, sparse_matrix tf.Output) (nnz tf.Output)

Returns the number of nonzeroes of `sparse_matrix`.

Arguments: 

sparse_matrix: A CSRSparseMatrix.

Returns The number of nonzeroes of `sparse_matrix`.

func SparseMatrixOrderingAMD 

func SparseMatrixOrderingAMD(scope *Scope, input tf.Output) (output tf.Output)

Computes the Approximate Minimum Degree (AMD) ordering of `input`.

Computes the Approximate Minimum Degree (AMD) ordering for a sparse matrix.

The returned permutation may be used to permute the rows and columns of the given sparse matrix. This typically results in permuted sparse matrix's sparse Cholesky (or other decompositions) in having fewer zero fill-in compared to decomposition of the original matrix.

The input sparse matrix may have rank 2 or rank 3. The output Tensor, representing would then have rank 1 or 2 respectively, with the same batch shape as the input.

Each component of the input sparse matrix must represent a square symmetric matrix; only the lower triangular part of the matrix is read. The values of the sparse matrix does not affect the returned permutation, only the sparsity pattern of the sparse matrix is used. Hence, a single AMD ordering may be reused for the Cholesky decompositions of sparse matrices with the same sparsity pattern but with possibly different values.

Each batch component of the output permutation represents a permutation of `N` elements, where the input sparse matrix components each have `N` rows. That is, the component contains each of the integers `{0, .. N-1}` exactly once. The `i`th element represents the row index that the `i`th row maps to.

Usage example:

```python 

from tensorflow.python.ops.linalg.sparse import sparse_csr_matrix_ops

a_indices = np.array([[0, 0], [1, 1], [2, 1], [2, 2], [3, 3]])
a_values = np.array([1.0, 2.0, 1.0, 3.0, 4.0], np.float32)
a_dense_shape = [4, 4]

with tf.Session() as sess:
  # Define (COO format) SparseTensor over Numpy array.
  a_st = tf.sparse.SparseTensor(a_indices, a_values, a_dense_shape)

  # Convert SparseTensors to CSR SparseMatrix.
  a_sm = sparse_csr_matrix_ops.sparse_tensor_to_csr_sparse_matrix(
      a_st.indices, a_st.values, a_st.dense_shape)

  # Obtain the AMD Ordering for the CSR SparseMatrix.
  ordering_amd = sparse_csr_matrix_ops.sparse_matrix_ordering_amd(sparse_matrix)

  ordering_amd_value = sess.run(ordering_amd)

```

`ordering_amd_value` stores the AMD ordering: `[1 2 3 0]`.

input: A `CSRSparseMatrix`.

Arguments: 

input: A `CSRSparseMatrix`.

Returns The Approximate Minimum Degree (AMD) ordering of `input`.

func SparseMatrixSoftmax 

func SparseMatrixSoftmax(scope *Scope, logits tf.Output, type_ tf.DataType) (softmax tf.Output)

Calculates the softmax of a CSRSparseMatrix.

Calculate the softmax of the innermost dimensions of a SparseMatrix.

Missing values are treated as `-inf` (i.e., logits of zero probability); and the output has the same sparsity structure as the input (though missing values in the output may now be treated as having probability zero).

Arguments: 

logits: A CSRSparseMatrix.

Returns A CSRSparseMatrix.

func SparseMatrixSoftmaxGrad 

func SparseMatrixSoftmaxGrad(scope *Scope, softmax tf.Output, grad_softmax tf.Output, type_ tf.DataType) (gradient tf.Output)

Calculates the gradient of the SparseMatrixSoftmax op.

Arguments: 

softmax: A CSRSparseMatrix.
grad_softmax: The gradient of `softmax`.

Returns The output gradient.

func SparseMatrixSparseCholesky 

func SparseMatrixSparseCholesky(scope *Scope, input tf.Output, permutation tf.Output, type_ tf.DataType) (output tf.Output)

Computes the sparse Cholesky decomposition of `input`.

Computes the Sparse Cholesky decomposition of a sparse matrix, with the given fill-in reducing permutation.

The input sparse matrix and the fill-in reducing permutation `permutation` must have compatible shapes. If the sparse matrix has rank 3; with the batch dimension `B`, then the `permutation` must be of rank 2; with the same batch dimension `B`. There is no support for broadcasting.

Furthermore, each component vector of `permutation` must be of length `N`, containing each of the integers {0, 1, ..., N - 1} exactly once, where `N` is the number of rows of each component of the sparse matrix.

Each component of the input sparse matrix must represent a symmetric positive definite (SPD) matrix; although only the lower triangular part of the matrix is read. If any individual component is not SPD, then an InvalidArgument error is thrown.

The returned sparse matrix has the same dense shape as the input sparse matrix. For each component `A` of the input sparse matrix, the corresponding output sparse matrix represents `L`, the lower triangular Cholesky factor satisfying the following identity:

``` 

A = L * Lt

```

where Lt denotes the transpose of L (or its conjugate transpose, if `type` is `complex64` or `complex128`).

The `type` parameter denotes the type of the matrix elements. The supported types are: `float32`, `float64`, `complex64` and `complex128`.

Usage example:

```python 

from tensorflow.python.ops.linalg.sparse import sparse_csr_matrix_ops

a_indices = np.array([[0, 0], [1, 1], [2, 1], [2, 2], [3, 3]])
a_values = np.array([1.0, 2.0, 1.0, 3.0, 4.0], np.float32)
a_dense_shape = [4, 4]

with tf.Session() as sess:
  # Define (COO format) SparseTensor over Numpy array.
  a_st = tf.sparse.SparseTensor(a_indices, a_values, a_dense_shape)

  # Convert SparseTensors to CSR SparseMatrix.
  a_sm = sparse_csr_matrix_ops.sparse_tensor_to_csr_sparse_matrix(
      a_st.indices, a_st.values, a_st.dense_shape)

  # Obtain the Sparse Cholesky factor using AMD Ordering for reducing zero
  # fill-in (number of structural non-zeros in the sparse Cholesky factor).
  ordering_amd = sparse_csr_matrix_ops.sparse_matrix_ordering_amd(sparse_matrix)
  cholesky_sparse_matrices = (
      sparse_csr_matrix_ops.sparse_matrix_sparse_cholesky(
          sparse_matrix, ordering_amd, type=tf.float32))

  # Convert the CSRSparseMatrix Cholesky factor to a dense Tensor
  dense_cholesky = sparse_csr_matrix_ops.csr_sparse_matrix_to_dense(
      cholesky_sparse_matrices, tf.float32)

  # Evaluate the dense Tensor value.
  dense_cholesky_value = sess.run(dense_cholesky)

```

`dense_cholesky_value` stores the dense Cholesky factor:

``` 

[[  1.  0.    0.    0.]
 [  0.  1.41  0.    0.]
 [  0.  0.70  1.58  0.]
 [  0.  0.    0.    2.]]

```

input: A `CSRSparseMatrix`. permutation: A `Tensor`. type: The type of `input`.

Arguments: 

input: A `CSRSparseMatrix`.
permutation: A fill-in reducing permutation matrix.

Returns The sparse Cholesky decompsition of `input`.

func SparseMatrixSparseMatMul 

func SparseMatrixSparseMatMul(scope *Scope, a tf.Output, b tf.Output, type_ tf.DataType, optional ...SparseMatrixSparseMatMulAttr) (c tf.Output)

Sparse-matrix-multiplies two CSR matrices `a` and `b`.

Performs a matrix multiplication of a sparse matrix `a` with a sparse matrix `b`; returns a sparse matrix `a * b`, unless either `a` or `b` is transposed or adjointed.

Each matrix may be transposed or adjointed (conjugated and transposed) according to the Boolean parameters `transpose_a`, `adjoint_a`, `transpose_b` and `adjoint_b`. At most one of `transpose_a` or `adjoint_a` may be True. Similarly, at most one of `transpose_b` or `adjoint_b` may be True.

The inputs must have compatible shapes. That is, the inner dimension of `a` must be equal to the outer dimension of `b`. This requirement is adjusted according to whether either `a` or `b` is transposed or adjointed.

The `type` parameter denotes the type of the matrix elements. Both `a` and `b` must have the same type. The supported types are: `float32`, `float64`, `complex64` and `complex128`.

Both `a` and `b` must have the same rank. Broadcasting is not supported. If they have rank 3, each batch of 2D CSRSparseMatrices within `a` and `b` must have the same dense shape.

The sparse matrix product may have numeric (non-structural) zeros. TODO(anudhyan): Consider adding a boolean attribute to control whether to prune zeros.

Usage example:

```python 

from tensorflow.python.ops.linalg.sparse import sparse_csr_matrix_ops

a_indices = np.array([[0, 0], [2, 3], [2, 4], [3, 0]])
a_values = np.array([1.0, 5.0, -1.0, -2.0], np.float32)
a_dense_shape = [4, 5]

b_indices = np.array([[0, 0], [3, 0], [3, 1]])
b_values = np.array([2.0, 7.0, 8.0], np.float32)
b_dense_shape = [5, 3]

with tf.Session() as sess:
  # Define (COO format) Sparse Tensors over Numpy arrays
  a_st = tf.sparse.SparseTensor(a_indices, a_values, a_dense_shape)
  b_st = tf.sparse.SparseTensor(b_indices, b_values, b_dense_shape)

  # Convert SparseTensors to CSR SparseMatrix
  a_sm = sparse_csr_matrix_ops.sparse_tensor_to_csr_sparse_matrix(
      a_st.indices, a_st.values, a_st.dense_shape)
  b_sm = sparse_csr_matrix_ops.sparse_tensor_to_csr_sparse_matrix(
      b_st.indices, b_st.values, b_st.dense_shape)

  # Compute the CSR SparseMatrix matrix multiplication
  c_sm = sparse_csr_matrix_ops.sparse_matrix_sparse_mat_mul(
      a=a_sm, b=b_sm, type=tf.float32)

  # Convert the CSR SparseMatrix product to a dense Tensor
  c_sm_dense = sparse_csr_matrix_ops.csr_sparse_matrix_to_dense(
      c_sm, tf.float32)
  # Evaluate the dense Tensor value
  c_sm_dense_value = sess.run(c_sm_dense)

```

`c_sm_dense_value` stores the dense matrix product:

``` 

[[  2.   0.   0.]
 [  0.   0.   0.]
 [ 35.  40.   0.]
 [ -4.   0.   0.]]

```

a: A `CSRSparseMatrix`. b: A `CSRSparseMatrix` with the same type and rank as `a`. type: The type of both `a` and `b`. transpose_a: If True, `a` transposed before multiplication. transpose_b: If True, `b` transposed before multiplication. adjoint_a: If True, `a` adjointed before multiplication. adjoint_b: If True, `b` adjointed before multiplication.

Arguments: 

a: A CSRSparseMatrix.
b: A CSRSparseMatrix.

Returns A CSRSparseMatrix.

func SparseMatrixTranspose 

func SparseMatrixTranspose(scope *Scope, input tf.Output, type_ tf.DataType, optional ...SparseMatrixTransposeAttr) (output tf.Output)

Transposes the inner (matrix) dimensions of a CSRSparseMatrix.

Transposes the inner (matrix) dimensions of a SparseMatrix and optionally conjugates its values.

Arguments: 

input: A CSRSparseMatrix.

Returns A CSRSparseMatrix.

func SparseMatrixZeros 

func SparseMatrixZeros(scope *Scope, dense_shape tf.Output, type_ tf.DataType) (sparse_matrix tf.Output)

Creates an all-zeros CSRSparseMatrix with shape `dense_shape`.

Arguments: 

dense_shape: The desired matrix shape.

Returns An empty CSR matrix with shape `dense_shape`.

func SparseReduceMax 

func SparseReduceMax(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceMaxAttr) (output tf.Output)

Computes the max of elements across dimensions of a SparseTensor.

This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_max()`. In particular, this Op also returns a dense `Tensor` instead of a sparse one.

Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.

Arguments: 

input_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

input_values: 1-D.  `N` non-empty values corresponding to `input_indices`.
input_shape: 1-D.  Shape of the input SparseTensor.
reduction_axes: 1-D.  Length-`K` vector containing the reduction axes.

Returns `R-K`-D. The reduced Tensor.

func SparseReduceMaxSparse 

func SparseReduceMaxSparse(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceMaxSparseAttr) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Computes the max of elements across dimensions of a SparseTensor.

This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_max()`. In contrast to SparseReduceMax, this Op returns a SparseTensor.

Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.

Arguments: 

input_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

input_values: 1-D.  `N` non-empty values corresponding to `input_indices`.
input_shape: 1-D.  Shape of the input SparseTensor.
reduction_axes: 1-D.  Length-`K` vector containing the reduction axes.

func SparseReduceSum 

func SparseReduceSum(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceSumAttr) (output tf.Output)

Computes the sum of elements across dimensions of a SparseTensor.

This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_sum()`. In particular, this Op also returns a dense `Tensor` instead of a sparse one.

Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.

Arguments: 

input_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

input_values: 1-D.  `N` non-empty values corresponding to `input_indices`.
input_shape: 1-D.  Shape of the input SparseTensor.
reduction_axes: 1-D.  Length-`K` vector containing the reduction axes.

Returns `R-K`-D. The reduced Tensor.

func SparseReduceSumSparse 

func SparseReduceSumSparse(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceSumSparseAttr) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Computes the sum of elements across dimensions of a SparseTensor.

This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_sum()`. In contrast to SparseReduceSum, this Op returns a SparseTensor.

Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.

Arguments: 

input_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

input_values: 1-D.  `N` non-empty values corresponding to `input_indices`.
input_shape: 1-D.  Shape of the input SparseTensor.
reduction_axes: 1-D.  Length-`K` vector containing the reduction axes.

func SparseReorder 

func SparseReorder(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output) (output_indices tf.Output, output_values tf.Output)

Reorders a SparseTensor into the canonical, row-major ordering.

Note that by convention, all sparse ops preserve the canonical ordering along increasing dimension number. The only time ordering can be violated is during manual manipulation of the indices and values vectors to add entries.

Reordering does not affect the shape of the SparseTensor.

If the tensor has rank `R` and `N` non-empty values, `input_indices` has shape `[N, R]`, input_values has length `N`, and input_shape has length `R`.

Arguments: 

input_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, possibly not in canonical ordering. 

input_values: 1-D.  `N` non-empty values corresponding to `input_indices`.
input_shape: 1-D.  Shape of the input SparseTensor.

Returns: 

output_indices: 2-D.  `N x R` matrix with the same indices as input_indices, but

in canonical row-major ordering. 

output_values: 1-D.  `N` non-empty values corresponding to `output_indices`.

func SparseReshape 

func SparseReshape(scope *Scope, input_indices tf.Output, input_shape tf.Output, new_shape tf.Output) (output_indices tf.Output, output_shape tf.Output)

Reshapes a SparseTensor to represent values in a new dense shape.

This operation has the same semantics as reshape on the represented dense tensor. The `input_indices` are recomputed based on the requested `new_shape`.

If one component of `new_shape` is the special value -1, the size of that dimension is computed so that the total dense size remains constant. At most one component of `new_shape` can be -1. The number of dense elements implied by `new_shape` must be the same as the number of dense elements originally implied by `input_shape`.

Reshaping does not affect the order of values in the SparseTensor.

If the input tensor has rank `R_in` and `N` non-empty values, and `new_shape` has length `R_out`, then `input_indices` has shape `[N, R_in]`, `input_shape` has length `R_in`, `output_indices` has shape `[N, R_out]`, and `output_shape` has length `R_out`.

Arguments: 

input_indices: 2-D.  `N x R_in` matrix with the indices of non-empty values in a

SparseTensor. 

input_shape: 1-D.  `R_in` vector with the input SparseTensor's dense shape.
new_shape: 1-D.  `R_out` vector with the requested new dense shape.

Returns: 

output_indices: 2-D.  `N x R_out` matrix with the updated indices of non-empty

values in the output SparseTensor. 

output_shape: 1-D.  `R_out` vector with the full dense shape of the output

SparseTensor. This is the same as `new_shape` but with any -1 dimensions filled in.

func SparseSegmentMean 

func SparseSegmentMean(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, optional ...SparseSegmentMeanAttr) (output tf.Output)

Computes the mean along sparse segments of a tensor.

See `tf.sparse.segment_sum` for usage examples.

Like `SegmentMean`, but `segment_ids` can have rank less than `data`'s first dimension, selecting a subset of dimension 0, specified by `indices`.

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SparseSegmentMeanGrad 

func SparseSegmentMeanGrad(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, output_dim0 tf.Output) (output tf.Output)

Computes gradients for SparseSegmentMean.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is output_dim0.

Arguments: 

grad: gradient propagated to the SparseSegmentMean op.
indices: indices passed to the corresponding SparseSegmentMean op.
segment_ids: segment_ids passed to the corresponding SparseSegmentMean op.
output_dim0: dimension 0 of "data" passed to SparseSegmentMean op.

func SparseSegmentMeanGradV2 added in v0.6.0 

func SparseSegmentMeanGradV2(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, dense_output_dim0 tf.Output) (output tf.Output, sorted_unique_indices tf.Output)

Computes gradients for SparseSegmentMean.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is the number of unique indexes in "indices". Also returns vector "sorted_unique_indices" containing the corresponding indexes from "indices".

Arguments: 

grad: gradient propagated to the SparseSegmentMean op.
indices: indices passed to the corresponding SparseSegmentMean op.
segment_ids: segment_ids passed to the corresponding SparseSegmentMean op.
dense_output_dim0: dimension 0 of "data" passed to SparseSegmentMean op.

func SparseSegmentMeanWithNumSegments 

func SparseSegmentMeanWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output, optional ...SparseSegmentMeanWithNumSegmentsAttr) (output tf.Output)

Computes the mean along sparse segments of a tensor.

Like `SparseSegmentMean`, but allows missing ids in `segment_ids`. If an id is missing, the `output` tensor at that position will be zeroed.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.
num_segments: Should equal the number of distinct segment IDs.

Returns Has same shape as data, except for dimension 0 which has size `num_segments`.

func SparseSegmentSqrtN 

func SparseSegmentSqrtN(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, optional ...SparseSegmentSqrtNAttr) (output tf.Output)

Computes the sum along sparse segments of a tensor divided by the sqrt of N.

N is the size of the segment being reduced.

See `tf.sparse.segment_sum` for usage examples.

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SparseSegmentSqrtNGrad 

func SparseSegmentSqrtNGrad(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, output_dim0 tf.Output) (output tf.Output)

Computes gradients for SparseSegmentSqrtN.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is output_dim0.

Arguments: 

grad: gradient propagated to the SparseSegmentSqrtN op.
indices: indices passed to the corresponding SparseSegmentSqrtN op.
segment_ids: segment_ids passed to the corresponding SparseSegmentSqrtN op.
output_dim0: dimension 0 of "data" passed to SparseSegmentSqrtN op.

func SparseSegmentSqrtNGradV2 added in v0.6.0 

func SparseSegmentSqrtNGradV2(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, dense_output_dim0 tf.Output) (output tf.Output, sorted_unique_indices tf.Output)

Computes gradients for SparseSegmentSqrtN.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is the number of unique indexes in "indices". Also returns vector "sorted_unique_indices" containing the corresponding indexes from "indices".

Arguments: 

grad: gradient propagated to the SparseSegmentSqrtN op.
indices: indices passed to the corresponding SparseSegmentSqrtN op.
segment_ids: segment_ids passed to the corresponding SparseSegmentSqrtN op.
dense_output_dim0: dimension 0 of "data" passed to SparseSegmentSqrtN op.

func SparseSegmentSqrtNWithNumSegments 

func SparseSegmentSqrtNWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output, optional ...SparseSegmentSqrtNWithNumSegmentsAttr) (output tf.Output)

Computes the sum along sparse segments of a tensor divided by the sqrt of N.

N is the size of the segment being reduced.

Like `SparseSegmentSqrtN`, but allows missing ids in `segment_ids`. If an id is missing, the `output` tensor at that position will be zeroed.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.
num_segments: Should equal the number of distinct segment IDs.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SparseSegmentSum 

func SparseSegmentSum(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, optional ...SparseSegmentSumAttr) (output tf.Output)

Computes the sum along sparse segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Like `SegmentSum`, but `segment_ids` can have rank less than `data`'s first dimension, selecting a subset of dimension 0, specified by `indices`.

For example:

```python c = tf.constant([[1,2,3,4], [-1,-2,-3,-4], [5,6,7,8]])

# Select two rows, one segment. tf.sparse_segment_sum(c, tf.constant([0, 1]), tf.constant([0, 0])) # => [[0 0 0 0]]

# Select two rows, two segment. tf.sparse_segment_sum(c, tf.constant([0, 1]), tf.constant([0, 1])) # => [[ 1 2 3 4] # [-1 -2 -3 -4]]

# Select all rows, two segments. tf.sparse_segment_sum(c, tf.constant([0, 1, 2]), tf.constant([0, 0, 1])) # => [[0 0 0 0] # [5 6 7 8]]

# Which is equivalent to: tf.segment_sum(c, tf.constant([0, 0, 1])) ```

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SparseSegmentSumGrad 

func SparseSegmentSumGrad(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, output_dim0 tf.Output) (output tf.Output)

Computes gradients for SparseSegmentSum.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is output_dim0.

Arguments: 

grad: gradient propagated to the SparseSegmentSum op.
indices: indices passed to the corresponding SparseSegmentSum op.
segment_ids: segment_ids passed to the corresponding SparseSegmentSum op.
output_dim0: dimension 0 of "data" passed to SparseSegmentSum op.

func SparseSegmentSumGradV2 added in v0.6.0 

func SparseSegmentSumGradV2(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, dense_output_dim0 tf.Output) (output tf.Output, sorted_unique_indices tf.Output)

Computes gradients for SparseSegmentSum.

Returns tensor "output" with same shape as grad, except for dimension 0 whose value is the number of unique indexes in "indices". Also returns vector "sorted_unique_indices" containing the corresponding indexes from "indices".

Arguments: 

grad: gradient propagated to the SparseSegmentSum op.
indices: indices passed to the corresponding SparseSegmentSum op.
segment_ids: segment_ids passed to the corresponding SparseSegmentSum op.
dense_output_dim0: dimension 0 of "data" passed to SparseSegmentSum op.

func SparseSegmentSumWithNumSegments 

func SparseSegmentSumWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output, optional ...SparseSegmentSumWithNumSegmentsAttr) (output tf.Output)

Computes the sum along sparse segments of a tensor.

Like `SparseSegmentSum`, but allows missing ids in `segment_ids`. If an id is missing, the `output` tensor at that position will be zeroed.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/sparse#Segmentation) for an explanation of segments.

For example:

```python c = tf.constant([[1,2,3,4], [-1,-2,-3,-4], [5,6,7,8]])

tf.sparse_segment_sum_with_num_segments( 

c, tf.constant([0, 1]), tf.constant([0, 0]), num_segments=3)

# => [[0 0 0 0] # [0 0 0 0] # [0 0 0 0]]

tf.sparse_segment_sum_with_num_segments(c, 

tf.constant([0, 1]),
tf.constant([0, 2],
num_segments=4))

# => [[ 1 2 3 4] # [ 0 0 0 0] # [-1 -2 -3 -4] # [ 0 0 0 0]] ```

Arguments: 

indices: A 1-D tensor. Has same rank as `segment_ids`.
segment_ids: A 1-D tensor. Values should be sorted and can be repeated.
num_segments: Should equal the number of distinct segment IDs.

Returns Has same shape as data, except for dimension 0 which has size `num_segments`.

func SparseSlice 

func SparseSlice(scope *Scope, indices tf.Output, values tf.Output, shape tf.Output, start tf.Output, size tf.Output) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)

Slice a `SparseTensor` based on the `start` and `size`.

For example, if the input is 

input_tensor = shape = [2, 7]
[    a   d e  ]
[b c          ]

Graphically the output tensors are: 

sparse_slice([0, 0], [2, 4]) = shape = [2, 4]
[    a  ]
[b c    ]

sparse_slice([0, 4], [2, 3]) = shape = [2, 3]
[ d e  ]
[      ]

Arguments: 

indices: 2-D tensor represents the indices of the sparse tensor.
values: 1-D tensor represents the values of the sparse tensor.
shape: 1-D. tensor represents the shape of the sparse tensor.
start: 1-D. tensor represents the start of the slice.
size: 1-D. tensor represents the size of the slice.

output indices: A list of 1-D tensors represents the indices of the output sparse tensors.

Returns: 

output_indices
output_values: A list of 1-D tensors represents the values of the output sparse

tensors. 

output_shape: A list of 1-D tensors represents the shape of the output sparse

tensors.

func SparseSliceGrad 

func SparseSliceGrad(scope *Scope, backprop_val_grad tf.Output, input_indices tf.Output, input_start tf.Output, output_indices tf.Output) (val_grad tf.Output)

The gradient operator for the SparseSlice op.

This op takes in the upstream gradient w.r.t. non-empty values of the sliced `SparseTensor`, and outputs the gradients w.r.t. the non-empty values of input `SparseTensor`.

Arguments: 

backprop_val_grad: 1-D. The gradient with respect to

the non-empty values of the sliced `SparseTensor`. 

input_indices: 2-D.  The `indices` of the input `SparseTensor`.
input_start: 1-D. tensor represents the start of the slice.
output_indices: 2-D.  The `indices` of the sliced `SparseTensor`.

Returns 1-D. The gradient with respect to the non-empty values of input `SparseTensor`.

func SparseSoftmax 

func SparseSoftmax(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output) (output tf.Output)

Applies softmax to a batched N-D `SparseTensor`.

The inputs represent an N-D SparseTensor with logical shape `[..., B, C]` (where `N >= 2`), and with indices sorted in the canonical lexicographic order.

This op is equivalent to applying the normal `tf.nn.softmax()` to each innermost logical submatrix with shape `[B, C]`, but with the catch that *the implicitly zero elements do not participate*. Specifically, the algorithm is equivalent to the following: 

(1) Applies `tf.nn.softmax()` to a densified view of each innermost submatrix
    with shape `[B, C]`, along the size-C dimension;
(2) Masks out the original implicitly-zero locations;
(3) Renormalizes the remaining elements.

Hence, the `SparseTensor` result has exactly the same non-zero indices and shape.

Arguments: 

sp_indices: 2-D.  `NNZ x R` matrix with the indices of non-empty values in a

SparseTensor, in canonical ordering. 

sp_values: 1-D.  `NNZ` non-empty values corresponding to `sp_indices`.
sp_shape: 1-D.  Shape of the input SparseTensor.

Returns 1-D. The `NNZ` values for the result `SparseTensor`.

func SparseSoftmaxCrossEntropyWithLogits 

func SparseSoftmaxCrossEntropyWithLogits(scope *Scope, features tf.Output, labels tf.Output) (loss tf.Output, backprop tf.Output)

Computes softmax cross entropy cost and gradients to backpropagate.

Unlike `SoftmaxCrossEntropyWithLogits`, this operation does not accept a matrix of label probabilities, but rather a single label per row of features. This label is considered to have probability 1.0 for the given row.

Inputs are the logits, not probabilities.

Arguments: 

features: batch_size x num_classes matrix
labels: batch_size vector with values in [0, num_classes).

This is the label for the given minibatch entry.

Returns: 

loss: Per example loss (batch_size vector).
backprop: backpropagated gradients (batch_size x num_classes matrix).

func SparseSparseMaximum 

func SparseSparseMaximum(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output) (output_indices tf.Output, output_values tf.Output)

Returns the element-wise max of two SparseTensors.

Assumes the two SparseTensors have the same shape, i.e., no broadcasting.

Arguments: 

a_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, in the canonical lexicographic ordering. 

a_values: 1-D.  `N` non-empty values corresponding to `a_indices`.
a_shape: 1-D.  Shape of the input SparseTensor.
b_indices: counterpart to `a_indices` for the other operand.
b_values: counterpart to `a_values` for the other operand; must be of the same dtype.
b_shape: counterpart to `a_shape` for the other operand; the two shapes must be equal.

Returns: 

output_indices: 2-D.  The indices of the output SparseTensor.
output_values: 1-D.  The values of the output SparseTensor.

func SparseSparseMinimum 

func SparseSparseMinimum(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output) (output_indices tf.Output, output_values tf.Output)

Returns the element-wise min of two SparseTensors.

Assumes the two SparseTensors have the same shape, i.e., no broadcasting.

Arguments: 

a_indices: 2-D.  `N x R` matrix with the indices of non-empty values in a

SparseTensor, in the canonical lexicographic ordering. 

a_values: 1-D.  `N` non-empty values corresponding to `a_indices`.
a_shape: 1-D.  Shape of the input SparseTensor.
b_indices: counterpart to `a_indices` for the other operand.
b_values: counterpart to `a_values` for the other operand; must be of the same dtype.
b_shape: counterpart to `a_shape` for the other operand; the two shapes must be equal.

Returns: 

output_indices: 2-D.  The indices of the output SparseTensor.
output_values: 1-D.  The values of the output SparseTensor.

func SparseSplit 

func SparseSplit(scope *Scope, split_dim tf.Output, indices tf.Output, values tf.Output, shape tf.Output, num_split int64) (output_indices []tf.Output, output_values []tf.Output, output_shape []tf.Output)

Split a `SparseTensor` into `num_split` tensors along one dimension.

If the `shape[split_dim]` is not an integer multiple of `num_split`. Slices `[0 : shape[split_dim] % num_split]` gets one extra dimension. For example, if `split_dim = 1` and `num_split = 2` and the input is 

input_tensor = shape = [2, 7]
[    a   d e  ]
[b c          ]

Graphically the output tensors are: 

output_tensor[0] = shape = [2, 4]
[    a  ]
[b c    ]

output_tensor[1] = shape = [2, 3]
[ d e  ]
[      ]

Arguments: 

split_dim: 0-D.  The dimension along which to split.  Must be in the range

`[0, rank(shape))`. 

indices: 2-D tensor represents the indices of the sparse tensor.
values: 1-D tensor represents the values of the sparse tensor.
shape: 1-D. tensor represents the shape of the sparse tensor.

output indices: A list of 1-D tensors represents the indices of the output sparse tensors. 

num_split: The number of ways to split.

Returns: 

output_indices
output_values: A list of 1-D tensors represents the values of the output sparse

tensors. 

output_shape: A list of 1-D tensors represents the shape of the output sparse

tensors.

func SparseTensorDenseAdd 

func SparseTensorDenseAdd(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b tf.Output) (output tf.Output)

Adds up a `SparseTensor` and a dense `Tensor`, producing a dense `Tensor`.

This Op does not require `a_indices` be sorted in standard lexicographic order.

Arguments: 

a_indices: 2-D.  The `indices` of the `SparseTensor`, with shape `[nnz, ndims]`.
a_values: 1-D.  The `values` of the `SparseTensor`, with shape `[nnz]`.
a_shape: 1-D.  The `shape` of the `SparseTensor`, with shape `[ndims]`.
b: `ndims`-D Tensor.  With shape `a_shape`.

func SparseTensorDenseMatMul 

func SparseTensorDenseMatMul(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b tf.Output, optional ...SparseTensorDenseMatMulAttr) (product tf.Output)

Multiply SparseTensor (of rank 2) "A" by dense matrix "B".

No validity checking is performed on the indices of A. However, the following input format is recommended for optimal behavior:

if adjoint_a == false: 

A should be sorted in lexicographically increasing order.  Use SparseReorder
if you're not sure.

if adjoint_a == true: 

A should be sorted in order of increasing dimension 1 (i.e., "column major"
order instead of "row major" order).

Arguments: 

a_indices: 2-D.  The `indices` of the `SparseTensor`, size `[nnz, 2]` Matrix.
a_values: 1-D.  The `values` of the `SparseTensor`, size `[nnz]` Vector.
a_shape: 1-D.  The `shape` of the `SparseTensor`, size `[2]` Vector.
b: 2-D.  A dense Matrix.

func SparseTensorSliceDataset 

func SparseTensorSliceDataset(scope *Scope, indices tf.Output, values tf.Output, dense_shape tf.Output) (handle tf.Output)

Creates a dataset that splits a SparseTensor into elements row-wise.

func SparseTensorToCSRSparseMatrix 

func SparseTensorToCSRSparseMatrix(scope *Scope, indices tf.Output, values tf.Output, dense_shape tf.Output) (sparse_matrix tf.Output)

Converts a SparseTensor to a (possibly batched) CSRSparseMatrix.

Arguments: 

indices: SparseTensor indices.
values: SparseTensor values.
dense_shape: SparseTensor dense shape.

Returns A (possibly batched) CSRSparseMatrix.

func SparseToDense 

func SparseToDense(scope *Scope, sparse_indices tf.Output, output_shape tf.Output, sparse_values tf.Output, default_value tf.Output, optional ...SparseToDenseAttr) (dense tf.Output)

Converts a sparse representation into a dense tensor.

Builds an array `dense` with shape `output_shape` such that

``` # If sparse_indices is scalar dense[i] = (i == sparse_indices ? sparse_values : default_value)

# If sparse_indices is a vector, then for each i dense[sparse_indices[i]] = sparse_values[i]

# If sparse_indices is an n by d matrix, then for each i in [0, n) dense[sparse_indices[i][0], ..., sparse_indices[i][d-1]] = sparse_values[i] ```

All other values in `dense` are set to `default_value`. If `sparse_values` is a scalar, all sparse indices are set to this single value.

Indices should be sorted in lexicographic order, and indices must not contain any repeats. If `validate_indices` is true, these properties are checked during execution.

Arguments: 

sparse_indices: 0-D, 1-D, or 2-D.  `sparse_indices[i]` contains the complete

index where `sparse_values[i]` will be placed. 

output_shape: 1-D.  Shape of the dense output tensor.
sparse_values: 1-D.  Values corresponding to each row of `sparse_indices`,

or a scalar value to be used for all sparse indices. 

default_value: Scalar value to set for indices not specified in

`sparse_indices`.

Returns Dense output tensor of shape `output_shape`.

func SparseToSparseSetOperation 

func SparseToSparseSetOperation(scope *Scope, set1_indices tf.Output, set1_values tf.Output, set1_shape tf.Output, set2_indices tf.Output, set2_values tf.Output, set2_shape tf.Output, set_operation string, optional ...SparseToSparseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)

Applies set operation along last dimension of 2 `SparseTensor` inputs.

See SetOperationOp::SetOperationFromContext for values of `set_operation`.

If `validate_indices` is `True`, `SparseToSparseSetOperation` validates the order and range of `set1` and `set2` indices.

Input `set1` is a `SparseTensor` represented by `set1_indices`, `set1_values`, and `set1_shape`. For `set1` ranked `n`, 1st `n-1` dimensions must be the same as `set2`. Dimension `n` contains values in a set, duplicates are allowed but ignored.

Input `set2` is a `SparseTensor` represented by `set2_indices`, `set2_values`, and `set2_shape`. For `set2` ranked `n`, 1st `n-1` dimensions must be the same as `set1`. Dimension `n` contains values in a set, duplicates are allowed but ignored.

If `validate_indices` is `True`, this op validates the order and range of `set1` and `set2` indices.

Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.

Arguments: 

set1_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major

order. 

set1_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major

order. 

set1_shape: 1D `Tensor`, shape of a `SparseTensor`. `set1_shape[0...n-1]` must

be the same as `set2_shape[0...n-1]`, `set1_shape[n]` is the max set size across `0...n-1` dimensions. 

set2_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major

order. 

set2_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major

order. 

set2_shape: 1D `Tensor`, shape of a `SparseTensor`. `set2_shape[0...n-1]` must

be the same as `set1_shape[0...n-1]`, `set2_shape[n]` is the max set size across `0...n-1` dimensions.

Returns: 

result_indices: 2D indices of a `SparseTensor`.
result_values: 1D values of a `SparseTensor`.
result_shape: 1D `Tensor` shape of a `SparseTensor`. `result_shape[0...n-1]` is

the same as the 1st `n-1` dimensions of `set1` and `set2`, `result_shape[n]` is the max result set size across all `0...n-1` dimensions.

func Split 

func Split(scope *Scope, axis tf.Output, value tf.Output, num_split int64) (output []tf.Output)

Splits a tensor into `num_split` tensors along one dimension.

Arguments: 

axis: 0-D.  The dimension along which to split.  Must be in the range

`[-rank(value), rank(value))`. 

value: The tensor to split.
num_split: The number of ways to split.  Must evenly divide

`value.shape[split_dim]`.

Returns They are identically shaped tensors, whose shape matches that of `value` except along `axis`, where their sizes are `values.shape[split_dim] / num_split`.

func SplitDedupData added in v0.4.0 

func SplitDedupData(scope *Scope, input tf.Output, integer_type tf.DataType, float_type tf.DataType, tuple_mask string, optional ...SplitDedupDataAttr) (integer_tensor tf.Output, float_tensor tf.Output)

An op splits input deduplication data XLA tuple into integer and floating point tensors.

Deduplication data is an XLA tuple, which consists of integer and floating point values. This op is to split these values into two groups for two types, and construct each group as one tensor to return.

Arguments: 

input: An XLA tuple including integer and float elements as deduplication data tuple.
integer_type: integer_tensor type. Allowed types: int32, int64, uint32, uint64.
float_type: float_tensor type. Allowed types: half, bfloat16, float.
tuple_mask: A serialized TensorProto string of output tuple mask. This mask is a 2-D tensor,

with first column as tuple element type, and second column as span of this type. For example, an output tuple of (1, 2, 0.1, 3), its mask is [[0, 2], [1, 1], [0, 1]]. We expect only two types of elements: integer(0) and float(1).

Returns: 

integer_tensor: A 1-D integer tensor, includes integer elements of deduplication data tuple.
float_tensor: A 1-D float tensor, includes float elements of deduplication data tuple.

func SplitV 

func SplitV(scope *Scope, value tf.Output, size_splits tf.Output, axis tf.Output, num_split int64) (output []tf.Output)

Splits a tensor into `num_split` tensors along one dimension.

Arguments: 

value: The tensor to split.
size_splits: list containing the sizes of each output tensor along the split

dimension. Must sum to the dimension of value along split_dim. Can contain one -1 indicating that dimension is to be inferred. 

axis: 0-D.  The dimension along which to split.  Must be in the range

`[-rank(value), rank(value))`.

Returns Tensors whose shape matches that of `value` except along `axis`, where their sizes are `size_splits[i]`.

func SqlDataset 

func SqlDataset(scope *Scope, driver_name tf.Output, data_source_name tf.Output, query tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that executes a SQL query and emits rows of the result set.

Arguments: 

driver_name: The database type. Currently, the only supported type is 'sqlite'.
data_source_name: A connection string to connect to the database.
query: A SQL query to execute.

func Sqrt 

func Sqrt(scope *Scope, x tf.Output) (y tf.Output)

Computes square root of x element-wise.

I.e., \\(y = \sqrt{x} = x^{1/2}\\).

func SqrtGrad 

func SqrtGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient for the sqrt of `x` wrt its input.

Specifically, `grad = dy * 0.5 / y`, where `y = sqrt(x)`, and `dy` is the corresponding input gradient.

func Square 

func Square(scope *Scope, x tf.Output) (y tf.Output)

Computes square of x element-wise.

I.e., \\(y = x * x = x^2\\).

func SquaredDifference 

func SquaredDifference(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns conj(x - y)(x - y) element-wise.

*NOTE*: `SquaredDifference` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func Squeeze 

func Squeeze(scope *Scope, input tf.Output, optional ...SqueezeAttr) (output tf.Output)

Removes dimensions of size 1 from the shape of a tensor.

Given a tensor `input`, this operation returns a tensor of the same type with all dimensions of size 1 removed. If you don't want to remove all size 1 dimensions, you can remove specific size 1 dimensions by specifying `axis`.

For example:

``` # 't' is a tensor of shape [1, 2, 1, 3, 1, 1] shape(squeeze(t)) ==> [2, 3] ```

Or, to remove specific size 1 dimensions:

``` # 't' is a tensor of shape [1, 2, 1, 3, 1, 1] shape(squeeze(t, [2, 4])) ==> [1, 2, 3, 1] ```

Arguments: 

input: The `input` to squeeze.

Returns Contains the same data as `input`, but has one or more dimensions of size 1 removed.

func StackCloseV2 

func StackCloseV2(scope *Scope, handle tf.Output) (o *tf.Operation)

Delete the stack from its resource container.

Arguments: 

handle: The handle to a stack.

Returns the created operation.

func StackPopV2 

func StackPopV2(scope *Scope, handle tf.Output, elem_type tf.DataType) (elem tf.Output)

Pop the element at the top of the stack.

Arguments: 

handle: The handle to a stack.
elem_type: The type of the elem that is popped.

Returns The tensor that is popped from the top of the stack.

func StackPushV2 

func StackPushV2(scope *Scope, handle tf.Output, elem tf.Output, optional ...StackPushV2Attr) (output tf.Output)

Push an element onto the stack.

Arguments: 

handle: The handle to a stack.
elem: The tensor to be pushed onto the stack.

Returns The same tensor as the input 'elem'.

func StackV2 

func StackV2(scope *Scope, max_size tf.Output, elem_type tf.DataType, optional ...StackV2Attr) (handle tf.Output)

A stack that produces elements in first-in last-out order.

Arguments: 

max_size: The maximum size of the stack if non-negative. If negative, the stack

size is unlimited. 

elem_type: The type of the elements on the stack.

Returns The handle to the stack.

func Stage 

func Stage(scope *Scope, values []tf.Output, optional ...StageAttr) (o *tf.Operation)

Stage values similar to a lightweight Enqueue.

The basic functionality of this Op is similar to a queue with many fewer capabilities and options. This Op is optimized for performance.

Arguments: 

values: a list of tensors

dtypes A list of data types that inserted values should adhere to.

Returns the created operation.

func StageClear 

func StageClear(scope *Scope, dtypes []tf.DataType, optional ...StageClearAttr) (o *tf.Operation)

Op removes all elements in the underlying container.

Returns the created operation.

func StagePeek 

func StagePeek(scope *Scope, index tf.Output, dtypes []tf.DataType, optional ...StagePeekAttr) (values []tf.Output)

Op peeks at the values at the specified index. If the

underlying container does not contain sufficient elements this op will block until it does. This Op is optimized for performance.

func StageSize 

func StageSize(scope *Scope, dtypes []tf.DataType, optional ...StageSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func StatefulStandardNormal 

func StatefulStandardNormal(scope *Scope, resource tf.Output, shape tf.Output, optional ...StatefulStandardNormalAttr) (output tf.Output)

Outputs random values from a normal distribution. This op is deprecated in favor of op 'StatefulStandardNormalV2'

DEPRECATED at GraphDef version 29: Use StatefulStandardNormalV2 instead

The generated values will have mean 0 and standard deviation 1.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
shape: The shape of the output tensor.

Returns A tensor of the specified shape filled with random normal values.

func StatefulStandardNormalV2 

func StatefulStandardNormalV2(scope *Scope, resource tf.Output, algorithm tf.Output, shape tf.Output, optional ...StatefulStandardNormalV2Attr) (output tf.Output)

Outputs random values from a normal distribution.

The generated values will have mean 0 and standard deviation 1.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
shape: The shape of the output tensor.

Returns A tensor of the specified shape filled with random normal values.

func StatefulTruncatedNormal 

func StatefulTruncatedNormal(scope *Scope, resource tf.Output, algorithm tf.Output, shape tf.Output, optional ...StatefulTruncatedNormalAttr) (output tf.Output)

Outputs random values from a truncated normal distribution.

The generated values follow a normal distribution with mean 0 and standard deviation 1, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
shape: The shape of the output tensor.

Returns Random values with specified shape.

func StatefulUniform 

func StatefulUniform(scope *Scope, resource tf.Output, algorithm tf.Output, shape tf.Output, optional ...StatefulUniformAttr) (output tf.Output)

Outputs random values from a uniform distribution.

The generated values follow a uniform distribution in the range `[0, 1)`. The lower bound 0 is included in the range, while the upper bound 1 is excluded.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
shape: The shape of the output tensor.

Returns Random values with specified shape.

func StatefulUniformFullInt 

func StatefulUniformFullInt(scope *Scope, resource tf.Output, algorithm tf.Output, shape tf.Output, optional ...StatefulUniformFullIntAttr) (output tf.Output)

Outputs random integers from a uniform distribution.

The generated values are uniform integers covering the whole range of `dtype`.

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
shape: The shape of the output tensor.

Returns Random values with specified shape.

func StatefulUniformInt 

func StatefulUniformInt(scope *Scope, resource tf.Output, algorithm tf.Output, shape tf.Output, minval tf.Output, maxval tf.Output) (output tf.Output)

Outputs random integers from a uniform distribution.

The generated values are uniform integers in the range `[minval, maxval)`. The lower bound `minval` is included in the range, while the upper bound `maxval` is excluded.

The random integers are slightly biased unless `maxval - minval` is an exact power of two. The bias is small for values of `maxval - minval` significantly smaller than the range of the output (either `2^32` or `2^64`).

Arguments: 

resource: The handle of the resource variable that stores the state of the RNG.
algorithm: The RNG algorithm.
shape: The shape of the output tensor.
minval: Minimum value (inclusive, scalar).
maxval: Maximum value (exclusive, scalar).

Returns Random values with specified shape.

func StatelessMultinomial 

func StatelessMultinomial(scope *Scope, logits tf.Output, num_samples tf.Output, seed tf.Output, optional ...StatelessMultinomialAttr) (output tf.Output)

Draws samples from a multinomial distribution.

Arguments: 

logits: 2-D Tensor with shape `[batch_size, num_classes]`.  Each slice `[i, :]`

represents the unnormalized log probabilities for all classes. 

num_samples: 0-D.  Number of independent samples to draw for each row slice.
seed: 2 seeds (shape [2]).

Returns 2-D Tensor with shape `[batch_size, num_samples]`. Each slice `[i, :]` contains the drawn class labels with range `[0, num_classes)`.

func StatelessRandomBinomial 

func StatelessRandomBinomial(scope *Scope, shape tf.Output, seed tf.Output, counts tf.Output, probs tf.Output, optional ...StatelessRandomBinomialAttr) (output tf.Output)

Outputs deterministic pseudorandom random numbers from a binomial distribution.

Outputs random values from a binomial distribution.

The outputs are a deterministic function of `shape`, `seed`, `counts`, and `probs`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).
counts: The counts of the binomial distribution. Must be broadcastable with `probs`,

and broadcastable with the rightmost dimensions of `shape`. 

probs: The probability of success for the binomial distribution. Must be broadcastable

with `counts` and broadcastable with the rightmost dimensions of `shape`.

Returns Random values with specified shape.

func StatelessRandomGammaV2 

func StatelessRandomGammaV2(scope *Scope, shape tf.Output, seed tf.Output, alpha tf.Output) (output tf.Output)

Outputs deterministic pseudorandom random numbers from a gamma distribution.

Outputs random values from a gamma distribution.

The outputs are a deterministic function of `shape`, `seed`, and `alpha`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).
alpha: The concentration of the gamma distribution. Shape must match the rightmost

dimensions of `shape`.

Returns Random values with specified shape.

func StatelessRandomGammaV3 added in v0.4.0 

func StatelessRandomGammaV3(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, alpha tf.Output) (output tf.Output)

Outputs deterministic pseudorandom random numbers from a gamma distribution.

Outputs random values from a gamma distribution.

The outputs are a deterministic function of the inputs.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).
alpha: The concentration of the gamma distribution. Shape must match the rightmost

dimensions of `shape`.

Returns Random values with specified shape.

func StatelessRandomGetAlg 

func StatelessRandomGetAlg(scope *Scope) (alg tf.Output)

Picks the best counter-based RNG algorithm based on device.

This op picks the best counter-based RNG algorithm based on device.

Returns The RNG algorithm (shape int32[]).

func StatelessRandomGetKeyCounter 

func StatelessRandomGetKeyCounter(scope *Scope, seed tf.Output) (key tf.Output, counter tf.Output)

Scrambles seed into key and counter, using the best algorithm based on device.

This op scrambles a shape-[2] seed into a key and a counter, both needed by counter-based RNG algorithms. The scrambing uses the best algorithm based on device. The scrambling is opaque but approximately satisfies the property that different seed results in different key/counter pair (which will in turn result in different random numbers).

Arguments: 

seed: 2 seeds (shape [2]).

Returns: 

key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Counter for the counter-based RNG algorithm. Since counter size is algorithm-dependent, this output will be right-padded with zeros to reach shape uint64[2] (the current maximal counter size among algorithms).

func StatelessRandomGetKeyCounterAlg 

func StatelessRandomGetKeyCounterAlg(scope *Scope, seed tf.Output) (key tf.Output, counter tf.Output, alg tf.Output)

Picks the best algorithm based on device, and scrambles seed into key and counter.

This op picks the best counter-based RNG algorithm based on device, and scrambles a shape-[2] seed into a key and a counter, both needed by the counter-based algorithm. The scrambling is opaque but approximately satisfies the property that different seed results in different key/counter pair (which will in turn result in different random numbers).

Arguments: 

seed: 2 seeds (shape [2]).

Returns: 

key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Counter for the counter-based RNG algorithm. Since counter size is algorithm-dependent, this output will be right-padded with zeros to reach shape uint64[2] (the current maximal counter size among algorithms).
alg: The RNG algorithm (shape int32[]).

func StatelessRandomNormal 

func StatelessRandomNormal(scope *Scope, shape tf.Output, seed tf.Output, optional ...StatelessRandomNormalAttr) (output tf.Output)

Outputs deterministic pseudorandom values from a normal distribution.

The generated values will have mean 0 and standard deviation 1.

The outputs are a deterministic function of `shape` and `seed`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).

Returns Random values with specified shape.

func StatelessRandomNormalV2 

func StatelessRandomNormalV2(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, optional ...StatelessRandomNormalV2Attr) (output tf.Output)

Outputs deterministic pseudorandom values from a normal distribution.

The generated values will have mean 0 and standard deviation 1.

The outputs are a deterministic function of `shape`, `key`, `counter` and `alg`.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).

Returns Random values with specified shape.

func StatelessRandomPoisson 

func StatelessRandomPoisson(scope *Scope, shape tf.Output, seed tf.Output, lam tf.Output, dtype tf.DataType) (output tf.Output)

Outputs deterministic pseudorandom random numbers from a Poisson distribution.

Outputs random values from a Poisson distribution.

The outputs are a deterministic function of `shape`, `seed`, and `lam`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).
lam: The rate of the Poisson distribution. Shape must match the rightmost dimensions

of `shape`. 

dtype: The type of the output.

Returns Random values with specified shape.

func StatelessRandomUniform 

func StatelessRandomUniform(scope *Scope, shape tf.Output, seed tf.Output, optional ...StatelessRandomUniformAttr) (output tf.Output)

Outputs deterministic pseudorandom random values from a uniform distribution.

The generated values follow a uniform distribution in the range `[0, 1)`. The lower bound 0 is included in the range, while the upper bound 1 is excluded.

The outputs are a deterministic function of `shape` and `seed`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).

Returns Random values with specified shape.

func StatelessRandomUniformFullInt 

func StatelessRandomUniformFullInt(scope *Scope, shape tf.Output, seed tf.Output, optional ...StatelessRandomUniformFullIntAttr) (output tf.Output)

Outputs deterministic pseudorandom random integers from a uniform distribution.

The generated values are uniform integers covering the whole range of `dtype`.

The outputs are a deterministic function of `shape` and `seed`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).

Returns Random values with specified shape.

func StatelessRandomUniformFullIntV2 

func StatelessRandomUniformFullIntV2(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, optional ...StatelessRandomUniformFullIntV2Attr) (output tf.Output)

Outputs deterministic pseudorandom random integers from a uniform distribution.

The generated values are uniform integers covering the whole range of `dtype`.

The outputs are a deterministic function of `shape`, `key`, `counter` and `alg`.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).

Returns Random values with specified shape.

func StatelessRandomUniformInt 

func StatelessRandomUniformInt(scope *Scope, shape tf.Output, seed tf.Output, minval tf.Output, maxval tf.Output) (output tf.Output)

Outputs deterministic pseudorandom random integers from a uniform distribution.

The generated values follow a uniform distribution in the range `[minval, maxval)`.

The outputs are a deterministic function of `shape`, `seed`, `minval`, and `maxval`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).
minval: Minimum value (inclusive, scalar).
maxval: Maximum value (exclusive, scalar).

Returns Random values with specified shape.

func StatelessRandomUniformIntV2 

func StatelessRandomUniformIntV2(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, minval tf.Output, maxval tf.Output) (output tf.Output)

Outputs deterministic pseudorandom random integers from a uniform distribution.

The generated values follow a uniform distribution in the range `[minval, maxval)`.

The outputs are a deterministic function of `shape`, `key`, `counter`, `alg`, `minval` and `maxval`.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).
minval: Minimum value (inclusive, scalar).
maxval: Maximum value (exclusive, scalar).

Returns Random values with specified shape.

func StatelessRandomUniformV2 

func StatelessRandomUniformV2(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, optional ...StatelessRandomUniformV2Attr) (output tf.Output)

Outputs deterministic pseudorandom random values from a uniform distribution.

The generated values follow a uniform distribution in the range `[0, 1)`. The lower bound 0 is included in the range, while the upper bound 1 is excluded.

The outputs are a deterministic function of `shape`, `key`, `counter` and `alg`.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).

Returns Random values with specified shape.

func StatelessSampleDistortedBoundingBox 

func StatelessSampleDistortedBoundingBox(scope *Scope, image_size tf.Output, bounding_boxes tf.Output, min_object_covered tf.Output, seed tf.Output, optional ...StatelessSampleDistortedBoundingBoxAttr) (begin tf.Output, size tf.Output, bboxes tf.Output)

Generate a randomly distorted bounding box for an image deterministically.

Bounding box annotations are often supplied in addition to ground-truth labels in image recognition or object localization tasks. A common technique for training such a system is to randomly distort an image while preserving its content, i.e. *data augmentation*. This Op, given the same `seed`, deterministically outputs a randomly distorted localization of an object, i.e. bounding box, given an `image_size`, `bounding_boxes` and a series of constraints.

The output of this Op is a single bounding box that may be used to crop the original image. The output is returned as 3 tensors: `begin`, `size` and `bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize what the bounding box looks like.

Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and the height of the underlying image.

The output of this Op is guaranteed to be the same given the same `seed` and is independent of how many times the function is called, and independent of global seed settings (e.g. `tf.random.set_seed`).

Example usage:

>>> image = np.array([[[1], [2], [3]], [[4], [5], [6]], [[7], [8], [9]]]) >>> bbox = tf.constant( ... [0.0, 0.0, 1.0, 1.0], dtype=tf.float32, shape=[1, 1, 4]) >>> seed = (1, 2) >>> # Generate a single distorted bounding box. >>> bbox_begin, bbox_size, bbox_draw = ( ... tf.image.stateless_sample_distorted_bounding_box( ... tf.shape(image), bounding_boxes=bbox, seed=seed)) >>> # Employ the bounding box to distort the image. >>> tf.slice(image, bbox_begin, bbox_size) <tf.Tensor: shape=(2, 2, 1), dtype=int64, numpy= array([[[1], 

 [2]],
[[4],
 [5]]])>

>>> # Draw the bounding box in an image summary. >>> colors = np.array([[1.0, 0.0, 0.0], [0.0, 0.0, 1.0]]) >>> tf.image.draw_bounding_boxes( ... tf.expand_dims(tf.cast(image, tf.float32),0), bbox_draw, colors) <tf.Tensor: shape=(1, 3, 3, 1), dtype=float32, numpy= array([[[[1.], 

 [1.],
 [3.]],
[[1.],
 [1.],
 [6.]],
[[7.],
 [8.],
 [9.]]]], dtype=float32)>

Note that if no bounding box information is available, setting `use_image_if_no_bounding_boxes = true` will assume there is a single implicit bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is false and no bounding boxes are supplied, an error is raised.

Arguments: 

image_size: 1-D, containing `[height, width, channels]`.
bounding_boxes: 3-D with shape `[batch, N, 4]` describing the N bounding boxes

associated with the image. 

min_object_covered: The cropped area of the image must contain at least this

fraction of any bounding box supplied. The value of this parameter should be non-negative. In the case of 0, the cropped area does not need to overlap any of the bounding boxes supplied. 

seed: 1-D with shape `[2]`. The seed to the random number generator. Must have dtype

`int32` or `int64`. (When using XLA, only `int32` is allowed.)

Returns: 

begin: 1-D, containing `[offset_height, offset_width, 0]`. Provide as input to

`tf.slice`. 

size: 1-D, containing `[target_height, target_width, -1]`. Provide as input to

`tf.slice`. 

bboxes: 3-D with shape `[1, 1, 4]` containing the distorted bounding box.

Provide as input to `tf.image.draw_bounding_boxes`.

func StatelessShuffle added in v0.2.0 

func StatelessShuffle(scope *Scope, value tf.Output, key tf.Output, counter tf.Output, alg tf.Output) (output tf.Output)

Randomly and deterministically shuffles a tensor along its first dimension.

The tensor is shuffled along dimension 0, such that each `value[j]` is mapped to one and only one `output[i]`. For example, a mapping that might occur for a 3x2 tensor is:

``` [[1, 2], [[5, 6], 

[3, 4],  ==>   [1, 2],
[5, 6]]        [3, 4]]

```

The outputs are a deterministic function of `value`, `key`, `counter` and `alg`.

Arguments: 

value: The tensor to be shuffled.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).

Returns A tensor of same shape and type as `value`, shuffled along its first dimension.

func StatelessTruncatedNormal 

func StatelessTruncatedNormal(scope *Scope, shape tf.Output, seed tf.Output, optional ...StatelessTruncatedNormalAttr) (output tf.Output)

Outputs deterministic pseudorandom values from a truncated normal distribution.

The generated values follow a normal distribution with mean 0 and standard deviation 1, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.

The outputs are a deterministic function of `shape` and `seed`.

Arguments: 

shape: The shape of the output tensor.
seed: 2 seeds (shape [2]).

Returns Random values with specified shape.

func StatelessTruncatedNormalV2 

func StatelessTruncatedNormalV2(scope *Scope, shape tf.Output, key tf.Output, counter tf.Output, alg tf.Output, optional ...StatelessTruncatedNormalV2Attr) (output tf.Output)

Outputs deterministic pseudorandom values from a truncated normal distribution.

The generated values follow a normal distribution with mean 0 and standard deviation 1, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.

The outputs are a deterministic function of `shape`, `key`, `counter` and `alg`.

Arguments: 

shape: The shape of the output tensor.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).

Returns Random values with specified shape.

func StaticRegexFullMatch 

func StaticRegexFullMatch(scope *Scope, input tf.Output, pattern string) (output tf.Output)

Check if the input matches the regex pattern.

The input is a string tensor of any shape. The pattern is the regular expression to be matched with every element of the input tensor. The boolean values (True or False) of the output tensor indicate if the input matches the regex pattern provided.

The pattern follows the re2 syntax (https://github.com/google/re2/wiki/Syntax)

Arguments: 

input: A string tensor of the text to be processed.
pattern: The regular expression to match the input.

Returns A bool tensor with the same shape as `input`.

func StaticRegexReplace 

func StaticRegexReplace(scope *Scope, input tf.Output, pattern string, rewrite string, optional ...StaticRegexReplaceAttr) (output tf.Output)

Replaces the match of pattern in input with rewrite.

It follows the re2 syntax (https://github.com/google/re2/wiki/Syntax)

Arguments: 

input: The text to be processed.
pattern: The regular expression to match the input.
rewrite: The rewrite to be applied to the matched expression.

Returns The text after applying pattern and rewrite.

func StatsAggregatorHandle 

func StatsAggregatorHandle(scope *Scope, optional ...StatsAggregatorHandleAttr) (handle tf.Output)

Creates a statistics manager resource.

func StatsAggregatorSetSummaryWriter 

func StatsAggregatorSetSummaryWriter(scope *Scope, stats_aggregator tf.Output, summary tf.Output) (o *tf.Operation)

Set a summary_writer_interface to record statistics using given stats_aggregator.

Returns the created operation.

func StatsAggregatorSummary 

func StatsAggregatorSummary(scope *Scope, iterator tf.Output) (summary tf.Output)

Produces a summary of any statistics recorded by the given statistics manager.

func StochasticCastToInt added in v0.5.0 

func StochasticCastToInt(scope *Scope, input tf.Output, key tf.Output, counter tf.Output, alg tf.Output, Tout tf.DataType) (output tf.Output)

Stochastically cast a given tensor from floats to ints.

The values are cast with a deterministic pseudo-random tensor from a uniform distribution generated from user given key, counter, algorithm. Values will saturate if out of the specified integer type range, and will become zero if inputs are NaN.

The outputs are a deterministic function of `input`, `key`, `counter`, `alg`.

Arguments: 

input: The operand to stochastically cast to int.
key: Key for the counter-based RNG algorithm (shape uint64[1]).
counter: Initial counter for the counter-based RNG algorithm (shape uint64[2] or uint64[1] depending on the algorithm). If a larger vector is given, only the needed portion on the left (i.e. [:N]) will be used.
alg: The RNG algorithm (shape int32[]).
Tout: The type of the output.

Returns The cast result with the same shape as the input.

func StopGradient 

func StopGradient(scope *Scope, input tf.Output) (output tf.Output)

Stops gradient computation.

When executed in a graph, this op outputs its input tensor as-is.

When building ops to compute gradients, this op prevents the contribution of its inputs to be taken into account. Normally, the gradient generator adds ops to a graph to compute the derivatives of a specified 'loss' by recursively finding out inputs that contributed to its computation. If you insert this op in the graph it inputs are masked from the gradient generator. They are not taken into account for computing gradients.

This is useful any time you want to compute a value with TensorFlow but need to pretend that the value was a constant. For example, the softmax function for a vector x can be written as

```python 

def softmax(x):
  numerator = tf.exp(x)
  denominator = tf.reduce_sum(numerator)
  return numerator / denominator

```

This however is susceptible to overflow if the values in x are large. An alternative more stable way is to subtract the maximum of x from each of the values.

```python 

def stable_softmax(x):
  z = x - tf.reduce_max(x)
  numerator = tf.exp(z)
  denominator = tf.reduce_sum(numerator)
  return numerator / denominator

```

However, when we backprop through the softmax to x, we dont want to backprop through the `tf.reduce_max(x)` (if the max values are not unique then the gradient could flow to the wrong input) calculation and treat that as a constant. Therefore, we should write this out as

```python 

def stable_softmax(x):
  z = x - tf.stop_gradient(tf.reduce_max(x))
  numerator = tf.exp(z)
  denominator = tf.reduce_sum(numerator)
  return numerator / denominator

```

Some other examples include:

  • The *EM* algorithm where the *M-step* should not involve backpropagation through the output of the *E-step*.
  • Contrastive divergence training of Boltzmann machines where, when differentiating the energy function, the training must not backpropagate through the graph that generated the samples from the model.
  • Adversarial training, where no backprop should happen through the adversarial example generation process.

func StridedSlice 

func StridedSlice(scope *Scope, input tf.Output, begin tf.Output, end tf.Output, strides tf.Output, optional ...StridedSliceAttr) (output tf.Output)

Return a strided slice from `input`.

Note, most python users will want to use the Python `Tensor.__getitem__` or `Variable.__getitem__` rather than this op directly.

The goal of this op is to produce a new tensor with a subset of the elements from the `n` dimensional `input` tensor. The subset is chosen using a sequence of `m` sparse range specifications encoded into the arguments of this function. Note, in some cases `m` could be equal to `n`, but this need not be the case. Each range specification entry can be one of the following:

  • An ellipsis (...). Ellipses are used to imply zero or more dimensions of full-dimension selection and are produced using `ellipsis_mask`. For example, `foo[...]` is the identity slice.

  • A new axis. This is used to insert a new shape=1 dimension and is produced using `new_axis_mask`. For example, `foo[:, ...]` where `foo` is shape `(3, 4)` produces a `(1, 3, 4)` tensor.

  • A range `begin:end:stride`. This is used to specify how much to choose from a given dimension. `stride` can be any integer but 0. `begin` is an integer which represents the index of the first value to select while `end` represents the index of the last value to select. The number of values selected in each dimension is `end - begin` if `stride > 0` and `begin - end` if `stride < 0`. `begin` and `end` can be negative where `-1` is the last element, `-2` is the second to last. `begin_mask` controls whether to replace the explicitly given `begin` with an implicit effective value of `0` if `stride > 0` and `-1` if `stride < 0`. `end_mask` is analogous but produces the number required to create the largest open interval. For example, given a shape `(3,)` tensor `foo[:]`, the effective `begin` and `end` are `0` and `3`. Do not assume this is equivalent to `foo[0:-1]` which has an effective `begin` and `end` of `0` and `2`. Another example is `foo[-2::-1]` which reverses the first dimension of a tensor while dropping the last two (in the original order elements). For example `foo = [1,2,3,4]; foo[-2::-1]` is `[4,3]`.

  • A single index. This is used to keep only elements that have a given index. For example (`foo[2, :]` on a shape `(5,6)` tensor produces a shape `(6,)` tensor. This is encoded in `begin` and `end` and `shrink_axis_mask`.

Each conceptual range specification is encoded in the op's argument. This encoding is best understand by considering a non-trivial example. In particular, `foo[1, 2:4, None, ..., :-3:-1, :]` will be encoded as

``` begin = [1, 2, x, x, 0, x] # x denotes don't care (usually 0) end = [2, 4, x, x, -3, x] strides = [1, 1, x, x, -1, 1] begin_mask = 1<<4 | 1<<5 = 48 end_mask = 1<<5 = 32 ellipsis_mask = 1<<3 = 8 new_axis_mask = 1<<2 = 4 shrink_axis_mask = 1<<0 = 1 ```

In this case if `foo.shape` is (5, 5, 5, 5, 5, 5) the final shape of the slice becomes (2, 1, 5, 5, 2, 5). Let us walk step by step through each argument specification.

1. The first argument in the example slice is turned into `begin = 1` and `end = begin + 1 = 2`. To disambiguate from the original spec `2:4` we also set the appropriate bit in `shrink_axis_mask`.

2. `2:4` is contributes 2, 4, 1 to begin, end, and stride. All masks have zero bits contributed.

3. None is a synonym for `tf.newaxis`. This means insert a dimension of size 1 dimension in the final shape. Dummy values are contributed to begin, end and stride, while the new_axis_mask bit is set.

4. `...` grab the full ranges from as many dimensions as needed to fully specify a slice for every dimension of the input shape.

5. `:-3:-1` shows the use of negative indices. A negative index `i` associated with a dimension that has shape `s` is converted to a positive index `s + i`. So `-1` becomes `s-1` (i.e. the last element). This conversion is done internally so begin, end and strides receive x, -3, and -1. The appropriate begin_mask bit is set to indicate the start range is the full range (ignoring the x).

6. `:` indicates that the entire contents of the corresponding dimension is selected. This is equivalent to `::` or `0::1`. begin, end, and strides receive 0, 0, and 1, respectively. The appropriate bits in `begin_mask` and `end_mask` are also set.

*Requirements*: 

`0 != strides[i] for i in [0, m)`
`ellipsis_mask must be a power of two (only one ellipsis)`

Arguments: 

begin: `begin[k]` specifies the offset into the `k`th range specification.

The exact dimension this corresponds to will be determined by context. Out-of-bounds values will be silently clamped. If the `k`th bit of `begin_mask` then `begin[k]` is ignored and the full range of the appropriate dimension is used instead. Negative values causes indexing to start from the highest element e.g. If `foo==[1,2,3]` then `foo[-1]==3`. 

end: `end[i]` is like `begin` with the exception that `end_mask` is

used to determine full ranges. 

strides: `strides[i]` specifies the increment in the `i`th specification

after extracting a given element. Negative indices will reverse the original order. Out or range values are clamped to `[0,dim[i]) if slice[i]>0` or `[-1,dim[i]-1] if slice[i] < 0`

func StridedSliceGrad 

func StridedSliceGrad(scope *Scope, shape tf.Output, begin tf.Output, end tf.Output, strides tf.Output, dy tf.Output, optional ...StridedSliceGradAttr) (output tf.Output)

Returns the gradient of `StridedSlice`.

Since `StridedSlice` cuts out pieces of its `input` which is size `shape`, its gradient will have the same shape (which is passed here as `shape`). The gradient will be zero in any element that the slice does not select.

Arguments are the same as StridedSliceGrad with the exception that `dy` is the input gradient to be propagated and `shape` is the shape of `StridedSlice`'s `input`.

func StringFormat 

func StringFormat(scope *Scope, inputs []tf.Output, optional ...StringFormatAttr) (output tf.Output)

Formats a string template using a list of tensors.

Formats a string template using a list of tensors, pretty-printing tensor summaries.

Arguments: 

inputs: The list of tensors to format into the placeholder string.

Returns = The resulting string scalar.

func StringJoin 

func StringJoin(scope *Scope, inputs []tf.Output, optional ...StringJoinAttr) (output tf.Output)

Joins the strings in the given list of string tensors into one tensor;

with the given separator (default is an empty separator).

Examples:

>>> s = ["hello", "world", "tensorflow"] >>> tf.strings.join(s, " ") <tf.Tensor: shape=(), dtype=string, numpy=b'hello world tensorflow'>

Arguments: 

inputs: A list of string tensors.  The tensors must all have the same shape,

or be scalars. Scalars may be mixed in; these will be broadcast to the shape of non-scalar inputs.

func StringLength 

func StringLength(scope *Scope, input tf.Output, optional ...StringLengthAttr) (output tf.Output)

String lengths of `input`.

Computes the length of each string given in the input tensor.

>>> strings = tf.constant(['Hello','TensorFlow', '\U0001F642']) >>> tf.strings.length(strings).numpy() # default counts bytes array([ 5, 10, 4], dtype=int32) >>> tf.strings.length(strings, unit="UTF8_CHAR").numpy() array([ 5, 10, 1], dtype=int32)

Arguments: 

input: The strings for which to compute the length for each element.

Returns Integer tensor that has the same shape as `input`. The output contains the element-wise string lengths of `input`.

func StringLower 

func StringLower(scope *Scope, input tf.Output, optional ...StringLowerAttr) (output tf.Output)

Converts all uppercase characters into their respective lowercase replacements.

Example:

>>> tf.strings.lower("CamelCase string and ALL CAPS") <tf.Tensor: shape=(), dtype=string, numpy=b'camelcase string and all caps'>

Arguments: 

input: The input to be lower-cased.

func StringNGrams 

func StringNGrams(scope *Scope, data tf.Output, data_splits tf.Output, separator string, ngram_widths []int64, left_pad string, right_pad string, pad_width int64, preserve_short_sequences bool) (ngrams tf.Output, ngrams_splits tf.Output)

Creates ngrams from ragged string data.

This op accepts a ragged tensor with 1 ragged dimension containing only strings and outputs a ragged tensor with 1 ragged dimension containing ngrams of that string, joined along the innermost axis.

Arguments: 

data: The values tensor of the ragged string tensor to make ngrams out of. Must be a

1D string tensor. 

data_splits: The splits tensor of the ragged string tensor to make ngrams out of.
separator: The string to append between elements of the token. Use "" for no separator.
ngram_widths: The sizes of the ngrams to create.
left_pad: The string to use to pad the left side of the ngram sequence. Only used if

pad_width != 0. 

right_pad: The string to use to pad the right side of the ngram sequence. Only used if

pad_width != 0. 

pad_width: The number of padding elements to add to each side of each

sequence. Note that padding will never be greater than 'ngram_widths'-1 regardless of this value. If `pad_width=-1`, then add `max(ngram_widths)-1` elements.

Returns: 

ngrams: The values tensor of the output ngrams ragged tensor.
ngrams_splits: The splits tensor of the output ngrams ragged tensor.

func StringSplit 

func StringSplit(scope *Scope, input tf.Output, delimiter tf.Output, optional ...StringSplitAttr) (indices tf.Output, values tf.Output, shape tf.Output)

Split elements of `input` based on `delimiter` into a `SparseTensor`.

Let N be the size of source (typically N will be the batch size). Split each element of `input` based on `delimiter` and return a `SparseTensor` containing the splitted tokens. Empty tokens are ignored.

`delimiter` can be empty, or a string of split characters. If `delimiter` is an 

empty string, each element of `input` is split into individual single-byte
character strings, including splitting of UTF-8 multibyte sequences. Otherwise
every character of `delimiter` is a potential split point.

For example: 

N = 2, input[0] is 'hello world' and input[1] is 'a b c', then the output
will be

indices = [0, 0;
           0, 1;
           1, 0;
           1, 1;
           1, 2]
shape = [2, 3]
values = ['hello', 'world', 'a', 'b', 'c']

Arguments: 

input: 1-D. Strings to split.
delimiter: 0-D. Delimiter characters (bytes), or empty string.

Returns: 

indices: A dense matrix of int64 representing the indices of the sparse tensor.
values: A vector of strings corresponding to the splited values.
shape: a length-2 vector of int64 representing the shape of the sparse

tensor, where the first value is N and the second value is the maximum number of tokens in a single input entry.

func StringSplitV2 

func StringSplitV2(scope *Scope, input tf.Output, sep tf.Output, optional ...StringSplitV2Attr) (indices tf.Output, values tf.Output, shape tf.Output)

Split elements of `source` based on `sep` into a `SparseTensor`.

Let N be the size of source (typically N will be the batch size). Split each element of `source` based on `sep` and return a `SparseTensor` containing the split tokens. Empty tokens are ignored.

For example, N = 2, source[0] is 'hello world' and source[1] is 'a b c', then the output will be ``` st.indices = [0, 0; 

0, 1;
1, 0;
1, 1;
1, 2]

st.shape = [2, 3] st.values = ['hello', 'world', 'a', 'b', 'c'] ```

If `sep` is given, consecutive delimiters are not grouped together and are deemed to delimit empty strings. For example, source of `"1<>2<><>3"` and sep of `"<>"` returns `["1", "2", "", "3"]`. If `sep` is None or an empty string, consecutive whitespace are regarded as a single separator, and the result will contain no empty strings at the startor end if the string has leading or trailing whitespace.

Note that the above mentioned behavior matches python's str.split.

Arguments: 

input: `1-D` string `Tensor`, the strings to split.
sep: `0-D` string `Tensor`, the delimiter character.

func StringStrip 

func StringStrip(scope *Scope, input tf.Output) (output tf.Output)

Strip leading and trailing whitespaces from the Tensor.

Examples:

>>> tf.strings.strip(["\nTensorFlow", " The python library "]).numpy() array([b'TensorFlow', b'The python library'], dtype=object)

Arguments: 

input: A string `Tensor` of any shape.

Returns A string `Tensor` of the same shape as the input.

func StringToHashBucket 

func StringToHashBucket(scope *Scope, string_tensor tf.Output, num_buckets int64) (output tf.Output)

Converts each string in the input Tensor to its hash mod by a number of buckets.

The hash function is deterministic on the content of the string within the process.

Note that the hash function may change from time to time. This functionality will be deprecated and it's recommended to use `tf.string_to_hash_bucket_fast()` or `tf.string_to_hash_bucket_strong()`.

Arguments: 

num_buckets: The number of buckets.

Returns A Tensor of the same shape as the input `string_tensor`.

func StringToHashBucketFast 

func StringToHashBucketFast(scope *Scope, input tf.Output, num_buckets int64) (output tf.Output)

Converts each string in the input Tensor to its hash mod by a number of buckets.

The hash function is deterministic on the content of the string within the process and will never change. However, it is not suitable for cryptography. This function may be used when CPU time is scarce and inputs are trusted or unimportant. There is a risk of adversaries constructing inputs that all hash to the same bucket. To prevent this problem, use a strong hash function with `tf.string_to_hash_bucket_strong`.

Examples:

>>> tf.strings.to_hash_bucket_fast(["Hello", "TensorFlow", "2.x"], 3).numpy() array([0, 2, 2])

Arguments: 

input: The strings to assign a hash bucket.
num_buckets: The number of buckets.

Returns A Tensor of the same shape as the input `string_tensor`.

func StringToHashBucketStrong 

func StringToHashBucketStrong(scope *Scope, input tf.Output, num_buckets int64, key []int64) (output tf.Output)

Converts each string in the input Tensor to its hash mod by a number of buckets.

The hash function is deterministic on the content of the string within the process. The hash function is a keyed hash function, where attribute `key` defines the key of the hash function. `key` is an array of 2 elements.

A strong hash is important when inputs may be malicious, e.g. URLs with additional components. Adversaries could try to make their inputs hash to the same bucket for a denial-of-service attack or to skew the results. A strong hash can be used to make it difficult to find inputs with a skewed hash value distribution over buckets. This requires that the hash function is seeded by a high-entropy (random) "key" unknown to the adversary.

The additional robustness comes at a cost of roughly 4x higher compute time than `tf.string_to_hash_bucket_fast`.

Examples:

>>> tf.strings.to_hash_bucket_strong(["Hello", "TF"], 3, [1, 2]).numpy() array([2, 0])

Arguments: 

input: The strings to assign a hash bucket.
num_buckets: The number of buckets.
key: The key used to seed the hash function, passed as a list of two uint64

elements.

Returns A Tensor of the same shape as the input `string_tensor`.

func StringToNumber 

func StringToNumber(scope *Scope, string_tensor tf.Output, optional ...StringToNumberAttr) (output tf.Output)

Converts each string in the input Tensor to the specified numeric type.

(Note that int32 overflow results in an error while float overflow results in a rounded value.)

Example:

>>> strings = ["5.0", "3.0", "7.0"] >>> tf.strings.to_number(strings) <tf.Tensor: shape=(3,), dtype=float32, numpy=array([5., 3., 7.], dtype=float32)>

Returns A Tensor of the same shape as the input `string_tensor`.

func StringUpper 

func StringUpper(scope *Scope, input tf.Output, optional ...StringUpperAttr) (output tf.Output)

Converts all lowercase characters into their respective uppercase replacements.

Example:

>>> tf.strings.upper("CamelCase string and ALL CAPS") <tf.Tensor: shape=(), dtype=string, numpy=b'CAMELCASE STRING AND ALL CAPS'>

Arguments: 

input: The input to be upper-cased.

func Sub 

func Sub(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x - y element-wise.

*NOTE*: `Subtract` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func Substr 

func Substr(scope *Scope, input tf.Output, pos tf.Output, len tf.Output, optional ...SubstrAttr) (output tf.Output)

Return substrings from `Tensor` of strings.

For each string in the input `Tensor`, creates a substring starting at index `pos` with a total length of `len`.

If `len` defines a substring that would extend beyond the length of the input string, or if `len` is negative, then as many characters as possible are used.

A negative `pos` indicates distance within the string backwards from the end.

If `pos` specifies an index which is out of range for any of the input strings, then an `InvalidArgumentError` is thrown.

`pos` and `len` must have the same shape, otherwise a `ValueError` is thrown on Op creation.

*NOTE*: `Substr` supports broadcasting up to two dimensions. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

---

Examples

Using scalar `pos` and `len`:

```python input = [b'Hello', b'World'] position = 1 length = 3

output = [b'ell', b'orl'] ```

Using `pos` and `len` with same shape as `input`:

```python input = [[b'ten', b'eleven', b'twelve'], 

[b'thirteen', b'fourteen', b'fifteen'],
[b'sixteen', b'seventeen', b'eighteen']]

position = [[1, 2, 3], 

[1, 2, 3],
[1, 2, 3]]

length = [[2, 3, 4], 

[4, 3, 2],
[5, 5, 5]]

output = [[b'en', b'eve', b'lve'], 

[b'hirt', b'urt', b'te'],
[b'ixtee', b'vente', b'hteen']]

```

Broadcasting `pos` and `len` onto `input`:

``` input = [[b'ten', b'eleven', b'twelve'], 

[b'thirteen', b'fourteen', b'fifteen'],
[b'sixteen', b'seventeen', b'eighteen'],
[b'nineteen', b'twenty', b'twentyone']]

position = [1, 2, 3] length = [1, 2, 3]

output = [[b'e', b'ev', b'lve'], 

[b'h', b'ur', b'tee'],
[b'i', b've', b'hte'],
[b'i', b'en', b'nty']]

```

Broadcasting `input` onto `pos` and `len`:

``` input = b'thirteen' position = [1, 5, 7] length = [3, 2, 1]

output = [b'hir', b'ee', b'n'] ```

Raises:

  • `ValueError`: If the first argument cannot be converted to a Tensor of `dtype string`.
  • `InvalidArgumentError`: If indices are out of range.
  • `ValueError`: If `pos` and `len` are not the same shape.

Arguments: 

input: Tensor of strings
pos: Scalar defining the position of first character in each substring
len: Scalar defining the number of characters to include in each substring

Returns Tensor of substrings

func Sum 

func Sum(scope *Scope, input tf.Output, axis tf.Output, optional ...SumAttr) (output tf.Output)

Computes the sum of elements across dimensions of a tensor.

Reduces `input` along the dimensions given in `axis`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `axis`. If `keep_dims` is true, the reduced dimensions are retained with length 1.

Arguments: 

input: The tensor to reduce.
axis: The dimensions to reduce. Must be in the range

`[-rank(input), rank(input))`.

Returns The reduced tensor.

func Svd 

func Svd(scope *Scope, input tf.Output, optional ...SvdAttr) (s tf.Output, u tf.Output, v tf.Output)

Computes the singular value decompositions of one or more matrices.

Computes the SVD of each inner matrix in `input` such that `input[..., :, :] = u[..., :, :] * diag(s[..., :, :]) * transpose(v[..., :, :])`

```python # a is a tensor containing a batch of matrices. # s is a tensor of singular values for each matrix. # u is the tensor containing the left singular vectors for each matrix. # v is the tensor containing the right singular vectors for each matrix. s, u, v = svd(a) s, _, _ = svd(a, compute_uv=False) ```

Arguments: 

input: A tensor of shape `[..., M, N]` whose inner-most 2 dimensions

form matrices of size `[M, N]`. Let `P` be the minimum of `M` and `N`.

Returns: 

s: Singular values. Shape is `[..., P]`.
u: Left singular vectors. If `full_matrices` is `False` then shape is

`[..., M, P]`; if `full_matrices` is `True` then shape is `[..., M, M]`. Undefined if `compute_uv` is `False`. 

v: Left singular vectors. If `full_matrices` is `False` then shape is

`[..., N, P]`. If `full_matrices` is `True` then shape is `[..., N, N]`. Undefined if `compute_uv` is false.

func Switch 

func Switch(scope *Scope, data tf.Output, pred tf.Output) (output_false tf.Output, output_true tf.Output)

Forwards `data` to the output port determined by `pred`.

If `pred` is true, the `data` input is forwarded to `output_true`. Otherwise, the data goes to `output_false`.

See also `RefSwitch` and `Merge`.

Arguments: 

data: The tensor to be forwarded to the appropriate output.
pred: A scalar that specifies which output port will receive data.

Returns: 

output_false: If `pred` is false, data will be forwarded to this output.
output_true: If `pred` is true, data will be forwarded to this output.

func SyncDevice added in v0.4.0 

func SyncDevice(scope *Scope) (o *tf.Operation)

Synchronizes the device this op is run on.

Only GPU ops are asynchrous in TensorFlow, and so this only has an effect when run on GPUs. On GPUs, this op synchronizes the GPU's compute stream.

Returns the created operation.

func TFRecordDataset 

func TFRecordDataset(scope *Scope, filenames tf.Output, compression_type tf.Output, buffer_size tf.Output, optional ...TFRecordDatasetAttr) (handle tf.Output)

Creates a dataset that emits the records from one or more TFRecord files.

Arguments: 

filenames: A scalar or vector containing the name(s) of the file(s) to be

read. 

compression_type: A scalar containing either (i) the empty string (no

compression), (ii) "ZLIB", or (iii) "GZIP". 

buffer_size: A scalar representing the number of bytes to buffer. A value of

0 means no buffering will be performed.

func TFRecordDatasetV2 added in v0.6.0 

func TFRecordDatasetV2(scope *Scope, filenames tf.Output, compression_type tf.Output, buffer_size tf.Output, byte_offsets tf.Output, optional ...TFRecordDatasetV2Attr) (handle tf.Output)

Creates a dataset that emits the records from one or more TFRecord files.

Arguments: 

filenames: A scalar or vector containing the name(s) of the file(s) to be

read. 

compression_type: A scalar containing either (i) the empty string (no

compression), (ii) "ZLIB", or (iii) "GZIP". 

buffer_size: A scalar representing the number of bytes to buffer. A value of

0 means no buffering will be performed. 

byte_offsets: A scalar or vector containing the number of bytes for each file

that will be skipped prior to reading.

func TFRecordReaderV2 

func TFRecordReaderV2(scope *Scope, optional ...TFRecordReaderV2Attr) (reader_handle tf.Output)

A Reader that outputs the records from a TensorFlow Records file.

Returns The handle to reference the Reader.

func TPUCompilationResult 

func TPUCompilationResult(scope *Scope) (output tf.Output)

Returns the result of a TPU compilation.

This operation returns the result of a TPU compilation as a serialized CompilationResultProto, which holds a status and an error message if an error occurred during compilation.

func TPUCompileSucceededAssert 

func TPUCompileSucceededAssert(scope *Scope, compilation_status tf.Output) (o *tf.Operation)

Asserts that compilation succeeded.

This op produces no output and closes the device during failure to ensure all pending device interactions fail.

'compilation_status' is a serialized CompilationResultProto.

Returns the created operation.

func TPUCopyWithDynamicShape added in v0.8.0 

func TPUCopyWithDynamicShape(scope *Scope, tensors []tf.Output, unpadded_sizes []tf.Output) (tpu_tensors []tf.Output)

Op that copies host tensor to device with dynamic shape support. For internal use only.

func TPUEmbeddingActivations 

func TPUEmbeddingActivations(scope *Scope, embedding_variable tf.Output, sliced_activations tf.Output, table_id int64, lookup_id int64) (output tf.Output)

An op enabling differentiation of TPU Embeddings.

This op simply returns its first input, which is assumed to have been sliced from the Tensors returned by TPUEmbeddingDequeueActivations. The presence of this op, and its first argument being a trainable Variable, enables automatic differentiation of graphs containing embeddings via the TPU Embedding Python libraries.

Arguments: 

embedding_variable: A trainable variable, enabling optimizers to find this op.
sliced_activations: The embedding activations Tensor to return.
table_id: The id of the table in the embedding layer configuration from which

these activations were computed. 

lookup_id: Identifier of the set of embedding indices which produced these

activations.

func TPUExecute 

func TPUExecute(scope *Scope, args []tf.Output, key tf.Output, Tresults []tf.DataType) (results []tf.Output)

Op that loads and executes a TPU program on a TPU device.

For the internal use of the distributed TPU compiler.

func TPUExecuteAndUpdateVariables 

func TPUExecuteAndUpdateVariables(scope *Scope, args []tf.Output, key tf.Output, Tresults []tf.DataType, device_var_reads_indices []int64, device_var_updates_indices []int64) (results []tf.Output)

Op that executes a program with optional in-place variable updates.

It (optionally) reads device variables, loads and executes a TPU program on a TPU device, and then (optionally) in-place updates variables using the program outputs, as specified in attributes device_var_reads_indices (program input indices from directly reading variables) and device_var_updates_indices (program output indices used to update variables, -1 means no-update/read-only). Such program outputs are consumed by these variables will not appear in the op output. For the internal use of the distributed TPU compiler.

func TPUOrdinalSelector 

func TPUOrdinalSelector(scope *Scope) (device_ordinals tf.Output)

A TPU core selector Op.

This Op produces a set of TPU cores (for warm-up) or a single TPU core (for regular inference) to execute the TPU program on. The output is consumed by TPUPartitionedCall.

Returns A vector 1 or more TPU cores.

func TPUPartitionedInput 

func TPUPartitionedInput(scope *Scope, inputs []tf.Output, optional ...TPUPartitionedInputAttr) (output tf.Output)

An op that groups a list of partitioned inputs together. This op

Arguments: 

inputs: A list of partitioned inputs which must have the same shape.

Returns A handle which represents the full shape of partitioned tensors.

func TPUPartitionedInputV2 added in v0.4.0 

func TPUPartitionedInputV2(scope *Scope, inputs []tf.Output, partition_dims []int64, optional ...TPUPartitionedInputV2Attr) (output tf.Output)

An op that groups a list of partitioned inputs together. Supports ND sharding.

Arguments: 

inputs: A list of partitioned inputs which must have the same shape.
partition_dims: A list of integers describing how each dimension is partitioned. Emptiness

indicates the inputs are replicated.

Returns A handle which represents the full shape of partitioned tensors.

func TPUPartitionedOutput 

func TPUPartitionedOutput(scope *Scope, inputs tf.Output, num_splits int64, optional ...TPUPartitionedOutputAttr) (output []tf.Output)

An op that demultiplexes a tensor to be sharded by XLA to a list of partitioned

outputs outside the XLA computation.

Arguments: 

inputs: A tensor which represents the full shape of partitioned tensors.

Returns A list of partitioned inputs which must have the same shape.

func TPUPartitionedOutputV2 added in v0.4.0 

func TPUPartitionedOutputV2(scope *Scope, inputs tf.Output, num_splits int64, partition_dims []int64) (output []tf.Output)

An op that demultiplexes a tensor to be sharded by XLA to a list of partitioned

outputs outside the XLA computation. Supports ND sharding.

Arguments: 

inputs: A tensor which represents the full shape of partitioned tensors.

partition_dims: A list of integers describing how each dimension is partitioned. Emptiness

indicates the inputs are replicated.

Returns A list of partitioned outputs which have the same shape.

func TPUReplicateMetadata 

func TPUReplicateMetadata(scope *Scope, num_replicas int64, optional ...TPUReplicateMetadataAttr) (o *tf.Operation)

Metadata indicating how the TPU computation should be replicated.

This operation holds the metadata common to operations of a `tpu.replicate()` computation subgraph.

Arguments: 

num_replicas: Number of replicas of the computation

Returns the created operation.

func TPUReplicatedInput 

func TPUReplicatedInput(scope *Scope, inputs []tf.Output, optional ...TPUReplicatedInputAttr) (output tf.Output)

Connects N inputs to an N-way replicated TPU computation.

This operation holds a replicated input to a `tpu.replicate()` computation subgraph. Each replicated input has the same shape and type alongside the output.

For example: ``` %a = "tf.opA"() %b = "tf.opB"() %replicated_input = "tf.TPUReplicatedInput"(%a, %b) %computation = "tf.Computation"(%replicated_input) ``` The above computation has a replicated input of two replicas.

func TPUReplicatedOutput 

func TPUReplicatedOutput(scope *Scope, input tf.Output, num_replicas int64) (outputs []tf.Output)

Connects N outputs from an N-way replicated TPU computation.

This operation holds a replicated output from a `tpu.replicate()` computation subgraph. Each replicated output has the same shape and type alongside the input.

For example: ``` %computation = "tf.Computation"() %replicated_output:2 = "tf.TPUReplicatedOutput"(%computation) ``` The above computation has a replicated output of two replicas.

func TPUReshardVariables 

func TPUReshardVariables(scope *Scope, vars []tf.Output, new_format_key tf.Output, format_state_var tf.Output) (o *tf.Operation)

Op that reshards on-device TPU variables to specified state.

Op that reshards on-device TPU variables to specified state. Internal use only.

The sharding state is represented as the key of the compilation that generated the sharding/unsharding programs along with the main program. new_format_key specifies the desired state, and format_state_var is the current state of the variables.

Returns the created operation.

func TPURoundRobin added in v0.2.0 

func TPURoundRobin(scope *Scope) (device_ordinal tf.Output)

Round-robin load balancing on TPU cores.

A load balancing op that round-robins among TPU cores.

This op round-robins between the integers in [0, NumTPUCoresVisiblePerHost]. It is useful for interfacing with TensorFlow ops that take as input a TPU core on which to execute computations, such as `TPUPartitionedCall`.

device_ordinal: An integer in [0, NumTPUCoresVisiblePerHost].

func TakeDataset 

func TakeDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...TakeDatasetAttr) (handle tf.Output)

Creates a dataset that contains `count` elements from the `input_dataset`.

Arguments: 

count: A scalar representing the number of elements from the `input_dataset`

that should be taken. A value of `-1` indicates that all of `input_dataset` is taken.

func TakeManySparseFromTensorsMap 

func TakeManySparseFromTensorsMap(scope *Scope, sparse_handles tf.Output, dtype tf.DataType, optional ...TakeManySparseFromTensorsMapAttr) (sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output)

Read `SparseTensors` from a `SparseTensorsMap` and concatenate them.

The input `sparse_handles` must be an `int64` matrix of shape `[N, 1]` where `N` is the minibatch size and the rows correspond to the output handles of `AddSparseToTensorsMap` or `AddManySparseToTensorsMap`. The ranks of the original `SparseTensor` objects that went into the given input ops must all match. When the final `SparseTensor` is created, it has rank one higher than the ranks of the incoming `SparseTensor` objects (they have been concatenated along a new row dimension on the left).

The output `SparseTensor` object's shape values for all dimensions but the first are the max across the input `SparseTensor` objects' shape values for the corresponding dimensions. Its first shape value is `N`, the minibatch size.

The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, after this step run `SparseReorder` to restore index ordering.

For example, if the handles represent an input, which is a `[2, 3]` matrix representing two original `SparseTensor` objects:

``` 

index = [ 0]
        [10]
        [20]
values = [1, 2, 3]
shape = [50]

```

and

``` 

index = [ 2]
        [10]
values = [4, 5]
shape = [30]

```

then the final `SparseTensor` will be:

``` 

index = [0  0]
        [0 10]
        [0 20]
        [1  2]
        [1 10]
values = [1, 2, 3, 4, 5]
shape = [2 50]

```

Arguments: 

sparse_handles: 1-D, The `N` serialized `SparseTensor` objects.

Shape: `[N]`. 

dtype: The `dtype` of the `SparseTensor` objects stored in the

`SparseTensorsMap`.

Returns: 

sparse_indices: 2-D.  The `indices` of the minibatch `SparseTensor`.
sparse_values: 1-D.  The `values` of the minibatch `SparseTensor`.
sparse_shape: 1-D.  The `shape` of the minibatch `SparseTensor`.

func Tan 

func Tan(scope *Scope, x tf.Output) (y tf.Output)

Computes tan of x element-wise. 

Given an input tensor, this function computes tangent of every
element in the tensor. Input range is `(-inf, inf)` and
output range is `(-inf, inf)`. If input lies outside the boundary, `nan`
is returned.

```python
x = tf.constant([-float("inf"), -9, -0.5, 1, 1.2, 200, 10000, float("inf")])
tf.math.tan(x) ==> [nan 0.45231566 -0.5463025 1.5574077 2.572152 -1.7925274 0.32097113 nan]
```

func Tanh 

func Tanh(scope *Scope, x tf.Output) (y tf.Output)

Computes hyperbolic tangent of `x` element-wise. 

Given an input tensor, this function computes hyperbolic tangent of every
element in the tensor. Input range is `[-inf, inf]` and
output range is `[-1,1]`.

>>> x = tf.constant([-float("inf"), -5, -0.5, 1, 1.2, 2, 3, float("inf")])
>>> tf.math.tanh(x)
<tf.Tensor: shape=(8,), dtype=float32, numpy=
array([-1.0, -0.99990916, -0.46211717,  0.7615942 ,  0.8336547 ,
        0.9640276 ,  0.9950547 ,  1.0], dtype=float32)>

func TanhGrad 

func TanhGrad(scope *Scope, y tf.Output, dy tf.Output) (z tf.Output)

Computes the gradient for the tanh of `x` wrt its input.

Specifically, `grad = dy * (1 - y*y)`, where `y = tanh(x)`, and `dy` is the corresponding input gradient.

func TensorArrayCloseV2 

func TensorArrayCloseV2(scope *Scope, handle tf.Output) (o *tf.Operation)

Deprecated. Use TensorArrayCloseV3

DEPRECATED at GraphDef version 26: Use TensorArrayCloseV3

Returns the created operation.

func TensorArrayCloseV3 

func TensorArrayCloseV3(scope *Scope, handle tf.Output) (o *tf.Operation)

Delete the TensorArray from its resource container.

This enables the user to close and release the resource in the middle of a step/run.

Arguments: 

handle: The handle to a TensorArray (output of TensorArray or TensorArrayGrad).

Returns the created operation.

func TensorArrayConcatV2 

func TensorArrayConcatV2(scope *Scope, handle tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayConcatV2Attr) (value tf.Output, lengths tf.Output)

Deprecated. Use TensorArrayConcatV3

func TensorArrayConcatV3 

func TensorArrayConcatV3(scope *Scope, handle tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayConcatV3Attr) (value tf.Output, lengths tf.Output)

Concat the elements from the TensorArray into value `value`.

Takes `T` elements of shapes 

```
(n0 x d0 x d1 x ...), (n1 x d0 x d1 x ...), ..., (n(T-1) x d0 x d1 x ...)
```

and concatenates them into a Tensor of shape: 

```
(n0 + n1 + ... + n(T-1) x d0 x d1 x ...)
```

All elements must have the same shape (excepting the first dimension).

Arguments: 

handle: The handle to a TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.
dtype: The type of the elem that is returned.

Returns: 

value: All of the elements in the TensorArray, concatenated along the first

axis. 

lengths: A vector of the row sizes of the original T elements in the

value output. In the example above, this would be the values: `(n1, n2, ..., n(T-1))`.

func TensorArrayGatherV2 

func TensorArrayGatherV2(scope *Scope, handle tf.Output, indices tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayGatherV2Attr) (value tf.Output)

Deprecated. Use TensorArrayGatherV3

DEPRECATED at GraphDef version 26: Use TensorArrayGatherV3

func TensorArrayGatherV3 

func TensorArrayGatherV3(scope *Scope, handle tf.Output, indices tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayGatherV3Attr) (value tf.Output)

Gather specific elements from the TensorArray into output `value`.

All elements selected by `indices` must have the same shape.

Arguments: 

handle: The handle to a TensorArray.
indices: The locations in the TensorArray from which to read tensor elements.
flow_in: A float scalar that enforces proper chaining of operations.
dtype: The type of the elem that is returned.

Returns All of the elements in the TensorArray, concatenated along a new axis (the new dimension 0).

func TensorArrayGradV2 

func TensorArrayGradV2(scope *Scope, handle tf.Output, flow_in tf.Output, source string) (grad_handle tf.Output)

Deprecated. Use TensorArrayGradV3

DEPRECATED at GraphDef version 26: Use TensorArrayGradV3

func TensorArrayGradV3 

func TensorArrayGradV3(scope *Scope, handle tf.Output, flow_in tf.Output, source string) (grad_handle tf.Output, flow_out tf.Output)

Creates a TensorArray for storing the gradients of values in the given handle.

If the given TensorArray gradient already exists, returns a reference to it.

Locks the size of the original TensorArray by disabling its dynamic size flag.

**A note about the input flow_in:**

The handle flow_in forces the execution of the gradient lookup to occur only after certain other operations have occurred. For example, when the forward TensorArray is dynamically sized, writes to this TensorArray may resize the object. The gradient TensorArray is statically sized based on the size of the forward TensorArray when this operation executes. Furthermore, the size of the forward TensorArray is frozen by this call. As a result, the flow is used to ensure that the call to generate the gradient TensorArray only happens after all writes are executed.

In the case of dynamically sized TensorArrays, gradient computation should only be performed on read operations that have themselves been chained via flow to occur only after all writes have executed. That way the final size of the forward TensorArray is known when this operation is called.

**A note about the source attribute:**

TensorArray gradient calls use an accumulator TensorArray object. If multiple gradients are calculated and run in the same session, the multiple gradient nodes may accidentally flow through the same accumulator TensorArray. This double counts and generally breaks the TensorArray gradient flow.

The solution is to identify which gradient call this particular TensorArray gradient is being called in. This is performed by identifying a unique string (e.g. "gradients", "gradients_1", ...) from the input gradient Tensor's name. This string is used as a suffix when creating the TensorArray gradient object here (the attribute `source`).

The attribute `source` is added as a suffix to the forward TensorArray's name when performing the creation / lookup, so that each separate gradient calculation gets its own TensorArray accumulator.

Arguments: 

handle: The handle to the forward TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.
source: The gradient source string, used to decide which gradient TensorArray

to return.

func TensorArrayGradWithShape 

func TensorArrayGradWithShape(scope *Scope, handle tf.Output, flow_in tf.Output, shape_to_prepend tf.Output, source string) (grad_handle tf.Output, flow_out tf.Output)

Creates a TensorArray for storing multiple gradients of values in the given handle.

Similar to TensorArrayGradV3. However it creates an accumulator with an expanded shape compared to the input TensorArray whose gradient is being computed. This enables multiple gradients for the same TensorArray to be calculated using the same accumulator.

Arguments: 

handle: The handle to the forward TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.
shape_to_prepend: An int32 vector representing a shape. Elements in the gradient accumulator will

have shape which is this shape_to_prepend value concatenated with shape of the elements in the TensorArray corresponding to the input handle. 

source: The gradient source string, used to decide which gradient TensorArray

to return.

func TensorArrayReadV2 

func TensorArrayReadV2(scope *Scope, handle tf.Output, index tf.Output, flow_in tf.Output, dtype tf.DataType) (value tf.Output)

Deprecated. Use TensorArrayReadV3

DEPRECATED at GraphDef version 26: Use TensorArrayReadV3

func TensorArrayReadV3 

func TensorArrayReadV3(scope *Scope, handle tf.Output, index tf.Output, flow_in tf.Output, dtype tf.DataType) (value tf.Output)

Read an element from the TensorArray into output `value`.

Arguments: 

handle: The handle to a TensorArray.

flow_in: A float scalar that enforces proper chaining of operations.
dtype: The type of the elem that is returned.

Returns The tensor that is read from the TensorArray.

func TensorArrayScatterV2 

func TensorArrayScatterV2(scope *Scope, handle tf.Output, indices tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)

Deprecated. Use TensorArrayScatterV3

DEPRECATED at GraphDef version 26: Use TensorArrayScatterV3

func TensorArrayScatterV3 

func TensorArrayScatterV3(scope *Scope, handle tf.Output, indices tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)

Scatter the data from the input value into specific TensorArray elements.

`indices` must be a vector, its length must match the first dim of `value`.

Arguments: 

handle: The handle to a TensorArray.
indices: The locations at which to write the tensor elements.
value: The concatenated tensor to write to the TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.

Returns A float scalar that enforces proper chaining of operations.

func TensorArraySizeV2 

func TensorArraySizeV2(scope *Scope, handle tf.Output, flow_in tf.Output) (size tf.Output)

Deprecated. Use TensorArraySizeV3

DEPRECATED at GraphDef version 26: Use TensorArraySizeV3

func TensorArraySizeV3 

func TensorArraySizeV3(scope *Scope, handle tf.Output, flow_in tf.Output) (size tf.Output)

Get the current size of the TensorArray.

Arguments: 

handle: The handle to a TensorArray (output of TensorArray or TensorArrayGrad).
flow_in: A float scalar that enforces proper chaining of operations.

Returns The current size of the TensorArray.

func TensorArraySplitV2 

func TensorArraySplitV2(scope *Scope, handle tf.Output, value tf.Output, lengths tf.Output, flow_in tf.Output) (flow_out tf.Output)

Deprecated. Use TensorArraySplitV3

DEPRECATED at GraphDef version 26: Use TensorArraySplitV3

func TensorArraySplitV3 

func TensorArraySplitV3(scope *Scope, handle tf.Output, value tf.Output, lengths tf.Output, flow_in tf.Output) (flow_out tf.Output)

Split the data from the input value into TensorArray elements.

Assuming that `lengths` takes on values 

```
(n0, n1, ..., n(T-1))
```

and that `value` has shape 

```
(n0 + n1 + ... + n(T-1) x d0 x d1 x ...),
```

this splits values into a TensorArray with T tensors.

TensorArray index t will be the subtensor of values with starting position 

```
(n0 + n1 + ... + n(t-1), 0, 0, ...)
```

and having size 

```
nt x d0 x d1 x ...
```

Arguments: 

handle: The handle to a TensorArray.
value: The concatenated tensor to write to the TensorArray.
lengths: The vector of lengths, how to split the rows of value into the

TensorArray. 

flow_in: A float scalar that enforces proper chaining of operations.

Returns A float scalar that enforces proper chaining of operations.

func TensorArrayV2 

func TensorArrayV2(scope *Scope, size tf.Output, dtype tf.DataType, optional ...TensorArrayV2Attr) (handle tf.Output)

Deprecated. Use TensorArrayV3

DEPRECATED at GraphDef version 26: Use TensorArrayV3

func TensorArrayV3 

func TensorArrayV3(scope *Scope, size tf.Output, dtype tf.DataType, optional ...TensorArrayV3Attr) (handle tf.Output, flow tf.Output)

An array of Tensors of given size.

Write data via Write and read via Read or Pack.

Arguments: 

size: The size of the array.
dtype: The type of the elements on the tensor_array.

Returns: 

handle: The handle to the TensorArray.
flow: A scalar used to control gradient flow.

func TensorArrayWriteV2 

func TensorArrayWriteV2(scope *Scope, handle tf.Output, index tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)

Deprecated. Use TensorArrayGradV3

DEPRECATED at GraphDef version 26: Use TensorArrayWriteV3

func TensorArrayWriteV3 

func TensorArrayWriteV3(scope *Scope, handle tf.Output, index tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)

Push an element onto the tensor_array.

Arguments: 

handle: The handle to a TensorArray.
index: The position to write to inside the TensorArray.
value: The tensor to write to the TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.

Returns A float scalar that enforces proper chaining of operations.

func TensorDataset 

func TensorDataset(scope *Scope, components []tf.Output, output_shapes []tf.Shape, optional ...TensorDatasetAttr) (handle tf.Output)

Creates a dataset that emits `components` as a tuple of tensors once.

func TensorListConcat 

func TensorListConcat(scope *Scope, input_handle tf.Output, element_dtype tf.DataType, optional ...TensorListConcatAttr) (tensor tf.Output, lengths tf.Output)

Concats all tensors in the list along the 0th dimension.

Requires that all tensors have the same shape except the first dimension.

input_handle: The input list. tensor: The concated result. lengths: Output tensor containing sizes of the 0th dimension of tensors in the list, used for computing the gradient.

func TensorListConcatV2 

func TensorListConcatV2(scope *Scope, input_handle tf.Output, element_shape tf.Output, leading_dims tf.Output, element_dtype tf.DataType) (tensor tf.Output, lengths tf.Output)

Concats all tensors in the list along the 0th dimension.

Requires that all tensors have the same shape except the first dimension.

input_handle: The input list. element_shape: The shape of the uninitialized elements in the list. If the first 

dimension is not -1, it is assumed that all list elements have the same
leading dim.

leading_dims: The list of leading dims of uninitialized list elements. Used if 

the leading dim of input_handle.element_shape or the element_shape input arg
is not already set.

tensor: The concated result. lengths: Output tensor containing sizes of the 0th dimension of tensors in the list, used for computing the gradient.

func TensorListElementShape 

func TensorListElementShape(scope *Scope, input_handle tf.Output, shape_type tf.DataType) (element_shape tf.Output)

The shape of the elements of the given list, as a tensor. 

input_handle: the list
element_shape: the shape of elements of the list

func TensorListFromTensor 

func TensorListFromTensor(scope *Scope, tensor tf.Output, element_shape tf.Output) (output_handle tf.Output)

Creates a TensorList which, when stacked, has the value of `tensor`.

Each tensor in the result list corresponds to one row of the input tensor.

tensor: The input tensor. output_handle: The list.

func TensorListGather 

func TensorListGather(scope *Scope, input_handle tf.Output, indices tf.Output, element_shape tf.Output, element_dtype tf.DataType) (values tf.Output)

Creates a Tensor by indexing into the TensorList.

Each row in the produced Tensor corresponds to the element in the TensorList specified by the given index (see `tf.gather`).

input_handle: The input tensor list. indices: The indices used to index into the list. values: The tensor.

func TensorListGetItem 

func TensorListGetItem(scope *Scope, input_handle tf.Output, index tf.Output, element_shape tf.Output, element_dtype tf.DataType) (item tf.Output)

Returns the item in the list with the given index.

input_handle: the list index: the position in the list from which an element will be retrieved item: the element at that position

func TensorListLength 

func TensorListLength(scope *Scope, input_handle tf.Output) (length tf.Output)

Returns the number of tensors in the input tensor list.

input_handle: the input list length: the number of tensors in the list

func TensorListPopBack 

func TensorListPopBack(scope *Scope, input_handle tf.Output, element_shape tf.Output, element_dtype tf.DataType) (output_handle tf.Output, tensor tf.Output)

Returns the last element of the input list as well as a list with all but that element.

Fails if the list is empty.

input_handle: the input list tensor: the withdrawn last element of the list element_dtype: the type of elements in the list element_shape: the shape of the output tensor

func TensorListPushBack 

func TensorListPushBack(scope *Scope, input_handle tf.Output, tensor tf.Output) (output_handle tf.Output)

Returns a list which has the passed-in `Tensor` as last element and the other elements of the given list in `input_handle`.

tensor: The tensor to put on the list. input_handle: The old list. output_handle: A list with the elements of the old list followed by tensor. element_dtype: the type of elements in the list. element_shape: a shape compatible with that of elements in the list.

func TensorListReserve 

func TensorListReserve(scope *Scope, element_shape tf.Output, num_elements tf.Output, element_dtype tf.DataType) (handle tf.Output)

List of the given size with empty elements.

element_shape: the shape of the future elements of the list num_elements: the number of elements to reserve handle: the output list element_dtype: the desired type of elements in the list.

func TensorListResize 

func TensorListResize(scope *Scope, input_handle tf.Output, size tf.Output) (output_handle tf.Output)

Resizes the list.

input_handle: the input list size: size of the output list

func TensorListScatter 

func TensorListScatter(scope *Scope, tensor tf.Output, indices tf.Output, element_shape tf.Output) (output_handle tf.Output)

Creates a TensorList by indexing into a Tensor.

Each member of the TensorList corresponds to one row of the input tensor, specified by the given index (see `tf.gather`).

tensor: The input tensor. indices: The indices used to index into the list. element_shape: The shape of the elements in the list (can be less specified than 

the shape of the tensor).

output_handle: The TensorList.

func TensorListScatterIntoExistingList 

func TensorListScatterIntoExistingList(scope *Scope, input_handle tf.Output, tensor tf.Output, indices tf.Output) (output_handle tf.Output)

Scatters tensor at indices in an input list.

Each member of the TensorList corresponds to one row of the input tensor, specified by the given index (see `tf.gather`).

input_handle: The list to scatter into. tensor: The input tensor. indices: The indices used to index into the list. output_handle: The TensorList.

func TensorListScatterV2 

func TensorListScatterV2(scope *Scope, tensor tf.Output, indices tf.Output, element_shape tf.Output, num_elements tf.Output) (output_handle tf.Output)

Creates a TensorList by indexing into a Tensor.

Each member of the TensorList corresponds to one row of the input tensor, specified by the given index (see `tf.gather`).

tensor: The input tensor. indices: The indices used to index into the list. element_shape: The shape of the elements in the list (can be less specified than 

the shape of the tensor).

num_elements: The size of the output list. Must be large enough to accommodate 

the largest index in indices. If -1, the list is just large enough to include
the largest index in indices.

output_handle: The TensorList.

func TensorListSetItem 

func TensorListSetItem(scope *Scope, input_handle tf.Output, index tf.Output, item tf.Output, optional ...TensorListSetItemAttr) (output_handle tf.Output)

Sets the index-th position of the list to contain the given tensor.

input_handle: the list index: the position in the list to which the tensor will be assigned item: the element to be assigned to that position output_handle: the new list, with the element in the proper position

func TensorListSplit 

func TensorListSplit(scope *Scope, tensor tf.Output, element_shape tf.Output, lengths tf.Output) (output_handle tf.Output)

Splits a tensor into a list.

list[i] corresponds to lengths[i] tensors from the input tensor. The tensor must have rank at least 1 and contain exactly sum(lengths) elements.

tensor: The input tensor. element_shape: A shape compatible with that of elements in the tensor. lengths: Vector of sizes of the 0th dimension of tensors in the list. output_handle: The list.

func TensorListStack 

func TensorListStack(scope *Scope, input_handle tf.Output, element_shape tf.Output, element_dtype tf.DataType, optional ...TensorListStackAttr) (tensor tf.Output)

Stacks all tensors in the list.

Requires that all tensors have the same shape.

input_handle: the input list tensor: the gathered result num_elements: optional. If not -1, the number of elements in the list.

func TensorMapErase 

func TensorMapErase(scope *Scope, input_handle tf.Output, key tf.Output, value_dtype tf.DataType) (output_handle tf.Output)

Returns a tensor map with item from given key erased.

input_handle: the original map output_handle: the map with value from given key removed key: the key of the value to be erased

func TensorMapHasKey 

func TensorMapHasKey(scope *Scope, input_handle tf.Output, key tf.Output) (has_key tf.Output)

Returns whether the given key exists in the map.

input_handle: the input map key: the key to check has_key: whether the key is already in the map or not

func TensorMapInsert 

func TensorMapInsert(scope *Scope, input_handle tf.Output, key tf.Output, value tf.Output) (output_handle tf.Output)

Returns a map that is the 'input_handle' with the given key-value pair inserted.

input_handle: the original map output_handle: the map with key and value inserted key: the key to be inserted value: the value to be inserted

func TensorMapLookup 

func TensorMapLookup(scope *Scope, input_handle tf.Output, key tf.Output, value_dtype tf.DataType) (value tf.Output)

Returns the value from a given key in a tensor map.

input_handle: the input map key: the key to be looked up value: the value found from the given key

func TensorMapSize 

func TensorMapSize(scope *Scope, input_handle tf.Output) (size tf.Output)

Returns the number of tensors in the input tensor map.

input_handle: the input map size: the number of tensors in the map

func TensorMapStackKeys 

func TensorMapStackKeys(scope *Scope, input_handle tf.Output, key_dtype tf.DataType) (keys tf.Output)

Returns a Tensor stack of all keys in a tensor map.

input_handle: the input map keys: the returned Tensor of all keys in the map

func TensorScatterAdd 

func TensorScatterAdd(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output, optional ...TensorScatterAddAttr) (output tf.Output)

Adds sparse `updates` to an existing tensor according to `indices`.

This operation creates a new tensor by adding sparse `updates` to the passed in `tensor`. This operation is very similar to `tf.compat.v1.scatter_nd_add`, except that the updates are added onto an existing tensor (as opposed to a variable). If the memory for the existing tensor cannot be re-used, a copy is made and updated.

`indices` is an integer tensor containing indices into a new tensor of shape `tensor.shape`. The last dimension of `indices` can be at most the rank of `tensor.shape`:

``` indices.shape[-1] <= tensor.shape.rank ```

The last dimension of `indices` corresponds to indices into elements (if `indices.shape[-1] = tensor.shape.rank`) or slices (if `indices.shape[-1] < tensor.shape.rank`) along dimension `indices.shape[-1]` of `tensor.shape`. `updates` is a tensor with shape

``` indices.shape[:-1] + tensor.shape[indices.shape[-1]:] ```

The simplest form of `tensor_scatter_nd_add` is to add individual elements to a tensor by index. For example, say we want to add 4 elements in a rank-1 tensor with 8 elements.

In Python, this scatter add operation would look like this:

>>> indices = tf.constant([[4], [3], [1], [7]]) >>> updates = tf.constant([9, 10, 11, 12]) >>> tensor = tf.ones([8], dtype=tf.int32) >>> updated = tf.tensor_scatter_nd_add(tensor, indices, updates) >>> updated <tf.Tensor: shape=(8,), dtype=int32, numpy=array([ 1, 12, 1, 11, 10, 1, 1, 13], dtype=int32)>

We can also, insert entire slices of a higher rank tensor all at once. For example, if we wanted to insert two slices in the first dimension of a rank-3 tensor with two matrices of new values.

In Python, this scatter add operation would look like this:

>>> indices = tf.constant([[0], [2]]) >>> updates = tf.constant([[[5, 5, 5, 5], [6, 6, 6, 6], ... [7, 7, 7, 7], [8, 8, 8, 8]], ... [[5, 5, 5, 5], [6, 6, 6, 6], ... [7, 7, 7, 7], [8, 8, 8, 8]]]) >>> tensor = tf.ones([4, 4, 4],dtype=tf.int32) >>> updated = tf.tensor_scatter_nd_add(tensor, indices, updates) >>> updated <tf.Tensor: shape=(4, 4, 4), dtype=int32, numpy=array([[[6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8], [9, 9, 9, 9]], 

[[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]],
[[6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8], [9, 9, 9, 9]],
[[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]]], dtype=int32)>

If `indices` contains any out-of-bound indices, depending on `bad_indices_policy`, the op will either return an error or ignore the out-of-bound indices. `bad_indices_policy` can be one of the following values:

  1. "" or "DEFAULT": raises on CPU and ignore on GPU. This is because historically on CPU and GPU we handle errors in different ways, and for backward compatibility we keep the default behavior.
  2. "ERROR": raises error; GPU does not support this value.
  3. "IGNORE": ignore the bad indices; supported on both CPU and GPU.

Arguments: 

tensor: Tensor to copy/update.
indices: Index tensor.
updates: Updates to scatter into output.

Returns A new tensor copied from tensor and updates added according to the indices.

func TensorScatterMax 

func TensorScatterMax(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output, optional ...TensorScatterMaxAttr) (output tf.Output)

Apply a sparse update to a tensor taking the element-wise maximum.

Returns a new tensor copied from `tensor` whose values are element-wise maximum between tensor and updates according to the indices.

>>> tensor = [0, 0, 0, 0, 0, 0, 0, 0] >>> indices = [[1], [4], [5]] >>> updates = [1, -1, 1] >>> tf.tensor_scatter_nd_max(tensor, indices, updates).numpy() array([0, 1, 0, 0, 0, 1, 0, 0], dtype=int32)

Refer to `tf.tensor_scatter_nd_update` for more details.

Arguments: 

tensor: Tensor to update.
indices: Index tensor.
updates: Updates to scatter into output.

Returns A new tensor copied from tensor whose values are element-wise maximum between tensor and updates according to the indices.

func TensorScatterSub 

func TensorScatterSub(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output, optional ...TensorScatterSubAttr) (output tf.Output)

Subtracts sparse `updates` from an existing tensor according to `indices`.

This operation creates a new tensor by subtracting sparse `updates` from the passed in `tensor`. This operation is very similar to `tf.scatter_nd_sub`, except that the updates are subtracted from an existing tensor (as opposed to a variable). If the memory for the existing tensor cannot be re-used, a copy is made and updated.

`indices` is an integer tensor containing indices into a new tensor of shape `shape`. The last dimension of `indices` can be at most the rank of `shape`: 

indices.shape[-1] <= shape.rank

The last dimension of `indices` corresponds to indices into elements (if `indices.shape[-1] = shape.rank`) or slices (if `indices.shape[-1] < shape.rank`) along dimension `indices.shape[-1]` of `shape`. `updates` is a tensor with shape 

indices.shape[:-1] + shape[indices.shape[-1]:]

The simplest form of tensor_scatter_sub is to subtract individual elements from a tensor by index. For example, say we want to insert 4 scattered elements in a rank-1 tensor with 8 elements.

In Python, this scatter subtract operation would look like this:

```python 

indices = tf.constant([[4], [3], [1], [7]])
updates = tf.constant([9, 10, 11, 12])
tensor = tf.ones([8], dtype=tf.int32)
updated = tf.tensor_scatter_nd_sub(tensor, indices, updates)
print(updated)

```

The resulting tensor would look like this: 

[1, -10, 1, -9, -8, 1, 1, -11]

We can also, insert entire slices of a higher rank tensor all at once. For example, if we wanted to insert two slices in the first dimension of a rank-3 tensor with two matrices of new values.

In Python, this scatter add operation would look like this:

```python 

indices = tf.constant([[0], [2]])
updates = tf.constant([[[5, 5, 5, 5], [6, 6, 6, 6],
                        [7, 7, 7, 7], [8, 8, 8, 8]],
                       [[5, 5, 5, 5], [6, 6, 6, 6],
                        [7, 7, 7, 7], [8, 8, 8, 8]]])
tensor = tf.ones([4, 4, 4],dtype=tf.int32)
updated = tf.tensor_scatter_nd_sub(tensor, indices, updates)
print(updated)

```

The resulting tensor would look like this: 

[[[-4, -4, -4, -4], [-5, -5, -5, -5], [-6, -6, -6, -6], [-7, -7, -7, -7]],
 [[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]],
 [[-4, -4, -4, -4], [-5, -5, -5, -5], [-6, -6, -6, -6], [-7, -7, -7, -7]],
 [[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]]]

Note that on CPU, if an out of bound index is found, an error is returned. On GPU, if an out of bound index is found, the index is ignored.

Arguments: 

tensor: Tensor to copy/update.
indices: Index tensor.
updates: Updates to scatter into output.

Returns A new tensor copied from tensor and updates subtracted according to the indices.

func TensorScatterUpdate 

func TensorScatterUpdate(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output, optional ...TensorScatterUpdateAttr) (output tf.Output)

Scatter `updates` into an existing tensor according to `indices`.

This operation creates a new tensor by applying sparse `updates` to the passed in `tensor`. This operation is very similar to `tf.scatter_nd`, except that the updates are scattered onto an existing tensor (as opposed to a zero-tensor). If the memory for the existing tensor cannot be re-used, a copy is made and updated.

If `indices` contains duplicates, then we pick the last update for the index.

**WARNING**: There are some GPU specific semantics for this operation. - If an out of bound index is found, the index is ignored. - The order in which updates are applied is nondeterministic, so the output will be nondeterministic if `indices` contains duplicates.

`indices` is an integer tensor containing indices into a new tensor of shape `shape`.

  • `indices` must have at least 2 axes: `(num_updates, index_depth)`.
  • The last axis of `indices` is how deep to index into `tensor` so this index depth must be less than the rank of `tensor`: `indices.shape[-1] <= tensor.ndim`

if `indices.shape[-1] = tensor.rank` this Op indexes and updates scalar elements. if `indices.shape[-1] < tensor.rank` it indexes and updates slices of the input `tensor`.

Each `update` has a rank of `tensor.rank - indices.shape[-1]`. The overall shape of `updates` is:

``` indices.shape[:-1] + tensor.shape[indices.shape[-1]:] ```

If `indices` contains any out-of-bound indices, depending on `bad_indices_policy`, the op will either return an error or ignore the out-of-bound indices. `bad_indices_policy` can be one of the following values:

  1. "" or "DEFAULT": raises on CPU and ignore on GPU. This is because historically on CPU and GPU we handle errors in different ways, and for backward compatibility we keep the default behavior.
  2. "ERROR": raises error; GPU does not support this value.
  3. "IGNORE": ignore the bad indices; supported on both CPU and GPU.

For usage examples see the python [tf.tensor_scatter_nd_update]( https://www.tensorflow.org/api_docs/python/tf/tensor_scatter_nd_update) function

Arguments: 

tensor: Tensor to copy/update.
indices: Index tensor.
updates: Updates to scatter into output.

Returns A new tensor with the given shape and updates applied according to the indices.

func TensorSliceDataset 

func TensorSliceDataset(scope *Scope, components []tf.Output, output_shapes []tf.Shape, optional ...TensorSliceDatasetAttr) (handle tf.Output)

Creates a dataset that emits each dim-0 slice of `components` once.

func TensorStridedSliceUpdate 

func TensorStridedSliceUpdate(scope *Scope, input tf.Output, begin tf.Output, end tf.Output, strides tf.Output, value tf.Output, optional ...TensorStridedSliceUpdateAttr) (output tf.Output)

Assign `value` to the sliced l-value reference of `input`.

The values of `value` are assigned to the positions in the tensor `input` that are selected by the slice parameters. The slice parameters `begin` `end` `strides` etc. work exactly as in `StridedSlice`.

NOTE this op currently does not support broadcasting and so `value`'s shape must be exactly the shape produced by the slice of `input`.

func TensorSummary 

func TensorSummary(scope *Scope, tensor tf.Output, optional ...TensorSummaryAttr) (summary tf.Output)

Outputs a `Summary` protocol buffer with a tensor.

This op is being phased out in favor of TensorSummaryV2, which lets callers pass a tag as well as a serialized SummaryMetadata proto string that contains plugin-specific data. We will keep this op to maintain backwards compatibility.

Arguments: 

tensor: A tensor to serialize.

func TensorSummaryV2 

func TensorSummaryV2(scope *Scope, tag tf.Output, tensor tf.Output, serialized_summary_metadata tf.Output) (summary tf.Output)

Outputs a `Summary` protocol buffer with a tensor and per-plugin data.

Arguments: 

tag: A string attached to this summary. Used for organization in TensorBoard.
tensor: A tensor to serialize.
serialized_summary_metadata: A serialized SummaryMetadata proto. Contains plugin

data.

func TextLineDataset 

func TextLineDataset(scope *Scope, filenames tf.Output, compression_type tf.Output, buffer_size tf.Output, optional ...TextLineDatasetAttr) (handle tf.Output)

Creates a dataset that emits the lines of one or more text files.

Arguments: 

filenames: A scalar or a vector containing the name(s) of the file(s) to be

read. 

compression_type: A scalar containing either (i) the empty string (no

compression), (ii) "ZLIB", or (iii) "GZIP". 

buffer_size: A scalar containing the number of bytes to buffer.

func TextLineReaderV2 

func TextLineReaderV2(scope *Scope, optional ...TextLineReaderV2Attr) (reader_handle tf.Output)

A Reader that outputs the lines of a file delimited by '\n'.

Returns The handle to reference the Reader.

func ThreadPoolDataset 

func ThreadPoolDataset(scope *Scope, input_dataset tf.Output, thread_pool tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

thread_pool: A resource produced by the ThreadPoolHandle op.

func ThreadPoolHandle 

func ThreadPoolHandle(scope *Scope, num_threads int64, display_name string, optional ...ThreadPoolHandleAttr) (handle tf.Output)

Creates a dataset that uses a custom thread pool to compute `input_dataset`.

Arguments: 

num_threads: The number of threads in the thread pool.
display_name: A human-readable name for the threads that may be visible in some

visualizations. threadpool.

Returns A resource that can be consumed by one or more ExperimentalThreadPoolDataset ops.

func ThreadUnsafeUnigramCandidateSampler 

func ThreadUnsafeUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...ThreadUnsafeUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a learned unigram distribution.

See explanations of candidate sampling and the data formats at go/candidate-sampling.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to randomly sample.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities. 

range_max: The sampler will sample integers from the interval [0, range_max).

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func Tile 

func Tile(scope *Scope, input tf.Output, multiples tf.Output) (output tf.Output)

Constructs a tensor by tiling a given tensor.

This operation creates a new tensor by replicating `input` `multiples` times. The output tensor's i'th dimension has `input.dims(i) * multiples[i]` elements, and the values of `input` are replicated `multiples[i]` times along the 'i'th dimension. For example, tiling `[a b c d]` by `[2]` produces `[a b c d a b c d]`.

>>> a = tf.constant([[1,2,3],[4,5,6]], tf.int32) >>> b = tf.constant([1,2], tf.int32) >>> tf.tile(a, b) <tf.Tensor: shape=(2, 6), dtype=int32, numpy= array([[1, 2, 3, 1, 2, 3], 

[4, 5, 6, 4, 5, 6]], dtype=int32)>

>>> c = tf.constant([2,1], tf.int32) >>> tf.tile(a, c) <tf.Tensor: shape=(4, 3), dtype=int32, numpy= array([[1, 2, 3], 

[4, 5, 6],
[1, 2, 3],
[4, 5, 6]], dtype=int32)>

>>> d = tf.constant([2,2], tf.int32) >>> tf.tile(a, d) <tf.Tensor: shape=(4, 6), dtype=int32, numpy= array([[1, 2, 3, 1, 2, 3], 

[4, 5, 6, 4, 5, 6],
[1, 2, 3, 1, 2, 3],
[4, 5, 6, 4, 5, 6]], dtype=int32)>

Arguments: 

input: Can be of any rank.
multiples: 1-D. Length must be the same as the number of dimensions in `input`

func TileGrad 

func TileGrad(scope *Scope, input tf.Output, multiples tf.Output) (output tf.Output)

Returns the gradient of `Tile`.

DEPRECATED at GraphDef version 3: TileGrad has been replaced with reduce_sum

Since `Tile` takes an input and repeats the input `multiples` times along each dimension, `TileGrad` takes in `multiples` and aggregates each repeated tile of `input` into `output`.

func Timestamp 

func Timestamp(scope *Scope) (ts tf.Output)

Provides the time since epoch in seconds.

Returns the timestamp as a `float64` for seconds since the Unix epoch.

Common usages include: * Logging * Providing a random number seed * Debugging graph execution * Generating timing information, mainly through comparison of timestamps

Note: In graph mode, the timestamp is computed when the op is executed, not when it is added to the graph. In eager mode, the timestamp is computed when the op is eagerly executed.

func ToBool 

func ToBool(scope *Scope, input tf.Output) (output tf.Output)

Converts a tensor to a scalar predicate.

Converts a tensor to a scalar predicate with the following rules:

  • For 0D tensors, truthiness is determined by comparing against a "zero" value. For numerical types it is the obvious zero. For strings it is the empty string.

  • For >0D tensors, truthiness is determined by looking at the number of elements. If has zero elements, then the result is false. Otherwise the result is true.

This matches the behavior of If and While for determining if a tensor counts as true/false for a branch condition.

func TopK 

func TopK(scope *Scope, input tf.Output, k int64, optional ...TopKAttr) (values tf.Output, indices tf.Output)

Finds values and indices of the `k` largest elements for the last dimension.

DEPRECATED at GraphDef version 7: Use TopKV2 instead

If the input is a vector (rank-1), finds the `k` largest entries in the vector and outputs their values and indices as vectors. Thus `values[j]` is the `j`-th largest entry in `input`, and its index is `indices[j]`.

For matrices (resp. higher rank input), computes the top `k` entries in each row (resp. vector along the last dimension). Thus, 

values.shape = indices.shape = input.shape[:-1] + [k]

If two elements are equal, the lower-index element appears first.

If `k` varies dynamically, use `TopKV2` below.

Arguments: 

input: 1-D or higher with last dimension at least `k`.
k: Number of top elements to look for along the last dimension (along each

row for matrices).

Returns: 

values: The `k` largest elements along each last dimensional slice.
indices: The indices of `values` within the last dimension of `input`.

func TopKUnique 

func TopKUnique(scope *Scope, input tf.Output, k int64) (topk tf.Output, topk_indices tf.Output)

Returns the TopK unique values in the array in sorted order.

The running time is proportional to the product of K and the input size. Sorting the whole array is more efficient for sufficiently large values of K. The median-of-medians algorithm is probably faster, but difficult to implement efficiently in XLA. If there are fewer than K unique numbers (not NANs), the results are padded with negative infinity. NaNs are never returned. Subnormal numbers are flushed to zero. If an element appears at multiple indices, the highest index is returned. If a TopK element never appears in the input due to padding values, the indices are padded with negative one. If a padding value appears in the input and padding is needed, the highest index of the padding value will be returned. The semantics are not the same as kth_order_statistic.

func TopKV2 

func TopKV2(scope *Scope, input tf.Output, k tf.Output, optional ...TopKV2Attr) (values tf.Output, indices tf.Output)

Finds values and indices of the `k` largest elements for the last dimension.

If the input is a vector (rank-1), finds the `k` largest entries in the vector and outputs their values and indices as vectors. Thus `values[j]` is the `j`-th largest entry in `input`, and its index is `indices[j]`.

For matrices (resp. higher rank input), computes the top `k` entries in each row (resp. vector along the last dimension). Thus, 

values.shape = indices.shape = input.shape[:-1] + [k]

If two elements are equal, the lower-index element appears first.

Arguments: 

input: 1-D or higher with last dimension at least `k`.
k: 0-D.  Number of top elements to look for along the last dimension (along each

row for matrices).

Returns: 

values: The `k` largest elements along each last dimensional slice.
indices: The indices of `values` within the last dimension of `input`.

func TopKWithUnique 

func TopKWithUnique(scope *Scope, input tf.Output, k int64) (topk tf.Output, topk_indices tf.Output)

Returns the TopK values in the array in sorted order.

This is a combination of MakeUnique and TopKUnique. The returned top-K will have its lower bits replaced by iota, thus it will be close to the original value but not exactly the same. The running time is proportional to the product of K and the input size. NaNs are never returned. Subnormal numbers are flushed to zero.

func TpuHandleToProtoKey added in v0.2.0 

func TpuHandleToProtoKey(scope *Scope, uid tf.Output) (proto_keys tf.Output)

Converts XRT's uid handles to TensorFlow-friendly input format.

Converts a uid handle for a compiled program into a vector of proto keys.

XRT compile ops return uids, and the TensorFlow execute op takes a proto key. This op enables a client to compile on TPU using XRT and execute using the standard TensorFlow execute op.

'uid' is the input handle. 'proto_keys' is a vector of proto keys, one for each core program.

func Transpose 

func Transpose(scope *Scope, x tf.Output, perm tf.Output) (y tf.Output)

Shuffle dimensions of x according to a permutation.

The output `y` has the same rank as `x`. The shapes of `x` and `y` satisfy: 

`y.shape[i] == x.shape[perm[i]] for i in [0, 1, ..., rank(x) - 1]`

func TridiagonalMatMul 

func TridiagonalMatMul(scope *Scope, superdiag tf.Output, maindiag tf.Output, subdiag tf.Output, rhs tf.Output) (output tf.Output)

Calculate product with tridiagonal matrix.

Calculates product of two matrices, where left matrix is a tridiagonal matrix.

Arguments: 

superdiag: Tensor of shape `[..., 1, M]`, representing superdiagonals of

tri-diagonal matrices to the left of multiplication. Last element is ignored. 

maindiag: Tensor of shape `[..., 1, M]`, representing main diagonals of tri-diagonal

matrices to the left of multiplication. 

subdiag: Tensor of shape `[..., 1, M]`, representing subdiagonals of tri-diagonal

matrices to the left of multiplication. First element is ignored. 

rhs: Tensor of shape `[..., M, N]`, representing MxN matrices to the right of

multiplication.

Returns Tensor of shape `[..., M, N]` containing the product.

func TridiagonalSolve 

func TridiagonalSolve(scope *Scope, diagonals tf.Output, rhs tf.Output, optional ...TridiagonalSolveAttr) (output tf.Output)

Solves tridiagonal systems of equations. 

Solves tridiagonal systems of equations.
Supports batch dimensions and multiple right-hand sides per each left-hand
side.
On CPU, solution is computed via Gaussian elimination with or without partial
pivoting, depending on `partial_pivoting` attribute. On GPU, Nvidia's cuSPARSE
library is used: https://docs.nvidia.com/cuda/cusparse/index.html#gtsv
Partial pivoting is not yet supported by XLA backends.

Arguments: 

diagonals: Tensor of shape `[..., 3, M]` whose innermost 2 dimensions represent the

tridiagonal matrices with three rows being the superdiagonal, diagonals, and subdiagonals, in order. The last element of the superdiagonal and the first element of the subdiagonal is ignored. 

rhs: Tensor of shape `[..., M, K]`, representing K right-hand sides per each

left-hand side.

Returns Tensor of shape `[..., M, K]` containing the solutions

func TruncateDiv 

func TruncateDiv(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns x / y element-wise, rounded towards zero.

Truncation designates that negative numbers will round fractional quantities toward zero. I.e. -7 / 5 = -1. This matches C semantics but it is different than Python semantics. See `FloorDiv` for a division function that matches Python Semantics.

*NOTE*: `TruncateDiv` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func TruncateMod 

func TruncateMod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns element-wise remainder of division. This emulates C semantics in that

the result here is consistent with a truncating divide. E.g. `truncate(x / y) * y + truncate_mod(x, y) = x`.

*NOTE*: `TruncateMod` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)

func TruncatedNormal 

func TruncatedNormal(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...TruncatedNormalAttr) (output tf.Output)

Outputs random values from a truncated normal distribution.

The generated values follow a normal distribution with mean 0 and standard deviation 1, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.

Arguments: 

shape: The shape of the output tensor.
dtype: The type of the output.

Returns A tensor of the specified shape filled with random truncated normal values.

func Unbatch 

func Unbatch(scope *Scope, batched_tensor tf.Output, batch_index tf.Output, id tf.Output, timeout_micros int64, optional ...UnbatchAttr) (unbatched_tensor tf.Output)

Reverses the operation of Batch for a single output Tensor.

An instance of Unbatch either receives an empty batched_tensor, in which case it asynchronously waits until the values become available from a concurrently running instance of Unbatch with the same container and shared_name, or receives a non-empty batched_tensor in which case it finalizes all other concurrently running instances and outputs its own element from the batch.

batched_tensor: The possibly transformed output of Batch. The size of the first 

dimension should remain unchanged by the transformations for the operation to
work.

batch_index: The matching batch_index obtained from Batch. id: The id scalar emitted by Batch. unbatched_tensor: The Tensor corresponding to this execution. timeout_micros: Maximum amount of time (in microseconds) to wait to receive the 

batched input tensor associated with a given invocation of the op.

container: Container to control resource sharing. shared_name: Instances of Unbatch with the same container and shared_name are 

assumed to possibly belong to the same batch. If left empty, the op name will
be used as the shared name.

func UnbatchDataset 

func UnbatchDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...UnbatchDatasetAttr) (handle tf.Output)

A dataset that splits the elements of its input into multiple elements.

func UnbatchGrad 

func UnbatchGrad(scope *Scope, original_input tf.Output, batch_index tf.Output, grad tf.Output, id tf.Output, optional ...UnbatchGradAttr) (batched_grad tf.Output)

Gradient of Unbatch.

Acts like Batch but using the given batch_index index of batching things as they become available. This ensures that the gradients are propagated back in the same session which did the forward pass.

original_input: The input to the Unbatch operation this is the gradient of. batch_index: The batch_index given to the Unbatch operation this is the gradient of. grad: The downstream gradient. id: The id scalar emitted by Batch. batched_grad: The return value, either an empty tensor or the batched gradient. container: Container to control resource sharing. shared_name: Instances of UnbatchGrad with the same container and shared_name 

are assumed to possibly belong to the same batch. If left empty, the op name
will be used as the shared name.

func UncompressElement 

func UncompressElement(scope *Scope, compressed tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (components []tf.Output)

Uncompresses a compressed dataset element.

func UnicodeDecode 

func UnicodeDecode(scope *Scope, input tf.Output, input_encoding string, optional ...UnicodeDecodeAttr) (row_splits tf.Output, char_values tf.Output)

Decodes each string in `input` into a sequence of Unicode code points.

The character codepoints for all strings are returned using a single vector `char_values`, with strings expanded to characters in row-major order.

The `row_splits` tensor indicates where the codepoints for each input string begin and end within the `char_values` tensor. In particular, the values for the `i`th string (in row-major order) are stored in the slice `[row_splits[i]:row_splits[i+1]]`. Thus:

  • `char_values[row_splits[i]+j]` is the Unicode codepoint for the `j`th character in the `i`th string (in row-major order).
  • `row_splits[i+1] - row_splits[i]` is the number of characters in the `i`th string (in row-major order).

Arguments: 

input: The text to be decoded. Can have any shape. Note that the output is flattened

to a vector of char values. 

input_encoding: Text encoding of the input strings. This is any of the encodings supported

by ICU ucnv algorithmic converters. Examples: `"UTF-16", "US ASCII", "UTF-8"`.

Returns: 

row_splits: A 1D int32 tensor containing the row splits.
char_values: A 1D int32 Tensor containing the decoded codepoints.

func UnicodeDecodeWithOffsets 

func UnicodeDecodeWithOffsets(scope *Scope, input tf.Output, input_encoding string, optional ...UnicodeDecodeWithOffsetsAttr) (row_splits tf.Output, char_values tf.Output, char_to_byte_starts tf.Output)

Decodes each string in `input` into a sequence of Unicode code points.

The character codepoints for all strings are returned using a single vector `char_values`, with strings expanded to characters in row-major order. Similarly, the character start byte offsets are returned using a single vector `char_to_byte_starts`, with strings expanded in row-major order.

The `row_splits` tensor indicates where the codepoints and start offsets for each input string begin and end within the `char_values` and `char_to_byte_starts` tensors. In particular, the values for the `i`th string (in row-major order) are stored in the slice `[row_splits[i]:row_splits[i+1]]`. Thus:

  • `char_values[row_splits[i]+j]` is the Unicode codepoint for the `j`th character in the `i`th string (in row-major order).
  • `char_to_bytes_starts[row_splits[i]+j]` is the start byte offset for the `j`th character in the `i`th string (in row-major order).
  • `row_splits[i+1] - row_splits[i]` is the number of characters in the `i`th string (in row-major order).

Arguments: 

input: The text to be decoded. Can have any shape. Note that the output is flattened

to a vector of char values. 

input_encoding: Text encoding of the input strings. This is any of the encodings supported

by ICU ucnv algorithmic converters. Examples: `"UTF-16", "US ASCII", "UTF-8"`.

Returns: 

row_splits: A 1D int32 tensor containing the row splits.
char_values: A 1D int32 Tensor containing the decoded codepoints.
char_to_byte_starts: A 1D int32 Tensor containing the byte index in the input string where each

character in `char_values` starts.

func UnicodeEncode 

func UnicodeEncode(scope *Scope, input_values tf.Output, input_splits tf.Output, output_encoding string, optional ...UnicodeEncodeAttr) (output tf.Output)

Encode a tensor of ints into unicode strings.

Returns a vector of strings, where `output[i]` is constructed by encoding the Unicode codepoints in `input_values[input_splits[i]:input_splits[i+1]]` using `output_encoding`.

---

Example:

``` input_values = [72, 101, 108, 108, 111, 87, 111, 114, 108, 100] input_splits = [0, 5, 10] output_encoding = 'UTF-8'

output = ['Hello', 'World'] ```

Arguments: 

input_values: A 1D tensor containing the unicode codepoints that should be encoded.
input_splits: A 1D tensor specifying how the unicode codepoints should be split into strings.

In particular, `output[i]` is constructed by encoding the codepoints in the slice `input_values[input_splits[i]:input_splits[i+1]]`. 

output_encoding: Unicode encoding of the output strings. Valid encodings are: `"UTF-8",

"UTF-16-BE", and "UTF-32-BE"`.

Returns The 1-D Tensor of strings encoded from the provided unicode codepoints.

func UnicodeScript 

func UnicodeScript(scope *Scope, input tf.Output) (output tf.Output)

Determine the script codes of a given tensor of Unicode integer code points.

This operation converts Unicode code points to script codes corresponding to each code point. Script codes correspond to International Components for Unicode (ICU) UScriptCode values.

See [ICU project docs](http://icu-project.org/apiref/icu4c/uscript_8h.html) for more details on script codes.

For an example, see the unicode strings guide on [unicode scripts] (https://www.tensorflow.org/tutorials/load_data/unicode#representing_unicode).

Returns -1 (USCRIPT_INVALID_CODE) for invalid codepoints. Output shape will match input shape.

Examples:

>>> tf.strings.unicode_script([1, 31, 38]) <tf.Tensor: shape=(3,), dtype=int32, numpy=array([0, 0, 0], dtype=int32)>

Arguments: 

input: A Tensor of int32 Unicode code points.

Returns A Tensor of int32 script codes corresponding to each input code point.

func UnicodeTranscode 

func UnicodeTranscode(scope *Scope, input tf.Output, input_encoding string, output_encoding string, optional ...UnicodeTranscodeAttr) (output tf.Output)

Transcode the input text from a source encoding to a destination encoding.

The input is a string tensor of any shape. The output is a string tensor of the same shape containing the transcoded strings. Output strings are always valid unicode. If the input contains invalid encoding positions, the `errors` attribute sets the policy for how to deal with them. If the default error-handling policy is used, invalid formatting will be substituted in the output by the `replacement_char`. If the errors policy is to `ignore`, any invalid encoding positions in the input are skipped and not included in the output. If it set to `strict` then any invalid formatting will result in an InvalidArgument error.

This operation can be used with `output_encoding = input_encoding` to enforce correct formatting for inputs even if they are already in the desired encoding.

If the input is prefixed by a Byte Order Mark needed to determine encoding (e.g. if the encoding is UTF-16 and the BOM indicates big-endian), then that BOM will be consumed and not emitted into the output. If the input encoding is marked with an explicit endianness (e.g. UTF-16-BE), then the BOM is interpreted as a non-breaking-space and is preserved in the output (including always for UTF-8).

The end result is that if the input is marked as an explicit endianness the transcoding is faithful to all codepoints in the source. If it is not marked with an explicit endianness, the BOM is not considered part of the string itself but as metadata, and so is not preserved in the output.

Examples:

>>> tf.strings.unicode_transcode(["Hello", "TensorFlow", "2.x"], "UTF-8", "UTF-16-BE") <tf.Tensor: shape=(3,), dtype=string, numpy= array([b'\x00H\x00e\x00l\x00l\x00o', 

b'\x00T\x00e\x00n\x00s\x00o\x00r\x00F\x00l\x00o\x00w',
b'\x002\x00.\x00x'], dtype=object)>

>>> tf.strings.unicode_transcode(["A", "B", "C"], "US ASCII", "UTF-8").numpy() array([b'A', b'B', b'C'], dtype=object)

Arguments: 

input: The text to be processed. Can have any shape.
input_encoding: Text encoding of the input strings. This is any of the encodings supported

by ICU ucnv algorithmic converters. Examples: `"UTF-16", "US ASCII", "UTF-8"`. 

output_encoding: The unicode encoding to use in the output. Must be one of

`"UTF-8", "UTF-16-BE", "UTF-32-BE"`. Multi-byte encodings will be big-endian.

Returns A string tensor containing unicode text encoded using `output_encoding`.

func UniformCandidateSampler 

func UniformCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...UniformCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)

Generates labels for candidate sampling with a uniform distribution.

See explanations of candidate sampling and the data formats at go/candidate-sampling.

For each batch, this op picks a single set of sampled candidate labels.

The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.

Arguments: 

true_classes: A batch_size * num_true matrix, in which each row contains the

IDs of the num_true target_classes in the corresponding original label. 

num_true: Number of true labels per context.
num_sampled: Number of candidates to randomly sample.
unique: If unique is true, we sample with rejection, so that all sampled

candidates in a batch are unique. This requires some approximation to estimate the post-rejection sampling probabilities. 

range_max: The sampler will sample integers from the interval [0, range_max).

Returns: 

sampled_candidates: A vector of length num_sampled, in which each element is

the ID of a sampled candidate. 

true_expected_count: A batch_size * num_true matrix, representing

the number of times each candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability. 

sampled_expected_count: A vector of length num_sampled, for each sampled

candidate representing the number of times the candidate is expected to occur in a batch of sampled candidates. If unique=true, then this is a probability.

func UniformDequantize added in v0.2.0 

func UniformDequantize(scope *Scope, input tf.Output, scales tf.Output, zero_points tf.Output, Tout tf.DataType, quantization_min_val int64, quantization_max_val int64, optional ...UniformDequantizeAttr) (output tf.Output)

Perform dequantization on the quantized Tensor `input`.

Given quantized `input` which was quantized using `scales` and `zero_points`, performs dequantization using the formula: dequantized_data = (quantized_data - zero_point) * scale.

Arguments: 

input: Must be a Tensor of Tin.
scales: The float value(s) used as scale(s) when quantizing original data that input represents.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (input.dim_size(quantization_axis),) (per-axis quantization). 

zero_points: The int32 value(s) used as zero_point(s) when quantizing original data that input represents.

Same shape condition as scales. 

Tout: The type of output Tensor. A tf.DType from: tf.qint8, tf.qint32
quantization_min_val: The quantization min value that was used when input was quantized.

The purpose of this attribute is typically (but not limited to) to indicate narrow range, where this is set to: `(Tin lowest) + 1` if narrow range, and `(Tin lowest)` otherwise. For example, if Tin is qint8, this is set to -127 if narrow range quantized or -128 if not. 

quantization_max_val: The quantization max value that was used when input was quantized.

The purpose of this attribute is typically (but not limited to) indicate narrow range, where this is set to: `(Tout max)` for both narrow range and not narrow range. For example, if Tin is qint8, this is set to 127.

Returns The output dequantized Tensor of Tout, whose shape is same as input.

func UniformQuantize added in v0.3.0 

func UniformQuantize(scope *Scope, input tf.Output, scales tf.Output, zero_points tf.Output, Tout tf.DataType, quantization_min_val int64, quantization_max_val int64, optional ...UniformQuantizeAttr) (output tf.Output)

Perform quantization on Tensor `input`.

Given `input`, `scales` and `zero_points`, performs quantization using the formula: quantized_data = floor(input_data * (1.0f / scale) + 0.5f) + zero_point

Arguments: 

input: Must be a Tensor of Tin.
scales: The float value(s) to use as scale(s) to quantize `input`.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (input.dim_size(quantization_axis),) (per-axis quantization). 

zero_points: The int32 value(s) to use as zero_point(s) to quantize `input`.

Same shape condition as scales. 

Tout: The type of output Tensor. A tf.DType from: tf.float32
quantization_min_val: The quantization min value to quantize `input`.

The purpose of this attribute is typically (but not limited to) to indicate narrow range, where this is set to: `(Tin lowest) + 1` if narrow range, and `(Tin lowest)` otherwise. For example, if Tin is qint8, this is set to -127 if narrow range quantized or -128 if not. 

quantization_max_val: The quantization max value to quantize `input`.

The purpose of this attribute is typically (but not limited to) indicate narrow range, where this is set to: `(Tout max)` for both narrow range and not narrow range. For example, if Tin is qint8, this is set to 127.

Returns The output quantized Tensor of Tout, whose shape is same as input.

func UniformQuantizedAdd added in v0.4.0 

func UniformQuantizedAdd(scope *Scope, lhs tf.Output, rhs tf.Output, lhs_scales tf.Output, lhs_zero_points tf.Output, rhs_scales tf.Output, rhs_zero_points tf.Output, output_scales tf.Output, output_zero_points tf.Output, lhs_quantization_min_val int64, lhs_quantization_max_val int64, rhs_quantization_min_val int64, rhs_quantization_max_val int64, output_quantization_min_val int64, output_quantization_max_val int64, optional ...UniformQuantizedAddAttr) (output tf.Output)

Perform quantized add of quantized Tensor `lhs` and quantized Tensor `rhs` to make quantized `output`.

Given quantized `lhs` and quantized `rhs`, performs quantized add on `lhs` and `rhs` to make quantized `output`.

`UniformQuantizedAdd` follows Numpy broadcasting rules. The two input array shapes are compared element-wise. Starting with the trailing dimensions, the two dimensions either have to be equal or one of them needs to be 1.

`lhs` and `rhs` must be quantized Tensor, where data value is quantized using the formula: ``` quantized_data = clip(original_data / scale + zero_point, quantization_min_val, quantization_max_val) ``` `output` is also quantized, using the same formula.

If `lhs` and `output` is both per-axis quantized, the quantization axis must match. Also, if `rhs` and `output` is both per-axis quantized, the quantization axis must match. *Match* means the axis must match when adding, regarding the broadcasting. i.e. For both operands `lhs` and `rhs`, if `operand.quantization_axis` >= 0 and `output.quantization_axis` >= 0, `operand.dims` - `operand.quantization_axis` must be equal to `output.dims` - `output.quantization_axis`.

Arguments: 

lhs: Must be a quantized tensor.
rhs: Must be a quantized tensor.
lhs_scales: The float value(s) used as scale factors when quantizing the original data that `lhs` represents.
lhs_zero_points: The int32 value(s) used as zero points when quantizing original data that `lhs` represents.

Must have same shape with `lhs_scales`. 

rhs_scales: The float value(s) used as scale factors when quantizing the original data that `rhs` represents.
rhs_zero_points: The int32 value(s) used as zero points when quantizing original data that `rhs` represents.

Must have same shape with `rhs_scales`. 

output_scales: The float value(s) to use as scale factors when quantizing original data that `output` represents.
output_zero_points: The int32 value(s) used as zero points when quantizing original data that output represents.

Must have same shape with `output_scales`. 

lhs_quantization_min_val: The min value of the quantized data stored in `lhs`.

For example, if `Tin` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

lhs_quantization_max_val: The max value of the quantized data stored in `lhs`.

For example, if `Tin` is `qint8`, this must be set to 127. 

rhs_quantization_min_val: The min value of the quantized data stored in `rhs`.

For example, if `Tin` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

rhs_quantization_max_val: The max value of the quantized data stored in `rhs`.

For example, if `Tin` is `qint8`, this must be set to 127. 

output_quantization_min_val: The min value of the quantized data stored in `output`.

For example, if `Tout` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

output_quantization_max_val: The max value of the quantized data stored in `output`.

For example, if `Tout` is `qint8`, this must be set to 127.

Returns The output quantized tensor.

func UniformQuantizedClipByValue added in v0.3.0 

func UniformQuantizedClipByValue(scope *Scope, operand tf.Output, min tf.Output, max tf.Output, scales tf.Output, zero_points tf.Output, quantization_min_val int64, quantization_max_val int64, optional ...UniformQuantizedClipByValueAttr) (output tf.Output)

Perform clip by value on the quantized Tensor `operand`.

Given quantized `operand` which was quantized using `scales` and `zero_points`, performs clip by value using `min` and `max` values. If quantization_axis is -1 (per-tensor quantized), the entire operand is clipped using scalar min, max. Otherwise (per-channel quantized), the clipping is also done per-channel.

Arguments: 

operand: Must be a Tensor of T.
min: The min value(s) to clip operand. Must be a Tensor of T.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (operand.dim_size(quantization_axis),) (per-axis quantization). 

max: The min value(s) to clip operand. Must be a Tensor of T.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (operand.dim_size(quantization_axis),) (per-axis quantization). 

scales: The float value(s) used as scale(s) when quantizing `operand`, `min` and `max`.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (operand.dim_size(quantization_axis),) (per-axis quantization). 

zero_points: The int32 value(s) used as zero_point(s) when quantizing `operand`, `min` and `max`.

Same shape condition as scales. 

quantization_min_val: The quantization min value that was used when operand was quantized.
quantization_max_val: The quantization max value that was used when operand was quantized.

Returns The output clipped Tensor of T, whose shape is same as operand.

func UniformQuantizedConvolution added in v0.4.0 

func UniformQuantizedConvolution(scope *Scope, lhs tf.Output, rhs tf.Output, lhs_scales tf.Output, lhs_zero_points tf.Output, rhs_scales tf.Output, rhs_zero_points tf.Output, output_scales tf.Output, output_zero_points tf.Output, Tout tf.DataType, padding string, lhs_quantization_min_val int64, lhs_quantization_max_val int64, rhs_quantization_min_val int64, rhs_quantization_max_val int64, output_quantization_min_val int64, output_quantization_max_val int64, optional ...UniformQuantizedConvolutionAttr) (output tf.Output)

Perform quantized convolution of quantized Tensor `lhs` and quantized Tensor `rhs`. to make quantized `output`.

Given quantized `lhs` and quantized `rhs`, performs quantized dot on `lhs` and `rhs` to make quantized `output`.

`lhs` and `rhs` must be Tensors of same rank, and meet following shape conditions. - `lhs_feature` % `feature_group_count` == 0 - `lhs_feature` % `rhs_input_feature` == 0 - `lhs_feature` / `feature_group_count` == `rhs_input_feature` - `rhs_output_feature` % `feature_group_count` == 0 - `lhs_batch` % `batch_group_count` == 0 - `rhs_output_feature` % `batch_group_count` == 0

`lhs` and `rhs` must be quantized Tensor, where data value is quantized using the formula: ``` quantized_data = clip(original_data / scale + zero_point, quantization_min_val, quantization_max_val) ``` `output` is also quantized, using the same formula. If `rhs` is per-tensor quantized, `output` must be also per-tensor quantized.

Arguments: 

lhs: Must be a quantized tensor, rank >= 3.
rhs: Must be a quantized tensor, same rank as `lhs`.
lhs_scales: The float value(s) used as scale factors when quantizing the original data that `lhs` represents.

Must be a scalar `Tensor` (`lhs` supports only per-tensor quantization). 

lhs_zero_points: The int32 value(s) used as zero points when quantizing original data that `lhs` represents.

Same shape condition as `lhs_scales`. 

rhs_scales: The float value(s) used as scale factors when quantizing the original data that `rhs` represents.

Must be a scalar `Tensor` for per-tensor quantization, or 1D `Tensor` of size `rhs.dim_size(kernel_output_feature_dimension)`, for per-channel quantization. 

rhs_zero_points: The int32 value(s) used as zero points when quantizing original data that `rhs` represents.

Same shape condition as `rhs_scales`. 

output_scales: The float value(s) to use as scale factors when quantizing original data that `output` represents.

Must be a scalar `Tensor` for per-tensor quantization, or 1D `Tensor` of size `rhs.dim_size(kernel_output_feature_dimension)` - which is equal to `output.dim_size(output_feature_dimension)`, for per-channel quantization. If `rhs` is per-tensor quantized, output must be also per-tensor quantized. This means that if `rhs_scales` and `rhs_zero_points` are scalar `Tensor`s, `output_scales` and `output_zero_points` must be scalar `Tensor`s as well. 

output_zero_points: The int32 value(s) used as zero points when quantizing original data that output represents.

Same shape condition as `output_scales`. 

Tout: The type of `output` `Tensor`.
padding: string from: `"SAME"`, `"VALID"`, or `"EXPLICIT"`, indicating the type of padding algorithm to use.
lhs_quantization_min_val: The min value of the quantized data stored in `lhs`.

For example, if `Tin` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

lhs_quantization_max_val: The max value of the quantized data stored in `lhs`.

For example, if `Tin` is `qint8`, this must be set to 127. 

rhs_quantization_min_val: The min value of the quantized data stored in `rhs`.

For example, if `Tin` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

rhs_quantization_max_val: The max value of the quantized data stored in `rhs`.

For example, if `Tin` is `qint8`, this must be set to 127. 

output_quantization_min_val: The min value of the quantized data stored in `output`.

For example, if `Tout` is `qint8`, this must be set to -127 if narrow range quantized or -128 if not. 

output_quantization_max_val: The max value of the quantized data stored in `output`.

For example, if `Tout` is `qint8`, this must be set to 127.

Returns The output quantized tensor of `Tout`, same rank as `lhs` and `rhs`.

func UniformQuantizedConvolutionHybrid added in v0.4.0 

func UniformQuantizedConvolutionHybrid(scope *Scope, lhs tf.Output, rhs tf.Output, rhs_scales tf.Output, rhs_zero_points tf.Output, Tout tf.DataType, padding string, rhs_quantization_min_val int64, rhs_quantization_max_val int64, optional ...UniformQuantizedConvolutionHybridAttr) (output tf.Output)

Perform hybrid quantized convolution of float Tensor `lhs` and quantized Tensor `rhs`.

Given float `lhs` and quantized `rhs`, internally performs quantization on `lhs`, and then performs quantized convolution on quantized `lhs` and `rhs`.

The internal quantization on `lhs` is a quantization to `Trhs`, dynamic range, per-batch (per-axis along axis `dimension_numbers.input_batch_dimension`), asymmetric, and not narrow range (the range is [Trhs_MIN, Trhs_MAX]).

`lhs` and `rhs` must be Tensors of same rank, and meet following shape conditions. - lhs_feature % feature_group_count == 0 - lhs_feature % rhs_input_feature == 0 - lhs_feature / feature_group_count == rhs_input_feature - rhs_output_feature % feature_group_count == 0 - lhs_batch % batch_group_count == 0 - rhs_output_feature % batch_group_count == 0

`rhs` must be quantized Tensor, where its data value is quantized using the formula: quantized_data = clip(original_data / scale + zero_point, quantization_min_val, quantization_max_val).

Arguments: 

lhs: Must be a non-quantized Tensor of `Tlhs`, rank >= 3.
rhs: Must be a quantized Tensor of `Trhs`, same rank as `lhs`.
rhs_scales: The float value(s) used as scale factors when quantizing the original data that `rhs` represents.

Must be a scalar Tensor for per-tensor quantization, or 1D Tensor of size `rhs.dim_size(kernel_output_feature_dimension)`, for per-channel quantization. 

rhs_zero_points: The int32 value(s) used as zero_point when quantizing original data that `rhs` represents.

Same shape condition as `rhs_scales`. 

Tout: The type of output Tensor.
padding: string from: `"SAME"`, `"VALID"`, or `"EXPLICIT"`, indicating the type of padding algorithm to use.
rhs_quantization_min_val: The min value of the quantized data stored in `rhs`.

For example, if `Trhs` is qint8, this must be set to -127 if narrow range quantized or -128 if not. 

rhs_quantization_max_val: The max value of the quantized data stored in `rhs`.

For example, if `Trhs` is qint8, this must be set to 127.

Returns The output Tensor of `Tout`, same rank as `lhs` and `rhs`. The output data is the non-quantized output data.

func UniformQuantizedDot added in v0.3.0 

func UniformQuantizedDot(scope *Scope, lhs tf.Output, rhs tf.Output, lhs_scales tf.Output, lhs_zero_points tf.Output, rhs_scales tf.Output, rhs_zero_points tf.Output, output_scales tf.Output, output_zero_points tf.Output, Tout tf.DataType, lhs_quantization_min_val int64, lhs_quantization_max_val int64, rhs_quantization_min_val int64, rhs_quantization_max_val int64, output_quantization_min_val int64, output_quantization_max_val int64, optional ...UniformQuantizedDotAttr) (output tf.Output)

Perform quantized dot of quantized Tensor `lhs` and quantized Tensor `rhs` to make quantized `output`.

Given quantized `lhs` and quantized `rhs`, performs quantized dot on `lhs` and `rhs` to make quantized `output`. `lhs` and `rhs` must be 2D Tensors and the lhs.dim_size(1) must match rhs.dim_size(0). `lhs` and `rhs` must be quantized Tensor, where data value is quantized using the formula: quantized_data = clip(original_data / scale + zero_point, quantization_min_val, quantization_max_val). `output` is also quantized, using the same formula. If `rhs` is per-tensor quantized, `output` must be also per-tensor quantized.

Arguments: 

lhs: Must be a 2D Tensor of Tin.
rhs: Must be a 2D Tensor of Tin.
lhs_scales: The float value(s) used as scale when quantizing original data that lhs represents.

Must be a scalar Tensor (lhs supports only per-tensor quantization). 

lhs_zero_points: The int32 value(s) used as zero_point when quantizing original data that lhs represents.

Same shape condition as lhs_scales. 

rhs_scales: The float value(s) used as scale when quantizing original data that rhs represents.

Must be a scalar Tensor (per-tensor quantization) or 1D Tensor of size (rhs.dim_size(1),) (per-channel quantization). 

rhs_zero_points: The int32 value(s) used as zero_point when quantizing original data that rhs represents.

Same shape condition as rhs_scales. 

output_scales: The float value(s) to use as scales when quantizing original data that output represents.

Must be a scalar Tensor (per-tensor quantization) or 1D Tensor of size (output.dim_size(1),) (per-channel quantization). If rhs is per-tensor quantized, output must be also per-tensor quantized. This means that if rhs_scales and rhs_zero_points are scalar Tensors, output_scales and output_zero_points must be scalar Tensors as well. 

output_zero_points: The int32 value(s) used as zero_point when quantizing original data that output represents.

Same shape condition as rhs_scales. 

Tout: The type of output Tensor.
lhs_quantization_min_val: The min value of the quantized data stored in lhs.

For example, if Tin is qint8, this must be set to -127 if narrow range quantized or -128 if not. 

lhs_quantization_max_val: The max value of the quantized data stored in rhs.

For example, if Tin is qint8, this must be set to 127. 

rhs_quantization_min_val: The min value of the quantized data stored in rhs.

For example, if Trhs is qint8, this must be set to -127 if narrow range quantized or -128 if not. 

rhs_quantization_max_val: The max value of the quantized data stored in rhs.

For example, if Trhs is qint8, this must be set to 127. 

output_quantization_min_val: The min value of the quantized data stored in output.

For example, if Tout is qint8, this must be set to -127 if narrow range quantized or -128 if not. 

output_quantization_max_val: The max value of the quantized data stored in output.

For example, if Tout is qint8, this must be set to 127.

Returns The output 2D Tensor of Tout, whose shape is (lhs.dim_size(0), rhs.dim_size(1)).

func UniformQuantizedDotHybrid added in v0.2.0 

func UniformQuantizedDotHybrid(scope *Scope, lhs tf.Output, rhs tf.Output, rhs_scales tf.Output, rhs_zero_points tf.Output, Tout tf.DataType, rhs_quantization_min_val int64, rhs_quantization_max_val int64, optional ...UniformQuantizedDotHybridAttr) (output tf.Output)

Perform hybrid quantized dot of float Tensor `lhs` and quantized Tensor `rhs`.

Given float `lhs` and quantized `rhs`, internally performs quantization on `lhs`, and then performs quantized dot on quantized lhs and `rhs`. The internal quantization on `lhs` is a quantization to qint8, dynamic range, per-batch (per-axis along axis 0), asymmetric, and not narrow range (the range is [-128, 127]). `lhs` and `rhs` must be 2D Tensors and the lhs.dim_size(1) must match rhs.dim_size(0). `rhs` must be quantized Tensor, where its data value is quantized using the formula: quantized_data = clip(original_data / scale + zero_point, quantization_min_val, quantization_max_val).

Arguments: 

lhs: Must be a 2D Tensor of Tlhs.
rhs: Must be a 2D Tensor of Trhs.
rhs_scales: The float value(s) used as scale when quantizing original data that rhs represents.

Must be a scalar Tensor (per-tensor quantization) or 1D Tensor of size (rhs.dim_size(1),) (per-channel quantization). 

rhs_zero_points: The int32 value(s) used as zero_point when quantizing original data that rhs represents.

Same shape condition as rhs_scales. 

Tout: The type of output Tensor.
rhs_quantization_min_val: The min value of the quantized data stored in rhs.

For example, if Trhs is qint8, this must be set to -127 if narrow range quantized or -128 if not. 

rhs_quantization_max_val: The max value of the quantized data stored in rhs.

For example, if Trhs is qint8, this must be set to 127.

Returns The output 2D Tensor of Tout, whose shape is (lhs.dim_size(0), rhs.dim_size(1)). The output data is the original output data itself (Not quantized).

func UniformRequantize added in v0.3.0 

func UniformRequantize(scope *Scope, input tf.Output, input_scales tf.Output, input_zero_points tf.Output, output_scales tf.Output, output_zero_points tf.Output, Tout tf.DataType, input_quantization_min_val int64, input_quantization_max_val int64, output_quantization_min_val int64, output_quantization_max_val int64, optional ...UniformRequantizeAttr) (output tf.Output)

Given quantized tensor `input`, requantize it with new quantization parameters.

Given quantized tensor `input`, which was quantized using {input_scales, input_zero_points, input_quantization_axis, input_quantization_min_val, input_quantization_max_val}, requantize it to a tensor, which is quantized using {output_scales, output_zero_points, output_quantization_axis, output_quantization_min_val, output_quantization_max_val}. The requantization is done by using the formula: output_quantized_data = clip( 

(input_quantized_data - input_zero_point) * (input_scale / output_scale) + output_zero_point,
output_quantization_min_val,
output_quantization_max_val)

Per-tensor and per-axis quantization supported cases are followings: * per-tensor -> per-tensor * per-tensor -> per-axis * per-axis -> per-axis where input_quantization_axis equals output_quantization_axis. i.e. At least one among input_quantization_axis and output_quantization_axis must be -1, or two must be equal.

Arguments: 

input: Must be a Tensor of Tin.
input_scales: The float value(s) used as scale(s) when quantizing original data that `input` represents.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (input.dim_size(quantization_axis),) (per-axis quantization). 

input_zero_points: The int32 value(s) used as zero_point(s) when quantizing original data that `input` represents.

Same shape condition as scales. 

output_scales: The float value(s) to use as new scale(s) to quantize original data that `input` represents.

Must be a scalar Tensor if quantization_axis is -1 (per-tensor quantization), otherwise 1D Tensor of size (input.dim_size(quantization_axis),) (per-axis quantization). 

output_zero_points: The int32 value(s) to use as new zero_point(s) to quantize original data that `input` represents.

Same shape condition as scales. 

Tout: The type of output Tensor. A tf.DType from: tf.qint8, tf.qint32
input_quantization_min_val: The quantization min value that was used when quantizing original data that `input` represents.

The purpose of this attribute is typically (but not limited to) to indicate narrow range, where this is set to: `(Tin lowest) + 1` if narrow range, and `(Tin lowest)` otherwise. For example, if Tin is qint8, this is set to -127 if narrow range quantized or -128 if not. 

input_quantization_max_val: The quantization max value that was used when quantizing original data that `input` represents.

The purpose of this attribute is typically (but not limited to) indicate narrow range, where this is set to: `(Tout max)` for both narrow range and not narrow range. For example, if Tin is qint8, this is set to 127. 

output_quantization_min_val: The new quantization min value to quantize original data that `input` represents.
output_quantization_max_val: The new quantization max value to quantize original data that `input` represents.

Returns The output quantized Tensor of Tout, whose shape is same as input.

func Unique 

func Unique(scope *Scope, x tf.Output, optional ...UniqueAttr) (y tf.Output, idx tf.Output)

Finds unique elements in a 1-D tensor.

This operation returns a tensor `y` containing all of the unique elements of `x` sorted in the same order that they occur in `x`; `x` does not need to be sorted. This operation also returns a tensor `idx` the same size as `x` that contains the index of each value of `x` in the unique output `y`. In other words:

`y[idx[i]] = x[i] for i in [0, 1,...,rank(x) - 1]`

Examples:

``` # tensor 'x' is [1, 1, 2, 4, 4, 4, 7, 8, 8] y, idx = unique(x) y ==> [1, 2, 4, 7, 8] idx ==> [0, 0, 1, 2, 2, 2, 3, 4, 4] ```

``` # tensor 'x' is [4, 5, 1, 2, 3, 3, 4, 5] y, idx = unique(x) y ==> [4, 5, 1, 2, 3] idx ==> [0, 1, 2, 3, 4, 4, 0, 1] ```

Arguments: 

x: 1-D.

Returns: 

y: 1-D.
idx: 1-D.

func UniqueDataset 

func UniqueDataset(scope *Scope, input_dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...UniqueDatasetAttr) (handle tf.Output)

Creates a dataset that contains the unique elements of `input_dataset`.

func UniqueV2 

func UniqueV2(scope *Scope, x tf.Output, axis tf.Output, optional ...UniqueV2Attr) (y tf.Output, idx tf.Output)

Finds unique elements along an axis of a tensor.

This operation either returns a tensor `y` containing unique elements along the `axis` of a tensor. The returned unique elements is sorted in the same order as they occur along `axis` in `x`. This operation also returns a tensor `idx` that is the same size as the number of the elements in `x` along the `axis` dimension. It contains the index in the unique output `y`. In other words, for an `1-D` tensor `x` with `axis = None:

`y[idx[i]] = x[i] for i in [0, 1,...,rank(x) - 1]`

For example:

``` # tensor 'x' is [1, 1, 2, 4, 4, 4, 7, 8, 8] y, idx = unique(x) y ==> [1, 2, 4, 7, 8] idx ==> [0, 0, 1, 2, 2, 2, 3, 4, 4] ```

For an `2-D` tensor `x` with `axis = 0`:

``` # tensor 'x' is [[1, 0, 0], # [1, 0, 0], # [2, 0, 0]] y, idx = unique(x, axis=0) y ==> [[1, 0, 0], 

[2, 0, 0]]

idx ==> [0, 0, 1] ```

For an `2-D` tensor `x` with `axis = 1`:

``` # tensor 'x' is [[1, 0, 0], # [1, 0, 0], # [2, 0, 0]] y, idx = unique(x, axis=1) y ==> [[1, 0], 

[1, 0],
[2, 0]]

idx ==> [0, 1, 1] ```

Arguments: 

x: A `Tensor`.
axis: A `Tensor` of type `int32` (default: None). The axis of the Tensor to

find the unique elements.

Returns: 

y: A `Tensor`. Unique elements along the `axis` of `Tensor` x.
idx: A 1-D Tensor. Has the same type as x that contains the index of each

value of x in the output y.

func UniqueWithCounts 

func UniqueWithCounts(scope *Scope, x tf.Output, optional ...UniqueWithCountsAttr) (y tf.Output, idx tf.Output, count tf.Output)

Finds unique elements in a 1-D tensor.

This operation returns a tensor `y` containing all of the unique elements of `x` sorted in the same order that they occur in `x`. This operation also returns a tensor `idx` the same size as `x` that contains the index of each value of `x` in the unique output `y`. Finally, it returns a third tensor `count` that contains the count of each element of `y` in `x`. In other words:

`y[idx[i]] = x[i] for i in [0, 1,...,rank(x) - 1]`

For example:

``` # tensor 'x' is [1, 1, 2, 4, 4, 4, 7, 8, 8] y, idx, count = unique_with_counts(x) y ==> [1, 2, 4, 7, 8] idx ==> [0, 0, 1, 2, 2, 2, 3, 4, 4] count ==> [2, 1, 3, 1, 2] ```

Arguments: 

x: 1-D.

Returns: 

y: 1-D.
idx: 1-D.
count: 1-D.

func UniqueWithCountsV2 

func UniqueWithCountsV2(scope *Scope, x tf.Output, axis tf.Output, optional ...UniqueWithCountsV2Attr) (y tf.Output, idx tf.Output, count tf.Output)

Finds unique elements along an axis of a tensor.

This operation either returns a tensor `y` containing unique elements along the `axis` of a tensor. The returned unique elements is sorted in the same order as they occur along `axis` in `x`. This operation also returns a tensor `idx` and a tensor `count` that are the same size as the number of the elements in `x` along the `axis` dimension. The `idx` contains the index in the unique output `y` and the `count` contains the count in the unique output `y`. In other words, for an `1-D` tensor `x` with `axis = None:

`y[idx[i]] = x[i] for i in [0, 1,...,rank(x) - 1]`

For example:

``` x = tf.constant([1, 1, 2, 4, 4, 4, 7, 8, 8]) y, idx, count = tf.raw_ops.UniqueWithCountsV2(x=x, axis = [0]) y ==> [1, 2, 4, 7, 8] idx ==> [0, 0, 1, 2, 2, 2, 3, 4, 4] count ==> [2, 1, 3, 1, 2] ```

For a `2-D` tensor `x` with `axis = 0`:

``` x = tf.constant([[1, 0, 0], 

[1, 0, 0],
[2, 0, 0]])

y, idx, count = tf.raw_ops.UniqueWithCountsV2(x=x, axis=[0]) y ==> [[1, 0, 0], 

[2, 0, 0]]

idx ==> [0, 0, 1] count ==> [2, 1] ```

For a `2-D` tensor `x` with `axis = 1`:

``` x = tf.constant([[1, 0, 0], 

[1, 0, 0],
[2, 0, 0]])

y, idx, count = tf.raw_ops.UniqueWithCountsV2(x=x, axis=[1]) y ==> [[1, 0], 

[1, 0],
[2, 0]]

idx ==> [0, 1, 1] count ==> [1, 2] ```

Arguments: 

x: A `Tensor`.
axis: A `Tensor` of type `int32` (default: None). The axis of the Tensor to

find the unique elements.

Returns: 

y: A `Tensor`. Unique elements along the `axis` of `Tensor` x.
idx: A 1-D Tensor. Has the same type as x that contains the index of each

value of x in the output y. 

count: A 1-D Tensor. The count of each value of x in the output y.

func Unpack 

func Unpack(scope *Scope, value tf.Output, num int64, optional ...UnpackAttr) (output []tf.Output)

Unpacks a given dimension of a rank-`R` tensor into `num` rank-`(R-1)` tensors.

Unpacks `num` tensors from `value` by chipping it along the `axis` dimension. For example, given a tensor of shape `(A, B, C, D)`;

If `axis == 0` then the i'th tensor in `output` is the slice `value[i, :, :, :]` 

and each tensor in `output` will have shape `(B, C, D)`. (Note that the
dimension unpacked along is gone, unlike `split`).

If `axis == 1` then the i'th tensor in `output` is the slice `value[:, i, :, :]` 

and each tensor in `output` will have shape `(A, C, D)`.

Etc.

This is the opposite of `pack`.

Arguments: 

value: 1-D or higher, with `axis` dimension size equal to `num`.

Returns The list of tensors unpacked from `value`.

func UnravelIndex 

func UnravelIndex(scope *Scope, indices tf.Output, dims tf.Output) (output tf.Output)

Converts an array of flat indices into a tuple of coordinate arrays.

Example:

``` y = tf.unravel_index(indices=[2, 5, 7], dims=[3, 3]) # 'dims' represent a hypothetical (3, 3) tensor of indices: # [[0, 1, *2*], # [3, 4, *5*], # [6, *7*, 8]] # For each entry from 'indices', this operation returns # its coordinates (marked with '*'), such as # 2 ==> (0, 2) # 5 ==> (1, 2) # 7 ==> (2, 1) y ==> [[0, 1, 2], [2, 2, 1]] ```

@compatibility(numpy) Equivalent to np.unravel_index @end_compatibility

Arguments: 

indices: An 0-D or 1-D `int` Tensor whose elements are indices into the

flattened version of an array of dimensions dims. 

dims: An 1-D `int` Tensor. The shape of the array to use for unraveling

indices.

Returns An 2-D (or 1-D if indices is 0-D) tensor where each row has the same shape as the indices array.

func UnsortedSegmentMax 

func UnsortedSegmentMax(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the maximum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

This operator is similar to `tf.math.unsorted_segment_sum`, Instead of computing the sum over segments, it computes the maximum such that:

\\(output_i = \max_{j...} data[j...]\\) where max is over tuples `j...` such that `segment_ids[j...] == i`.

If the maximum is empty for a given segment ID `i`, it outputs the smallest possible value for the specific numeric type, `output[i] = numeric_limits<T>::lowest()`.

If the given segment ID `i` is negative, then the corresponding value is dropped, and will not be included in the result.

Caution: On CPU, values in `segment_ids` are always validated to be less than `num_segments`, and an error is thrown for out-of-bound indices. On GPU, this does not throw an error for out-of-bound indices. On Gpu, out-of-bound indices result in safe but unspecified behavior, which may include ignoring out-of-bound indices or outputting a tensor with a 0 stored in the first dimension of its shape if `num_segments` is 0.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/UnsortedSegmentMax.png" alt> </div>

For example:

>>> c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) >>> tf.math.unsorted_segment_max(c, tf.constant([0, 1, 0]), num_segments=2).numpy() array([[4, 3, 3, 4], 

[5,  6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A tensor whose shape is a prefix of `data.shape`.

The values must be less than `num_segments`.

Caution: The values are always validated to be in range on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimension which has size `num_segments`.

func UnsortedSegmentMin 

func UnsortedSegmentMin(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the minimum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

This operator is similar to `tf.math.unsorted_segment_sum`, Instead of computing the sum over segments, it computes the minimum such that:

\\(output_i = \min_{j...} data_[j...]\\) where min is over tuples `j...` such that `segment_ids[j...] == i`.

If the minimum is empty for a given segment ID `i`, it outputs the largest possible value for the specific numeric type, `output[i] = numeric_limits<T>::max()`.

For example:

>>> c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) >>> tf.math.unsorted_segment_min(c, tf.constant([0, 1, 0]), num_segments=2).numpy() array([[1, 2, 2, 1], 

[5, 6, 7, 8]], dtype=int32)

If the given segment ID `i` is negative, then the corresponding value is dropped, and will not be included in the result.

Caution: On CPU, values in `segment_ids` are always validated to be less than `num_segments`, and an error is thrown for out-of-bound indices. On GPU, this does not throw an error for out-of-bound indices. On Gpu, out-of-bound indices result in safe but unspecified behavior, which may include ignoring out-of-bound indices or outputting a tensor with a 0 stored in the first dimension of its shape if `num_segments` is 0.

Arguments: 

segment_ids: A tensor whose shape is a prefix of `data.shape`.

The values must be less than `num_segments`.

Caution: The values are always validated to be in range on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimension which has size `num_segments`.

func UnsortedSegmentProd 

func UnsortedSegmentProd(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the product along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

This operator is similar to `tf.math.unsorted_segment_sum`, Instead of computing the sum over segments, it computes the product of all entries belonging to a segment such that:

\\(output_i = \prod_{j...} data[j...]\\) where the product is over tuples `j...` such that `segment_ids[j...] == i`.

For example:

>>> c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) >>> tf.math.unsorted_segment_prod(c, tf.constant([0, 1, 0]), num_segments=2).numpy() array([[4, 6, 6, 4], 

[5, 6, 7, 8]], dtype=int32)

If there is no entry for a given segment ID `i`, it outputs 1.

If the given segment ID `i` is negative, then the corresponding value is dropped, and will not be included in the result. Caution: On CPU, values in `segment_ids` are always validated to be less than `num_segments`, and an error is thrown for out-of-bound indices. On GPU, this does not throw an error for out-of-bound indices. On Gpu, out-of-bound indices result in safe but unspecified behavior, which may include ignoring out-of-bound indices or outputting a tensor with a 0 stored in the first dimension of its shape if `num_segments` is 0.

Arguments: 

segment_ids: A tensor whose shape is a prefix of `data.shape`.

The values must be less than `num_segments`.

Caution: The values are always validated to be in range on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimension which has size `num_segments`.

func UnsortedSegmentSum 

func UnsortedSegmentSum(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)

Computes the sum along segments of a tensor.

Read [the section on segmentation](https://tensorflow.org/api_docs/python/tf/math#Segmentation) for an explanation of segments.

Computes a tensor such that \\(output[i] = \sum_{j...} data[j...]\\) where the sum is over tuples `j...` such that `segment_ids[j...] == i`. Unlike `SegmentSum`, `segment_ids` need not be sorted and need not cover all values in the full range of valid values.

If the sum is empty for a given segment ID `i`, `output[i] = 0`. If the given segment ID `i` is negative, the value is dropped and will not be added to the sum of the segment.

`num_segments` should equal the number of distinct segment IDs.

Caution: On CPU, values in `segment_ids` are always validated to be less than `num_segments`, and an error is thrown for out-of-bound indices. On GPU, this does not throw an error for out-of-bound indices. On Gpu, out-of-bound indices result in safe but unspecified behavior, which may include ignoring out-of-bound indices or outputting a tensor with a 0 stored in the first dimension of its shape if `num_segments` is 0.

<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="https://www.tensorflow.org/images/UnsortedSegmentSum.png" alt> </div>

>>> c = [[1,2,3,4], [5,6,7,8], [4,3,2,1]] >>> tf.math.unsorted_segment_sum(c, [0, 1, 0], num_segments=2).numpy() array([[5, 5, 5, 5], 

[5, 6, 7, 8]], dtype=int32)

Arguments: 

segment_ids: A tensor whose shape is a prefix of `data.shape`.

The values must be less than `num_segments`.

Caution: The values are always validated to be in range on CPU, never validated on GPU.

Returns Has same shape as data, except for the first `segment_ids.rank` dimensions, which are replaced with a single dimension which has size `num_segments`.

func Unstage 

func Unstage(scope *Scope, dtypes []tf.DataType, optional ...UnstageAttr) (values []tf.Output)

Op is similar to a lightweight Dequeue.

The basic functionality is similar to dequeue with many fewer capabilities and options. This Op is optimized for performance.

func UpdateTaskIdAndGlobalCoreArray added in v0.8.2 

func UpdateTaskIdAndGlobalCoreArray(scope *Scope, tpu_task_id_to_shard_id []tf.Output) (o *tf.Operation)

An op to update the task ID and global core array.

This op is to update the task ID and global core array.

Arguments: 

tpu_task_id_to_shard_id: An array of int32 that maps TPU task ID to shard ID.

Returns the created operation.

func UpperBound 

func UpperBound(scope *Scope, sorted_inputs tf.Output, values tf.Output, optional ...UpperBoundAttr) (output tf.Output)

Applies upper_bound(sorted_search_values, values) along each row.

Each set of rows with the same index in (sorted_inputs, values) is treated independently. The resulting row is the equivalent of calling `np.searchsorted(sorted_inputs, values, side='right')`.

The result is not a global index to the entire `Tensor`, but rather just the index in the last dimension.

A 2-D example: 

sorted_sequence = [[0, 3, 9, 9, 10],
                   [1, 2, 3, 4, 5]]
values = [[2, 4, 9],
          [0, 2, 6]]

result = UpperBound(sorted_sequence, values)

result == [[1, 2, 4],
           [0, 2, 5]]

Arguments: 

sorted_inputs: 2-D Tensor where each row is ordered.
values: 2-D Tensor with the same numbers of rows as `sorted_search_values`. Contains

the values that will be searched for in `sorted_search_values`.

Returns A `Tensor` with the same shape as `values`. It contains the last scalar index into the last dimension where values can be inserted without changing the ordered property.

func VarHandleOp 

func VarHandleOp(scope *Scope, dtype tf.DataType, shape tf.Shape, optional ...VarHandleOpAttr) (resource tf.Output)

Creates a handle to a Variable resource.

Arguments: 

dtype: the type of this variable. Must agree with the dtypes

of all ops using this variable. 

shape: The (possibly partially specified) shape of this variable.

func VarIsInitializedOp 

func VarIsInitializedOp(scope *Scope, resource tf.Output) (is_initialized tf.Output)

Checks whether a resource handle-based variable has been initialized.

Arguments: 

resource: the input resource handle.

Returns a scalar boolean which is true if the variable has been initialized.

func VariableShape 

func VariableShape(scope *Scope, input tf.Output, optional ...VariableShapeAttr) (output tf.Output)

Returns the shape of the variable pointed to by `resource`.

This operation returns a 1-D integer tensor representing the shape of `input`.

For example:

``` # 't' is [[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]] shape(t) ==> [2, 2, 3] ```

func Where 

func Where(scope *Scope, condition tf.Output) (index tf.Output)

Returns locations of nonzero / true values in a tensor.

This operation returns the coordinates of true elements in `condition`. The coordinates are returned in a 2-D tensor where the first dimension (rows) represents the number of true elements, and the second dimension (columns) represents the coordinates of the true elements. Keep in mind, the shape of the output tensor can vary depending on how many true values there are in `condition`. Indices are output in row-major order.

For example:

``` # 'input' tensor is [[True, False] # [True, False]] # 'input' has two true values, so output has two coordinates. # 'input' has rank of 2, so coordinates have two indices. where(input) ==> [[0, 0], 

[1, 0]]

# `condition` tensor is [[[True, False] # [True, False]] # [[False, True] # [False, True]] # [[False, False] # [False, True]]] # 'input' has 5 true values, so output has 5 coordinates. # 'input' has rank of 3, so coordinates have three indices. where(input) ==> [[0, 0, 0], 

[0, 1, 0],
[1, 0, 1],
[1, 1, 1],
[2, 1, 1]]

# `condition` tensor is [[[1.5, 0.0] # [-0.5, 0.0]] # [[0.0, 0.25] # [0.0, 0.75]] # [[0.0, 0.0] # [0.0, 0.01]]] # 'input' has 5 nonzero values, so output has 5 coordinates. # 'input' has rank of 3, so coordinates have three indices. where(input) ==> [[0, 0, 0], 

[0, 1, 0],
[1, 0, 1],
[1, 1, 1],
[2, 1, 1]]

# `condition` tensor is [[[1.5 + 0.0j, 0.0 + 0.0j] # [0.0 + 0.5j, 0.0 + 0.0j]] # [[0.0 + 0.0j, 0.25 + 1.5j] # [0.0 + 0.0j, 0.75 + 0.0j]] # [[0.0 + 0.0j, 0.0 + 0.0j] # [0.0 + 0.0j, 0.01 + 0.0j]]] # 'input' has 5 nonzero magnitude values, so output has 5 coordinates. # 'input' has rank of 3, so coordinates have three indices. where(input) ==> [[0, 0, 0], 

[0, 1, 0],
[1, 0, 1],
[1, 1, 1],
[2, 1, 1]]

```

func WholeFileReaderV2 

func WholeFileReaderV2(scope *Scope, optional ...WholeFileReaderV2Attr) (reader_handle tf.Output)

A Reader that outputs the entire contents of a file as a value.

To use, enqueue filenames in a Queue. The output of ReaderRead will be a filename (key) and the contents of that file (value).

Returns The handle to reference the Reader.

func WindowDataset 

func WindowDataset(scope *Scope, input_dataset tf.Output, size tf.Output, shift tf.Output, stride tf.Output, drop_remainder tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...WindowDatasetAttr) (handle tf.Output)
Combines (nests of) input elements into a dataset of (nests of) windows.

A "window" is a finite dataset of flat elements of size `size` (or possibly
fewer if there are not enough input elements to fill the window and
`drop_remainder` evaluates to false).

The `shift` argument determines the number of input elements by which
the window moves on each iteration.  The first element in the `k`th window
will be element

```
1 + (k-1) * shift
```

of the input dataset. In particular, the first element of the first window
will always be the first element of the input dataset.

If the `stride` parameter is greater than 1, then each window will skip
`(stride - 1)` input elements between each element that appears in the
window. Output windows will still contain `size` elements regardless of
the value of `stride`.

The `stride` argument determines the stride of the input elements, and the
`shift` argument determines the shift of the window.

For example, letting `{...}` to represent a Dataset:

- `tf.data.Dataset.range(7).window(2)` produces
  `{{0, 1}, {2, 3}, {4, 5}, {6}}`
- `tf.data.Dataset.range(7).window(3, 2, 1, True)` produces
  `{{0, 1, 2}, {2, 3, 4}, {4, 5, 6}}`
- `tf.data.Dataset.range(7).window(3, 1, 2, True)` produces
  `{{0, 2, 4}, {1, 3, 5}, {2, 4, 6}}`

Note that when the `window` transformation is applied to a dataset of
nested elements, it produces a dataset of nested windows.

For example:

- `tf.data.Dataset.from_tensor_slices((range(4), range(4))).window(2)`
  produces `{({0, 1}, {0, 1}), ({2, 3}, {2, 3})}`
- `tf.data.Dataset.from_tensor_slices({"a": range(4)}).window(2)`
  produces `{{"a": {0, 1}}, {"a": {2, 3}}}`

Arguments: 

size: An integer scalar, representing the number of elements

of the input dataset to combine into a window. Must be positive. 

shift: An integer scalar, representing the number of input elements

by which the window moves in each iteration. Defaults to `size`. Must be positive. 

stride: An integer scalar, representing the stride of the input elements

in the sliding window. Must be positive. The default value of 1 means "retain every input element". 

drop_remainder: A Boolean scalar, representing whether the last window should be

dropped if its size is smaller than `window_size`.

func WorkerHeartbeat 

func WorkerHeartbeat(scope *Scope, request tf.Output) (response tf.Output)

Worker heartbeat op.

Heartbeats may be sent periodically to indicate the coordinator is still active, to retrieve the current worker status and to expedite shutdown when necessary.

Arguments: 

request: A string tensor containing a serialized WorkerHeartbeatRequest

Returns A string tensor containing a serialized WorkerHeartbeatResponse

func WriteAudioSummary 

func WriteAudioSummary(scope *Scope, writer tf.Output, step tf.Output, tag tf.Output, tensor tf.Output, sample_rate tf.Output, optional ...WriteAudioSummaryAttr) (o *tf.Operation)

Writes an audio summary.

Writes encoded audio summary `tensor` at `step` with `tag` using summary `writer`. `sample_rate` is the audio sample rate is Hz.

Returns the created operation.

func WriteFile 

func WriteFile(scope *Scope, filename tf.Output, contents tf.Output) (o *tf.Operation)

Writes `contents` to the file at input `filename`.

Creates the file and recursively creates directory if it does not exist.

Arguments: 

filename: scalar. The name of the file to which we write the contents.
contents: scalar. The content to be written to the output file.

Returns the created operation.

func WriteGraphSummary 

func WriteGraphSummary(scope *Scope, writer tf.Output, step tf.Output, tensor tf.Output) (o *tf.Operation)

Writes a graph summary.

Writes TensorFlow graph `tensor` at `step` using summary `writer`.

Returns the created operation.

func WriteHistogramSummary 

func WriteHistogramSummary(scope *Scope, writer tf.Output, step tf.Output, tag tf.Output, values tf.Output) (o *tf.Operation)

Writes a histogram summary.

Writes histogram `values` at `step` with `tag` using summary `writer`.

Returns the created operation.

func WriteImageSummary 

func WriteImageSummary(scope *Scope, writer tf.Output, step tf.Output, tag tf.Output, tensor tf.Output, bad_color tf.Output, optional ...WriteImageSummaryAttr) (o *tf.Operation)

Writes an image summary.

Writes image `tensor` at `step` with `tag` using summary `writer`. `tensor` is image with shape [height, width, channels].

Returns the created operation.

func WriteRawProtoSummary 

func WriteRawProtoSummary(scope *Scope, writer tf.Output, step tf.Output, tensor tf.Output) (o *tf.Operation)

Writes a serialized proto summary.

Writes `tensor`, a serialized proto at `step` using summary `writer`.

Returns the created operation.

func WriteScalarSummary 

func WriteScalarSummary(scope *Scope, writer tf.Output, step tf.Output, tag tf.Output, value tf.Output) (o *tf.Operation)

Writes a scalar summary.

Writes scalar `value` at `step` with `tag` using summary `writer`.

Returns the created operation.

func WriteSummary 

func WriteSummary(scope *Scope, writer tf.Output, step tf.Output, tensor tf.Output, tag tf.Output, summary_metadata tf.Output) (o *tf.Operation)

Writes a tensor summary.

Writes `tensor` at `step` with `tag` using summary `writer`.

Returns the created operation.

func Xdivy 

func Xdivy(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns 0 if x == 0, and x / y otherwise, elementwise.

func XlaAllReduce 

func XlaAllReduce(scope *Scope, input tf.Output, group_assignment tf.Output, reduce_op string, mode string) (output tf.Output)

Wraps the XLA AllReduce operator 

documented at https://www.tensorflow.org/xla/operation_semantics#allreduce.

Arguments: 

input: Array or a non-empty tuple of arrays to reduce across replicas.
group_assignment: Groups between which the reductions are performed.
reduce_op: Reduction computation.
mode: group mode.

CrossReplica: group_assignment contains replica_id. Each group contains the 

replicas for the current partition.

CrossReplicaAndPartition: group_assignment contains replica_id. Each group 

contains the replicas for all partitions.

func XlaBroadcastHelper 

func XlaBroadcastHelper(scope *Scope, lhs tf.Output, rhs tf.Output, broadcast_dims tf.Output) (lhs_output tf.Output, rhs_output tf.Output)

Helper operator for performing XLA-style broadcasts

Broadcasts `lhs` and `rhs` to the same rank, by adding size 1 dimensions to whichever of `lhs` and `rhs` has the lower rank, using XLA's broadcasting rules for binary operators.

Arguments: 

lhs: the LHS input tensor
rhs: the RHS input tensor
broadcast_dims: an XLA-style broadcast dimension specification

Returns: 

lhs_output: the broadcasted LHS tensor
rhs_output: the broadcasted RHS tensor

func XlaConcatND 

func XlaConcatND(scope *Scope, inputs []tf.Output, num_concats []int64, optional ...XlaConcatNDAttr) (output tf.Output)

Concats input tensor across all dimensions.

An op which merges slices the input tensor based on the given num_splits attribute, strips paddings optionally, and returns the merged tensor without paddings.

This op may be generated via the TPU bridge.

For example, with `input` tensor: ``` [[0, 1], 

[4, 5]]

[[2, 3], 

[6, 7]]

[[8, 9], 

[12, 13]]

[[10, 11], 

[14, 15]]

``` `num_splits`: ``` [2, 2] ``` and `paddings`: ``` [1, 1] ``` the expected `outputs` is: ``` [[0, 1, 2], 

[4, 5, 6],
[8, 9, 10]]

```

Arguments: 

inputs: Input tensor slices in row-major order to merge across all dimensions. All

inputs must have the same shape. 

num_concats: Number of ways to merge per dimension.

Returns Output tensor formed from merging input slices based on num_concats defined.

func XlaConv 

func XlaConv(scope *Scope, lhs tf.Output, rhs tf.Output, window_strides tf.Output, padding tf.Output, lhs_dilation tf.Output, rhs_dilation tf.Output, feature_group_count tf.Output, dimension_numbers string, precision_config string) (output tf.Output)

Wraps the XLA ConvGeneralDilated operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#conv_convolution

.

Arguments: 

lhs: the input tensor
rhs: the kernel tensor
window_strides: the inter-window strides
padding: the padding to apply at the start and end of each input dimensions
lhs_dilation: dilation to apply between input elements
rhs_dilation: dilation to apply between kernel elements
feature_group_count: number of feature groups for grouped convolution.
dimension_numbers: a serialized xla::ConvolutionDimensionNumbers proto.
precision_config: a serialized xla::PrecisionConfig proto.

func XlaConvV2 

func XlaConvV2(scope *Scope, lhs tf.Output, rhs tf.Output, window_strides tf.Output, padding tf.Output, lhs_dilation tf.Output, rhs_dilation tf.Output, feature_group_count tf.Output, dimension_numbers string, precision_config string, preferred_element_type tf.DataType, optional ...XlaConvV2Attr) (output tf.Output)

Wraps the XLA ConvGeneralDilated operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#conv_convolution

.

Arguments: 

lhs: input tensor
rhs: kernel tensor
window_strides: inter-window strides
padding: padding to apply at the start and end of each input dimensions
lhs_dilation: dilation to apply between input elements
rhs_dilation: dilation to apply between kernel elements
feature_group_count: number of feature groups for grouped convolution.
dimension_numbers: serialized xla::ConvolutionDimensionNumbers proto.
precision_config: serialized xla::PrecisionConfig proto.
preferred_element_type: type of the tensor.

func XlaCustomCall 

func XlaCustomCall(scope *Scope, args []tf.Output, target_name string, backend_config string, dtype tf.DataType, shape tf.Shape) (output tf.Output)

Wraps the XLA CustomCall operator 

documented at https://www.tensorflow.org/xla/operation_semantics#customcall.

Arguments: 

args: A list of `Tensor` with possibly different types.
target_name: Name of the function. A call instruction will be emitted which

targets this symbol name. 

backend_config: String, used to encode serialized metadata to the backend.
dtype: Output tensor data type.
shape: Output tensor shape.

func XlaCustomCallV2 added in v0.3.0 

func XlaCustomCallV2(scope *Scope, operands []tf.Output, call_target_name string, backend_config string, has_side_effect bool, result_dtypes []tf.DataType, result_shapes []tf.Shape) (results []tf.Output)

Emits an HLO `CustomCall` operation with multiple outputs.

As opposed to `XlaCustomCall`, this operation supports multiple outputs.

See `CustomCall` specification at 

https://tensorflow.org/xla/operation_semantics#customcall,

and `mhlo.custom_call` specification at 

https://tensorflow.org/mlir/hlo_ops#mhlocustom_call_mlirmhlocustomcallop.

Arguments: 

operands: A sequence of tensors with possibly different types.
call_target_name: Name of the user function. The function signature must conform

to version 3 of the API, see `API_VERSION_STATUS_RETURNING_UNIFIED`. All operands and results assumed to be in the default layout. 

backend_config: A string that encodes a metadata for the backend.
has_side_effect: Indicates whether the custom call has side effects.
result_dtypes: Types of all results.
result_shapes: Shapes of all results.

func XlaDequantize 

func XlaDequantize(scope *Scope, input tf.Output, min_range float32, max_range float32, mode string, transpose_output bool) (output tf.Output)

Takes the packed uint32 input and unpacks the input to uint8 to do

Dequantization on device.

Arguments: 

input: Input tensors whose types is uint32, shape is [d0, ..., dn].
min_range: The minimum scalar value possibly produced for the input.
max_range: The maximum scalar value possibly produced for the input.
mode: String to determine the dequantize mode in {"MIN_COMBINED", "MIN_FIRST", "SCALED"}.
transpose_output: Boolean to determine if output is transposed. transpose_output

is faster when input is large and rank of input is higher than 1.

Returns Output tensors whose types is bfloat16. If transpose_output is true, output shape is [dn * 4, dn-1, ..., d1, d0]. If transpose_output is false, output shape is [d0,..., dn * 4].

func XlaDot 

func XlaDot(scope *Scope, lhs tf.Output, rhs tf.Output, dimension_numbers string, precision_config string) (output tf.Output)

Wraps the XLA DotGeneral operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#dotgeneral

.

Arguments: 

lhs: the LHS tensor
rhs: the RHS tensor
dimension_numbers: a serialized xla::DotDimensionNumbers proto.
precision_config: a serialized xla::PrecisionConfig proto.

func XlaDotV2 

func XlaDotV2(scope *Scope, lhs tf.Output, rhs tf.Output, dimension_numbers string, precision_config string, preferred_element_type tf.DataType) (output tf.Output)

Wraps the XLA DotGeneral operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#dotgeneral

.

Arguments: 

lhs: the LHS tensor
rhs: the RHS tensor
dimension_numbers: a serialized xla::DotDimensionNumbers proto.
precision_config: a serialized xla::PrecisionConfig proto.
preferred_element_type: The type of the tensor.

func XlaDynamicSlice 

func XlaDynamicSlice(scope *Scope, input tf.Output, start_indices tf.Output, size_indices tf.Output) (output tf.Output)

Wraps the XLA DynamicSlice operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#dynamicslice

.

DynamicSlice extracts a sub-array from the input array at dynamic start_indices. The size of the slice in each dimension is passed in size_indices, which specify the end point of exclusive slice intervals in each dimension -- [start, start + size). The shape of start_indices must have rank 1, with dimension size equal to the rank of operand.

Arguments: 

input: A `Tensor` of type T.
start_indices: List of N integers containing the slice size for each

dimension. Each value must be strictly greater than zero, and start + size must be less than or equal to the size of the dimension to avoid implementation defined behavior.

func XlaDynamicUpdateSlice 

func XlaDynamicUpdateSlice(scope *Scope, input tf.Output, update tf.Output, indices tf.Output) (output tf.Output)

Wraps the XLA DynamicUpdateSlice operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#dynamicupdateslice

.

XlaDynamicUpdateSlice generates a result which is the value of the `input` operand, with a slice update overwritten at `indices`. The shape of `update` determines the shape of the sub-array of the result which is updated. The shape of indices must be rank == 1, with dimension size equal to the rank of `input`.

Handling of out-of-bounds slice indices is implementation-defined.

Arguments: 

input: A `Tensor` of type T.
update: A `Tensor` of type T. Same rank as `input`.
indices: A vector of indices into `input`. Must have length equal to the rank of

`input`.

Returns A `Tensor` of type T.

func XlaEinsum 

func XlaEinsum(scope *Scope, a tf.Output, b tf.Output, equation string) (product tf.Output)

An op which supports basic einsum op with 2 inputs and 1 output.

This op has better TPU performance since it doesn't have explicitly reshape and transpose operations as tf.einsum does.

func XlaGather 

func XlaGather(scope *Scope, operand tf.Output, start_indices tf.Output, slice_sizes tf.Output, dimension_numbers string, indices_are_sorted bool) (output tf.Output)

Wraps the XLA Gather operator documented at 

https://www.tensorflow.org/xla/operation_semantics#gather

Arguments: 

operand: The array we're gathering from.
start_indices: Array containing the starting indices of the slices we gather.
slice_sizes: slice_sizes[i] is the bounds for the slice on dimension i.
dimension_numbers: A serialized xla::GatherDimensionNumbers proto.
indices_are_sorted: Boolean indicating if the indices are sorted.

func XlaKeyValueSort 

func XlaKeyValueSort(scope *Scope, keys tf.Output, values tf.Output) (sorted_keys tf.Output, sorted_values tf.Output)

Wraps the XLA Sort operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#sort

.

Sorts a tensor. Currently only sorts in ascending order are supported.

Arguments: 

keys: A `Tensor` of type K.
values: A `Tensor` of type V.

Returns: 

sorted_keys: A `Tensor` of type K.
sorted_values: A `Tensor` of type V.

func XlaOptimizationBarrier 

func XlaOptimizationBarrier(scope *Scope, input []tf.Output) (output []tf.Output)

Wraps the XLA OptimizationBarrier operator.

Documented at https://www.tensorflow.org/xla/operation_semantics#optimizationbarrier.

Arguments: 

input: A Tuple of Arrays of any type.

func XlaPad 

func XlaPad(scope *Scope, input tf.Output, padding_value tf.Output, padding_low tf.Output, padding_high tf.Output, padding_interior tf.Output) (output tf.Output)

Wraps the XLA Pad operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#pad

.

Arguments: 

input: A `Tensor` of type T.
padding_value: A scalar `Tensor` of type T.
padding_low: the padding to apply at the start of each input dimensions. Must

be a compile-time constant 1D tensor of length equal to rank of input. 

padding_high: the padding to apply at the end of each input dimension. Must

be a compile-time constant 1D tensor of length equal to rank of input. 

padding_interior: the padding to apply between each input element. Must

be a compile-time constant 1D tensor of length equal to rank of input, containing only non-negative values.

Returns A `Tensor` of type T.

func XlaRecv 

func XlaRecv(scope *Scope, dtype tf.DataType, tensor_name string, shape tf.Shape) (tensor tf.Output)

Receives the named tensor from another XLA computation. Wraps the XLA Recv

operator documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#recv .

Arguments: 

dtype: The type of the tensor.
tensor_name: A string key that identifies the channel.
shape: The shape of the tensor.

Returns The tensor to receive.

func XlaRecvFromHost 

func XlaRecvFromHost(scope *Scope, Toutput tf.DataType, shape tf.Shape, key string) (output tf.Output)

An op to receive a tensor from the host.

output: the tensor that will be received from the host. Toutput: element type for output. shape: shape for output. key: A unique identifier for this region used to match up host transfers.

func XlaRecvTPUEmbeddingActivations added in v0.2.0 

func XlaRecvTPUEmbeddingActivations(scope *Scope, deduplication_data tf.Output, num_tables int64, config string) (outputs []tf.Output)

An op that receives embedding activations on the TPU.

The TPU system performs the embedding lookups and aggregations. The results of these aggregations are visible to the Tensorflow Graph as the outputs of a XlaRecvTPUEmbeddingActivations Op. This op returns a list containing one Tensor of activations per table specified in the model.

Arguments: 

deduplication_data: A Tensor with type=DT_VARIANT containing the deduplication

data. The tensor is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation. 

num_tables: The number of output activation tensors. If feature descriptor is

present in the tpu embedding config, it is equal to the number of features otherwise equal to number of embedding tables in the model. 

config: Serialized TPUEmbeddingConfiguration proto.

Returns A TensorList of embedding activations containing one Tensor per embedding table in the model.

func XlaRecvTPUEmbeddingActivationsV2 added in v0.8.2 

func XlaRecvTPUEmbeddingActivationsV2(scope *Scope, deduplication_data tf.Output, num_tables int64, config string, embedding_partitions string, hbm_buffers_config string, tpu_topology string) (outputs []tf.Output)

An op that receives embedding activations on the TPU.

The TPU system performs the embedding lookups and aggregations. The results of these aggregations are visible to the Tensorflow Graph as the outputs of a XlaRecvTPUEmbeddingActivations Op. This op returns a list containing one Tensor of activations per table specified in the model.

Arguments: 

deduplication_data: A Tensor with type=DT_VARIANT containing the deduplication

data. The tensor is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation. 

num_tables: The number of output activation tensors. If feature descriptor is

present in the tpu embedding config, it is equal to the number of features otherwise equal to number of embedding tables in the model. 

config: Serialized TPUEmbeddingConfiguration proto.
embedding_partitions: Serialized EmbeddingPartitionsProto proto.
hbm_buffers_config: Serialized HbmBuffersConfig proto.
tpu_topology: Serialized TpuTopologyArgsProto proto.

Returns A TensorList of embedding activations containing one Tensor per embedding table in the model.

func XlaRecvTPUEmbeddingDeduplicationData added in v0.2.0 

func XlaRecvTPUEmbeddingDeduplicationData(scope *Scope, config string) (output tf.Output)

Receives deduplication data (indices and weights) from the embedding core.

The deduplication data is a Tensor with type=DT_VARIANT. The tensor itself is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.

func XlaRecvTPUEmbeddingDeduplicationDataV2 added in v0.8.2 

func XlaRecvTPUEmbeddingDeduplicationDataV2(scope *Scope, config string, embedding_partitions string, hbm_buffers_config string, tpu_topology string) (output tf.Output)

Receives deduplication data (indices and weights) from the embedding core.

The deduplication data is a Tensor with type=DT_VARIANT. The tensor itself is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation.

Arguments: 

config: Serialized TPUEmbeddingConfiguration proto.
embedding_partitions: Serialized EmbeddingPartitionsProto proto.
hbm_buffers_config: Serialized HbmBuffersConfig proto.
tpu_topology: Serialized TpuTopologyArgsProto proto.

func XlaReducePrecision added in v0.3.0 

func XlaReducePrecision(scope *Scope, operand tf.Output, exponent_bits int64, mantissa_bits int64) (output tf.Output)

Wraps the XLA ReducePrecision operator 

documented at https://www.tensorflow.org/xla/operation_semantics#reduceprecision.

Arguments: 

operand: array of floating-point type.
exponent_bits: number of exponent bits in lower-precision format
mantissa_bits: number of mantissa bits in lower-precision format

func XlaReduceScatter 

func XlaReduceScatter(scope *Scope, input tf.Output, group_assignment tf.Output, scatter_dimension tf.Output, reduce_op string) (output tf.Output)

Wraps the XLA ReduceScatter operator 

documented at https://www.tensorflow.org/xla/operation_semantics#reducescatter.

Arguments: 

input: Array or a non-empty tuple of arrays to reduce across replicas.
group_assignment: Groups between which the reductions are performed.
scatter_dimension: Dimension to scatter.
reduce_op: Reduction computation.

func XlaRemoveDynamicDimensionSize 

func XlaRemoveDynamicDimensionSize(scope *Scope, input tf.Output, dim_index tf.Output) (output tf.Output)

Inverse of XlaSetDynamicDimensionSize.

Make an xla bounded dynamic dimension into a static dimension. The bound of the size of dimension `dim_index` becomes the static dimension size.

func XlaReplicaId 

func XlaReplicaId(scope *Scope) (id tf.Output)

Replica ID.

func XlaRngBitGenerator 

func XlaRngBitGenerator(scope *Scope, algorithm tf.Output, initial_state tf.Output, shape tf.Output, optional ...XlaRngBitGeneratorAttr) (output_key tf.Output, output tf.Output)

Stateless PRNG bit generator.

Wraps the XLA RngBitGenerator operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#rngbitgenerator.

Arguments: 

algorithm: The PRNG algorithm to use, one of

tf.random.Algorithm.{PHILOX, THREEFRY, AUTO_SELECT}. 

initial_state: Initial state for the PRNG algorithm. For THREEFRY, it should be

a u64[2] and for PHILOX a u64[3]. 

shape: The output shape of the generated data.

func XlaSelfAdjointEig 

func XlaSelfAdjointEig(scope *Scope, a tf.Output, lower bool, max_iter int64, epsilon float32) (w tf.Output, v tf.Output)

Computes the eigen decomposition of a batch of self-adjoint matrices

(Note: Only real inputs are supported).

Computes the eigenvalues and eigenvectors of the innermost N-by-N matrices in tensor such that tensor[...,:,:] * v[..., :,i] = e[..., i] * v[...,:,i], for i=0...N-1.

Arguments: 

a: the input tensor.
lower: a boolean specifies whether the calculation is done with the lower

triangular part or the upper triangular part. 

max_iter: maximum number of sweep update, i.e., the whole lower triangular

part or upper triangular part based on parameter lower. Heuristically, it has been argued that approximately logN sweeps are needed in practice (Ref: Golub & van Loan "Matrix Computation"). 

epsilon: the tolerance ratio.

Returns: 

w: The eigenvalues in ascending order, each repeated according to its

multiplicity. 

v: The column v[..., :, i] is the normalized eigenvector corresponding to the

eigenvalue w[..., i].

func XlaSend 

func XlaSend(scope *Scope, tensor tf.Output, tensor_name string) (o *tf.Operation)

Sends the named tensor to another XLA computation. Wraps the XLA Send operator

documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#send .

Arguments: 

tensor: The tensor to send.
tensor_name: A string key that identifies the channel.

Returns the created operation.

func XlaSendTPUEmbeddingGradients added in v0.2.0 

func XlaSendTPUEmbeddingGradients(scope *Scope, gradients []tf.Output, learning_rates []tf.Output, deduplication_data tf.Output, config string) (o *tf.Operation)

An op that performs gradient updates of embedding tables.

The gradients argument is a TensorList having the same length and shapes as the return value of XlaRecvTPUEmbeddingActivations, but contains gradients of the model's loss with respect to the embedding activations. The embedding tables are updated from these gradients via the optimizer specified in the TPUEmbeddingConfiguration proto given to tpu.initialize_system.

Arguments: 

gradients: A TensorList of gradients with which to update embedding tables.
learning_rates: A TensorList of learning rates used for updating the embedding

tables via the optimizer. The length of the TensorList must be equal to the number of dynamic learning rate tags specified in the TPUEmbeddingConfiguration proto. 

deduplication_data: A Tensor with type=DT_VARIANT containing the deduplication

data. The tensor is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation. 

config: Serialized TPUEmbeddingConfiguration proto.

Returns the created operation.

func XlaSendTPUEmbeddingGradientsV2 added in v0.8.2 

func XlaSendTPUEmbeddingGradientsV2(scope *Scope, gradients []tf.Output, learning_rates []tf.Output, deduplication_data tf.Output, config string, embedding_partitions string, hbm_buffers_config string, tpu_topology string) (o *tf.Operation)

An op that performs gradient updates of embedding tables.

The gradients argument is a TensorList having the same length and shapes as the return value of XlaRecvTPUEmbeddingActivations, but contains gradients of the model's loss with respect to the embedding activations. The embedding tables are updated from these gradients via the optimizer specified in the TPUEmbeddingConfiguration proto given to tpu.initialize_system.

Arguments: 

gradients: A TensorList of gradients with which to update embedding tables.
learning_rates: A TensorList of learning rates used for updating the embedding

tables via the optimizer. The length of the TensorList must be equal to the number of dynamic learning rate tags specified in the TPUEmbeddingConfiguration proto. 

deduplication_data: A Tensor with type=DT_VARIANT containing the deduplication

data. The tensor is an XLA nested tuple containing N elements (where N is the ratio of the number of embedding to tensor cores per TPU chip). Each element of the nested tuple is a tuple of rank 1 tensors. Each tensor either contains indices (DT_UINT32) for embedding lookup on the TensorCore or weights (DT_FLOAT) to apply to the output of the embedding lookup operation. 

config: Serialized TPUEmbeddingConfiguration proto.
embedding_partitions: Serialized EmbeddingPartitionsProto proto.
hbm_buffers_config: Serialized HbmBuffersConfig proto.
tpu_topology: Serialized TpuTopologyArgsProto proto.

Returns the created operation.

func XlaSendToHost 

func XlaSendToHost(scope *Scope, input tf.Output, key string) (o *tf.Operation)

An op to send a tensor to the host.

input: the tensor that will be sent to the host. Tinput: element type for input. key: A unique identifier for this region used to match up host transfers.

Returns the created operation.

func XlaSetBound 

func XlaSetBound(scope *Scope, input tf.Output, bound tf.Output) (output tf.Output)

Set a bound for the given input value as a hint to Xla compiler, 

returns the same value.

func XlaSetDynamicDimensionSize 

func XlaSetDynamicDimensionSize(scope *Scope, input tf.Output, dim_index tf.Output, size tf.Output) (output tf.Output)

Make a static dimension into a xla bounded dynamic dimension. 

The current static dimension size will become the bound and the second
operand becomes the dynamic size of the dimension.

func XlaSharding 

func XlaSharding(scope *Scope, input tf.Output, optional ...XlaShardingAttr) (output tf.Output)

An op which shards the input based on the given sharding attribute. It can

selectively annotate a subset of tensor dimensions by skipping unspecified_dims, and the sharding annotation should be replicated in those dims.

func XlaSort 

func XlaSort(scope *Scope, input tf.Output) (output tf.Output)

Wraps the XLA Sort operator, documented at 

https://www.tensorflow.org/performance/xla/operation_semantics#sort

.

Sorts a tensor. Currently only sorts in ascending order are supported.

Arguments: 

input: A `Tensor` of type T.

Returns A `Tensor` of type T.

func XlaSplitND 

func XlaSplitND(scope *Scope, input tf.Output, N int64, num_splits []int64, optional ...XlaSplitNDAttr) (outputs []tf.Output)

Splits input tensor across all dimensions.

An op which slices the input tensor based on the given num_splits attribute, pads slices optionally, and returned the slices. Slices are returned in row-major order.

This op may be generated via the TPU bridge.

For example, with `input` tensor: ``` [[0, 1, 2], 

[3, 4, 5],
[6, 7, 8]]

``` `num_splits`: ``` [2, 2] ``` and `paddings`: ``` [1, 1] ``` the expected `outputs` is: ``` [[0, 1], 

[3, 4]]

[[2, 0], 

[5, 0]]

[[6, 7], 

[0, 0]]

[[8, 0], 

[0, 0]]

```

Arguments: 

input: Input tensor to split across all dimensions.

num_splits: Number of ways to split per dimension. Shape dimensions must be evenly

divisible.

Returns Output slices based on input and num_splits defined, in row-major order.

func XlaSpmdFullToShardShape 

func XlaSpmdFullToShardShape(scope *Scope, input tf.Output, manual_sharding string, optional ...XlaSpmdFullToShardShapeAttr) (output tf.Output)

An op used by XLA SPMD partitioner to switch from automatic partitioning to

manual partitioning. It annotates the input (full-shape, to be automatically partitioned) with the same sharding used by manual partitioning, and outputs a shard-shaped tensor to be consumed by later manually-partitioned ops. If the shape is not evenly partitionable, the padding region will be masked with 0s. The conversion can happen partially in subgroups, by specifying the dim attribute, where only that dim will be converted.

func XlaSpmdShardToFullShape 

func XlaSpmdShardToFullShape(scope *Scope, input tf.Output, manual_sharding string, full_shape tf.Shape, optional ...XlaSpmdShardToFullShapeAttr) (output tf.Output)

An op used by XLA SPMD partitioner to switch from manual partitioning to

automatic partitioning. It converts the shard-shaped, manually partitioned input into full-shaped tensor to be partitioned automatically with the same sharding used by manual partitioning. The conversion can happen partially in subgroups, by specifying the dim attribute, where only that dim will be converted.

func XlaSvd 

func XlaSvd(scope *Scope, a tf.Output, max_iter int64, epsilon float32, precision_config string) (s tf.Output, u tf.Output, v tf.Output)

Computes the eigen decomposition of a batch of self-adjoint matrices

(Note: Only real inputs are supported).

Computes the eigenvalues and eigenvectors of the innermost M-by-N matrices in tensor such that tensor[...,:,:] = u[..., :, :] * Diag(s[..., :]) * Transpose(v[...,:,:]).

Arguments: 

a: the input tensor.
max_iter: maximum number of sweep update, i.e., the whole lower triangular

part or upper triangular part based on parameter lower. Heuristically, it has been argued that approximately log(min (M, N)) sweeps are needed in practice (Ref: Golub & van Loan "Matrix Computation"). 

epsilon: the tolerance ratio.
precision_config: a serialized xla::PrecisionConfig proto.

Returns: 

s: Singular values. The values are sorted in reverse order of magnitude, so

s[..., 0] is the largest value, s[..., 1] is the second largest, etc. 

u: Left singular vectors.
v: Right singular vectors.

func Xlog1py 

func Xlog1py(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns 0 if x == 0, and x * log1p(y) otherwise, elementwise.

func Xlogy 

func Xlogy(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)

Returns 0 if x == 0, and x * log(y) otherwise, elementwise.

func ZerosLike 

func ZerosLike(scope *Scope, x tf.Output) (y tf.Output)

Returns a tensor of zeros with the same shape and type as x.

Arguments: 

x: a tensor of type T.

Returns a tensor of the same shape and type as x but filled with zeros.

func Zeta 

func Zeta(scope *Scope, x tf.Output, q tf.Output) (z tf.Output)

Compute the Hurwitz zeta function \\(\zeta(x, q)\\).

The Hurwitz zeta function is defined as:

\\(\zeta(x, q) = \sum_{n=0}^{\infty} (q + n)^{-x}\\)

func ZipDataset 

func ZipDataset(scope *Scope, input_datasets []tf.Output, output_types []tf.DataType, output_shapes []tf.Shape, optional ...ZipDatasetAttr) (handle tf.Output)

Creates a dataset that zips together `input_datasets`.

The elements of the resulting dataset are created by zipping corresponding elements from each of the input datasets.

The size of the resulting dataset will match the size of the smallest input dataset, and no error will be raised if input datasets have different sizes.

Arguments: 

input_datasets: List of `N` variant Tensors representing datasets to be zipped together.

Types

type AbortAttr 

type AbortAttr func(optionalAttr)

AbortAttr is an optional argument to Abort.

func AbortErrorMsg 

func AbortErrorMsg(value string) AbortAttr

AbortErrorMsg sets the optional error_msg attribute to value.

value: A string which is the message associated with the exception. If not specified, defaults to ""

func AbortExitWithoutError 

func AbortExitWithoutError(value bool) AbortAttr

AbortExitWithoutError sets the optional exit_without_error attribute to value. If not specified, defaults to false

type AddManySparseToTensorsMapAttr 

type AddManySparseToTensorsMapAttr func(optionalAttr)

AddManySparseToTensorsMapAttr is an optional argument to AddManySparseToTensorsMap.

func AddManySparseToTensorsMapContainer 

func AddManySparseToTensorsMapContainer(value string) AddManySparseToTensorsMapAttr

AddManySparseToTensorsMapContainer sets the optional container attribute to value.

value: The container name for the `SparseTensorsMap` created by this op. If not specified, defaults to ""

func AddManySparseToTensorsMapSharedName 

func AddManySparseToTensorsMapSharedName(value string) AddManySparseToTensorsMapAttr

AddManySparseToTensorsMapSharedName sets the optional shared_name attribute to value.

value: The shared name for the `SparseTensorsMap` created by this op. If blank, the new Operation's unique name is used. If not specified, defaults to ""

type AddSparseToTensorsMapAttr 

type AddSparseToTensorsMapAttr func(optionalAttr)

AddSparseToTensorsMapAttr is an optional argument to AddSparseToTensorsMap.

func AddSparseToTensorsMapContainer 

func AddSparseToTensorsMapContainer(value string) AddSparseToTensorsMapAttr

AddSparseToTensorsMapContainer sets the optional container attribute to value.

value: The container name for the `SparseTensorsMap` created by this op. If not specified, defaults to ""

func AddSparseToTensorsMapSharedName 

func AddSparseToTensorsMapSharedName(value string) AddSparseToTensorsMapAttr

AddSparseToTensorsMapSharedName sets the optional shared_name attribute to value.

value: The shared name for the `SparseTensorsMap` created by this op. If blank, the new Operation's unique name is used. If not specified, defaults to ""

type AllAttr 

type AllAttr func(optionalAttr)

AllAttr is an optional argument to All.

func AllKeepDims 

func AllKeepDims(value bool) AllAttr

AllKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type AllCandidateSamplerAttr 

type AllCandidateSamplerAttr func(optionalAttr)

AllCandidateSamplerAttr is an optional argument to AllCandidateSampler.

func AllCandidateSamplerSeed 

func AllCandidateSamplerSeed(value int64) AllCandidateSamplerAttr

AllCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func AllCandidateSamplerSeed2 

func AllCandidateSamplerSeed2(value int64) AllCandidateSamplerAttr

AllCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type AngleAttr 

type AngleAttr func(optionalAttr)

AngleAttr is an optional argument to Angle.

func AngleTout 

func AngleTout(value tf.DataType) AngleAttr

AngleTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type AnonymousMutableDenseHashTableAttr 

type AnonymousMutableDenseHashTableAttr func(optionalAttr)

AnonymousMutableDenseHashTableAttr is an optional argument to AnonymousMutableDenseHashTable.

func AnonymousMutableDenseHashTableInitialNumBuckets 

func AnonymousMutableDenseHashTableInitialNumBuckets(value int64) AnonymousMutableDenseHashTableAttr

AnonymousMutableDenseHashTableInitialNumBuckets sets the optional initial_num_buckets attribute to value.

value: The initial number of hash table buckets. Must be a power to 2. If not specified, defaults to 131072

func AnonymousMutableDenseHashTableMaxLoadFactor 

func AnonymousMutableDenseHashTableMaxLoadFactor(value float32) AnonymousMutableDenseHashTableAttr

AnonymousMutableDenseHashTableMaxLoadFactor sets the optional max_load_factor attribute to value.

value: The maximum ratio between number of entries and number of buckets before growing the table. Must be between 0 and 1. If not specified, defaults to 0.8

func AnonymousMutableDenseHashTableValueShape 

func AnonymousMutableDenseHashTableValueShape(value tf.Shape) AnonymousMutableDenseHashTableAttr

AnonymousMutableDenseHashTableValueShape sets the optional value_shape attribute to value.

value: The shape of each value. If not specified, defaults to {}

type AnonymousMutableHashTableOfTensorsAttr 

type AnonymousMutableHashTableOfTensorsAttr func(optionalAttr)

AnonymousMutableHashTableOfTensorsAttr is an optional argument to AnonymousMutableHashTableOfTensors.

func AnonymousMutableHashTableOfTensorsValueShape 

func AnonymousMutableHashTableOfTensorsValueShape(value tf.Shape) AnonymousMutableHashTableOfTensorsAttr

AnonymousMutableHashTableOfTensorsValueShape sets the optional value_shape attribute to value. If not specified, defaults to {}

type AnyAttr 

type AnyAttr func(optionalAttr)

AnyAttr is an optional argument to Any.

func AnyKeepDims 

func AnyKeepDims(value bool) AnyAttr

AnyKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type ApproxTopKAttr added in v0.2.0 

type ApproxTopKAttr func(optionalAttr)

ApproxTopKAttr is an optional argument to ApproxTopK.

func ApproxTopKAggregateToTopk added in v0.2.0 

func ApproxTopKAggregateToTopk(value bool) ApproxTopKAttr

ApproxTopKAggregateToTopk sets the optional aggregate_to_topk attribute to value.

value: When true, aggregates approximate results to top-k. When false, returns the approximate results. The number of the approximate results is implementation defined and is greater equals to the specified `k`. If not specified, defaults to true

func ApproxTopKIsMaxK added in v0.2.0 

func ApproxTopKIsMaxK(value bool) ApproxTopKAttr

ApproxTopKIsMaxK sets the optional is_max_k attribute to value.

value: When true, computes max-k; otherwise computes min-k. If not specified, defaults to true

func ApproxTopKRecallTarget added in v0.2.0 

func ApproxTopKRecallTarget(value float32) ApproxTopKAttr

ApproxTopKRecallTarget sets the optional recall_target attribute to value.

value: Recall target for the approximation. Range in (0,1] If not specified, defaults to 0.95

func ApproxTopKReductionDimension added in v0.2.0 

func ApproxTopKReductionDimension(value int64) ApproxTopKAttr

ApproxTopKReductionDimension sets the optional reduction_dimension attribute to value.

value: Integer dimension along which to search. Default: -1. If not specified, defaults to -1

func ApproxTopKReductionInputSizeOverride added in v0.2.0 

func ApproxTopKReductionInputSizeOverride(value int64) ApproxTopKAttr

ApproxTopKReductionInputSizeOverride sets the optional reduction_input_size_override attribute to value.

value: When set to a positive value, it overrides the size determined by `input[reduction_dim]` for evaluating the recall. This option is useful when the given `input` is only a subset of the overall computation in SPMD or distributed pipelines, where the true input size cannot be deferred by the `input` shape. If not specified, defaults to -1

type ApproximateEqualAttr 

type ApproximateEqualAttr func(optionalAttr)

ApproximateEqualAttr is an optional argument to ApproximateEqual.

func ApproximateEqualTolerance 

func ApproximateEqualTolerance(value float32) ApproximateEqualAttr

ApproximateEqualTolerance sets the optional tolerance attribute to value. If not specified, defaults to 1e-05

type ArgMaxAttr 

type ArgMaxAttr func(optionalAttr)

ArgMaxAttr is an optional argument to ArgMax.

func ArgMaxOutputType 

func ArgMaxOutputType(value tf.DataType) ArgMaxAttr

ArgMaxOutputType sets the optional output_type attribute to value. If not specified, defaults to DT_INT64

type ArgMinAttr 

type ArgMinAttr func(optionalAttr)

ArgMinAttr is an optional argument to ArgMin.

func ArgMinOutputType 

func ArgMinOutputType(value tf.DataType) ArgMinAttr

ArgMinOutputType sets the optional output_type attribute to value. If not specified, defaults to DT_INT64

type AsStringAttr 

type AsStringAttr func(optionalAttr)

AsStringAttr is an optional argument to AsString.

func AsStringFill 

func AsStringFill(value string) AsStringAttr

AsStringFill sets the optional fill attribute to value.

value: The value to pad if width > -1. If empty, pads with spaces. Another typical value is '0'. String cannot be longer than 1 character. If not specified, defaults to ""

func AsStringPrecision 

func AsStringPrecision(value int64) AsStringAttr

AsStringPrecision sets the optional precision attribute to value.

value: The post-decimal precision to use for floating point numbers. Only used if precision > -1. If not specified, defaults to -1

func AsStringScientific 

func AsStringScientific(value bool) AsStringAttr

AsStringScientific sets the optional scientific attribute to value.

value: Use scientific notation for floating point numbers. If not specified, defaults to false

func AsStringShortest 

func AsStringShortest(value bool) AsStringAttr

AsStringShortest sets the optional shortest attribute to value.

value: Use shortest representation (either scientific or standard) for floating point numbers. If not specified, defaults to false

func AsStringWidth 

func AsStringWidth(value int64) AsStringAttr

AsStringWidth sets the optional width attribute to value.

value: Pad pre-decimal numbers to this width. Applies to both floating point and integer numbers. Only used if width > -1. If not specified, defaults to -1

type AssertAttr 

type AssertAttr func(optionalAttr)

AssertAttr is an optional argument to Assert.

func AssertSummarize 

func AssertSummarize(value int64) AssertAttr

AssertSummarize sets the optional summarize attribute to value.

value: Print this many entries of each tensor. If not specified, defaults to 3

type AssignVariableOpAttr 

type AssignVariableOpAttr func(optionalAttr)

AssignVariableOpAttr is an optional argument to AssignVariableOp.

func AssignVariableOpValidateShape 

func AssignVariableOpValidateShape(value bool) AssignVariableOpAttr

AssignVariableOpValidateShape sets the optional validate_shape attribute to value. If not specified, defaults to false

type AssignVariableXlaConcatNDAttr 

type AssignVariableXlaConcatNDAttr func(optionalAttr)

AssignVariableXlaConcatNDAttr is an optional argument to AssignVariableXlaConcatND.

func AssignVariableXlaConcatNDPaddings 

func AssignVariableXlaConcatNDPaddings(value []int64) AssignVariableXlaConcatNDAttr

AssignVariableXlaConcatNDPaddings sets the optional paddings attribute to value.

value: Optional list of right paddings per dimension to strip from the final merged tensor. These paddings must not exceed the dimension size of the merged result prior to stripping paddings. If not specified, defaults to {}

type AudioSpectrogramAttr 

type AudioSpectrogramAttr func(optionalAttr)

AudioSpectrogramAttr is an optional argument to AudioSpectrogram.

func AudioSpectrogramMagnitudeSquared 

func AudioSpectrogramMagnitudeSquared(value bool) AudioSpectrogramAttr

AudioSpectrogramMagnitudeSquared sets the optional magnitude_squared attribute to value.

value: Whether to return the squared magnitude or just the magnitude. Using squared magnitude can avoid extra calculations. If not specified, defaults to false

type AudioSummaryAttr 

type AudioSummaryAttr func(optionalAttr)

AudioSummaryAttr is an optional argument to AudioSummary.

func AudioSummaryMaxOutputs 

func AudioSummaryMaxOutputs(value int64) AudioSummaryAttr

AudioSummaryMaxOutputs sets the optional max_outputs attribute to value.

value: Max number of batch elements to generate audio for. If not specified, defaults to 3

REQUIRES: value >= 1

type AudioSummaryV2Attr 

type AudioSummaryV2Attr func(optionalAttr)

AudioSummaryV2Attr is an optional argument to AudioSummaryV2.

func AudioSummaryV2MaxOutputs 

func AudioSummaryV2MaxOutputs(value int64) AudioSummaryV2Attr

AudioSummaryV2MaxOutputs sets the optional max_outputs attribute to value.

value: Max number of batch elements to generate audio for. If not specified, defaults to 3

REQUIRES: value >= 1

type AutoShardDatasetAttr 

type AutoShardDatasetAttr func(optionalAttr)

AutoShardDatasetAttr is an optional argument to AutoShardDataset.

func AutoShardDatasetAutoShardPolicy 

func AutoShardDatasetAutoShardPolicy(value int64) AutoShardDatasetAttr

AutoShardDatasetAutoShardPolicy sets the optional auto_shard_policy attribute to value. If not specified, defaults to 0

func AutoShardDatasetNumReplicas 

func AutoShardDatasetNumReplicas(value int64) AutoShardDatasetAttr

AutoShardDatasetNumReplicas sets the optional num_replicas attribute to value. If not specified, defaults to 0

type AvgPool3DAttr 

type AvgPool3DAttr func(optionalAttr)

AvgPool3DAttr is an optional argument to AvgPool3D.

func AvgPool3DDataFormat 

func AvgPool3DDataFormat(value string) AvgPool3DAttr

AvgPool3DDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

type AvgPool3DGradAttr 

type AvgPool3DGradAttr func(optionalAttr)

AvgPool3DGradAttr is an optional argument to AvgPool3DGrad.

func AvgPool3DGradDataFormat 

func AvgPool3DGradDataFormat(value string) AvgPool3DGradAttr

AvgPool3DGradDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

type AvgPoolAttr 

type AvgPoolAttr func(optionalAttr)

AvgPoolAttr is an optional argument to AvgPool.

func AvgPoolDataFormat 

func AvgPoolDataFormat(value string) AvgPoolAttr

AvgPoolDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type AvgPoolGradAttr 

type AvgPoolGradAttr func(optionalAttr)

AvgPoolGradAttr is an optional argument to AvgPoolGrad.

func AvgPoolGradDataFormat 

func AvgPoolGradDataFormat(value string) AvgPoolGradAttr

AvgPoolGradDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type BatchAttr 

type BatchAttr func(optionalAttr)

BatchAttr is an optional argument to Batch.

func BatchAllowedBatchSizes 

func BatchAllowedBatchSizes(value []int64) BatchAttr

BatchAllowedBatchSizes sets the optional allowed_batch_sizes attribute to value. If not specified, defaults to {}

func BatchBatchingQueue 

func BatchBatchingQueue(value string) BatchAttr

BatchBatchingQueue sets the optional batching_queue attribute to value. If not specified, defaults to ""

func BatchContainer 

func BatchContainer(value string) BatchAttr

BatchContainer sets the optional container attribute to value. If not specified, defaults to ""

func BatchMaxEnqueuedBatches 

func BatchMaxEnqueuedBatches(value int64) BatchAttr

BatchMaxEnqueuedBatches sets the optional max_enqueued_batches attribute to value. If not specified, defaults to 10

func BatchSharedName 

func BatchSharedName(value string) BatchAttr

BatchSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BatchDatasetAttr 

type BatchDatasetAttr func(optionalAttr)

BatchDatasetAttr is an optional argument to BatchDataset.

func BatchDatasetMetadata 

func BatchDatasetMetadata(value string) BatchDatasetAttr

BatchDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type BatchDatasetV2Attr 

type BatchDatasetV2Attr func(optionalAttr)

BatchDatasetV2Attr is an optional argument to BatchDatasetV2.

func BatchDatasetV2Metadata 

func BatchDatasetV2Metadata(value string) BatchDatasetV2Attr

BatchDatasetV2Metadata sets the optional metadata attribute to value. If not specified, defaults to ""

func BatchDatasetV2ParallelCopy 

func BatchDatasetV2ParallelCopy(value bool) BatchDatasetV2Attr

BatchDatasetV2ParallelCopy sets the optional parallel_copy attribute to value. If not specified, defaults to false

type BatchMatMulAttr 

type BatchMatMulAttr func(optionalAttr)

BatchMatMulAttr is an optional argument to BatchMatMul.

func BatchMatMulAdjX 

func BatchMatMulAdjX(value bool) BatchMatMulAttr

BatchMatMulAdjX sets the optional adj_x attribute to value.

value: If `True`, adjoint the slices of `x`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulAdjY 

func BatchMatMulAdjY(value bool) BatchMatMulAttr

BatchMatMulAdjY sets the optional adj_y attribute to value.

value: If `True`, adjoint the slices of `y`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulGradX added in v0.8.0 

func BatchMatMulGradX(value bool) BatchMatMulAttr

BatchMatMulGradX sets the optional grad_x attribute to value. If not specified, defaults to false

func BatchMatMulGradY added in v0.8.0 

func BatchMatMulGradY(value bool) BatchMatMulAttr

BatchMatMulGradY sets the optional grad_y attribute to value. If not specified, defaults to false

type BatchMatMulV2Attr 

type BatchMatMulV2Attr func(optionalAttr)

BatchMatMulV2Attr is an optional argument to BatchMatMulV2.

func BatchMatMulV2AdjX 

func BatchMatMulV2AdjX(value bool) BatchMatMulV2Attr

BatchMatMulV2AdjX sets the optional adj_x attribute to value.

value: If `True`, adjoint the slices of `x`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulV2AdjY 

func BatchMatMulV2AdjY(value bool) BatchMatMulV2Attr

BatchMatMulV2AdjY sets the optional adj_y attribute to value.

value: If `True`, adjoint the slices of `y`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulV2GradX added in v0.8.0 

func BatchMatMulV2GradX(value bool) BatchMatMulV2Attr

BatchMatMulV2GradX sets the optional grad_x attribute to value. If not specified, defaults to false

func BatchMatMulV2GradY added in v0.8.0 

func BatchMatMulV2GradY(value bool) BatchMatMulV2Attr

BatchMatMulV2GradY sets the optional grad_y attribute to value. If not specified, defaults to false

type BatchMatMulV3Attr 

type BatchMatMulV3Attr func(optionalAttr)

BatchMatMulV3Attr is an optional argument to BatchMatMulV3.

func BatchMatMulV3AdjX 

func BatchMatMulV3AdjX(value bool) BatchMatMulV3Attr

BatchMatMulV3AdjX sets the optional adj_x attribute to value.

value: If `True`, adjoint the slices of `x`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulV3AdjY 

func BatchMatMulV3AdjY(value bool) BatchMatMulV3Attr

BatchMatMulV3AdjY sets the optional adj_y attribute to value.

value: If `True`, adjoint the slices of `y`. Defaults to `False`. If not specified, defaults to false

func BatchMatMulV3GradX added in v0.8.0 

func BatchMatMulV3GradX(value bool) BatchMatMulV3Attr

BatchMatMulV3GradX sets the optional grad_x attribute to value. If not specified, defaults to false

func BatchMatMulV3GradY added in v0.8.0 

func BatchMatMulV3GradY(value bool) BatchMatMulV3Attr

BatchMatMulV3GradY sets the optional grad_y attribute to value. If not specified, defaults to false

type BiasAddAttr 

type BiasAddAttr func(optionalAttr)

BiasAddAttr is an optional argument to BiasAdd.

func BiasAddDataFormat 

func BiasAddDataFormat(value string) BiasAddAttr

BiasAddDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the bias tensor will be added to the last dimension of the value tensor. Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

The tensor will be added to "in_channels", the third-to-the-last 

dimension.

If not specified, defaults to "NHWC"

type BiasAddGradAttr 

type BiasAddGradAttr func(optionalAttr)

BiasAddGradAttr is an optional argument to BiasAddGrad.

func BiasAddGradDataFormat 

func BiasAddGradDataFormat(value string) BiasAddGradAttr

BiasAddGradDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the bias tensor will be added to the last dimension of the value tensor. Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

The tensor will be added to "in_channels", the third-to-the-last 

dimension.

If not specified, defaults to "NHWC"

type BlockLSTMAttr 

type BlockLSTMAttr func(optionalAttr)

BlockLSTMAttr is an optional argument to BlockLSTM.

func BlockLSTMCellClip 

func BlockLSTMCellClip(value float32) BlockLSTMAttr

BlockLSTMCellClip sets the optional cell_clip attribute to value.

value: Value to clip the 'cs' value to. If not specified, defaults to 3

func BlockLSTMForgetBias 

func BlockLSTMForgetBias(value float32) BlockLSTMAttr

BlockLSTMForgetBias sets the optional forget_bias attribute to value.

value: The forget gate bias. If not specified, defaults to 1

func BlockLSTMUsePeephole 

func BlockLSTMUsePeephole(value bool) BlockLSTMAttr

BlockLSTMUsePeephole sets the optional use_peephole attribute to value.

value: Whether to use peephole weights. If not specified, defaults to false

type BlockLSTMV2Attr 

type BlockLSTMV2Attr func(optionalAttr)

BlockLSTMV2Attr is an optional argument to BlockLSTMV2.

func BlockLSTMV2CellClip 

func BlockLSTMV2CellClip(value float32) BlockLSTMV2Attr

BlockLSTMV2CellClip sets the optional cell_clip attribute to value.

value: Value to clip the 'cs' value to. If not specified, defaults to 0

func BlockLSTMV2UsePeephole 

func BlockLSTMV2UsePeephole(value bool) BlockLSTMV2Attr

BlockLSTMV2UsePeephole sets the optional use_peephole attribute to value.

value: Whether to use peephole weights. If not specified, defaults to false

type BoostedTreesCalculateBestFeatureSplitAttr 

type BoostedTreesCalculateBestFeatureSplitAttr func(optionalAttr)

BoostedTreesCalculateBestFeatureSplitAttr is an optional argument to BoostedTreesCalculateBestFeatureSplit.

func BoostedTreesCalculateBestFeatureSplitSplitType 

func BoostedTreesCalculateBestFeatureSplitSplitType(value string) BoostedTreesCalculateBestFeatureSplitAttr

BoostedTreesCalculateBestFeatureSplitSplitType sets the optional split_type attribute to value.

value: A string indicating if this Op should perform inequality split or equality split. If not specified, defaults to "inequality"

type BoostedTreesCreateQuantileStreamResourceAttr 

type BoostedTreesCreateQuantileStreamResourceAttr func(optionalAttr)

BoostedTreesCreateQuantileStreamResourceAttr is an optional argument to BoostedTreesCreateQuantileStreamResource.

func BoostedTreesCreateQuantileStreamResourceMaxElements 

func BoostedTreesCreateQuantileStreamResourceMaxElements(value int64) BoostedTreesCreateQuantileStreamResourceAttr

BoostedTreesCreateQuantileStreamResourceMaxElements sets the optional max_elements attribute to value.

value: int; The maximum number of data points that can be fed to the stream. If not specified, defaults to 1099511627776

type BoostedTreesEnsembleResourceHandleOpAttr 

type BoostedTreesEnsembleResourceHandleOpAttr func(optionalAttr)

BoostedTreesEnsembleResourceHandleOpAttr is an optional argument to BoostedTreesEnsembleResourceHandleOp.

func BoostedTreesEnsembleResourceHandleOpContainer 

func BoostedTreesEnsembleResourceHandleOpContainer(value string) BoostedTreesEnsembleResourceHandleOpAttr

BoostedTreesEnsembleResourceHandleOpContainer sets the optional container attribute to value. If not specified, defaults to ""

func BoostedTreesEnsembleResourceHandleOpSharedName 

func BoostedTreesEnsembleResourceHandleOpSharedName(value string) BoostedTreesEnsembleResourceHandleOpAttr

BoostedTreesEnsembleResourceHandleOpSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BoostedTreesQuantileStreamResourceFlushAttr 

type BoostedTreesQuantileStreamResourceFlushAttr func(optionalAttr)

BoostedTreesQuantileStreamResourceFlushAttr is an optional argument to BoostedTreesQuantileStreamResourceFlush.

func BoostedTreesQuantileStreamResourceFlushGenerateQuantiles 

func BoostedTreesQuantileStreamResourceFlushGenerateQuantiles(value bool) BoostedTreesQuantileStreamResourceFlushAttr

BoostedTreesQuantileStreamResourceFlushGenerateQuantiles sets the optional generate_quantiles attribute to value.

value: bool; If True, the output will be the num_quantiles for each stream where the ith entry is the ith quantile of the input with an approximation error of epsilon. Duplicate values may be present. If False, the output will be the points in the histogram that we got which roughly translates to 1/epsilon boundaries and without any duplicates. Default to False. If not specified, defaults to false

type BoostedTreesQuantileStreamResourceHandleOpAttr 

type BoostedTreesQuantileStreamResourceHandleOpAttr func(optionalAttr)

BoostedTreesQuantileStreamResourceHandleOpAttr is an optional argument to BoostedTreesQuantileStreamResourceHandleOp.

func BoostedTreesQuantileStreamResourceHandleOpContainer 

func BoostedTreesQuantileStreamResourceHandleOpContainer(value string) BoostedTreesQuantileStreamResourceHandleOpAttr

BoostedTreesQuantileStreamResourceHandleOpContainer sets the optional container attribute to value. If not specified, defaults to ""

func BoostedTreesQuantileStreamResourceHandleOpSharedName 

func BoostedTreesQuantileStreamResourceHandleOpSharedName(value string) BoostedTreesQuantileStreamResourceHandleOpAttr

BoostedTreesQuantileStreamResourceHandleOpSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BoostedTreesSparseCalculateBestFeatureSplitAttr 

type BoostedTreesSparseCalculateBestFeatureSplitAttr func(optionalAttr)

BoostedTreesSparseCalculateBestFeatureSplitAttr is an optional argument to BoostedTreesSparseCalculateBestFeatureSplit.

func BoostedTreesSparseCalculateBestFeatureSplitSplitType 

func BoostedTreesSparseCalculateBestFeatureSplitSplitType(value string) BoostedTreesSparseCalculateBestFeatureSplitAttr

BoostedTreesSparseCalculateBestFeatureSplitSplitType sets the optional split_type attribute to value.

value: A string indicating if this Op should perform inequality split or equality split. If not specified, defaults to "inequality"

type BoostedTreesUpdateEnsembleV2Attr 

type BoostedTreesUpdateEnsembleV2Attr func(optionalAttr)

BoostedTreesUpdateEnsembleV2Attr is an optional argument to BoostedTreesUpdateEnsembleV2.

func BoostedTreesUpdateEnsembleV2LogitsDimension 

func BoostedTreesUpdateEnsembleV2LogitsDimension(value int64) BoostedTreesUpdateEnsembleV2Attr

BoostedTreesUpdateEnsembleV2LogitsDimension sets the optional logits_dimension attribute to value.

value: scalar, dimension of the logits If not specified, defaults to 1

type CTCBeamSearchDecoderAttr 

type CTCBeamSearchDecoderAttr func(optionalAttr)

CTCBeamSearchDecoderAttr is an optional argument to CTCBeamSearchDecoder.

func CTCBeamSearchDecoderMergeRepeated 

func CTCBeamSearchDecoderMergeRepeated(value bool) CTCBeamSearchDecoderAttr

CTCBeamSearchDecoderMergeRepeated sets the optional merge_repeated attribute to value.

value: If true, merge repeated classes in output. If not specified, defaults to true

type CTCGreedyDecoderAttr 

type CTCGreedyDecoderAttr func(optionalAttr)

CTCGreedyDecoderAttr is an optional argument to CTCGreedyDecoder.

func CTCGreedyDecoderBlankIndex 

func CTCGreedyDecoderBlankIndex(value int64) CTCGreedyDecoderAttr

CTCGreedyDecoderBlankIndex sets the optional blank_index attribute to value. If not specified, defaults to -1

func CTCGreedyDecoderMergeRepeated 

func CTCGreedyDecoderMergeRepeated(value bool) CTCGreedyDecoderAttr

CTCGreedyDecoderMergeRepeated sets the optional merge_repeated attribute to value.

value: If True, merge repeated classes in output. If not specified, defaults to false

type CTCLossAttr 

type CTCLossAttr func(optionalAttr)

CTCLossAttr is an optional argument to CTCLoss.

func CTCLossCtcMergeRepeated 

func CTCLossCtcMergeRepeated(value bool) CTCLossAttr

CTCLossCtcMergeRepeated sets the optional ctc_merge_repeated attribute to value.

value: Scalar. If set to false, *during* CTC calculation repeated non-blank labels will not be merged and are interpreted as individual labels. This is a simplified version of CTC. If not specified, defaults to true

func CTCLossIgnoreLongerOutputsThanInputs 

func CTCLossIgnoreLongerOutputsThanInputs(value bool) CTCLossAttr

CTCLossIgnoreLongerOutputsThanInputs sets the optional ignore_longer_outputs_than_inputs attribute to value.

value: Scalar. If set to true, during CTC calculation, items that have longer output sequences than input sequences are skipped: they don't contribute to the loss term and have zero-gradient. If not specified, defaults to false

func CTCLossPreprocessCollapseRepeated 

func CTCLossPreprocessCollapseRepeated(value bool) CTCLossAttr

CTCLossPreprocessCollapseRepeated sets the optional preprocess_collapse_repeated attribute to value.

value: Scalar, if true then repeated labels are collapsed prior to the CTC calculation. If not specified, defaults to false

type CTCLossV2Attr 

type CTCLossV2Attr func(optionalAttr)

CTCLossV2Attr is an optional argument to CTCLossV2.

func CTCLossV2CtcMergeRepeated 

func CTCLossV2CtcMergeRepeated(value bool) CTCLossV2Attr

CTCLossV2CtcMergeRepeated sets the optional ctc_merge_repeated attribute to value.

value: Scalar. If set to false, *during* CTC calculation repeated non-blank labels will not be merged and are interpreted as individual labels. This is a simplified version of CTC. If not specified, defaults to true

func CTCLossV2IgnoreLongerOutputsThanInputs 

func CTCLossV2IgnoreLongerOutputsThanInputs(value bool) CTCLossV2Attr

CTCLossV2IgnoreLongerOutputsThanInputs sets the optional ignore_longer_outputs_than_inputs attribute to value.

value: Scalar. If set to true, during CTC calculation, items that have longer output sequences than input sequences are skipped: they don't contribute to the loss term and have zero-gradient. If not specified, defaults to false

func CTCLossV2PreprocessCollapseRepeated 

func CTCLossV2PreprocessCollapseRepeated(value bool) CTCLossV2Attr

CTCLossV2PreprocessCollapseRepeated sets the optional preprocess_collapse_repeated attribute to value.

value: Scalar, if true then repeated labels are collapsed prior to the CTC calculation. If not specified, defaults to false

type CacheDatasetAttr 

type CacheDatasetAttr func(optionalAttr)

CacheDatasetAttr is an optional argument to CacheDataset.

func CacheDatasetMetadata 

func CacheDatasetMetadata(value string) CacheDatasetAttr

CacheDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type CastAttr 

type CastAttr func(optionalAttr)

CastAttr is an optional argument to Cast.

func CastTruncate 

func CastTruncate(value bool) CastAttr

CastTruncate sets the optional Truncate attribute to value. If not specified, defaults to false

type CollectiveAllToAllV2Attr added in v0.5.0 

type CollectiveAllToAllV2Attr func(optionalAttr)

CollectiveAllToAllV2Attr is an optional argument to CollectiveAllToAllV2.

func CollectiveAllToAllV2CommunicationHint added in v0.5.0 

func CollectiveAllToAllV2CommunicationHint(value string) CollectiveAllToAllV2Attr

CollectiveAllToAllV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveAllToAllV2IsStateless added in v0.7.0 

func CollectiveAllToAllV2IsStateless(value bool) CollectiveAllToAllV2Attr

CollectiveAllToAllV2IsStateless sets the optional is_stateless attribute to value. If not specified, defaults to false

func CollectiveAllToAllV2TimeoutSeconds added in v0.5.0 

func CollectiveAllToAllV2TimeoutSeconds(value float32) CollectiveAllToAllV2Attr

CollectiveAllToAllV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveAllToAllV3Attr 

type CollectiveAllToAllV3Attr func(optionalAttr)

CollectiveAllToAllV3Attr is an optional argument to CollectiveAllToAllV3.

func CollectiveAllToAllV3TimeoutSeconds 

func CollectiveAllToAllV3TimeoutSeconds(value float32) CollectiveAllToAllV3Attr

CollectiveAllToAllV3TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveBcastRecvAttr 

type CollectiveBcastRecvAttr func(optionalAttr)

CollectiveBcastRecvAttr is an optional argument to CollectiveBcastRecv.

func CollectiveBcastRecvCommunicationHint 

func CollectiveBcastRecvCommunicationHint(value string) CollectiveBcastRecvAttr

CollectiveBcastRecvCommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveBcastRecvTimeoutSeconds 

func CollectiveBcastRecvTimeoutSeconds(value float32) CollectiveBcastRecvAttr

CollectiveBcastRecvTimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveBcastRecvV2Attr 

type CollectiveBcastRecvV2Attr func(optionalAttr)

CollectiveBcastRecvV2Attr is an optional argument to CollectiveBcastRecvV2.

func CollectiveBcastRecvV2CommunicationHint 

func CollectiveBcastRecvV2CommunicationHint(value string) CollectiveBcastRecvV2Attr

CollectiveBcastRecvV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveBcastRecvV2TimeoutSeconds 

func CollectiveBcastRecvV2TimeoutSeconds(value float32) CollectiveBcastRecvV2Attr

CollectiveBcastRecvV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveBcastSendAttr 

type CollectiveBcastSendAttr func(optionalAttr)

CollectiveBcastSendAttr is an optional argument to CollectiveBcastSend.

func CollectiveBcastSendCommunicationHint 

func CollectiveBcastSendCommunicationHint(value string) CollectiveBcastSendAttr

CollectiveBcastSendCommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveBcastSendTimeoutSeconds 

func CollectiveBcastSendTimeoutSeconds(value float32) CollectiveBcastSendAttr

CollectiveBcastSendTimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveBcastSendV2Attr 

type CollectiveBcastSendV2Attr func(optionalAttr)

CollectiveBcastSendV2Attr is an optional argument to CollectiveBcastSendV2.

func CollectiveBcastSendV2CommunicationHint 

func CollectiveBcastSendV2CommunicationHint(value string) CollectiveBcastSendV2Attr

CollectiveBcastSendV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveBcastSendV2TimeoutSeconds 

func CollectiveBcastSendV2TimeoutSeconds(value float32) CollectiveBcastSendV2Attr

CollectiveBcastSendV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveGatherAttr 

type CollectiveGatherAttr func(optionalAttr)

CollectiveGatherAttr is an optional argument to CollectiveGather.

func CollectiveGatherCommunicationHint 

func CollectiveGatherCommunicationHint(value string) CollectiveGatherAttr

CollectiveGatherCommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveGatherTimeoutSeconds 

func CollectiveGatherTimeoutSeconds(value float32) CollectiveGatherAttr

CollectiveGatherTimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveGatherV2Attr 

type CollectiveGatherV2Attr func(optionalAttr)

CollectiveGatherV2Attr is an optional argument to CollectiveGatherV2.

func CollectiveGatherV2CommunicationHint 

func CollectiveGatherV2CommunicationHint(value string) CollectiveGatherV2Attr

CollectiveGatherV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveGatherV2IsStateless added in v0.7.0 

func CollectiveGatherV2IsStateless(value bool) CollectiveGatherV2Attr

CollectiveGatherV2IsStateless sets the optional is_stateless attribute to value. If not specified, defaults to false

func CollectiveGatherV2TimeoutSeconds 

func CollectiveGatherV2TimeoutSeconds(value float32) CollectiveGatherV2Attr

CollectiveGatherV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveInitializeCommunicatorAttr 

type CollectiveInitializeCommunicatorAttr func(optionalAttr)

CollectiveInitializeCommunicatorAttr is an optional argument to CollectiveInitializeCommunicator.

func CollectiveInitializeCommunicatorCommunicationHint 

func CollectiveInitializeCommunicatorCommunicationHint(value string) CollectiveInitializeCommunicatorAttr

CollectiveInitializeCommunicatorCommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveInitializeCommunicatorTimeoutSeconds 

func CollectiveInitializeCommunicatorTimeoutSeconds(value float32) CollectiveInitializeCommunicatorAttr

CollectiveInitializeCommunicatorTimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveReduceAttr 

type CollectiveReduceAttr func(optionalAttr)

CollectiveReduceAttr is an optional argument to CollectiveReduce.

func CollectiveReduceCommunicationHint 

func CollectiveReduceCommunicationHint(value string) CollectiveReduceAttr

CollectiveReduceCommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveReduceTimeoutSeconds 

func CollectiveReduceTimeoutSeconds(value float32) CollectiveReduceAttr

CollectiveReduceTimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

func CollectiveReduceWaitFor 

func CollectiveReduceWaitFor(value []int64) CollectiveReduceAttr

CollectiveReduceWaitFor sets the optional wait_for attribute to value. If not specified, defaults to {}

type CollectiveReduceScatterV2Attr added in v0.4.0 

type CollectiveReduceScatterV2Attr func(optionalAttr)

CollectiveReduceScatterV2Attr is an optional argument to CollectiveReduceScatterV2.

func CollectiveReduceScatterV2CommunicationHint added in v0.4.0 

func CollectiveReduceScatterV2CommunicationHint(value string) CollectiveReduceScatterV2Attr

CollectiveReduceScatterV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveReduceScatterV2IsStateless added in v0.7.0 

func CollectiveReduceScatterV2IsStateless(value bool) CollectiveReduceScatterV2Attr

CollectiveReduceScatterV2IsStateless sets the optional is_stateless attribute to value. If not specified, defaults to false

func CollectiveReduceScatterV2MaxSubdivsPerDevice added in v0.4.0 

func CollectiveReduceScatterV2MaxSubdivsPerDevice(value int64) CollectiveReduceScatterV2Attr

CollectiveReduceScatterV2MaxSubdivsPerDevice sets the optional max_subdivs_per_device attribute to value. If not specified, defaults to -1

func CollectiveReduceScatterV2TimeoutSeconds added in v0.4.0 

func CollectiveReduceScatterV2TimeoutSeconds(value float32) CollectiveReduceScatterV2Attr

CollectiveReduceScatterV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveReduceV2Attr 

type CollectiveReduceV2Attr func(optionalAttr)

CollectiveReduceV2Attr is an optional argument to CollectiveReduceV2.

func CollectiveReduceV2CommunicationHint 

func CollectiveReduceV2CommunicationHint(value string) CollectiveReduceV2Attr

CollectiveReduceV2CommunicationHint sets the optional communication_hint attribute to value. If not specified, defaults to "auto"

func CollectiveReduceV2IsStateless added in v0.7.0 

func CollectiveReduceV2IsStateless(value bool) CollectiveReduceV2Attr

CollectiveReduceV2IsStateless sets the optional is_stateless attribute to value. If not specified, defaults to false

func CollectiveReduceV2MaxSubdivsPerDevice 

func CollectiveReduceV2MaxSubdivsPerDevice(value int64) CollectiveReduceV2Attr

CollectiveReduceV2MaxSubdivsPerDevice sets the optional max_subdivs_per_device attribute to value. If not specified, defaults to -1

func CollectiveReduceV2TimeoutSeconds 

func CollectiveReduceV2TimeoutSeconds(value float32) CollectiveReduceV2Attr

CollectiveReduceV2TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CollectiveReduceV3Attr 

type CollectiveReduceV3Attr func(optionalAttr)

CollectiveReduceV3Attr is an optional argument to CollectiveReduceV3.

func CollectiveReduceV3TimeoutSeconds 

func CollectiveReduceV3TimeoutSeconds(value float32) CollectiveReduceV3Attr

CollectiveReduceV3TimeoutSeconds sets the optional timeout_seconds attribute to value. If not specified, defaults to 0

type CombinedNonMaxSuppressionAttr 

type CombinedNonMaxSuppressionAttr func(optionalAttr)

CombinedNonMaxSuppressionAttr is an optional argument to CombinedNonMaxSuppression.

func CombinedNonMaxSuppressionClipBoxes 

func CombinedNonMaxSuppressionClipBoxes(value bool) CombinedNonMaxSuppressionAttr

CombinedNonMaxSuppressionClipBoxes sets the optional clip_boxes attribute to value.

value: If true, assume the box coordinates are between [0, 1] and clip the output boxes if they fall beyond [0, 1]. If false, do not do clipping and output the box coordinates as it is. If not specified, defaults to true

func CombinedNonMaxSuppressionPadPerClass 

func CombinedNonMaxSuppressionPadPerClass(value bool) CombinedNonMaxSuppressionAttr

CombinedNonMaxSuppressionPadPerClass sets the optional pad_per_class attribute to value.

value: If false, the output nmsed boxes, scores and classes are padded/clipped to `max_total_size`. If true, the output nmsed boxes, scores and classes are padded to be of length `max_size_per_class`*`num_classes`, unless it exceeds `max_total_size` in which case it is clipped to `max_total_size`. Defaults to false. If not specified, defaults to false

type ComplexAbsAttr 

type ComplexAbsAttr func(optionalAttr)

ComplexAbsAttr is an optional argument to ComplexAbs.

func ComplexAbsTout 

func ComplexAbsTout(value tf.DataType) ComplexAbsAttr

ComplexAbsTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type ComplexAttr 

type ComplexAttr func(optionalAttr)

ComplexAttr is an optional argument to Complex.

func ComplexTout 

func ComplexTout(value tf.DataType) ComplexAttr

ComplexTout sets the optional Tout attribute to value. If not specified, defaults to DT_COMPLEX64

type ComputeAccidentalHitsAttr 

type ComputeAccidentalHitsAttr func(optionalAttr)

ComputeAccidentalHitsAttr is an optional argument to ComputeAccidentalHits.

func ComputeAccidentalHitsSeed 

func ComputeAccidentalHitsSeed(value int64) ComputeAccidentalHitsAttr

ComputeAccidentalHitsSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func ComputeAccidentalHitsSeed2 

func ComputeAccidentalHitsSeed2(value int64) ComputeAccidentalHitsAttr

ComputeAccidentalHitsSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type ConcatenateDatasetAttr 

type ConcatenateDatasetAttr func(optionalAttr)

ConcatenateDatasetAttr is an optional argument to ConcatenateDataset.

func ConcatenateDatasetMetadata 

func ConcatenateDatasetMetadata(value string) ConcatenateDatasetAttr

ConcatenateDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type ConfigureAndInitializeGlobalTPUAttr added in v0.3.0 

type ConfigureAndInitializeGlobalTPUAttr func(optionalAttr)

ConfigureAndInitializeGlobalTPUAttr is an optional argument to ConfigureAndInitializeGlobalTPU.

func ConfigureAndInitializeGlobalTPUUseTfrtHostRuntime added in v0.3.0 

func ConfigureAndInitializeGlobalTPUUseTfrtHostRuntime(value bool) ConfigureAndInitializeGlobalTPUAttr

ConfigureAndInitializeGlobalTPUUseTfrtHostRuntime sets the optional use_tfrt_host_runtime attribute to value. If not specified, defaults to true

type ConfigureDistributedTPUAttr 

type ConfigureDistributedTPUAttr func(optionalAttr)

ConfigureDistributedTPUAttr is an optional argument to ConfigureDistributedTPU.

func ConfigureDistributedTPUCompilationFailureClosesChips 

func ConfigureDistributedTPUCompilationFailureClosesChips(value bool) ConfigureDistributedTPUAttr

ConfigureDistributedTPUCompilationFailureClosesChips sets the optional compilation_failure_closes_chips attribute to value. If not specified, defaults to true

func ConfigureDistributedTPUEmbeddingConfig 

func ConfigureDistributedTPUEmbeddingConfig(value string) ConfigureDistributedTPUAttr

ConfigureDistributedTPUEmbeddingConfig sets the optional embedding_config attribute to value.

value: Reserved. Do not use. If not specified, defaults to ""

func ConfigureDistributedTPUEnableWholeMeshCompilations 

func ConfigureDistributedTPUEnableWholeMeshCompilations(value bool) ConfigureDistributedTPUAttr

ConfigureDistributedTPUEnableWholeMeshCompilations sets the optional enable_whole_mesh_compilations attribute to value. If not specified, defaults to false

func ConfigureDistributedTPUIsGlobalInit 

func ConfigureDistributedTPUIsGlobalInit(value bool) ConfigureDistributedTPUAttr

ConfigureDistributedTPUIsGlobalInit sets the optional is_global_init attribute to value.

value: Reserved. Do not use. If not specified, defaults to false

func ConfigureDistributedTPUTpuCancellationClosesChips 

func ConfigureDistributedTPUTpuCancellationClosesChips(value int64) ConfigureDistributedTPUAttr

ConfigureDistributedTPUTpuCancellationClosesChips sets the optional tpu_cancellation_closes_chips attribute to value. If not specified, defaults to 0

func ConfigureDistributedTPUTpuEmbeddingConfig 

func ConfigureDistributedTPUTpuEmbeddingConfig(value string) ConfigureDistributedTPUAttr

ConfigureDistributedTPUTpuEmbeddingConfig sets the optional tpu_embedding_config attribute to value.

value: Serialized tensorflow.tpu.TPUEmbeddingConfiguration that describes the embedding lookups of the program. If not specified, defaults to ""

type Conv2DAttr 

type Conv2DAttr func(optionalAttr)

Conv2DAttr is an optional argument to Conv2D.

func Conv2DDataFormat 

func Conv2DDataFormat(value string) Conv2DAttr

Conv2DDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, height, width, channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, channels, height, width].

If not specified, defaults to "NHWC"

func Conv2DDilations 

func Conv2DDilations(value []int64) Conv2DAttr

Conv2DDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func Conv2DExplicitPaddings 

func Conv2DExplicitPaddings(value []int64) Conv2DAttr

Conv2DExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func Conv2DUseCudnnOnGpu 

func Conv2DUseCudnnOnGpu(value bool) Conv2DAttr

Conv2DUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropFilterAttr 

type Conv2DBackpropFilterAttr func(optionalAttr)

Conv2DBackpropFilterAttr is an optional argument to Conv2DBackpropFilter.

func Conv2DBackpropFilterDataFormat 

func Conv2DBackpropFilterDataFormat(value string) Conv2DBackpropFilterAttr

Conv2DBackpropFilterDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func Conv2DBackpropFilterDilations 

func Conv2DBackpropFilterDilations(value []int64) Conv2DBackpropFilterAttr

Conv2DBackpropFilterDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func Conv2DBackpropFilterExplicitPaddings 

func Conv2DBackpropFilterExplicitPaddings(value []int64) Conv2DBackpropFilterAttr

Conv2DBackpropFilterExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func Conv2DBackpropFilterUseCudnnOnGpu 

func Conv2DBackpropFilterUseCudnnOnGpu(value bool) Conv2DBackpropFilterAttr

Conv2DBackpropFilterUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropFilterV2Attr added in v0.4.0 

type Conv2DBackpropFilterV2Attr func(optionalAttr)

Conv2DBackpropFilterV2Attr is an optional argument to Conv2DBackpropFilterV2.

func Conv2DBackpropFilterV2DataFormat added in v0.4.0 

func Conv2DBackpropFilterV2DataFormat(value string) Conv2DBackpropFilterV2Attr

Conv2DBackpropFilterV2DataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func Conv2DBackpropFilterV2Dilations added in v0.4.0 

func Conv2DBackpropFilterV2Dilations(value []int64) Conv2DBackpropFilterV2Attr

Conv2DBackpropFilterV2Dilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func Conv2DBackpropFilterV2ExplicitPaddings added in v0.4.0 

func Conv2DBackpropFilterV2ExplicitPaddings(value []int64) Conv2DBackpropFilterV2Attr

Conv2DBackpropFilterV2ExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func Conv2DBackpropFilterV2UseCudnnOnGpu added in v0.4.0 

func Conv2DBackpropFilterV2UseCudnnOnGpu(value bool) Conv2DBackpropFilterV2Attr

Conv2DBackpropFilterV2UseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropInputAttr 

type Conv2DBackpropInputAttr func(optionalAttr)

Conv2DBackpropInputAttr is an optional argument to Conv2DBackpropInput.

func Conv2DBackpropInputDataFormat 

func Conv2DBackpropInputDataFormat(value string) Conv2DBackpropInputAttr

Conv2DBackpropInputDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func Conv2DBackpropInputDilations 

func Conv2DBackpropInputDilations(value []int64) Conv2DBackpropInputAttr

Conv2DBackpropInputDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func Conv2DBackpropInputExplicitPaddings 

func Conv2DBackpropInputExplicitPaddings(value []int64) Conv2DBackpropInputAttr

Conv2DBackpropInputExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func Conv2DBackpropInputUseCudnnOnGpu 

func Conv2DBackpropInputUseCudnnOnGpu(value bool) Conv2DBackpropInputAttr

Conv2DBackpropInputUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropInputV2Attr added in v0.4.0 

type Conv2DBackpropInputV2Attr func(optionalAttr)

Conv2DBackpropInputV2Attr is an optional argument to Conv2DBackpropInputV2.

func Conv2DBackpropInputV2DataFormat added in v0.4.0 

func Conv2DBackpropInputV2DataFormat(value string) Conv2DBackpropInputV2Attr

Conv2DBackpropInputV2DataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func Conv2DBackpropInputV2Dilations added in v0.4.0 

func Conv2DBackpropInputV2Dilations(value []int64) Conv2DBackpropInputV2Attr

Conv2DBackpropInputV2Dilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func Conv2DBackpropInputV2ExplicitPaddings added in v0.4.0 

func Conv2DBackpropInputV2ExplicitPaddings(value []int64) Conv2DBackpropInputV2Attr

Conv2DBackpropInputV2ExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func Conv2DBackpropInputV2UseCudnnOnGpu added in v0.4.0 

func Conv2DBackpropInputV2UseCudnnOnGpu(value bool) Conv2DBackpropInputV2Attr

Conv2DBackpropInputV2UseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv3DAttr 

type Conv3DAttr func(optionalAttr)

Conv3DAttr is an optional argument to Conv3D.

func Conv3DDataFormat 

func Conv3DDataFormat(value string) Conv3DAttr

Conv3DDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

func Conv3DDilations 

func Conv3DDilations(value []int64) Conv3DAttr

Conv3DDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 5. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1 i:1}

type Conv3DBackpropFilterAttr 

type Conv3DBackpropFilterAttr func(optionalAttr)

Conv3DBackpropFilterAttr is an optional argument to Conv3DBackpropFilter.

func Conv3DBackpropFilterDilations 

func Conv3DBackpropFilterDilations(value []int64) Conv3DBackpropFilterAttr

Conv3DBackpropFilterDilations sets the optional dilations attribute to value. If not specified, defaults to {i:1 i:1 i:1 i:1 i:1}

type Conv3DBackpropFilterV2Attr 

type Conv3DBackpropFilterV2Attr func(optionalAttr)

Conv3DBackpropFilterV2Attr is an optional argument to Conv3DBackpropFilterV2.

func Conv3DBackpropFilterV2DataFormat 

func Conv3DBackpropFilterV2DataFormat(value string) Conv3DBackpropFilterV2Attr

Conv3DBackpropFilterV2DataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

func Conv3DBackpropFilterV2Dilations 

func Conv3DBackpropFilterV2Dilations(value []int64) Conv3DBackpropFilterV2Attr

Conv3DBackpropFilterV2Dilations sets the optional dilations attribute to value.

value: 1-D tensor of length 5. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1 i:1}

type Conv3DBackpropInputAttr 

type Conv3DBackpropInputAttr func(optionalAttr)

Conv3DBackpropInputAttr is an optional argument to Conv3DBackpropInput.

func Conv3DBackpropInputDilations 

func Conv3DBackpropInputDilations(value []int64) Conv3DBackpropInputAttr

Conv3DBackpropInputDilations sets the optional dilations attribute to value. If not specified, defaults to {i:1 i:1 i:1 i:1 i:1}

type Conv3DBackpropInputV2Attr 

type Conv3DBackpropInputV2Attr func(optionalAttr)

Conv3DBackpropInputV2Attr is an optional argument to Conv3DBackpropInputV2.

func Conv3DBackpropInputV2DataFormat 

func Conv3DBackpropInputV2DataFormat(value string) Conv3DBackpropInputV2Attr

Conv3DBackpropInputV2DataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

func Conv3DBackpropInputV2Dilations 

func Conv3DBackpropInputV2Dilations(value []int64) Conv3DBackpropInputV2Attr

Conv3DBackpropInputV2Dilations sets the optional dilations attribute to value.

value: 1-D tensor of length 5. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1 i:1}

type ConvAttr added in v0.6.0 

type ConvAttr func(optionalAttr)

ConvAttr is an optional argument to Conv.

func ConvBatchDims added in v0.6.0 

func ConvBatchDims(value int64) ConvAttr

ConvBatchDims sets the optional batch_dims attribute to value.

value: A positive integer specifying the number of batch dimensions for the input tensor. Should be less than the rank of the input tensor. If not specified, defaults to 1

func ConvDataFormat added in v0.6.0 

func ConvDataFormat(value string) ConvAttr

ConvDataFormat sets the optional data_format attribute to value.

value: Used to set the data format. By default `CHANNELS_FIRST`, uses `NHWC (2D) / NDHWC (3D)` or if `CHANNELS_LAST`, uses `NCHW (2D) / NCDHW (3D)`. If not specified, defaults to "CHANNELS_LAST"

func ConvDilations added in v0.6.0 

func ConvDilations(value []int64) ConvAttr

ConvDilations sets the optional dilations attribute to value.

value: 1-D tensor of length `N+2`. The dilation factor for each dimension of `input`. If set to `k > 1`, there will be `k-1` skipped cells between each filter element on that dimension. The dimension order is determined by the value of `channels_last_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {}

func ConvExplicitPaddings added in v0.6.0 

func ConvExplicitPaddings(value []int64) ConvAttr

ConvExplicitPaddings sets the optional explicit_paddings attribute to value.

value: If `padding` is `"EXPLICIT"`, the list of explicit padding amounts. For the ith dimension, the amount of padding inserted before and after the dimension is `explicit_paddings[2 * i]` and `explicit_paddings[2 * i + 1]`, respectively. If `padding` is not `"EXPLICIT"`, `explicit_paddings` must be empty. If not specified, defaults to {}

func ConvGroups added in v0.6.0 

func ConvGroups(value int64) ConvAttr

ConvGroups sets the optional groups attribute to value.

value: A positive integer specifying the number of groups in which the input is split along the channel axis. Each group is convolved separately with `filters / groups` filters. The output is the concatenation of all the groups results along the channel axis. Input channels and filters must both be divisible by groups. If not specified, defaults to 1

type CopyAttr 

type CopyAttr func(optionalAttr)

CopyAttr is an optional argument to Copy.

func CopyDebugOpsSpec 

func CopyDebugOpsSpec(value []string) CopyAttr

CopyDebugOpsSpec sets the optional debug_ops_spec attribute to value.

value: A list of debug op spec (op, url, gated_grpc) for attached debug ops. Each element of the list has the format <debug_op>;<grpc_url>;<gated_grpc>, wherein gated_grpc is boolean represented as 0/1. E.g., "DebugIdentity;grpc://foo:3333;1", "DebugIdentity;file:///tmp/tfdbg_1;0". If not specified, defaults to {}

func CopyTensorName 

func CopyTensorName(value string) CopyAttr

CopyTensorName sets the optional tensor_name attribute to value.

value: The name of the input tensor. If not specified, defaults to ""

type CopyHostAttr 

type CopyHostAttr func(optionalAttr)

CopyHostAttr is an optional argument to CopyHost.

func CopyHostDebugOpsSpec 

func CopyHostDebugOpsSpec(value []string) CopyHostAttr

CopyHostDebugOpsSpec sets the optional debug_ops_spec attribute to value.

value: A list of debug op spec (op, url, gated_grpc) for attached debug ops. Each element of the list has the format <debug_op>;<grpc_url>;<gated_grpc>, wherein gated_grpc is boolean represented as 0/1. E.g., "DebugIdentity;grpc://foo:3333;1", "DebugIdentity;file:///tmp/tfdbg_1;0". If not specified, defaults to {}

func CopyHostTensorName 

func CopyHostTensorName(value string) CopyHostAttr

CopyHostTensorName sets the optional tensor_name attribute to value.

value: The name of the input tensor. If not specified, defaults to ""

type CropAndResizeAttr 

type CropAndResizeAttr func(optionalAttr)

CropAndResizeAttr is an optional argument to CropAndResize.

func CropAndResizeExtrapolationValue 

func CropAndResizeExtrapolationValue(value float32) CropAndResizeAttr

CropAndResizeExtrapolationValue sets the optional extrapolation_value attribute to value.

value: Value used for extrapolation, when applicable. If not specified, defaults to 0

func CropAndResizeMethod 

func CropAndResizeMethod(value string) CropAndResizeAttr

CropAndResizeMethod sets the optional method attribute to value.

value: A string specifying the sampling method for resizing. It can be either `"bilinear"` or `"nearest"` and default to `"bilinear"`. Currently two sampling methods are supported: Bilinear and Nearest Neighbor. If not specified, defaults to "bilinear"

type CropAndResizeGradBoxesAttr 

type CropAndResizeGradBoxesAttr func(optionalAttr)

CropAndResizeGradBoxesAttr is an optional argument to CropAndResizeGradBoxes.

func CropAndResizeGradBoxesMethod 

func CropAndResizeGradBoxesMethod(value string) CropAndResizeGradBoxesAttr

CropAndResizeGradBoxesMethod sets the optional method attribute to value.

value: A string specifying the interpolation method. Only 'bilinear' is supported for now. If not specified, defaults to "bilinear"

type CropAndResizeGradImageAttr 

type CropAndResizeGradImageAttr func(optionalAttr)

CropAndResizeGradImageAttr is an optional argument to CropAndResizeGradImage.

func CropAndResizeGradImageMethod 

func CropAndResizeGradImageMethod(value string) CropAndResizeGradImageAttr

CropAndResizeGradImageMethod sets the optional method attribute to value.

value: A string specifying the interpolation method. Only 'bilinear' is supported for now. If not specified, defaults to "bilinear"

type CudnnRNNAttr 

type CudnnRNNAttr func(optionalAttr)

CudnnRNNAttr is an optional argument to CudnnRNN.

func CudnnRNNDirection 

func CudnnRNNDirection(value string) CudnnRNNAttr

CudnnRNNDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNDropout 

func CudnnRNNDropout(value float32) CudnnRNNAttr

CudnnRNNDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNInputMode 

func CudnnRNNInputMode(value string) CudnnRNNAttr

CudnnRNNInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNIsTraining 

func CudnnRNNIsTraining(value bool) CudnnRNNAttr

CudnnRNNIsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNRnnMode 

func CudnnRNNRnnMode(value string) CudnnRNNAttr

CudnnRNNRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNSeed 

func CudnnRNNSeed(value int64) CudnnRNNAttr

CudnnRNNSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNSeed2 

func CudnnRNNSeed2(value int64) CudnnRNNAttr

CudnnRNNSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropAttr 

type CudnnRNNBackpropAttr func(optionalAttr)

CudnnRNNBackpropAttr is an optional argument to CudnnRNNBackprop.

func CudnnRNNBackpropDirection 

func CudnnRNNBackpropDirection(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropDropout 

func CudnnRNNBackpropDropout(value float32) CudnnRNNBackpropAttr

CudnnRNNBackpropDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropInputMode 

func CudnnRNNBackpropInputMode(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropRnnMode 

func CudnnRNNBackpropRnnMode(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropSeed 

func CudnnRNNBackpropSeed(value int64) CudnnRNNBackpropAttr

CudnnRNNBackpropSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropSeed2 

func CudnnRNNBackpropSeed2(value int64) CudnnRNNBackpropAttr

CudnnRNNBackpropSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropV2Attr 

type CudnnRNNBackpropV2Attr func(optionalAttr)

CudnnRNNBackpropV2Attr is an optional argument to CudnnRNNBackpropV2.

func CudnnRNNBackpropV2Direction 

func CudnnRNNBackpropV2Direction(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropV2Dropout 

func CudnnRNNBackpropV2Dropout(value float32) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV2InputMode 

func CudnnRNNBackpropV2InputMode(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropV2RnnMode 

func CudnnRNNBackpropV2RnnMode(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropV2Seed 

func CudnnRNNBackpropV2Seed(value int64) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV2Seed2 

func CudnnRNNBackpropV2Seed2(value int64) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropV3Attr 

type CudnnRNNBackpropV3Attr func(optionalAttr)

CudnnRNNBackpropV3Attr is an optional argument to CudnnRNNBackpropV3.

func CudnnRNNBackpropV3Direction 

func CudnnRNNBackpropV3Direction(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropV3Dropout 

func CudnnRNNBackpropV3Dropout(value float32) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3InputMode 

func CudnnRNNBackpropV3InputMode(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropV3NumProj 

func CudnnRNNBackpropV3NumProj(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3RnnMode 

func CudnnRNNBackpropV3RnnMode(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropV3Seed 

func CudnnRNNBackpropV3Seed(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3Seed2 

func CudnnRNNBackpropV3Seed2(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3TimeMajor 

func CudnnRNNBackpropV3TimeMajor(value bool) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3TimeMajor sets the optional time_major attribute to value. If not specified, defaults to true

type CudnnRNNCanonicalToParamsAttr 

type CudnnRNNCanonicalToParamsAttr func(optionalAttr)

CudnnRNNCanonicalToParamsAttr is an optional argument to CudnnRNNCanonicalToParams.

func CudnnRNNCanonicalToParamsDirection 

func CudnnRNNCanonicalToParamsDirection(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNCanonicalToParamsDropout 

func CudnnRNNCanonicalToParamsDropout(value float32) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsInputMode 

func CudnnRNNCanonicalToParamsInputMode(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNCanonicalToParamsRnnMode 

func CudnnRNNCanonicalToParamsRnnMode(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNCanonicalToParamsSeed 

func CudnnRNNCanonicalToParamsSeed(value int64) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsSeed2 

func CudnnRNNCanonicalToParamsSeed2(value int64) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNCanonicalToParamsV2Attr 

type CudnnRNNCanonicalToParamsV2Attr func(optionalAttr)

CudnnRNNCanonicalToParamsV2Attr is an optional argument to CudnnRNNCanonicalToParamsV2.

func CudnnRNNCanonicalToParamsV2Direction 

func CudnnRNNCanonicalToParamsV2Direction(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNCanonicalToParamsV2Dropout 

func CudnnRNNCanonicalToParamsV2Dropout(value float32) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2InputMode 

func CudnnRNNCanonicalToParamsV2InputMode(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNCanonicalToParamsV2NumProj 

func CudnnRNNCanonicalToParamsV2NumProj(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2RnnMode 

func CudnnRNNCanonicalToParamsV2RnnMode(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNCanonicalToParamsV2Seed 

func CudnnRNNCanonicalToParamsV2Seed(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2Seed2 

func CudnnRNNCanonicalToParamsV2Seed2(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsSizeAttr 

type CudnnRNNParamsSizeAttr func(optionalAttr)

CudnnRNNParamsSizeAttr is an optional argument to CudnnRNNParamsSize.

func CudnnRNNParamsSizeDirection 

func CudnnRNNParamsSizeDirection(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsSizeDropout 

func CudnnRNNParamsSizeDropout(value float32) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeInputMode 

func CudnnRNNParamsSizeInputMode(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsSizeNumProj 

func CudnnRNNParamsSizeNumProj(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeNumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeRnnMode 

func CudnnRNNParamsSizeRnnMode(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsSizeSeed 

func CudnnRNNParamsSizeSeed(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeSeed2 

func CudnnRNNParamsSizeSeed2(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsToCanonicalAttr 

type CudnnRNNParamsToCanonicalAttr func(optionalAttr)

CudnnRNNParamsToCanonicalAttr is an optional argument to CudnnRNNParamsToCanonical.

func CudnnRNNParamsToCanonicalDirection 

func CudnnRNNParamsToCanonicalDirection(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsToCanonicalDropout 

func CudnnRNNParamsToCanonicalDropout(value float32) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalInputMode 

func CudnnRNNParamsToCanonicalInputMode(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsToCanonicalRnnMode 

func CudnnRNNParamsToCanonicalRnnMode(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsToCanonicalSeed 

func CudnnRNNParamsToCanonicalSeed(value int64) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalSeed2 

func CudnnRNNParamsToCanonicalSeed2(value int64) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsToCanonicalV2Attr 

type CudnnRNNParamsToCanonicalV2Attr func(optionalAttr)

CudnnRNNParamsToCanonicalV2Attr is an optional argument to CudnnRNNParamsToCanonicalV2.

func CudnnRNNParamsToCanonicalV2Direction 

func CudnnRNNParamsToCanonicalV2Direction(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsToCanonicalV2Dropout 

func CudnnRNNParamsToCanonicalV2Dropout(value float32) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2InputMode 

func CudnnRNNParamsToCanonicalV2InputMode(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsToCanonicalV2NumProj 

func CudnnRNNParamsToCanonicalV2NumProj(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2RnnMode 

func CudnnRNNParamsToCanonicalV2RnnMode(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsToCanonicalV2Seed 

func CudnnRNNParamsToCanonicalV2Seed(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2Seed2 

func CudnnRNNParamsToCanonicalV2Seed2(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNV2Attr 

type CudnnRNNV2Attr func(optionalAttr)

CudnnRNNV2Attr is an optional argument to CudnnRNNV2.

func CudnnRNNV2Direction 

func CudnnRNNV2Direction(value string) CudnnRNNV2Attr

CudnnRNNV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNV2Dropout 

func CudnnRNNV2Dropout(value float32) CudnnRNNV2Attr

CudnnRNNV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNV2InputMode 

func CudnnRNNV2InputMode(value string) CudnnRNNV2Attr

CudnnRNNV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNV2IsTraining 

func CudnnRNNV2IsTraining(value bool) CudnnRNNV2Attr

CudnnRNNV2IsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNV2RnnMode 

func CudnnRNNV2RnnMode(value string) CudnnRNNV2Attr

CudnnRNNV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNV2Seed 

func CudnnRNNV2Seed(value int64) CudnnRNNV2Attr

CudnnRNNV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNV2Seed2 

func CudnnRNNV2Seed2(value int64) CudnnRNNV2Attr

CudnnRNNV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNV3Attr 

type CudnnRNNV3Attr func(optionalAttr)

CudnnRNNV3Attr is an optional argument to CudnnRNNV3.

func CudnnRNNV3Direction 

func CudnnRNNV3Direction(value string) CudnnRNNV3Attr

CudnnRNNV3Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNV3Dropout 

func CudnnRNNV3Dropout(value float32) CudnnRNNV3Attr

CudnnRNNV3Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNV3InputMode 

func CudnnRNNV3InputMode(value string) CudnnRNNV3Attr

CudnnRNNV3InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNV3IsTraining 

func CudnnRNNV3IsTraining(value bool) CudnnRNNV3Attr

CudnnRNNV3IsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNV3NumProj 

func CudnnRNNV3NumProj(value int64) CudnnRNNV3Attr

CudnnRNNV3NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNV3RnnMode 

func CudnnRNNV3RnnMode(value string) CudnnRNNV3Attr

CudnnRNNV3RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNV3Seed 

func CudnnRNNV3Seed(value int64) CudnnRNNV3Attr

CudnnRNNV3Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNV3Seed2 

func CudnnRNNV3Seed2(value int64) CudnnRNNV3Attr

CudnnRNNV3Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

func CudnnRNNV3TimeMajor 

func CudnnRNNV3TimeMajor(value bool) CudnnRNNV3Attr

CudnnRNNV3TimeMajor sets the optional time_major attribute to value. If not specified, defaults to true

type CumprodAttr 

type CumprodAttr func(optionalAttr)

CumprodAttr is an optional argument to Cumprod.

func CumprodExclusive 

func CumprodExclusive(value bool) CumprodAttr

CumprodExclusive sets the optional exclusive attribute to value.

value: If `True`, perform exclusive cumprod. If not specified, defaults to false

func CumprodReverse 

func CumprodReverse(value bool) CumprodAttr

CumprodReverse sets the optional reverse attribute to value.

value: A `bool` (default: False). If not specified, defaults to false

type CumsumAttr 

type CumsumAttr func(optionalAttr)

CumsumAttr is an optional argument to Cumsum.

func CumsumExclusive 

func CumsumExclusive(value bool) CumsumAttr

CumsumExclusive sets the optional exclusive attribute to value.

value: If `True`, perform exclusive cumsum. If not specified, defaults to false

func CumsumReverse 

func CumsumReverse(value bool) CumsumAttr

CumsumReverse sets the optional reverse attribute to value.

value: A `bool` (default: False). If not specified, defaults to false

type CumulativeLogsumexpAttr 

type CumulativeLogsumexpAttr func(optionalAttr)

CumulativeLogsumexpAttr is an optional argument to CumulativeLogsumexp.

func CumulativeLogsumexpExclusive 

func CumulativeLogsumexpExclusive(value bool) CumulativeLogsumexpAttr

CumulativeLogsumexpExclusive sets the optional exclusive attribute to value.

value: If `True`, perform exclusive cumulative log-sum-exp. If not specified, defaults to false

func CumulativeLogsumexpReverse 

func CumulativeLogsumexpReverse(value bool) CumulativeLogsumexpAttr

CumulativeLogsumexpReverse sets the optional reverse attribute to value.

value: A `bool` (default: False). If not specified, defaults to false

type DataFormatDimMapAttr 

type DataFormatDimMapAttr func(optionalAttr)

DataFormatDimMapAttr is an optional argument to DataFormatDimMap.

func DataFormatDimMapDstFormat 

func DataFormatDimMapDstFormat(value string) DataFormatDimMapAttr

DataFormatDimMapDstFormat sets the optional dst_format attribute to value.

value: destination data format. If not specified, defaults to "NCHW"

func DataFormatDimMapSrcFormat 

func DataFormatDimMapSrcFormat(value string) DataFormatDimMapAttr

DataFormatDimMapSrcFormat sets the optional src_format attribute to value.

value: source data format. If not specified, defaults to "NHWC"

type DataFormatVecPermuteAttr 

type DataFormatVecPermuteAttr func(optionalAttr)

DataFormatVecPermuteAttr is an optional argument to DataFormatVecPermute.

func DataFormatVecPermuteDstFormat 

func DataFormatVecPermuteDstFormat(value string) DataFormatVecPermuteAttr

DataFormatVecPermuteDstFormat sets the optional dst_format attribute to value.

value: destination data format. If not specified, defaults to "NCHW"

func DataFormatVecPermuteSrcFormat 

func DataFormatVecPermuteSrcFormat(value string) DataFormatVecPermuteAttr

DataFormatVecPermuteSrcFormat sets the optional src_format attribute to value.

value: source data format. If not specified, defaults to "NHWC"

type DataServiceDatasetAttr 

type DataServiceDatasetAttr func(optionalAttr)

DataServiceDatasetAttr is an optional argument to DataServiceDataset.

func DataServiceDatasetCrossTrainerCacheOptions added in v0.2.0 

func DataServiceDatasetCrossTrainerCacheOptions(value string) DataServiceDatasetAttr

DataServiceDatasetCrossTrainerCacheOptions sets the optional cross_trainer_cache_options attribute to value. If not specified, defaults to ""

func DataServiceDatasetDataTransferProtocol 

func DataServiceDatasetDataTransferProtocol(value string) DataServiceDatasetAttr

DataServiceDatasetDataTransferProtocol sets the optional data_transfer_protocol attribute to value. If not specified, defaults to ""

func DataServiceDatasetTargetWorkers 

func DataServiceDatasetTargetWorkers(value string) DataServiceDatasetAttr

DataServiceDatasetTargetWorkers sets the optional target_workers attribute to value. If not specified, defaults to "AUTO"

func DataServiceDatasetTaskRefreshIntervalHintMs 

func DataServiceDatasetTaskRefreshIntervalHintMs(value int64) DataServiceDatasetAttr

DataServiceDatasetTaskRefreshIntervalHintMs sets the optional task_refresh_interval_hint_ms attribute to value. If not specified, defaults to -1

type DataServiceDatasetV2Attr 

type DataServiceDatasetV2Attr func(optionalAttr)

DataServiceDatasetV2Attr is an optional argument to DataServiceDatasetV2.

func DataServiceDatasetV2CrossTrainerCacheOptions added in v0.2.0 

func DataServiceDatasetV2CrossTrainerCacheOptions(value string) DataServiceDatasetV2Attr

DataServiceDatasetV2CrossTrainerCacheOptions sets the optional cross_trainer_cache_options attribute to value. If not specified, defaults to ""

func DataServiceDatasetV2DataTransferProtocol 

func DataServiceDatasetV2DataTransferProtocol(value string) DataServiceDatasetV2Attr

DataServiceDatasetV2DataTransferProtocol sets the optional data_transfer_protocol attribute to value. If not specified, defaults to ""

func DataServiceDatasetV2TargetWorkers 

func DataServiceDatasetV2TargetWorkers(value string) DataServiceDatasetV2Attr

DataServiceDatasetV2TargetWorkers sets the optional target_workers attribute to value. If not specified, defaults to "AUTO"

func DataServiceDatasetV2TaskRefreshIntervalHintMs 

func DataServiceDatasetV2TaskRefreshIntervalHintMs(value int64) DataServiceDatasetV2Attr

DataServiceDatasetV2TaskRefreshIntervalHintMs sets the optional task_refresh_interval_hint_ms attribute to value. If not specified, defaults to -1

type DatasetCardinalityAttr added in v0.5.0 

type DatasetCardinalityAttr func(optionalAttr)

DatasetCardinalityAttr is an optional argument to DatasetCardinality.

func DatasetCardinalityCardinalityOptions added in v0.5.0 

func DatasetCardinalityCardinalityOptions(value string) DatasetCardinalityAttr

DatasetCardinalityCardinalityOptions sets the optional cardinality_options attribute to value. If not specified, defaults to ""

type DatasetToGraphAttr 

type DatasetToGraphAttr func(optionalAttr)

DatasetToGraphAttr is an optional argument to DatasetToGraph.

func DatasetToGraphAllowStateful 

func DatasetToGraphAllowStateful(value bool) DatasetToGraphAttr

DatasetToGraphAllowStateful sets the optional allow_stateful attribute to value. If not specified, defaults to false

func DatasetToGraphStatefulWhitelist 

func DatasetToGraphStatefulWhitelist(value []string) DatasetToGraphAttr

DatasetToGraphStatefulWhitelist sets the optional stateful_whitelist attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func DatasetToGraphStripDeviceAssignment 

func DatasetToGraphStripDeviceAssignment(value bool) DatasetToGraphAttr

DatasetToGraphStripDeviceAssignment sets the optional strip_device_assignment attribute to value. If not specified, defaults to false

type DatasetToGraphV2Attr 

type DatasetToGraphV2Attr func(optionalAttr)

DatasetToGraphV2Attr is an optional argument to DatasetToGraphV2.

func DatasetToGraphV2ExternalStatePolicy 

func DatasetToGraphV2ExternalStatePolicy(value int64) DatasetToGraphV2Attr

DatasetToGraphV2ExternalStatePolicy sets the optional external_state_policy attribute to value. If not specified, defaults to 0

func DatasetToGraphV2StripDeviceAssignment 

func DatasetToGraphV2StripDeviceAssignment(value bool) DatasetToGraphV2Attr

DatasetToGraphV2StripDeviceAssignment sets the optional strip_device_assignment attribute to value. If not specified, defaults to false

type DatasetToSingleElementAttr 

type DatasetToSingleElementAttr func(optionalAttr)

DatasetToSingleElementAttr is an optional argument to DatasetToSingleElement.

func DatasetToSingleElementMetadata 

func DatasetToSingleElementMetadata(value string) DatasetToSingleElementAttr

DatasetToSingleElementMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type DebugIdentityAttr 

type DebugIdentityAttr func(optionalAttr)

DebugIdentityAttr is an optional argument to DebugIdentity.

func DebugIdentityDebugUrls 

func DebugIdentityDebugUrls(value []string) DebugIdentityAttr

DebugIdentityDebugUrls sets the optional debug_urls attribute to value.

value: List of URLs to debug targets, e.g., 

file:///foo/tfdbg_dump, grpc:://localhost:11011

If not specified, defaults to {}

func DebugIdentityDeviceName 

func DebugIdentityDeviceName(value string) DebugIdentityAttr

DebugIdentityDeviceName sets the optional device_name attribute to value.

value: Name of the device on which the tensor resides. If not specified, defaults to ""

func DebugIdentityGatedGrpc 

func DebugIdentityGatedGrpc(value bool) DebugIdentityAttr

DebugIdentityGatedGrpc sets the optional gated_grpc attribute to value.

value: Whether this op will be gated. If any of the debug_urls of this 

debug node is of the grpc:// scheme, when the value of this attribute is set
to True, the data will not actually be sent via the grpc stream unless this
debug op has been enabled at the debug_url. If all of the debug_urls of this
debug node are of the grpc:// scheme and the debug op is enabled at none of
them, the output will be an empty Tensor.

If not specified, defaults to false

func DebugIdentityTensorName 

func DebugIdentityTensorName(value string) DebugIdentityAttr

DebugIdentityTensorName sets the optional tensor_name attribute to value.

value: Name of the input tensor. If not specified, defaults to ""

type DebugIdentityV2Attr 

type DebugIdentityV2Attr func(optionalAttr)

DebugIdentityV2Attr is an optional argument to DebugIdentityV2.

func DebugIdentityV2CircularBufferSize 

func DebugIdentityV2CircularBufferSize(value int64) DebugIdentityV2Attr

DebugIdentityV2CircularBufferSize sets the optional circular_buffer_size attribute to value. If not specified, defaults to 1000

func DebugIdentityV2DebugUrls 

func DebugIdentityV2DebugUrls(value []string) DebugIdentityV2Attr

DebugIdentityV2DebugUrls sets the optional debug_urls attribute to value.

value: List of URLs to debug targets, e.g., file:///foo/tfdbg_dump. If not specified, defaults to {}

func DebugIdentityV2OpName 

func DebugIdentityV2OpName(value string) DebugIdentityV2Attr

DebugIdentityV2OpName sets the optional op_name attribute to value.

value: Optional. Name of the op that the debug op is concerned with. 

Used only for single-tensor trace.

If not specified, defaults to ""

func DebugIdentityV2OutputSlot 

func DebugIdentityV2OutputSlot(value int64) DebugIdentityV2Attr

DebugIdentityV2OutputSlot sets the optional output_slot attribute to value.

value: Optional. Output slot index of the tensor that the debug op 

is concerned with. Used only for single-tensor trace.

If not specified, defaults to -1

func DebugIdentityV2TensorDebugMode 

func DebugIdentityV2TensorDebugMode(value int64) DebugIdentityV2Attr

DebugIdentityV2TensorDebugMode sets the optional tensor_debug_mode attribute to value.

value: TensorDebugMode enum value. See debug_event.proto for details. If not specified, defaults to -1

func DebugIdentityV2TfdbgContextId 

func DebugIdentityV2TfdbgContextId(value string) DebugIdentityV2Attr

DebugIdentityV2TfdbgContextId sets the optional tfdbg_context_id attribute to value.

value: A tfdbg-generated ID for the context that the op belongs to, 

e.g., a concrete compiled tf.function.

If not specified, defaults to ""

func DebugIdentityV2TfdbgRunId 

func DebugIdentityV2TfdbgRunId(value string) DebugIdentityV2Attr

DebugIdentityV2TfdbgRunId sets the optional tfdbg_run_id attribute to value. If not specified, defaults to ""

type DebugIdentityV3Attr added in v0.5.0 

type DebugIdentityV3Attr func(optionalAttr)

DebugIdentityV3Attr is an optional argument to DebugIdentityV3.

func DebugIdentityV3DebugUrls added in v0.5.0 

func DebugIdentityV3DebugUrls(value []string) DebugIdentityV3Attr

DebugIdentityV3DebugUrls sets the optional debug_urls attribute to value.

value: List of URLs to debug targets, e.g., 

file:///foo/tfdbg_dump, grpc:://localhost:11011

If not specified, defaults to {}

func DebugIdentityV3DeviceName added in v0.5.0 

func DebugIdentityV3DeviceName(value string) DebugIdentityV3Attr

DebugIdentityV3DeviceName sets the optional device_name attribute to value.

value: Name of the device on which the tensor resides. If not specified, defaults to ""

func DebugIdentityV3GatedGrpc added in v0.5.0 

func DebugIdentityV3GatedGrpc(value bool) DebugIdentityV3Attr

DebugIdentityV3GatedGrpc sets the optional gated_grpc attribute to value.

value: Whether this op will be gated. If any of the debug_urls of this 

debug node is of the grpc:// scheme, when the value of this attribute is set
to True, the data will not actually be sent via the grpc stream unless this
debug op has been enabled at the debug_url. If all of the debug_urls of this
debug node are of the grpc:// scheme and the debug op is enabled at none of
them, the output will be an empty Tensor.

If not specified, defaults to false

func DebugIdentityV3IoIndex added in v0.5.0 

func DebugIdentityV3IoIndex(value int64) DebugIdentityV3Attr

DebugIdentityV3IoIndex sets the optional io_index attribute to value.

value: The index of which the tensor is an input or output of the node. If not specified, defaults to -1

func DebugIdentityV3IoOfNode added in v0.5.0 

func DebugIdentityV3IoOfNode(value string) DebugIdentityV3Attr

DebugIdentityV3IoOfNode sets the optional io_of_node attribute to value.

value: Name of the node of which the tensor is an input or output. If not specified, defaults to ""

func DebugIdentityV3IsInput added in v0.5.0 

func DebugIdentityV3IsInput(value bool) DebugIdentityV3Attr

DebugIdentityV3IsInput sets the optional is_input attribute to value.

value: If true, the tensor is an input of the node; otherwise the output. If not specified, defaults to false

func DebugIdentityV3TensorName added in v0.5.0 

func DebugIdentityV3TensorName(value string) DebugIdentityV3Attr

DebugIdentityV3TensorName sets the optional tensor_name attribute to value.

value: Name of the input tensor. If not specified, defaults to ""

type DebugNanCountAttr 

type DebugNanCountAttr func(optionalAttr)

DebugNanCountAttr is an optional argument to DebugNanCount.

func DebugNanCountDebugUrls 

func DebugNanCountDebugUrls(value []string) DebugNanCountAttr

DebugNanCountDebugUrls sets the optional debug_urls attribute to value.

value: List of URLs to debug targets, e.g., 

file:///foo/tfdbg_dump, grpc:://localhost:11011.

If not specified, defaults to {}

func DebugNanCountDeviceName 

func DebugNanCountDeviceName(value string) DebugNanCountAttr

DebugNanCountDeviceName sets the optional device_name attribute to value. If not specified, defaults to ""

func DebugNanCountGatedGrpc 

func DebugNanCountGatedGrpc(value bool) DebugNanCountAttr

DebugNanCountGatedGrpc sets the optional gated_grpc attribute to value.

value: Whether this op will be gated. If any of the debug_urls of this 

debug node is of the grpc:// scheme, when the value of this attribute is set
to True, the data will not actually be sent via the grpc stream unless this
debug op has been enabled at the debug_url. If all of the debug_urls of this
debug node are of the grpc:// scheme and the debug op is enabled at none of
them, the output will be an empty Tensor.

If not specified, defaults to false

func DebugNanCountTensorName 

func DebugNanCountTensorName(value string) DebugNanCountAttr

DebugNanCountTensorName sets the optional tensor_name attribute to value.

value: Name of the input tensor. If not specified, defaults to ""

type DebugNumericSummaryAttr 

type DebugNumericSummaryAttr func(optionalAttr)

DebugNumericSummaryAttr is an optional argument to DebugNumericSummary.

func DebugNumericSummaryDebugUrls 

func DebugNumericSummaryDebugUrls(value []string) DebugNumericSummaryAttr

DebugNumericSummaryDebugUrls sets the optional debug_urls attribute to value.

value: List of URLs to debug targets, e.g., 

file:///foo/tfdbg_dump, grpc:://localhost:11011.

If not specified, defaults to {}

func DebugNumericSummaryDeviceName 

func DebugNumericSummaryDeviceName(value string) DebugNumericSummaryAttr

DebugNumericSummaryDeviceName sets the optional device_name attribute to value. If not specified, defaults to ""

func DebugNumericSummaryGatedGrpc 

func DebugNumericSummaryGatedGrpc(value bool) DebugNumericSummaryAttr

DebugNumericSummaryGatedGrpc sets the optional gated_grpc attribute to value.

value: Whether this op will be gated. If any of the debug_urls of this 

debug node is of the grpc:// scheme, when the value of this attribute is set
to True, the data will not actually be sent via the grpc stream unless this
debug op has been enabled at the debug_url. If all of the debug_urls of this
debug node are of the grpc:// scheme and the debug op is enabled at none of
them, the output will be an empty Tensor.

If not specified, defaults to false

func DebugNumericSummaryLowerBound 

func DebugNumericSummaryLowerBound(value float32) DebugNumericSummaryAttr

DebugNumericSummaryLowerBound sets the optional lower_bound attribute to value.

value: (float) The lower bound <= which values will be included in the 

generalized -inf count. Default: -inf.

If not specified, defaults to -inf

func DebugNumericSummaryMuteIfHealthy 

func DebugNumericSummaryMuteIfHealthy(value bool) DebugNumericSummaryAttr

DebugNumericSummaryMuteIfHealthy sets the optional mute_if_healthy attribute to value.

value: (bool) Do not send data to the debug URLs unless at least one 

of elements [2], [3] and [7] (i.e., the nan count and the generalized -inf and
inf counts) is non-zero.

If not specified, defaults to false

func DebugNumericSummaryTensorName 

func DebugNumericSummaryTensorName(value string) DebugNumericSummaryAttr

DebugNumericSummaryTensorName sets the optional tensor_name attribute to value.

value: Name of the input tensor. If not specified, defaults to ""

func DebugNumericSummaryUpperBound 

func DebugNumericSummaryUpperBound(value float32) DebugNumericSummaryAttr

DebugNumericSummaryUpperBound sets the optional upper_bound attribute to value.

value: (float) The upper bound >= which values will be included in the 

generalized +inf count. Default: +inf.

If not specified, defaults to inf

type DebugNumericSummaryV2Attr 

type DebugNumericSummaryV2Attr func(optionalAttr)

DebugNumericSummaryV2Attr is an optional argument to DebugNumericSummaryV2.

func DebugNumericSummaryV2OutputDtype 

func DebugNumericSummaryV2OutputDtype(value tf.DataType) DebugNumericSummaryV2Attr

DebugNumericSummaryV2OutputDtype sets the optional output_dtype attribute to value.

value: Optional. The type of the output. Can be float32 or float64 (default: float32). If not specified, defaults to DT_FLOAT

func DebugNumericSummaryV2TensorDebugMode 

func DebugNumericSummaryV2TensorDebugMode(value int64) DebugNumericSummaryV2Attr

DebugNumericSummaryV2TensorDebugMode sets the optional tensor_debug_mode attribute to value.

value: Tensor debug mode: the mode in which the input tensor is summarized 

by the op. See the TensorDebugMode enum in
tensorflow/core/protobuf/debug_event.proto for details.

Supported values: 

2 (CURT_HEALTH): Output a float32/64 tensor of shape [2]. The 1st
element is the tensor_id, if provided, and -1 otherwise. The 2nd
element is a bit which is set to 1 if the input tensor has an
infinity or nan value, or zero otherwise.

3 (CONCISE_HEALTH): Output a float32/64 tensor of shape [5]. The 1st
element is the tensor_id, if provided, and -1 otherwise. The
remaining four slots are the total number of elements, -infs,
+infs, and nans in the input tensor respectively.

4 (FULL_HEALTH): Output a float32/64 tensor of shape [11]. The 1st
element is the tensor_id, if provided, and -1 otherwise. The 2nd
element is the device_id, if provided, and -1 otherwise. The 3rd
element holds the datatype value of the input tensor as according
to the enumerated type in tensorflow/core/framework/types.proto.
The remaining elements hold the total number of elements, -infs,
+infs, nans, negative finite numbers, zeros, and positive finite
numbers in the input tensor respectively.

5 (SHAPE): Output a float32/64 tensor of shape [10]. The 1st
element is the tensor_id, if provided, and -1 otherwise. The 2nd
element holds the datatype value of the input tensor as according
to the enumerated type in tensorflow/core/framework/types.proto.
The 3rd element holds the rank of the tensor. The 4th element holds
the number of elements within the tensor. Finally the remaining 6
elements hold the shape of the tensor. If the rank of the tensor
is lower than 6, the shape is right padded with zeros. If the rank
is greater than 6, the head of the shape is truncated.

6 (FULL_NUMERICS): Output a float32/64 tensor of shape [22]. The 1st
element is the tensor_id, if provided, and -1 otherwise. The 2nd
element is the device_id, if provided, and -1 otherwise. The 3rd
element holds the datatype value of the input tensor as according
to the enumerated type in tensorflow/core/framework/types.proto.
The 4th element holds the rank of the tensor. The 5th to 11th
elements hold the shape of the tensor. If the rank of the tensor
is lower than 6, the shape is right padded with zeros. If the rank
is greater than 6, the head of the shape is truncated. The 12th to
18th elements hold the number of elements, -infs, +infs, nans,
denormal floats, negative finite numbers, zeros, and positive
finite numbers in the input tensor respectively. The final four
elements hold the min value, max value, mean, and variance of the
input tensor.

8 (REDUCE_INF_NAN_THREE_SLOTS): Output a float32/64 tensor of shape
[3]. The 1st element is -inf if any elements of the input tensor
is -inf, or zero otherwise. The 2nd element is +inf if any elements
of the input tensor is +inf, or zero otherwise.  The 3rd element is
nan if any element of the input tensor is nan, or zero otherwise.

If not specified, defaults to -1

func DebugNumericSummaryV2TensorId 

func DebugNumericSummaryV2TensorId(value int64) DebugNumericSummaryV2Attr

DebugNumericSummaryV2TensorId sets the optional tensor_id attribute to value.

value: Optional. An integer identifier for the tensor being summarized by this op. If not specified, defaults to -1

type DecodeAndCropJpegAttr 

type DecodeAndCropJpegAttr func(optionalAttr)

DecodeAndCropJpegAttr is an optional argument to DecodeAndCropJpeg.

func DecodeAndCropJpegAcceptableFraction 

func DecodeAndCropJpegAcceptableFraction(value float32) DecodeAndCropJpegAttr

DecodeAndCropJpegAcceptableFraction sets the optional acceptable_fraction attribute to value.

value: The minimum required fraction of lines before a truncated input is accepted. If not specified, defaults to 1

func DecodeAndCropJpegChannels 

func DecodeAndCropJpegChannels(value int64) DecodeAndCropJpegAttr

DecodeAndCropJpegChannels sets the optional channels attribute to value.

value: Number of color channels for the decoded image. If not specified, defaults to 0

func DecodeAndCropJpegDctMethod 

func DecodeAndCropJpegDctMethod(value string) DecodeAndCropJpegAttr

DecodeAndCropJpegDctMethod sets the optional dct_method attribute to value.

value: string specifying a hint about the algorithm used for decompression. Defaults to "" which maps to a system-specific default. Currently valid values are ["INTEGER_FAST", "INTEGER_ACCURATE"]. The hint may be ignored (e.g., the internal jpeg library changes to a version that does not have that specific option.) If not specified, defaults to ""

func DecodeAndCropJpegFancyUpscaling 

func DecodeAndCropJpegFancyUpscaling(value bool) DecodeAndCropJpegAttr

DecodeAndCropJpegFancyUpscaling sets the optional fancy_upscaling attribute to value.

value: If true use a slower but nicer upscaling of the chroma planes (yuv420/422 only). If not specified, defaults to true

func DecodeAndCropJpegRatio 

func DecodeAndCropJpegRatio(value int64) DecodeAndCropJpegAttr

DecodeAndCropJpegRatio sets the optional ratio attribute to value.

value: Downscaling ratio. If not specified, defaults to 1

func DecodeAndCropJpegTryRecoverTruncated 

func DecodeAndCropJpegTryRecoverTruncated(value bool) DecodeAndCropJpegAttr

DecodeAndCropJpegTryRecoverTruncated sets the optional try_recover_truncated attribute to value.

value: If true try to recover an image from truncated input. If not specified, defaults to false

type DecodeBmpAttr 

type DecodeBmpAttr func(optionalAttr)

DecodeBmpAttr is an optional argument to DecodeBmp.

func DecodeBmpChannels 

func DecodeBmpChannels(value int64) DecodeBmpAttr

DecodeBmpChannels sets the optional channels attribute to value. If not specified, defaults to 0

type DecodeCSVAttr 

type DecodeCSVAttr func(optionalAttr)

DecodeCSVAttr is an optional argument to DecodeCSV.

func DecodeCSVFieldDelim 

func DecodeCSVFieldDelim(value string) DecodeCSVAttr

DecodeCSVFieldDelim sets the optional field_delim attribute to value.

value: char delimiter to separate fields in a record. If not specified, defaults to ","

func DecodeCSVNaValue 

func DecodeCSVNaValue(value string) DecodeCSVAttr

DecodeCSVNaValue sets the optional na_value attribute to value.

value: Additional string to recognize as NA/NaN. If not specified, defaults to ""

func DecodeCSVSelectCols 

func DecodeCSVSelectCols(value []int64) DecodeCSVAttr

DecodeCSVSelectCols sets the optional select_cols attribute to value. If not specified, defaults to {}

func DecodeCSVUseQuoteDelim 

func DecodeCSVUseQuoteDelim(value bool) DecodeCSVAttr

DecodeCSVUseQuoteDelim sets the optional use_quote_delim attribute to value.

value: If false, treats double quotation marks as regular characters inside of the string fields (ignoring RFC 4180, Section 2, Bullet 5). If not specified, defaults to true

type DecodeCompressedAttr 

type DecodeCompressedAttr func(optionalAttr)

DecodeCompressedAttr is an optional argument to DecodeCompressed.

func DecodeCompressedCompressionType 

func DecodeCompressedCompressionType(value string) DecodeCompressedAttr

DecodeCompressedCompressionType sets the optional compression_type attribute to value.

value: A scalar containing either (i) the empty string (no compression), (ii) "ZLIB", or (iii) "GZIP". If not specified, defaults to ""

type DecodeImageAttr 

type DecodeImageAttr func(optionalAttr)

DecodeImageAttr is an optional argument to DecodeImage.

func DecodeImageChannels 

func DecodeImageChannels(value int64) DecodeImageAttr

DecodeImageChannels sets the optional channels attribute to value.

value: Number of color channels for the decoded image. If not specified, defaults to 0

func DecodeImageDtype 

func DecodeImageDtype(value tf.DataType) DecodeImageAttr

DecodeImageDtype sets the optional dtype attribute to value.

value: The desired DType of the returned Tensor. If not specified, defaults to DT_UINT8

func DecodeImageExpandAnimations 

func DecodeImageExpandAnimations(value bool) DecodeImageAttr

DecodeImageExpandAnimations sets the optional expand_animations attribute to value.

value: Controls the output shape of the returned op. If True, the returned op will produce a 3-D tensor for PNG, JPEG, and BMP files; and a 4-D tensor for all GIFs, whether animated or not. If, False, the returned op will produce a 3-D tensor for all file types and will truncate animated GIFs to the first frame. If not specified, defaults to true

type DecodeJpegAttr 

type DecodeJpegAttr func(optionalAttr)

DecodeJpegAttr is an optional argument to DecodeJpeg.

func DecodeJpegAcceptableFraction 

func DecodeJpegAcceptableFraction(value float32) DecodeJpegAttr

DecodeJpegAcceptableFraction sets the optional acceptable_fraction attribute to value.

value: The minimum required fraction of lines before a truncated input is accepted. If not specified, defaults to 1

func DecodeJpegChannels 

func DecodeJpegChannels(value int64) DecodeJpegAttr

DecodeJpegChannels sets the optional channels attribute to value.

value: Number of color channels for the decoded image. If not specified, defaults to 0

func DecodeJpegDctMethod 

func DecodeJpegDctMethod(value string) DecodeJpegAttr

DecodeJpegDctMethod sets the optional dct_method attribute to value.

value: string specifying a hint about the algorithm used for decompression. Defaults to "" which maps to a system-specific default. Currently valid values are ["INTEGER_FAST", "INTEGER_ACCURATE"]. The hint may be ignored (e.g., the internal jpeg library changes to a version that does not have that specific option.) If not specified, defaults to ""

func DecodeJpegFancyUpscaling 

func DecodeJpegFancyUpscaling(value bool) DecodeJpegAttr

DecodeJpegFancyUpscaling sets the optional fancy_upscaling attribute to value.

value: If true use a slower but nicer upscaling of the chroma planes (yuv420/422 only). If not specified, defaults to true

func DecodeJpegRatio 

func DecodeJpegRatio(value int64) DecodeJpegAttr

DecodeJpegRatio sets the optional ratio attribute to value.

value: Downscaling ratio. If not specified, defaults to 1

func DecodeJpegTryRecoverTruncated 

func DecodeJpegTryRecoverTruncated(value bool) DecodeJpegAttr

DecodeJpegTryRecoverTruncated sets the optional try_recover_truncated attribute to value.

value: If true try to recover an image from truncated input. If not specified, defaults to false

type DecodePaddedRawAttr 

type DecodePaddedRawAttr func(optionalAttr)

DecodePaddedRawAttr is an optional argument to DecodePaddedRaw.

func DecodePaddedRawLittleEndian 

func DecodePaddedRawLittleEndian(value bool) DecodePaddedRawAttr

DecodePaddedRawLittleEndian sets the optional little_endian attribute to value.

value: Whether the input `input_bytes` is in little-endian order. Ignored for `out_type` values that are stored in a single byte, like `uint8` If not specified, defaults to true

type DecodePngAttr 

type DecodePngAttr func(optionalAttr)

DecodePngAttr is an optional argument to DecodePng.

func DecodePngChannels 

func DecodePngChannels(value int64) DecodePngAttr

DecodePngChannels sets the optional channels attribute to value.

value: Number of color channels for the decoded image. If not specified, defaults to 0

func DecodePngDtype 

func DecodePngDtype(value tf.DataType) DecodePngAttr

DecodePngDtype sets the optional dtype attribute to value. If not specified, defaults to DT_UINT8

type DecodeProtoV2Attr 

type DecodeProtoV2Attr func(optionalAttr)

DecodeProtoV2Attr is an optional argument to DecodeProtoV2.

func DecodeProtoV2DescriptorSource 

func DecodeProtoV2DescriptorSource(value string) DecodeProtoV2Attr

DecodeProtoV2DescriptorSource sets the optional descriptor_source attribute to value.

value: Either the special value `local://` or a path to a file containing a serialized `FileDescriptorSet`. If not specified, defaults to "local://"

func DecodeProtoV2MessageFormat 

func DecodeProtoV2MessageFormat(value string) DecodeProtoV2Attr

DecodeProtoV2MessageFormat sets the optional message_format attribute to value.

value: Either `binary` or `text`. If not specified, defaults to "binary"

func DecodeProtoV2Sanitize 

func DecodeProtoV2Sanitize(value bool) DecodeProtoV2Attr

DecodeProtoV2Sanitize sets the optional sanitize attribute to value.

value: Whether to sanitize the result or not. If not specified, defaults to false

type DecodeRawAttr 

type DecodeRawAttr func(optionalAttr)

DecodeRawAttr is an optional argument to DecodeRaw.

func DecodeRawLittleEndian 

func DecodeRawLittleEndian(value bool) DecodeRawAttr

DecodeRawLittleEndian sets the optional little_endian attribute to value.

value: Whether the input `bytes` are in little-endian order. Ignored for `out_type` values that are stored in a single byte like `uint8`. If not specified, defaults to true

type DecodeWavAttr 

type DecodeWavAttr func(optionalAttr)

DecodeWavAttr is an optional argument to DecodeWav.

func DecodeWavDesiredChannels 

func DecodeWavDesiredChannels(value int64) DecodeWavAttr

DecodeWavDesiredChannels sets the optional desired_channels attribute to value.

value: Number of sample channels wanted. If not specified, defaults to -1

func DecodeWavDesiredSamples 

func DecodeWavDesiredSamples(value int64) DecodeWavAttr

DecodeWavDesiredSamples sets the optional desired_samples attribute to value.

value: Length of audio requested. If not specified, defaults to -1

type DenseBincountAttr 

type DenseBincountAttr func(optionalAttr)

DenseBincountAttr is an optional argument to DenseBincount.

func DenseBincountBinaryOutput 

func DenseBincountBinaryOutput(value bool) DenseBincountAttr

DenseBincountBinaryOutput sets the optional binary_output attribute to value.

value: bool; Whether the kernel should count the appearance or number of occurrences. If not specified, defaults to false

type DenseCountSparseOutputAttr 

type DenseCountSparseOutputAttr func(optionalAttr)

DenseCountSparseOutputAttr is an optional argument to DenseCountSparseOutput.

func DenseCountSparseOutputMaxlength 

func DenseCountSparseOutputMaxlength(value int64) DenseCountSparseOutputAttr

DenseCountSparseOutputMaxlength sets the optional maxlength attribute to value.

value: Maximum value to count. Can be set to -1 for no maximum. If not specified, defaults to -1

REQUIRES: value >= -1

func DenseCountSparseOutputMinlength 

func DenseCountSparseOutputMinlength(value int64) DenseCountSparseOutputAttr

DenseCountSparseOutputMinlength sets the optional minlength attribute to value.

value: Minimum value to count. Can be set to -1 for no minimum. If not specified, defaults to -1

REQUIRES: value >= -1

type DenseToDenseSetOperationAttr 

type DenseToDenseSetOperationAttr func(optionalAttr)

DenseToDenseSetOperationAttr is an optional argument to DenseToDenseSetOperation.

func DenseToDenseSetOperationValidateIndices 

func DenseToDenseSetOperationValidateIndices(value bool) DenseToDenseSetOperationAttr

DenseToDenseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type DenseToSparseSetOperationAttr 

type DenseToSparseSetOperationAttr func(optionalAttr)

DenseToSparseSetOperationAttr is an optional argument to DenseToSparseSetOperation.

func DenseToSparseSetOperationValidateIndices 

func DenseToSparseSetOperationValidateIndices(value bool) DenseToSparseSetOperationAttr

DenseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type DepthToSpaceAttr 

type DepthToSpaceAttr func(optionalAttr)

DepthToSpaceAttr is an optional argument to DepthToSpace.

func DepthToSpaceDataFormat 

func DepthToSpaceDataFormat(value string) DepthToSpaceAttr

DepthToSpaceDataFormat sets the optional data_format attribute to value. If not specified, defaults to "NHWC"

type DepthwiseConv2dNativeAttr 

type DepthwiseConv2dNativeAttr func(optionalAttr)

DepthwiseConv2dNativeAttr is an optional argument to DepthwiseConv2dNative.

func DepthwiseConv2dNativeDataFormat 

func DepthwiseConv2dNativeDataFormat(value string) DepthwiseConv2dNativeAttr

DepthwiseConv2dNativeDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, height, width, channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, channels, height, width].

If not specified, defaults to "NHWC"

func DepthwiseConv2dNativeDilations 

func DepthwiseConv2dNativeDilations(value []int64) DepthwiseConv2dNativeAttr

DepthwiseConv2dNativeDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func DepthwiseConv2dNativeExplicitPaddings 

func DepthwiseConv2dNativeExplicitPaddings(value []int64) DepthwiseConv2dNativeAttr

DepthwiseConv2dNativeExplicitPaddings sets the optional explicit_paddings attribute to value. If not specified, defaults to {}

type DepthwiseConv2dNativeBackpropFilterAttr 

type DepthwiseConv2dNativeBackpropFilterAttr func(optionalAttr)

DepthwiseConv2dNativeBackpropFilterAttr is an optional argument to DepthwiseConv2dNativeBackpropFilter.

func DepthwiseConv2dNativeBackpropFilterDataFormat 

func DepthwiseConv2dNativeBackpropFilterDataFormat(value string) DepthwiseConv2dNativeBackpropFilterAttr

DepthwiseConv2dNativeBackpropFilterDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, height, width, channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, channels, height, width].

If not specified, defaults to "NHWC"

func DepthwiseConv2dNativeBackpropFilterDilations 

func DepthwiseConv2dNativeBackpropFilterDilations(value []int64) DepthwiseConv2dNativeBackpropFilterAttr

DepthwiseConv2dNativeBackpropFilterDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func DepthwiseConv2dNativeBackpropFilterExplicitPaddings 

func DepthwiseConv2dNativeBackpropFilterExplicitPaddings(value []int64) DepthwiseConv2dNativeBackpropFilterAttr

DepthwiseConv2dNativeBackpropFilterExplicitPaddings sets the optional explicit_paddings attribute to value. If not specified, defaults to {}

type DepthwiseConv2dNativeBackpropInputAttr 

type DepthwiseConv2dNativeBackpropInputAttr func(optionalAttr)

DepthwiseConv2dNativeBackpropInputAttr is an optional argument to DepthwiseConv2dNativeBackpropInput.

func DepthwiseConv2dNativeBackpropInputDataFormat 

func DepthwiseConv2dNativeBackpropInputDataFormat(value string) DepthwiseConv2dNativeBackpropInputAttr

DepthwiseConv2dNativeBackpropInputDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, height, width, channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, channels, height, width].

If not specified, defaults to "NHWC"

func DepthwiseConv2dNativeBackpropInputDilations 

func DepthwiseConv2dNativeBackpropInputDilations(value []int64) DepthwiseConv2dNativeBackpropInputAttr

DepthwiseConv2dNativeBackpropInputDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func DepthwiseConv2dNativeBackpropInputExplicitPaddings 

func DepthwiseConv2dNativeBackpropInputExplicitPaddings(value []int64) DepthwiseConv2dNativeBackpropInputAttr

DepthwiseConv2dNativeBackpropInputExplicitPaddings sets the optional explicit_paddings attribute to value. If not specified, defaults to {}

type DequantizeAttr 

type DequantizeAttr func(optionalAttr)

DequantizeAttr is an optional argument to Dequantize.

func DequantizeAxis 

func DequantizeAxis(value int64) DequantizeAttr

DequantizeAxis sets the optional axis attribute to value. If not specified, defaults to -1

func DequantizeDtype 

func DequantizeDtype(value tf.DataType) DequantizeAttr

DequantizeDtype sets the optional dtype attribute to value.

value: Type of the output tensor. Currently Dequantize supports float and bfloat16. If 'dtype' is 'bfloat16', it only supports 'MIN_COMBINED' mode. If not specified, defaults to DT_FLOAT

func DequantizeMode 

func DequantizeMode(value string) DequantizeAttr

DequantizeMode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"

func DequantizeNarrowRange 

func DequantizeNarrowRange(value bool) DequantizeAttr

DequantizeNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

type DestroyResourceOpAttr 

type DestroyResourceOpAttr func(optionalAttr)

DestroyResourceOpAttr is an optional argument to DestroyResourceOp.

func DestroyResourceOpIgnoreLookupError 

func DestroyResourceOpIgnoreLookupError(value bool) DestroyResourceOpAttr

DestroyResourceOpIgnoreLookupError sets the optional ignore_lookup_error attribute to value.

value: whether to ignore the error when the resource doesn't exist. If not specified, defaults to true

type DirectedInterleaveDatasetAttr 

type DirectedInterleaveDatasetAttr func(optionalAttr)

DirectedInterleaveDatasetAttr is an optional argument to DirectedInterleaveDataset.

func DirectedInterleaveDatasetStopOnEmptyDataset 

func DirectedInterleaveDatasetStopOnEmptyDataset(value bool) DirectedInterleaveDatasetAttr

DirectedInterleaveDatasetStopOnEmptyDataset sets the optional stop_on_empty_dataset attribute to value. If not specified, defaults to false

type DynamicEnqueueTPUEmbeddingArbitraryTensorBatchAttr 

type DynamicEnqueueTPUEmbeddingArbitraryTensorBatchAttr func(optionalAttr)

DynamicEnqueueTPUEmbeddingArbitraryTensorBatchAttr is an optional argument to DynamicEnqueueTPUEmbeddingArbitraryTensorBatch.

func DynamicEnqueueTPUEmbeddingArbitraryTensorBatchCombiners 

func DynamicEnqueueTPUEmbeddingArbitraryTensorBatchCombiners(value []string) DynamicEnqueueTPUEmbeddingArbitraryTensorBatchAttr

DynamicEnqueueTPUEmbeddingArbitraryTensorBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

type EagerPyFuncAttr 

type EagerPyFuncAttr func(optionalAttr)

EagerPyFuncAttr is an optional argument to EagerPyFunc.

func EagerPyFuncIsAsync 

func EagerPyFuncIsAsync(value bool) EagerPyFuncAttr

EagerPyFuncIsAsync sets the optional is_async attribute to value. If not specified, defaults to false

type EditDistanceAttr 

type EditDistanceAttr func(optionalAttr)

EditDistanceAttr is an optional argument to EditDistance.

func EditDistanceNormalize 

func EditDistanceNormalize(value bool) EditDistanceAttr

EditDistanceNormalize sets the optional normalize attribute to value.

value: boolean (if true, edit distances are normalized by length of truth).

The output is: If not specified, defaults to true

type EigAttr 

type EigAttr func(optionalAttr)

EigAttr is an optional argument to Eig.

func EigComputeV 

func EigComputeV(value bool) EigAttr

EigComputeV sets the optional compute_v attribute to value.

value: If `True` then eigenvectors will be computed and returned in `v`. Otherwise, only the eigenvalues will be computed. If not specified, defaults to true

type EmptyAttr 

type EmptyAttr func(optionalAttr)

EmptyAttr is an optional argument to Empty.

func EmptyInit 

func EmptyInit(value bool) EmptyAttr

EmptyInit sets the optional init attribute to value.

value: If True, initialize the returned tensor with the default value of dtype. Otherwise, the implementation is free not to initializethe tensor's content. If not specified, defaults to false

type EncodeBase64Attr 

type EncodeBase64Attr func(optionalAttr)

EncodeBase64Attr is an optional argument to EncodeBase64.

func EncodeBase64Pad 

func EncodeBase64Pad(value bool) EncodeBase64Attr

EncodeBase64Pad sets the optional pad attribute to value.

value: Bool whether padding is applied at the ends. If not specified, defaults to false

type EncodeJpegAttr 

type EncodeJpegAttr func(optionalAttr)

EncodeJpegAttr is an optional argument to EncodeJpeg.

func EncodeJpegChromaDownsampling 

func EncodeJpegChromaDownsampling(value bool) EncodeJpegAttr

EncodeJpegChromaDownsampling sets the optional chroma_downsampling attribute to value.

value: See http://en.wikipedia.org/wiki/Chroma_subsampling. If not specified, defaults to true

func EncodeJpegDensityUnit 

func EncodeJpegDensityUnit(value string) EncodeJpegAttr

EncodeJpegDensityUnit sets the optional density_unit attribute to value.

value: Unit used to specify `x_density` and `y_density`: pixels per inch (`'in'`) or centimeter (`'cm'`). If not specified, defaults to "in"

func EncodeJpegFormat 

func EncodeJpegFormat(value string) EncodeJpegAttr

EncodeJpegFormat sets the optional format attribute to value.

value: Per pixel image format. If not specified, defaults to ""

func EncodeJpegOptimizeSize 

func EncodeJpegOptimizeSize(value bool) EncodeJpegAttr

EncodeJpegOptimizeSize sets the optional optimize_size attribute to value.

value: If True, spend CPU/RAM to reduce size with no quality change. If not specified, defaults to false

func EncodeJpegProgressive 

func EncodeJpegProgressive(value bool) EncodeJpegAttr

EncodeJpegProgressive sets the optional progressive attribute to value.

value: If True, create a JPEG that loads progressively (coarse to fine). If not specified, defaults to false

func EncodeJpegQuality 

func EncodeJpegQuality(value int64) EncodeJpegAttr

EncodeJpegQuality sets the optional quality attribute to value.

value: Quality of the compression from 0 to 100 (higher is better and slower). If not specified, defaults to 95

func EncodeJpegXDensity 

func EncodeJpegXDensity(value int64) EncodeJpegAttr

EncodeJpegXDensity sets the optional x_density attribute to value.

value: Horizontal pixels per density unit. If not specified, defaults to 300

func EncodeJpegXmpMetadata 

func EncodeJpegXmpMetadata(value string) EncodeJpegAttr

EncodeJpegXmpMetadata sets the optional xmp_metadata attribute to value.

value: If not empty, embed this XMP metadata in the image header. If not specified, defaults to ""

func EncodeJpegYDensity 

func EncodeJpegYDensity(value int64) EncodeJpegAttr

EncodeJpegYDensity sets the optional y_density attribute to value.

value: Vertical pixels per density unit. If not specified, defaults to 300

type EncodePngAttr 

type EncodePngAttr func(optionalAttr)

EncodePngAttr is an optional argument to EncodePng.

func EncodePngCompression 

func EncodePngCompression(value int64) EncodePngAttr

EncodePngCompression sets the optional compression attribute to value.

value: Compression level. If not specified, defaults to -1

type EncodeProtoAttr 

type EncodeProtoAttr func(optionalAttr)

EncodeProtoAttr is an optional argument to EncodeProto.

func EncodeProtoDescriptorSource 

func EncodeProtoDescriptorSource(value string) EncodeProtoAttr

EncodeProtoDescriptorSource sets the optional descriptor_source attribute to value. If not specified, defaults to "local://"

type EnqueueTPUEmbeddingArbitraryTensorBatchAttr 

type EnqueueTPUEmbeddingArbitraryTensorBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingArbitraryTensorBatchAttr is an optional argument to EnqueueTPUEmbeddingArbitraryTensorBatch.

func EnqueueTPUEmbeddingArbitraryTensorBatchCombiners 

func EnqueueTPUEmbeddingArbitraryTensorBatchCombiners(value []string) EnqueueTPUEmbeddingArbitraryTensorBatchAttr

EnqueueTPUEmbeddingArbitraryTensorBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

func EnqueueTPUEmbeddingArbitraryTensorBatchDeviceOrdinal 

func EnqueueTPUEmbeddingArbitraryTensorBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingArbitraryTensorBatchAttr

EnqueueTPUEmbeddingArbitraryTensorBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. Should be >= 0 and less than the number of TPU cores in the task on which the node is placed. If not specified, defaults to -1

type EnqueueTPUEmbeddingBatchAttr 

type EnqueueTPUEmbeddingBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingBatchAttr is an optional argument to EnqueueTPUEmbeddingBatch.

func EnqueueTPUEmbeddingBatchCombiners 

func EnqueueTPUEmbeddingBatchCombiners(value []string) EnqueueTPUEmbeddingBatchAttr

EnqueueTPUEmbeddingBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

func EnqueueTPUEmbeddingBatchDeviceOrdinal 

func EnqueueTPUEmbeddingBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingBatchAttr

EnqueueTPUEmbeddingBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. If not specified, defaults to -1

type EnqueueTPUEmbeddingIntegerBatchAttr 

type EnqueueTPUEmbeddingIntegerBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingIntegerBatchAttr is an optional argument to EnqueueTPUEmbeddingIntegerBatch.

func EnqueueTPUEmbeddingIntegerBatchDeviceOrdinal 

func EnqueueTPUEmbeddingIntegerBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingIntegerBatchAttr

EnqueueTPUEmbeddingIntegerBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. Should be >= 0 and less than the number of TPU cores in the task on which the node is placed. If not specified, defaults to -1

type EnqueueTPUEmbeddingRaggedTensorBatchAttr 

type EnqueueTPUEmbeddingRaggedTensorBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingRaggedTensorBatchAttr is an optional argument to EnqueueTPUEmbeddingRaggedTensorBatch.

func EnqueueTPUEmbeddingRaggedTensorBatchCombiners 

func EnqueueTPUEmbeddingRaggedTensorBatchCombiners(value []string) EnqueueTPUEmbeddingRaggedTensorBatchAttr

EnqueueTPUEmbeddingRaggedTensorBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

func EnqueueTPUEmbeddingRaggedTensorBatchDeviceOrdinal 

func EnqueueTPUEmbeddingRaggedTensorBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingRaggedTensorBatchAttr

EnqueueTPUEmbeddingRaggedTensorBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. Should be >= 0 and less than the number of TPU cores in the task on which the node is placed. If not specified, defaults to -1

func EnqueueTPUEmbeddingRaggedTensorBatchMaxSequenceLengths 

func EnqueueTPUEmbeddingRaggedTensorBatchMaxSequenceLengths(value []int64) EnqueueTPUEmbeddingRaggedTensorBatchAttr

EnqueueTPUEmbeddingRaggedTensorBatchMaxSequenceLengths sets the optional max_sequence_lengths attribute to value. If not specified, defaults to {}

func EnqueueTPUEmbeddingRaggedTensorBatchNumFeatures 

func EnqueueTPUEmbeddingRaggedTensorBatchNumFeatures(value []int64) EnqueueTPUEmbeddingRaggedTensorBatchAttr

EnqueueTPUEmbeddingRaggedTensorBatchNumFeatures sets the optional num_features attribute to value. If not specified, defaults to {}

type EnqueueTPUEmbeddingSparseBatchAttr 

type EnqueueTPUEmbeddingSparseBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingSparseBatchAttr is an optional argument to EnqueueTPUEmbeddingSparseBatch.

func EnqueueTPUEmbeddingSparseBatchCombiners 

func EnqueueTPUEmbeddingSparseBatchCombiners(value []string) EnqueueTPUEmbeddingSparseBatchAttr

EnqueueTPUEmbeddingSparseBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

func EnqueueTPUEmbeddingSparseBatchDeviceOrdinal 

func EnqueueTPUEmbeddingSparseBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingSparseBatchAttr

EnqueueTPUEmbeddingSparseBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. Should be >= 0 and less than the number of TPU cores in the task on which the node is placed. If not specified, defaults to -1

type EnqueueTPUEmbeddingSparseTensorBatchAttr 

type EnqueueTPUEmbeddingSparseTensorBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingSparseTensorBatchAttr is an optional argument to EnqueueTPUEmbeddingSparseTensorBatch.

func EnqueueTPUEmbeddingSparseTensorBatchCombiners 

func EnqueueTPUEmbeddingSparseTensorBatchCombiners(value []string) EnqueueTPUEmbeddingSparseTensorBatchAttr

EnqueueTPUEmbeddingSparseTensorBatchCombiners sets the optional combiners attribute to value.

value: A list of string scalars, one for each embedding table that specify how to normalize the embedding activations after weighted summation. Supported combiners are 'mean', 'sum', or 'sqrtn'. It is invalid to have the sum of the weights be 0 for 'mean' or the sum of the squared weights be 0 for 'sqrtn'. If combiners isn't passed, the default is to use 'sum' for all tables. If not specified, defaults to {}

func EnqueueTPUEmbeddingSparseTensorBatchDeviceOrdinal 

func EnqueueTPUEmbeddingSparseTensorBatchDeviceOrdinal(value int64) EnqueueTPUEmbeddingSparseTensorBatchAttr

EnqueueTPUEmbeddingSparseTensorBatchDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. Should be >= 0 and less than the number of TPU cores in the task on which the node is placed. If not specified, defaults to -1

func EnqueueTPUEmbeddingSparseTensorBatchMaxSequenceLengths 

func EnqueueTPUEmbeddingSparseTensorBatchMaxSequenceLengths(value []int64) EnqueueTPUEmbeddingSparseTensorBatchAttr

EnqueueTPUEmbeddingSparseTensorBatchMaxSequenceLengths sets the optional max_sequence_lengths attribute to value. If not specified, defaults to {}

func EnqueueTPUEmbeddingSparseTensorBatchNumFeatures 

func EnqueueTPUEmbeddingSparseTensorBatchNumFeatures(value []int64) EnqueueTPUEmbeddingSparseTensorBatchAttr

EnqueueTPUEmbeddingSparseTensorBatchNumFeatures sets the optional num_features attribute to value. If not specified, defaults to {}

type EnterAttr 

type EnterAttr func(optionalAttr)

EnterAttr is an optional argument to Enter.

func EnterIsConstant 

func EnterIsConstant(value bool) EnterAttr

EnterIsConstant sets the optional is_constant attribute to value.

value: If true, the output is constant within the child frame. If not specified, defaults to false

func EnterParallelIterations 

func EnterParallelIterations(value int64) EnterAttr

EnterParallelIterations sets the optional parallel_iterations attribute to value.

value: The number of iterations allowed to run in parallel. If not specified, defaults to 10

type EqualAttr 

type EqualAttr func(optionalAttr)

EqualAttr is an optional argument to Equal.

func EqualIncompatibleShapeError 

func EqualIncompatibleShapeError(value bool) EqualAttr

EqualIncompatibleShapeError sets the optional incompatible_shape_error attribute to value. If not specified, defaults to true

type EuclideanNormAttr 

type EuclideanNormAttr func(optionalAttr)

EuclideanNormAttr is an optional argument to EuclideanNorm.

func EuclideanNormKeepDims 

func EuclideanNormKeepDims(value bool) EuclideanNormAttr

EuclideanNormKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type ExperimentalAutoShardDatasetAttr 

type ExperimentalAutoShardDatasetAttr func(optionalAttr)

ExperimentalAutoShardDatasetAttr is an optional argument to ExperimentalAutoShardDataset.

func ExperimentalAutoShardDatasetAutoShardPolicy 

func ExperimentalAutoShardDatasetAutoShardPolicy(value int64) ExperimentalAutoShardDatasetAttr

ExperimentalAutoShardDatasetAutoShardPolicy sets the optional auto_shard_policy attribute to value. If not specified, defaults to 0

type ExperimentalIgnoreErrorsDatasetAttr 

type ExperimentalIgnoreErrorsDatasetAttr func(optionalAttr)

ExperimentalIgnoreErrorsDatasetAttr is an optional argument to ExperimentalIgnoreErrorsDataset.

func ExperimentalIgnoreErrorsDatasetLogWarning 

func ExperimentalIgnoreErrorsDatasetLogWarning(value bool) ExperimentalIgnoreErrorsDatasetAttr

ExperimentalIgnoreErrorsDatasetLogWarning sets the optional log_warning attribute to value. If not specified, defaults to false

type ExperimentalParseExampleDatasetAttr 

type ExperimentalParseExampleDatasetAttr func(optionalAttr)

ExperimentalParseExampleDatasetAttr is an optional argument to ExperimentalParseExampleDataset.

func ExperimentalParseExampleDatasetSloppy 

func ExperimentalParseExampleDatasetSloppy(value bool) ExperimentalParseExampleDatasetAttr

ExperimentalParseExampleDatasetSloppy sets the optional sloppy attribute to value. If not specified, defaults to false

type ExperimentalRebatchDatasetAttr 

type ExperimentalRebatchDatasetAttr func(optionalAttr)

ExperimentalRebatchDatasetAttr is an optional argument to ExperimentalRebatchDataset.

func ExperimentalRebatchDatasetUseFallback 

func ExperimentalRebatchDatasetUseFallback(value bool) ExperimentalRebatchDatasetAttr

ExperimentalRebatchDatasetUseFallback sets the optional use_fallback attribute to value. If not specified, defaults to true

type ExperimentalStatsAggregatorHandleAttr 

type ExperimentalStatsAggregatorHandleAttr func(optionalAttr)

ExperimentalStatsAggregatorHandleAttr is an optional argument to ExperimentalStatsAggregatorHandle.

func ExperimentalStatsAggregatorHandleContainer 

func ExperimentalStatsAggregatorHandleContainer(value string) ExperimentalStatsAggregatorHandleAttr

ExperimentalStatsAggregatorHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ExperimentalStatsAggregatorHandleSharedName 

func ExperimentalStatsAggregatorHandleSharedName(value string) ExperimentalStatsAggregatorHandleAttr

ExperimentalStatsAggregatorHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ExperimentalThreadPoolHandleAttr 

type ExperimentalThreadPoolHandleAttr func(optionalAttr)

ExperimentalThreadPoolHandleAttr is an optional argument to ExperimentalThreadPoolHandle.

func ExperimentalThreadPoolHandleContainer 

func ExperimentalThreadPoolHandleContainer(value string) ExperimentalThreadPoolHandleAttr

ExperimentalThreadPoolHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ExperimentalThreadPoolHandleMaxIntraOpParallelism 

func ExperimentalThreadPoolHandleMaxIntraOpParallelism(value int64) ExperimentalThreadPoolHandleAttr

ExperimentalThreadPoolHandleMaxIntraOpParallelism sets the optional max_intra_op_parallelism attribute to value.

value: The maximum degree of parallelism to use within operations that execute on this threadpool. If not specified, defaults to 1

func ExperimentalThreadPoolHandleSharedName 

func ExperimentalThreadPoolHandleSharedName(value string) ExperimentalThreadPoolHandleAttr

ExperimentalThreadPoolHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ExtractGlimpseAttr 

type ExtractGlimpseAttr func(optionalAttr)

ExtractGlimpseAttr is an optional argument to ExtractGlimpse.

func ExtractGlimpseCentered 

func ExtractGlimpseCentered(value bool) ExtractGlimpseAttr

ExtractGlimpseCentered sets the optional centered attribute to value.

value: indicates if the offset coordinates are centered relative to the image, in which case the (0, 0) offset is relative to the center of the input images. If false, the (0,0) offset corresponds to the upper left corner of the input images. If not specified, defaults to true

func ExtractGlimpseNoise 

func ExtractGlimpseNoise(value string) ExtractGlimpseAttr

ExtractGlimpseNoise sets the optional noise attribute to value.

value: indicates if the noise should `uniform`, `gaussian`, or `zero`. The default is `uniform` which means the noise type will be decided by `uniform_noise`. If not specified, defaults to "uniform"

func ExtractGlimpseNormalized 

func ExtractGlimpseNormalized(value bool) ExtractGlimpseAttr

ExtractGlimpseNormalized sets the optional normalized attribute to value.

value: indicates if the offset coordinates are normalized. If not specified, defaults to true

func ExtractGlimpseUniformNoise 

func ExtractGlimpseUniformNoise(value bool) ExtractGlimpseAttr

ExtractGlimpseUniformNoise sets the optional uniform_noise attribute to value.

value: indicates if the noise should be generated using a uniform distribution or a Gaussian distribution. If not specified, defaults to true

type ExtractGlimpseV2Attr 

type ExtractGlimpseV2Attr func(optionalAttr)

ExtractGlimpseV2Attr is an optional argument to ExtractGlimpseV2.

func ExtractGlimpseV2Centered 

func ExtractGlimpseV2Centered(value bool) ExtractGlimpseV2Attr

ExtractGlimpseV2Centered sets the optional centered attribute to value.

value: indicates if the offset coordinates are centered relative to the image, in which case the (0, 0) offset is relative to the center of the input images. If false, the (0,0) offset corresponds to the upper left corner of the input images. If not specified, defaults to true

func ExtractGlimpseV2Noise 

func ExtractGlimpseV2Noise(value string) ExtractGlimpseV2Attr

ExtractGlimpseV2Noise sets the optional noise attribute to value.

value: indicates if the noise should `uniform`, `gaussian`, or `zero`. The default is `uniform` which means the noise type will be decided by `uniform_noise`. If not specified, defaults to "uniform"

func ExtractGlimpseV2Normalized 

func ExtractGlimpseV2Normalized(value bool) ExtractGlimpseV2Attr

ExtractGlimpseV2Normalized sets the optional normalized attribute to value.

value: indicates if the offset coordinates are normalized. If not specified, defaults to true

func ExtractGlimpseV2UniformNoise 

func ExtractGlimpseV2UniformNoise(value bool) ExtractGlimpseV2Attr

ExtractGlimpseV2UniformNoise sets the optional uniform_noise attribute to value.

value: indicates if the noise should be generated using a uniform distribution or a Gaussian distribution. If not specified, defaults to true

type ExtractJpegShapeAttr 

type ExtractJpegShapeAttr func(optionalAttr)

ExtractJpegShapeAttr is an optional argument to ExtractJpegShape.

func ExtractJpegShapeOutputType 

func ExtractJpegShapeOutputType(value tf.DataType) ExtractJpegShapeAttr

ExtractJpegShapeOutputType sets the optional output_type attribute to value.

value: (Optional) The output type of the operation (int32 or int64). Defaults to int32. If not specified, defaults to DT_INT32

type FIFOQueueV2Attr 

type FIFOQueueV2Attr func(optionalAttr)

FIFOQueueV2Attr is an optional argument to FIFOQueueV2.

func FIFOQueueV2Capacity 

func FIFOQueueV2Capacity(value int64) FIFOQueueV2Attr

FIFOQueueV2Capacity sets the optional capacity attribute to value.

value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1

func FIFOQueueV2Container 

func FIFOQueueV2Container(value string) FIFOQueueV2Attr

FIFOQueueV2Container sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func FIFOQueueV2Shapes 

func FIFOQueueV2Shapes(value []tf.Shape) FIFOQueueV2Attr

FIFOQueueV2Shapes sets the optional shapes attribute to value.

value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func FIFOQueueV2SharedName 

func FIFOQueueV2SharedName(value string) FIFOQueueV2Attr

FIFOQueueV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""

type FakeQuantWithMinMaxArgsAttr 

type FakeQuantWithMinMaxArgsAttr func(optionalAttr)

FakeQuantWithMinMaxArgsAttr is an optional argument to FakeQuantWithMinMaxArgs.

func FakeQuantWithMinMaxArgsMax 

func FakeQuantWithMinMaxArgsMax(value float32) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsMax sets the optional max attribute to value. If not specified, defaults to 6

func FakeQuantWithMinMaxArgsMin 

func FakeQuantWithMinMaxArgsMin(value float32) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsMin sets the optional min attribute to value. If not specified, defaults to -6

func FakeQuantWithMinMaxArgsNarrowRange 

func FakeQuantWithMinMaxArgsNarrowRange(value bool) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxArgsNumBits 

func FakeQuantWithMinMaxArgsNumBits(value int64) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxArgsGradientAttr 

type FakeQuantWithMinMaxArgsGradientAttr func(optionalAttr)

FakeQuantWithMinMaxArgsGradientAttr is an optional argument to FakeQuantWithMinMaxArgsGradient.

func FakeQuantWithMinMaxArgsGradientMax 

func FakeQuantWithMinMaxArgsGradientMax(value float32) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientMax sets the optional max attribute to value. If not specified, defaults to 6

func FakeQuantWithMinMaxArgsGradientMin 

func FakeQuantWithMinMaxArgsGradientMin(value float32) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientMin sets the optional min attribute to value. If not specified, defaults to -6

func FakeQuantWithMinMaxArgsGradientNarrowRange 

func FakeQuantWithMinMaxArgsGradientNarrowRange(value bool) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxArgsGradientNumBits 

func FakeQuantWithMinMaxArgsGradientNumBits(value int64) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsAttr 

type FakeQuantWithMinMaxVarsAttr func(optionalAttr)

FakeQuantWithMinMaxVarsAttr is an optional argument to FakeQuantWithMinMaxVars.

func FakeQuantWithMinMaxVarsNarrowRange 

func FakeQuantWithMinMaxVarsNarrowRange(value bool) FakeQuantWithMinMaxVarsAttr

FakeQuantWithMinMaxVarsNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxVarsNumBits 

func FakeQuantWithMinMaxVarsNumBits(value int64) FakeQuantWithMinMaxVarsAttr

FakeQuantWithMinMaxVarsNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsGradientAttr 

type FakeQuantWithMinMaxVarsGradientAttr func(optionalAttr)

FakeQuantWithMinMaxVarsGradientAttr is an optional argument to FakeQuantWithMinMaxVarsGradient.

func FakeQuantWithMinMaxVarsGradientNarrowRange 

func FakeQuantWithMinMaxVarsGradientNarrowRange(value bool) FakeQuantWithMinMaxVarsGradientAttr

FakeQuantWithMinMaxVarsGradientNarrowRange sets the optional narrow_range attribute to value.

value: Whether to quantize into 2^num_bits - 1 distinct values. If not specified, defaults to false

func FakeQuantWithMinMaxVarsGradientNumBits 

func FakeQuantWithMinMaxVarsGradientNumBits(value int64) FakeQuantWithMinMaxVarsGradientAttr

FakeQuantWithMinMaxVarsGradientNumBits sets the optional num_bits attribute to value.

value: The bitwidth of the quantization; between 2 and 8, inclusive. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsPerChannelAttr 

type FakeQuantWithMinMaxVarsPerChannelAttr func(optionalAttr)

FakeQuantWithMinMaxVarsPerChannelAttr is an optional argument to FakeQuantWithMinMaxVarsPerChannel.

func FakeQuantWithMinMaxVarsPerChannelNarrowRange 

func FakeQuantWithMinMaxVarsPerChannelNarrowRange(value bool) FakeQuantWithMinMaxVarsPerChannelAttr

FakeQuantWithMinMaxVarsPerChannelNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxVarsPerChannelNumBits 

func FakeQuantWithMinMaxVarsPerChannelNumBits(value int64) FakeQuantWithMinMaxVarsPerChannelAttr

FakeQuantWithMinMaxVarsPerChannelNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsPerChannelGradientAttr 

type FakeQuantWithMinMaxVarsPerChannelGradientAttr func(optionalAttr)

FakeQuantWithMinMaxVarsPerChannelGradientAttr is an optional argument to FakeQuantWithMinMaxVarsPerChannelGradient.

func FakeQuantWithMinMaxVarsPerChannelGradientNarrowRange 

func FakeQuantWithMinMaxVarsPerChannelGradientNarrowRange(value bool) FakeQuantWithMinMaxVarsPerChannelGradientAttr

FakeQuantWithMinMaxVarsPerChannelGradientNarrowRange sets the optional narrow_range attribute to value.

value: Whether to quantize into 2^num_bits - 1 distinct values. If not specified, defaults to false

func FakeQuantWithMinMaxVarsPerChannelGradientNumBits 

func FakeQuantWithMinMaxVarsPerChannelGradientNumBits(value int64) FakeQuantWithMinMaxVarsPerChannelGradientAttr

FakeQuantWithMinMaxVarsPerChannelGradientNumBits sets the optional num_bits attribute to value.

value: The bitwidth of the quantization; between 2 and 16, inclusive. If not specified, defaults to 8

type FinalizeDatasetAttr 

type FinalizeDatasetAttr func(optionalAttr)

FinalizeDatasetAttr is an optional argument to FinalizeDataset.

func FinalizeDatasetHasCapturedRef 

func FinalizeDatasetHasCapturedRef(value bool) FinalizeDatasetAttr

FinalizeDatasetHasCapturedRef sets the optional has_captured_ref attribute to value. If not specified, defaults to false

type FixedLengthRecordDatasetAttr 

type FixedLengthRecordDatasetAttr func(optionalAttr)

FixedLengthRecordDatasetAttr is an optional argument to FixedLengthRecordDataset.

func FixedLengthRecordDatasetMetadata 

func FixedLengthRecordDatasetMetadata(value string) FixedLengthRecordDatasetAttr

FixedLengthRecordDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type FixedLengthRecordReaderV2Attr 

type FixedLengthRecordReaderV2Attr func(optionalAttr)

FixedLengthRecordReaderV2Attr is an optional argument to FixedLengthRecordReaderV2.

func FixedLengthRecordReaderV2Container 

func FixedLengthRecordReaderV2Container(value string) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2Container sets the optional container attribute to value.

value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func FixedLengthRecordReaderV2Encoding 

func FixedLengthRecordReaderV2Encoding(value string) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2Encoding sets the optional encoding attribute to value.

value: The type of encoding for the file. Currently ZLIB and GZIP are supported. Defaults to none. If not specified, defaults to ""

func FixedLengthRecordReaderV2FooterBytes 

func FixedLengthRecordReaderV2FooterBytes(value int64) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2FooterBytes sets the optional footer_bytes attribute to value.

value: Number of bytes in the footer, defaults to 0. If not specified, defaults to 0

func FixedLengthRecordReaderV2HeaderBytes 

func FixedLengthRecordReaderV2HeaderBytes(value int64) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2HeaderBytes sets the optional header_bytes attribute to value.

value: Number of bytes in the header, defaults to 0. If not specified, defaults to 0

func FixedLengthRecordReaderV2HopBytes 

func FixedLengthRecordReaderV2HopBytes(value int64) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2HopBytes sets the optional hop_bytes attribute to value.

value: Number of bytes to hop before each read. Default of 0 means using record_bytes. If not specified, defaults to 0

func FixedLengthRecordReaderV2SharedName 

func FixedLengthRecordReaderV2SharedName(value string) FixedLengthRecordReaderV2Attr

FixedLengthRecordReaderV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

type FixedUnigramCandidateSamplerAttr 

type FixedUnigramCandidateSamplerAttr func(optionalAttr)

FixedUnigramCandidateSamplerAttr is an optional argument to FixedUnigramCandidateSampler.

func FixedUnigramCandidateSamplerDistortion 

func FixedUnigramCandidateSamplerDistortion(value float32) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerDistortion sets the optional distortion attribute to value.

value: The distortion is used to skew the unigram probability distribution. Each weight is first raised to the distortion's power before adding to the internal unigram distribution. As a result, distortion = 1.0 gives regular unigram sampling (as defined by the vocab file), and distortion = 0.0 gives a uniform distribution. If not specified, defaults to 1

func FixedUnigramCandidateSamplerNumReservedIds 

func FixedUnigramCandidateSamplerNumReservedIds(value int64) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerNumReservedIds sets the optional num_reserved_ids attribute to value.

value: Optionally some reserved IDs can be added in the range [0, ..., num_reserved_ids) by the users. One use case is that a special unknown word token is used as ID 0. These IDs will have a sampling probability of 0. If not specified, defaults to 0

func FixedUnigramCandidateSamplerNumShards 

func FixedUnigramCandidateSamplerNumShards(value int64) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerNumShards sets the optional num_shards attribute to value.

value: A sampler can be used to sample from a subset of the original range in order to speed up the whole computation through parallelism. This parameter (together with 'shard') indicates the number of partitions that are being used in the overall computation. If not specified, defaults to 1

REQUIRES: value >= 1

func FixedUnigramCandidateSamplerSeed 

func FixedUnigramCandidateSamplerSeed(value int64) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func FixedUnigramCandidateSamplerSeed2 

func FixedUnigramCandidateSamplerSeed2(value int64) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

func FixedUnigramCandidateSamplerShard 

func FixedUnigramCandidateSamplerShard(value int64) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerShard sets the optional shard attribute to value.

value: A sampler can be used to sample from a subset of the original range in order to speed up the whole computation through parallelism. This parameter (together with 'num_shards') indicates the particular partition number of a sampler op, when partitioning is being used. If not specified, defaults to 0

REQUIRES: value >= 0

func FixedUnigramCandidateSamplerUnigrams 

func FixedUnigramCandidateSamplerUnigrams(value []float32) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerUnigrams sets the optional unigrams attribute to value.

value: A list of unigram counts or probabilities, one per ID in sequential order. Exactly one of vocab_file and unigrams should be passed to this op. If not specified, defaults to {}

func FixedUnigramCandidateSamplerVocabFile 

func FixedUnigramCandidateSamplerVocabFile(value string) FixedUnigramCandidateSamplerAttr

FixedUnigramCandidateSamplerVocabFile sets the optional vocab_file attribute to value.

value: Each valid line in this file (which should have a CSV-like format) corresponds to a valid word ID. IDs are in sequential order, starting from num_reserved_ids. The last entry in each line is expected to be a value corresponding to the count or relative probability. Exactly one of vocab_file and unigrams needs to be passed to this op. If not specified, defaults to ""

type FractionalAvgPoolAttr 

type FractionalAvgPoolAttr func(optionalAttr)

FractionalAvgPoolAttr is an optional argument to FractionalAvgPool.

func FractionalAvgPoolDeterministic 

func FractionalAvgPoolDeterministic(value bool) FractionalAvgPoolAttr

FractionalAvgPoolDeterministic sets the optional deterministic attribute to value.

value: When set to True, a fixed pooling region will be used when iterating over a FractionalAvgPool node in the computation graph. Mainly used in unit test to make FractionalAvgPool deterministic. If not specified, defaults to false

func FractionalAvgPoolOverlapping 

func FractionalAvgPoolOverlapping(value bool) FractionalAvgPoolAttr

FractionalAvgPoolOverlapping sets the optional overlapping attribute to value.

value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:

`index 0 1 2 3 4`

`value 20 5 16 3 7`

If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [41/3, 26/3] for fractional avg pooling. If not specified, defaults to false

func FractionalAvgPoolPseudoRandom 

func FractionalAvgPoolPseudoRandom(value bool) FractionalAvgPoolAttr

FractionalAvgPoolPseudoRandom sets the optional pseudo_random attribute to value.

value: When set to True, generates the pooling sequence in a pseudorandom fashion, otherwise, in a random fashion. Check paper [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071) for difference between pseudorandom and random. If not specified, defaults to false

func FractionalAvgPoolSeed 

func FractionalAvgPoolSeed(value int64) FractionalAvgPoolAttr

FractionalAvgPoolSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func FractionalAvgPoolSeed2 

func FractionalAvgPoolSeed2(value int64) FractionalAvgPoolAttr

FractionalAvgPoolSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type FractionalAvgPoolGradAttr 

type FractionalAvgPoolGradAttr func(optionalAttr)

FractionalAvgPoolGradAttr is an optional argument to FractionalAvgPoolGrad.

func FractionalAvgPoolGradOverlapping 

func FractionalAvgPoolGradOverlapping(value bool) FractionalAvgPoolGradAttr

FractionalAvgPoolGradOverlapping sets the optional overlapping attribute to value.

value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:

`index 0 1 2 3 4`

`value 20 5 16 3 7`

If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [41/3, 26/3] for fractional avg pooling. If not specified, defaults to false

type FractionalMaxPoolAttr 

type FractionalMaxPoolAttr func(optionalAttr)

FractionalMaxPoolAttr is an optional argument to FractionalMaxPool.

func FractionalMaxPoolDeterministic 

func FractionalMaxPoolDeterministic(value bool) FractionalMaxPoolAttr

FractionalMaxPoolDeterministic sets the optional deterministic attribute to value.

value: When set to True, a fixed pooling region will be used when iterating over a FractionalMaxPool node in the computation graph. Mainly used in unit test to make FractionalMaxPool deterministic. If not specified, defaults to false

func FractionalMaxPoolOverlapping 

func FractionalMaxPoolOverlapping(value bool) FractionalMaxPoolAttr

FractionalMaxPoolOverlapping sets the optional overlapping attribute to value.

value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:

`index 0 1 2 3 4`

`value 20 5 16 3 7`

If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [20, 16] for fractional max pooling. If not specified, defaults to false

func FractionalMaxPoolPseudoRandom 

func FractionalMaxPoolPseudoRandom(value bool) FractionalMaxPoolAttr

FractionalMaxPoolPseudoRandom sets the optional pseudo_random attribute to value.

value: When set to True, generates the pooling sequence in a pseudorandom fashion, otherwise, in a random fashion. Check paper [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071) for difference between pseudorandom and random. If not specified, defaults to false

func FractionalMaxPoolSeed 

func FractionalMaxPoolSeed(value int64) FractionalMaxPoolAttr

FractionalMaxPoolSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func FractionalMaxPoolSeed2 

func FractionalMaxPoolSeed2(value int64) FractionalMaxPoolAttr

FractionalMaxPoolSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type FractionalMaxPoolGradAttr 

type FractionalMaxPoolGradAttr func(optionalAttr)

FractionalMaxPoolGradAttr is an optional argument to FractionalMaxPoolGrad.

func FractionalMaxPoolGradOverlapping 

func FractionalMaxPoolGradOverlapping(value bool) FractionalMaxPoolGradAttr

FractionalMaxPoolGradOverlapping sets the optional overlapping attribute to value.

value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:

`index 0 1 2 3 4`

`value 20 5 16 3 7`

If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [20, 16] for fractional max pooling. If not specified, defaults to false

type FusedBatchNormAttr 

type FusedBatchNormAttr func(optionalAttr)

FusedBatchNormAttr is an optional argument to FusedBatchNorm.

func FusedBatchNormDataFormat 

func FusedBatchNormDataFormat(value string) FusedBatchNormAttr

FusedBatchNormDataFormat sets the optional data_format attribute to value.

value: The data format for x and y. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormEpsilon 

func FusedBatchNormEpsilon(value float32) FusedBatchNormAttr

FusedBatchNormEpsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormExponentialAvgFactor 

func FusedBatchNormExponentialAvgFactor(value float32) FusedBatchNormAttr

FusedBatchNormExponentialAvgFactor sets the optional exponential_avg_factor attribute to value. If not specified, defaults to 1

func FusedBatchNormIsTraining 

func FusedBatchNormIsTraining(value bool) FusedBatchNormAttr

FusedBatchNormIsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedBatchNormGradAttr 

type FusedBatchNormGradAttr func(optionalAttr)

FusedBatchNormGradAttr is an optional argument to FusedBatchNormGrad.

func FusedBatchNormGradDataFormat 

func FusedBatchNormGradDataFormat(value string) FusedBatchNormGradAttr

FusedBatchNormGradDataFormat sets the optional data_format attribute to value.

value: The data format for y_backprop, x, x_backprop. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormGradEpsilon 

func FusedBatchNormGradEpsilon(value float32) FusedBatchNormGradAttr

FusedBatchNormGradEpsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormGradIsTraining 

func FusedBatchNormGradIsTraining(value bool) FusedBatchNormGradAttr

FusedBatchNormGradIsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedBatchNormGradV2Attr 

type FusedBatchNormGradV2Attr func(optionalAttr)

FusedBatchNormGradV2Attr is an optional argument to FusedBatchNormGradV2.

func FusedBatchNormGradV2DataFormat 

func FusedBatchNormGradV2DataFormat(value string) FusedBatchNormGradV2Attr

FusedBatchNormGradV2DataFormat sets the optional data_format attribute to value.

value: The data format for y_backprop, x, x_backprop. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormGradV2Epsilon 

func FusedBatchNormGradV2Epsilon(value float32) FusedBatchNormGradV2Attr

FusedBatchNormGradV2Epsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormGradV2IsTraining 

func FusedBatchNormGradV2IsTraining(value bool) FusedBatchNormGradV2Attr

FusedBatchNormGradV2IsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedBatchNormGradV3Attr 

type FusedBatchNormGradV3Attr func(optionalAttr)

FusedBatchNormGradV3Attr is an optional argument to FusedBatchNormGradV3.

func FusedBatchNormGradV3DataFormat 

func FusedBatchNormGradV3DataFormat(value string) FusedBatchNormGradV3Attr

FusedBatchNormGradV3DataFormat sets the optional data_format attribute to value.

value: The data format for y_backprop, x, x_backprop. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormGradV3Epsilon 

func FusedBatchNormGradV3Epsilon(value float32) FusedBatchNormGradV3Attr

FusedBatchNormGradV3Epsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormGradV3IsTraining 

func FusedBatchNormGradV3IsTraining(value bool) FusedBatchNormGradV3Attr

FusedBatchNormGradV3IsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedBatchNormV2Attr 

type FusedBatchNormV2Attr func(optionalAttr)

FusedBatchNormV2Attr is an optional argument to FusedBatchNormV2.

func FusedBatchNormV2DataFormat 

func FusedBatchNormV2DataFormat(value string) FusedBatchNormV2Attr

FusedBatchNormV2DataFormat sets the optional data_format attribute to value.

value: The data format for x and y. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormV2Epsilon 

func FusedBatchNormV2Epsilon(value float32) FusedBatchNormV2Attr

FusedBatchNormV2Epsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormV2ExponentialAvgFactor 

func FusedBatchNormV2ExponentialAvgFactor(value float32) FusedBatchNormV2Attr

FusedBatchNormV2ExponentialAvgFactor sets the optional exponential_avg_factor attribute to value. If not specified, defaults to 1

func FusedBatchNormV2IsTraining 

func FusedBatchNormV2IsTraining(value bool) FusedBatchNormV2Attr

FusedBatchNormV2IsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedBatchNormV3Attr 

type FusedBatchNormV3Attr func(optionalAttr)

FusedBatchNormV3Attr is an optional argument to FusedBatchNormV3.

func FusedBatchNormV3DataFormat 

func FusedBatchNormV3DataFormat(value string) FusedBatchNormV3Attr

FusedBatchNormV3DataFormat sets the optional data_format attribute to value.

value: The data format for x and y. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"

func FusedBatchNormV3Epsilon 

func FusedBatchNormV3Epsilon(value float32) FusedBatchNormV3Attr

FusedBatchNormV3Epsilon sets the optional epsilon attribute to value.

value: A small float number added to the variance of x. If not specified, defaults to 0.0001

func FusedBatchNormV3ExponentialAvgFactor 

func FusedBatchNormV3ExponentialAvgFactor(value float32) FusedBatchNormV3Attr

FusedBatchNormV3ExponentialAvgFactor sets the optional exponential_avg_factor attribute to value. If not specified, defaults to 1

func FusedBatchNormV3IsTraining 

func FusedBatchNormV3IsTraining(value bool) FusedBatchNormV3Attr

FusedBatchNormV3IsTraining sets the optional is_training attribute to value.

value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true

type FusedResizeAndPadConv2DAttr 

type FusedResizeAndPadConv2DAttr func(optionalAttr)

FusedResizeAndPadConv2DAttr is an optional argument to FusedResizeAndPadConv2D.

func FusedResizeAndPadConv2DResizeAlignCorners 

func FusedResizeAndPadConv2DResizeAlignCorners(value bool) FusedResizeAndPadConv2DAttr

FusedResizeAndPadConv2DResizeAlignCorners sets the optional resize_align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

type GatherAttr 

type GatherAttr func(optionalAttr)

GatherAttr is an optional argument to Gather.

func GatherValidateIndices 

func GatherValidateIndices(value bool) GatherAttr

GatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type GatherNdAttr added in v0.8.2 

type GatherNdAttr func(optionalAttr)

GatherNdAttr is an optional argument to GatherNd.

func GatherNdBadIndicesPolicy added in v0.8.2 

func GatherNdBadIndicesPolicy(value string) GatherNdAttr

GatherNdBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type GatherV2Attr 

type GatherV2Attr func(optionalAttr)

GatherV2Attr is an optional argument to GatherV2.

func GatherV2BatchDims 

func GatherV2BatchDims(value int64) GatherV2Attr

GatherV2BatchDims sets the optional batch_dims attribute to value. If not specified, defaults to 0

type GenerateBoundingBoxProposalsAttr 

type GenerateBoundingBoxProposalsAttr func(optionalAttr)

GenerateBoundingBoxProposalsAttr is an optional argument to GenerateBoundingBoxProposals.

func GenerateBoundingBoxProposalsPostNmsTopn 

func GenerateBoundingBoxProposalsPostNmsTopn(value int64) GenerateBoundingBoxProposalsAttr

GenerateBoundingBoxProposalsPostNmsTopn sets the optional post_nms_topn attribute to value.

value: An integer. Maximum number of rois in the output. If not specified, defaults to 300

type GenerateVocabRemappingAttr 

type GenerateVocabRemappingAttr func(optionalAttr)

GenerateVocabRemappingAttr is an optional argument to GenerateVocabRemapping.

func GenerateVocabRemappingOldVocabSize 

func GenerateVocabRemappingOldVocabSize(value int64) GenerateVocabRemappingAttr

GenerateVocabRemappingOldVocabSize sets the optional old_vocab_size attribute to value.

value: Number of entries in the old vocab file to consider. If -1, use the entire old vocabulary. If not specified, defaults to -1

REQUIRES: value >= -1

type HashTableV2Attr 

type HashTableV2Attr func(optionalAttr)

HashTableV2Attr is an optional argument to HashTableV2.

func HashTableV2Container 

func HashTableV2Container(value string) HashTableV2Attr

HashTableV2Container sets the optional container attribute to value.

value: If non-empty, this table is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func HashTableV2SharedName 

func HashTableV2SharedName(value string) HashTableV2Attr

HashTableV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this table is shared under the given name across multiple sessions. If not specified, defaults to ""

func HashTableV2UseNodeNameSharing 

func HashTableV2UseNodeNameSharing(value bool) HashTableV2Attr

HashTableV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value.

value: If true and shared_name is empty, the table is shared using the node name. If not specified, defaults to false

type HistogramFixedWidthAttr 

type HistogramFixedWidthAttr func(optionalAttr)

HistogramFixedWidthAttr is an optional argument to HistogramFixedWidth.

func HistogramFixedWidthDtype 

func HistogramFixedWidthDtype(value tf.DataType) HistogramFixedWidthAttr

HistogramFixedWidthDtype sets the optional dtype attribute to value. If not specified, defaults to DT_INT32

type IRFFT2DAttr 

type IRFFT2DAttr func(optionalAttr)

IRFFT2DAttr is an optional argument to IRFFT2D.

func IRFFT2DTreal 

func IRFFT2DTreal(value tf.DataType) IRFFT2DAttr

IRFFT2DTreal sets the optional Treal attribute to value. If not specified, defaults to DT_FLOAT

type IRFFT3DAttr 

type IRFFT3DAttr func(optionalAttr)

IRFFT3DAttr is an optional argument to IRFFT3D.

func IRFFT3DTreal 

func IRFFT3DTreal(value tf.DataType) IRFFT3DAttr

IRFFT3DTreal sets the optional Treal attribute to value. If not specified, defaults to DT_FLOAT

type IRFFTAttr 

type IRFFTAttr func(optionalAttr)

IRFFTAttr is an optional argument to IRFFT.

func IRFFTTreal 

func IRFFTTreal(value tf.DataType) IRFFTAttr

IRFFTTreal sets the optional Treal attribute to value. If not specified, defaults to DT_FLOAT

type IRFFTNDAttr added in v0.7.0 

type IRFFTNDAttr func(optionalAttr)

IRFFTNDAttr is an optional argument to IRFFTND.

func IRFFTNDTreal added in v0.7.0 

func IRFFTNDTreal(value tf.DataType) IRFFTNDAttr

IRFFTNDTreal sets the optional Treal attribute to value. If not specified, defaults to DT_FLOAT

type IdentityReaderV2Attr 

type IdentityReaderV2Attr func(optionalAttr)

IdentityReaderV2Attr is an optional argument to IdentityReaderV2.

func IdentityReaderV2Container 

func IdentityReaderV2Container(value string) IdentityReaderV2Attr

IdentityReaderV2Container sets the optional container attribute to value.

value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func IdentityReaderV2SharedName 

func IdentityReaderV2SharedName(value string) IdentityReaderV2Attr

IdentityReaderV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

type IgnoreErrorsDatasetAttr 

type IgnoreErrorsDatasetAttr func(optionalAttr)

IgnoreErrorsDatasetAttr is an optional argument to IgnoreErrorsDataset.

func IgnoreErrorsDatasetLogWarning 

func IgnoreErrorsDatasetLogWarning(value bool) IgnoreErrorsDatasetAttr

IgnoreErrorsDatasetLogWarning sets the optional log_warning attribute to value. If not specified, defaults to false

type ImagAttr 

type ImagAttr func(optionalAttr)

ImagAttr is an optional argument to Imag.

func ImagTout 

func ImagTout(value tf.DataType) ImagAttr

ImagTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type ImageProjectiveTransformV2Attr 

type ImageProjectiveTransformV2Attr func(optionalAttr)

ImageProjectiveTransformV2Attr is an optional argument to ImageProjectiveTransformV2.

func ImageProjectiveTransformV2FillMode 

func ImageProjectiveTransformV2FillMode(value string) ImageProjectiveTransformV2Attr

ImageProjectiveTransformV2FillMode sets the optional fill_mode attribute to value.

value: Fill mode, "REFLECT", "WRAP", or "CONSTANT". If not specified, defaults to "CONSTANT"

type ImageProjectiveTransformV3Attr 

type ImageProjectiveTransformV3Attr func(optionalAttr)

ImageProjectiveTransformV3Attr is an optional argument to ImageProjectiveTransformV3.

func ImageProjectiveTransformV3FillMode 

func ImageProjectiveTransformV3FillMode(value string) ImageProjectiveTransformV3Attr

ImageProjectiveTransformV3FillMode sets the optional fill_mode attribute to value.

value: Fill mode, "REFLECT", "WRAP", "CONSTANT", or "NEAREST". If not specified, defaults to "CONSTANT"

type ImageSummaryAttr 

type ImageSummaryAttr func(optionalAttr)

ImageSummaryAttr is an optional argument to ImageSummary.

func ImageSummaryBadColor 

func ImageSummaryBadColor(value tf.Tensor) ImageSummaryAttr

ImageSummaryBadColor sets the optional bad_color attribute to value.

value: Color to use for pixels with non-finite values. If not specified, defaults to {dtype:DT_UINT8 tensor_shape:{dim:{size:4}} int_val:255 int_val:0 int_val:0 int_val:255}

func ImageSummaryMaxImages 

func ImageSummaryMaxImages(value int64) ImageSummaryAttr

ImageSummaryMaxImages sets the optional max_images attribute to value.

value: Max number of batch elements to generate images for. If not specified, defaults to 3

REQUIRES: value >= 1

type InfeedEnqueueAttr 

type InfeedEnqueueAttr func(optionalAttr)

InfeedEnqueueAttr is an optional argument to InfeedEnqueue.

func InfeedEnqueueDeviceOrdinal 

func InfeedEnqueueDeviceOrdinal(value int64) InfeedEnqueueAttr

InfeedEnqueueDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. If not specified, defaults to -1

func InfeedEnqueueLayout 

func InfeedEnqueueLayout(value []int64) InfeedEnqueueAttr

InfeedEnqueueLayout sets the optional layout attribute to value.

value: A vector holding the requested layout in minor-to-major sequence. If a layout attribute is passed, but its values are all -1, the layout will be computed by the infeed operation. If not specified, defaults to {}

func InfeedEnqueueShape 

func InfeedEnqueueShape(value tf.Shape) InfeedEnqueueAttr

InfeedEnqueueShape sets the optional shape attribute to value.

value: The shape of the tensor. If not specified, defaults to {}

type InfeedEnqueuePrelinearizedBufferAttr 

type InfeedEnqueuePrelinearizedBufferAttr func(optionalAttr)

InfeedEnqueuePrelinearizedBufferAttr is an optional argument to InfeedEnqueuePrelinearizedBuffer.

func InfeedEnqueuePrelinearizedBufferDeviceOrdinal 

func InfeedEnqueuePrelinearizedBufferDeviceOrdinal(value int64) InfeedEnqueuePrelinearizedBufferAttr

InfeedEnqueuePrelinearizedBufferDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device and = 0 when the Op is running on the CPU device. If not specified, defaults to -1

type InfeedEnqueueTupleAttr 

type InfeedEnqueueTupleAttr func(optionalAttr)

InfeedEnqueueTupleAttr is an optional argument to InfeedEnqueueTuple.

func InfeedEnqueueTupleDeviceOrdinal 

func InfeedEnqueueTupleDeviceOrdinal(value int64) InfeedEnqueueTupleAttr

InfeedEnqueueTupleDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. If not specified, defaults to -1

func InfeedEnqueueTupleLayouts 

func InfeedEnqueueTupleLayouts(value []int64) InfeedEnqueueTupleAttr

InfeedEnqueueTupleLayouts sets the optional layouts attribute to value.

value: A vector holding the requested layout in minor-to-major sequence for all the tuple shapes, in the order the shapes appear in the "shapes" input. The layout elements for a sub-shape can be set to -1, in which case the corresponding layout will be computed by the infeed operation. If not specified, defaults to {}

type InitializeTableFromTextFileV2Attr 

type InitializeTableFromTextFileV2Attr func(optionalAttr)

InitializeTableFromTextFileV2Attr is an optional argument to InitializeTableFromTextFileV2.

func InitializeTableFromTextFileV2Delimiter 

func InitializeTableFromTextFileV2Delimiter(value string) InitializeTableFromTextFileV2Attr

InitializeTableFromTextFileV2Delimiter sets the optional delimiter attribute to value.

value: Delimiter to separate fields in a line. If not specified, defaults to "\t"

func InitializeTableFromTextFileV2Offset 

func InitializeTableFromTextFileV2Offset(value int64) InitializeTableFromTextFileV2Attr

InitializeTableFromTextFileV2Offset sets the optional offset attribute to value. If not specified, defaults to 0

func InitializeTableFromTextFileV2VocabSize 

func InitializeTableFromTextFileV2VocabSize(value int64) InitializeTableFromTextFileV2Attr

InitializeTableFromTextFileV2VocabSize sets the optional vocab_size attribute to value.

value: Number of elements of the file, use -1 if unknown. If not specified, defaults to -1

REQUIRES: value >= -1

type IsTPUEmbeddingInitializedAttr 

type IsTPUEmbeddingInitializedAttr func(optionalAttr)

IsTPUEmbeddingInitializedAttr is an optional argument to IsTPUEmbeddingInitialized.

func IsTPUEmbeddingInitializedConfig 

func IsTPUEmbeddingInitializedConfig(value string) IsTPUEmbeddingInitializedAttr

IsTPUEmbeddingInitializedConfig sets the optional config attribute to value. If not specified, defaults to ""

type IsotonicRegressionAttr 

type IsotonicRegressionAttr func(optionalAttr)

IsotonicRegressionAttr is an optional argument to IsotonicRegression.

func IsotonicRegressionOutputDtype 

func IsotonicRegressionOutputDtype(value tf.DataType) IsotonicRegressionAttr

IsotonicRegressionOutputDtype sets the optional output_dtype attribute to value.

value: Dtype of output. If not specified, defaults to DT_FLOAT

type IteratorFromStringHandleAttr 

type IteratorFromStringHandleAttr func(optionalAttr)

IteratorFromStringHandleAttr is an optional argument to IteratorFromStringHandle.

func IteratorFromStringHandleOutputShapes 

func IteratorFromStringHandleOutputShapes(value []tf.Shape) IteratorFromStringHandleAttr

IteratorFromStringHandleOutputShapes sets the optional output_shapes attribute to value.

value: If specified, defines the shape of each tuple component in an element produced by the resulting iterator. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func IteratorFromStringHandleOutputTypes 

func IteratorFromStringHandleOutputTypes(value []tf.DataType) IteratorFromStringHandleAttr

IteratorFromStringHandleOutputTypes sets the optional output_types attribute to value.

value: If specified, defines the type of each tuple component in an element produced by the resulting iterator. If not specified, defaults to {}

REQUIRES: len(value) >= 0

type LRNAttr 

type LRNAttr func(optionalAttr)

LRNAttr is an optional argument to LRN.

func LRNAlpha 

func LRNAlpha(value float32) LRNAttr

LRNAlpha sets the optional alpha attribute to value.

value: A scale factor, usually positive. If not specified, defaults to 1

func LRNBeta 

func LRNBeta(value float32) LRNAttr

LRNBeta sets the optional beta attribute to value.

value: An exponent. If not specified, defaults to 0.5

func LRNBias 

func LRNBias(value float32) LRNAttr

LRNBias sets the optional bias attribute to value.

value: An offset (usually positive to avoid dividing by 0). If not specified, defaults to 1

func LRNDepthRadius 

func LRNDepthRadius(value int64) LRNAttr

LRNDepthRadius sets the optional depth_radius attribute to value.

value: 0-D. Half-width of the 1-D normalization window. If not specified, defaults to 5

type LRNGradAttr 

type LRNGradAttr func(optionalAttr)

LRNGradAttr is an optional argument to LRNGrad.

func LRNGradAlpha 

func LRNGradAlpha(value float32) LRNGradAttr

LRNGradAlpha sets the optional alpha attribute to value.

value: A scale factor, usually positive. If not specified, defaults to 1

func LRNGradBeta 

func LRNGradBeta(value float32) LRNGradAttr

LRNGradBeta sets the optional beta attribute to value.

value: An exponent. If not specified, defaults to 0.5

func LRNGradBias 

func LRNGradBias(value float32) LRNGradAttr

LRNGradBias sets the optional bias attribute to value.

value: An offset (usually > 0 to avoid dividing by 0). If not specified, defaults to 1

func LRNGradDepthRadius 

func LRNGradDepthRadius(value int64) LRNGradAttr

LRNGradDepthRadius sets the optional depth_radius attribute to value.

value: A depth radius. If not specified, defaults to 5

type LSTMBlockCellAttr 

type LSTMBlockCellAttr func(optionalAttr)

LSTMBlockCellAttr is an optional argument to LSTMBlockCell.

func LSTMBlockCellCellClip 

func LSTMBlockCellCellClip(value float32) LSTMBlockCellAttr

LSTMBlockCellCellClip sets the optional cell_clip attribute to value.

value: Value to clip the 'cs' value to. If not specified, defaults to 3

func LSTMBlockCellForgetBias 

func LSTMBlockCellForgetBias(value float32) LSTMBlockCellAttr

LSTMBlockCellForgetBias sets the optional forget_bias attribute to value.

value: The forget gate bias. If not specified, defaults to 1

func LSTMBlockCellUsePeephole 

func LSTMBlockCellUsePeephole(value bool) LSTMBlockCellAttr

LSTMBlockCellUsePeephole sets the optional use_peephole attribute to value.

value: Whether to use peephole weights. If not specified, defaults to false

type LeakyReluAttr 

type LeakyReluAttr func(optionalAttr)

LeakyReluAttr is an optional argument to LeakyRelu.

func LeakyReluAlpha 

func LeakyReluAlpha(value float32) LeakyReluAttr

LeakyReluAlpha sets the optional alpha attribute to value. If not specified, defaults to 0.2

type LeakyReluGradAttr 

type LeakyReluGradAttr func(optionalAttr)

LeakyReluGradAttr is an optional argument to LeakyReluGrad.

func LeakyReluGradAlpha 

func LeakyReluGradAlpha(value float32) LeakyReluGradAttr

LeakyReluGradAlpha sets the optional alpha attribute to value. If not specified, defaults to 0.2

type LearnedUnigramCandidateSamplerAttr 

type LearnedUnigramCandidateSamplerAttr func(optionalAttr)

LearnedUnigramCandidateSamplerAttr is an optional argument to LearnedUnigramCandidateSampler.

func LearnedUnigramCandidateSamplerSeed 

func LearnedUnigramCandidateSamplerSeed(value int64) LearnedUnigramCandidateSamplerAttr

LearnedUnigramCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func LearnedUnigramCandidateSamplerSeed2 

func LearnedUnigramCandidateSamplerSeed2(value int64) LearnedUnigramCandidateSamplerAttr

LearnedUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type ListDatasetAttr added in v0.2.0 

type ListDatasetAttr func(optionalAttr)

ListDatasetAttr is an optional argument to ListDataset.

func ListDatasetMetadata added in v0.2.0 

func ListDatasetMetadata(value string) ListDatasetAttr

ListDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type ListDiffAttr 

type ListDiffAttr func(optionalAttr)

ListDiffAttr is an optional argument to ListDiff.

func ListDiffOutIdx 

func ListDiffOutIdx(value tf.DataType) ListDiffAttr

ListDiffOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type LoadAndRemapMatrixAttr 

type LoadAndRemapMatrixAttr func(optionalAttr)

LoadAndRemapMatrixAttr is an optional argument to LoadAndRemapMatrix.

func LoadAndRemapMatrixMaxRowsInMemory 

func LoadAndRemapMatrixMaxRowsInMemory(value int64) LoadAndRemapMatrixAttr

LoadAndRemapMatrixMaxRowsInMemory sets the optional max_rows_in_memory attribute to value.

value: The maximum number of rows to load from the checkpoint at once. If less than or equal to 0, the entire matrix will be loaded into memory. Setting this arg trades increased disk reads for lower memory usage. If not specified, defaults to -1

type LoadTPUEmbeddingADAMParametersAttr 

type LoadTPUEmbeddingADAMParametersAttr func(optionalAttr)

LoadTPUEmbeddingADAMParametersAttr is an optional argument to LoadTPUEmbeddingADAMParameters.

func LoadTPUEmbeddingADAMParametersConfig 

func LoadTPUEmbeddingADAMParametersConfig(value string) LoadTPUEmbeddingADAMParametersAttr

LoadTPUEmbeddingADAMParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingADAMParametersTableId 

func LoadTPUEmbeddingADAMParametersTableId(value int64) LoadTPUEmbeddingADAMParametersAttr

LoadTPUEmbeddingADAMParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingADAMParametersTableName 

func LoadTPUEmbeddingADAMParametersTableName(value string) LoadTPUEmbeddingADAMParametersAttr

LoadTPUEmbeddingADAMParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdadeltaParametersAttr 

type LoadTPUEmbeddingAdadeltaParametersAttr func(optionalAttr)

LoadTPUEmbeddingAdadeltaParametersAttr is an optional argument to LoadTPUEmbeddingAdadeltaParameters.

func LoadTPUEmbeddingAdadeltaParametersConfig 

func LoadTPUEmbeddingAdadeltaParametersConfig(value string) LoadTPUEmbeddingAdadeltaParametersAttr

LoadTPUEmbeddingAdadeltaParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingAdadeltaParametersTableId 

func LoadTPUEmbeddingAdadeltaParametersTableId(value int64) LoadTPUEmbeddingAdadeltaParametersAttr

LoadTPUEmbeddingAdadeltaParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingAdadeltaParametersTableName 

func LoadTPUEmbeddingAdadeltaParametersTableName(value string) LoadTPUEmbeddingAdadeltaParametersAttr

LoadTPUEmbeddingAdadeltaParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdagradMomentumParametersAttr 

type LoadTPUEmbeddingAdagradMomentumParametersAttr func(optionalAttr)

LoadTPUEmbeddingAdagradMomentumParametersAttr is an optional argument to LoadTPUEmbeddingAdagradMomentumParameters.

func LoadTPUEmbeddingAdagradMomentumParametersConfig 

func LoadTPUEmbeddingAdagradMomentumParametersConfig(value string) LoadTPUEmbeddingAdagradMomentumParametersAttr

LoadTPUEmbeddingAdagradMomentumParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingAdagradMomentumParametersTableId 

func LoadTPUEmbeddingAdagradMomentumParametersTableId(value int64) LoadTPUEmbeddingAdagradMomentumParametersAttr

LoadTPUEmbeddingAdagradMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingAdagradMomentumParametersTableName 

func LoadTPUEmbeddingAdagradMomentumParametersTableName(value string) LoadTPUEmbeddingAdagradMomentumParametersAttr

LoadTPUEmbeddingAdagradMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdagradParametersAttr 

type LoadTPUEmbeddingAdagradParametersAttr func(optionalAttr)

LoadTPUEmbeddingAdagradParametersAttr is an optional argument to LoadTPUEmbeddingAdagradParameters.

func LoadTPUEmbeddingAdagradParametersConfig 

func LoadTPUEmbeddingAdagradParametersConfig(value string) LoadTPUEmbeddingAdagradParametersAttr

LoadTPUEmbeddingAdagradParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingAdagradParametersTableId 

func LoadTPUEmbeddingAdagradParametersTableId(value int64) LoadTPUEmbeddingAdagradParametersAttr

LoadTPUEmbeddingAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingAdagradParametersTableName 

func LoadTPUEmbeddingAdagradParametersTableName(value string) LoadTPUEmbeddingAdagradParametersAttr

LoadTPUEmbeddingAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingCenteredRMSPropParametersAttr 

type LoadTPUEmbeddingCenteredRMSPropParametersAttr func(optionalAttr)

LoadTPUEmbeddingCenteredRMSPropParametersAttr is an optional argument to LoadTPUEmbeddingCenteredRMSPropParameters.

func LoadTPUEmbeddingCenteredRMSPropParametersConfig 

func LoadTPUEmbeddingCenteredRMSPropParametersConfig(value string) LoadTPUEmbeddingCenteredRMSPropParametersAttr

LoadTPUEmbeddingCenteredRMSPropParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingCenteredRMSPropParametersTableId 

func LoadTPUEmbeddingCenteredRMSPropParametersTableId(value int64) LoadTPUEmbeddingCenteredRMSPropParametersAttr

LoadTPUEmbeddingCenteredRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingCenteredRMSPropParametersTableName 

func LoadTPUEmbeddingCenteredRMSPropParametersTableName(value string) LoadTPUEmbeddingCenteredRMSPropParametersAttr

LoadTPUEmbeddingCenteredRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingFTRLParametersAttr 

type LoadTPUEmbeddingFTRLParametersAttr func(optionalAttr)

LoadTPUEmbeddingFTRLParametersAttr is an optional argument to LoadTPUEmbeddingFTRLParameters.

func LoadTPUEmbeddingFTRLParametersConfig 

func LoadTPUEmbeddingFTRLParametersConfig(value string) LoadTPUEmbeddingFTRLParametersAttr

LoadTPUEmbeddingFTRLParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingFTRLParametersTableId 

func LoadTPUEmbeddingFTRLParametersTableId(value int64) LoadTPUEmbeddingFTRLParametersAttr

LoadTPUEmbeddingFTRLParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingFTRLParametersTableName 

func LoadTPUEmbeddingFTRLParametersTableName(value string) LoadTPUEmbeddingFTRLParametersAttr

LoadTPUEmbeddingFTRLParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingFrequencyEstimatorParametersAttr 

type LoadTPUEmbeddingFrequencyEstimatorParametersAttr func(optionalAttr)

LoadTPUEmbeddingFrequencyEstimatorParametersAttr is an optional argument to LoadTPUEmbeddingFrequencyEstimatorParameters.

func LoadTPUEmbeddingFrequencyEstimatorParametersConfig 

func LoadTPUEmbeddingFrequencyEstimatorParametersConfig(value string) LoadTPUEmbeddingFrequencyEstimatorParametersAttr

LoadTPUEmbeddingFrequencyEstimatorParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingFrequencyEstimatorParametersTableId 

func LoadTPUEmbeddingFrequencyEstimatorParametersTableId(value int64) LoadTPUEmbeddingFrequencyEstimatorParametersAttr

LoadTPUEmbeddingFrequencyEstimatorParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingFrequencyEstimatorParametersTableName 

func LoadTPUEmbeddingFrequencyEstimatorParametersTableName(value string) LoadTPUEmbeddingFrequencyEstimatorParametersAttr

LoadTPUEmbeddingFrequencyEstimatorParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingMDLAdagradLightParametersAttr 

type LoadTPUEmbeddingMDLAdagradLightParametersAttr func(optionalAttr)

LoadTPUEmbeddingMDLAdagradLightParametersAttr is an optional argument to LoadTPUEmbeddingMDLAdagradLightParameters.

func LoadTPUEmbeddingMDLAdagradLightParametersConfig 

func LoadTPUEmbeddingMDLAdagradLightParametersConfig(value string) LoadTPUEmbeddingMDLAdagradLightParametersAttr

LoadTPUEmbeddingMDLAdagradLightParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingMDLAdagradLightParametersTableId 

func LoadTPUEmbeddingMDLAdagradLightParametersTableId(value int64) LoadTPUEmbeddingMDLAdagradLightParametersAttr

LoadTPUEmbeddingMDLAdagradLightParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingMDLAdagradLightParametersTableName 

func LoadTPUEmbeddingMDLAdagradLightParametersTableName(value string) LoadTPUEmbeddingMDLAdagradLightParametersAttr

LoadTPUEmbeddingMDLAdagradLightParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingMomentumParametersAttr 

type LoadTPUEmbeddingMomentumParametersAttr func(optionalAttr)

LoadTPUEmbeddingMomentumParametersAttr is an optional argument to LoadTPUEmbeddingMomentumParameters.

func LoadTPUEmbeddingMomentumParametersConfig 

func LoadTPUEmbeddingMomentumParametersConfig(value string) LoadTPUEmbeddingMomentumParametersAttr

LoadTPUEmbeddingMomentumParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingMomentumParametersTableId 

func LoadTPUEmbeddingMomentumParametersTableId(value int64) LoadTPUEmbeddingMomentumParametersAttr

LoadTPUEmbeddingMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingMomentumParametersTableName 

func LoadTPUEmbeddingMomentumParametersTableName(value string) LoadTPUEmbeddingMomentumParametersAttr

LoadTPUEmbeddingMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingProximalAdagradParametersAttr 

type LoadTPUEmbeddingProximalAdagradParametersAttr func(optionalAttr)

LoadTPUEmbeddingProximalAdagradParametersAttr is an optional argument to LoadTPUEmbeddingProximalAdagradParameters.

func LoadTPUEmbeddingProximalAdagradParametersConfig 

func LoadTPUEmbeddingProximalAdagradParametersConfig(value string) LoadTPUEmbeddingProximalAdagradParametersAttr

LoadTPUEmbeddingProximalAdagradParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingProximalAdagradParametersTableId 

func LoadTPUEmbeddingProximalAdagradParametersTableId(value int64) LoadTPUEmbeddingProximalAdagradParametersAttr

LoadTPUEmbeddingProximalAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingProximalAdagradParametersTableName 

func LoadTPUEmbeddingProximalAdagradParametersTableName(value string) LoadTPUEmbeddingProximalAdagradParametersAttr

LoadTPUEmbeddingProximalAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingRMSPropParametersAttr 

type LoadTPUEmbeddingRMSPropParametersAttr func(optionalAttr)

LoadTPUEmbeddingRMSPropParametersAttr is an optional argument to LoadTPUEmbeddingRMSPropParameters.

func LoadTPUEmbeddingRMSPropParametersConfig 

func LoadTPUEmbeddingRMSPropParametersConfig(value string) LoadTPUEmbeddingRMSPropParametersAttr

LoadTPUEmbeddingRMSPropParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingRMSPropParametersTableId 

func LoadTPUEmbeddingRMSPropParametersTableId(value int64) LoadTPUEmbeddingRMSPropParametersAttr

LoadTPUEmbeddingRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingRMSPropParametersTableName 

func LoadTPUEmbeddingRMSPropParametersTableName(value string) LoadTPUEmbeddingRMSPropParametersAttr

LoadTPUEmbeddingRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingStochasticGradientDescentParametersAttr 

type LoadTPUEmbeddingStochasticGradientDescentParametersAttr func(optionalAttr)

LoadTPUEmbeddingStochasticGradientDescentParametersAttr is an optional argument to LoadTPUEmbeddingStochasticGradientDescentParameters.

func LoadTPUEmbeddingStochasticGradientDescentParametersConfig 

func LoadTPUEmbeddingStochasticGradientDescentParametersConfig(value string) LoadTPUEmbeddingStochasticGradientDescentParametersAttr

LoadTPUEmbeddingStochasticGradientDescentParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func LoadTPUEmbeddingStochasticGradientDescentParametersTableId 

func LoadTPUEmbeddingStochasticGradientDescentParametersTableId(value int64) LoadTPUEmbeddingStochasticGradientDescentParametersAttr

LoadTPUEmbeddingStochasticGradientDescentParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func LoadTPUEmbeddingStochasticGradientDescentParametersTableName 

func LoadTPUEmbeddingStochasticGradientDescentParametersTableName(value string) LoadTPUEmbeddingStochasticGradientDescentParametersAttr

LoadTPUEmbeddingStochasticGradientDescentParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LogUniformCandidateSamplerAttr 

type LogUniformCandidateSamplerAttr func(optionalAttr)

LogUniformCandidateSamplerAttr is an optional argument to LogUniformCandidateSampler.

func LogUniformCandidateSamplerSeed 

func LogUniformCandidateSamplerSeed(value int64) LogUniformCandidateSamplerAttr

LogUniformCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func LogUniformCandidateSamplerSeed2 

func LogUniformCandidateSamplerSeed2(value int64) LogUniformCandidateSamplerAttr

LogUniformCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type LowerBoundAttr 

type LowerBoundAttr func(optionalAttr)

LowerBoundAttr is an optional argument to LowerBound.

func LowerBoundOutType 

func LowerBoundOutType(value tf.DataType) LowerBoundAttr

LowerBoundOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type LuAttr 

type LuAttr func(optionalAttr)

LuAttr is an optional argument to Lu.

func LuOutputIdxType 

func LuOutputIdxType(value tf.DataType) LuAttr

LuOutputIdxType sets the optional output_idx_type attribute to value. If not specified, defaults to DT_INT32

type MapClearAttr 

type MapClearAttr func(optionalAttr)

MapClearAttr is an optional argument to MapClear.

func MapClearCapacity 

func MapClearCapacity(value int64) MapClearAttr

MapClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapClearContainer 

func MapClearContainer(value string) MapClearAttr

MapClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapClearMemoryLimit 

func MapClearMemoryLimit(value int64) MapClearAttr

MapClearMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapClearSharedName 

func MapClearSharedName(value string) MapClearAttr

MapClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapIncompleteSizeAttr 

type MapIncompleteSizeAttr func(optionalAttr)

MapIncompleteSizeAttr is an optional argument to MapIncompleteSize.

func MapIncompleteSizeCapacity 

func MapIncompleteSizeCapacity(value int64) MapIncompleteSizeAttr

MapIncompleteSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapIncompleteSizeContainer 

func MapIncompleteSizeContainer(value string) MapIncompleteSizeAttr

MapIncompleteSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapIncompleteSizeMemoryLimit 

func MapIncompleteSizeMemoryLimit(value int64) MapIncompleteSizeAttr

MapIncompleteSizeMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapIncompleteSizeSharedName 

func MapIncompleteSizeSharedName(value string) MapIncompleteSizeAttr

MapIncompleteSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapPeekAttr 

type MapPeekAttr func(optionalAttr)

MapPeekAttr is an optional argument to MapPeek.

func MapPeekCapacity 

func MapPeekCapacity(value int64) MapPeekAttr

MapPeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapPeekContainer 

func MapPeekContainer(value string) MapPeekAttr

MapPeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapPeekMemoryLimit 

func MapPeekMemoryLimit(value int64) MapPeekAttr

MapPeekMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapPeekSharedName 

func MapPeekSharedName(value string) MapPeekAttr

MapPeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapSizeAttr 

type MapSizeAttr func(optionalAttr)

MapSizeAttr is an optional argument to MapSize.

func MapSizeCapacity 

func MapSizeCapacity(value int64) MapSizeAttr

MapSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapSizeContainer 

func MapSizeContainer(value string) MapSizeAttr

MapSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapSizeMemoryLimit 

func MapSizeMemoryLimit(value int64) MapSizeAttr

MapSizeMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapSizeSharedName 

func MapSizeSharedName(value string) MapSizeAttr

MapSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapStageAttr 

type MapStageAttr func(optionalAttr)

MapStageAttr is an optional argument to MapStage.

func MapStageCapacity 

func MapStageCapacity(value int64) MapStageAttr

MapStageCapacity sets the optional capacity attribute to value.

value: Maximum number of elements in the Staging Area. If > 0, inserts on the container will block when the capacity is reached. If not specified, defaults to 0

REQUIRES: value >= 0

func MapStageContainer 

func MapStageContainer(value string) MapStageAttr

MapStageContainer sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func MapStageMemoryLimit 

func MapStageMemoryLimit(value int64) MapStageAttr

MapStageMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapStageSharedName 

func MapStageSharedName(value string) MapStageAttr

MapStageSharedName sets the optional shared_name attribute to value.

value: It is necessary to match this name to the matching Unstage Op. If not specified, defaults to ""

type MapUnstageAttr 

type MapUnstageAttr func(optionalAttr)

MapUnstageAttr is an optional argument to MapUnstage.

func MapUnstageCapacity 

func MapUnstageCapacity(value int64) MapUnstageAttr

MapUnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageContainer 

func MapUnstageContainer(value string) MapUnstageAttr

MapUnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapUnstageMemoryLimit 

func MapUnstageMemoryLimit(value int64) MapUnstageAttr

MapUnstageMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageSharedName 

func MapUnstageSharedName(value string) MapUnstageAttr

MapUnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapUnstageNoKeyAttr 

type MapUnstageNoKeyAttr func(optionalAttr)

MapUnstageNoKeyAttr is an optional argument to MapUnstageNoKey.

func MapUnstageNoKeyCapacity 

func MapUnstageNoKeyCapacity(value int64) MapUnstageNoKeyAttr

MapUnstageNoKeyCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageNoKeyContainer 

func MapUnstageNoKeyContainer(value string) MapUnstageNoKeyAttr

MapUnstageNoKeyContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapUnstageNoKeyMemoryLimit 

func MapUnstageNoKeyMemoryLimit(value int64) MapUnstageNoKeyAttr

MapUnstageNoKeyMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageNoKeySharedName 

func MapUnstageNoKeySharedName(value string) MapUnstageNoKeyAttr

MapUnstageNoKeySharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MatMulAttr 

type MatMulAttr func(optionalAttr)

MatMulAttr is an optional argument to MatMul.

func MatMulGradA added in v0.8.0 

func MatMulGradA(value bool) MatMulAttr

MatMulGradA sets the optional grad_a attribute to value. If not specified, defaults to false

func MatMulGradB added in v0.8.0 

func MatMulGradB(value bool) MatMulAttr

MatMulGradB sets the optional grad_b attribute to value. If not specified, defaults to false

func MatMulTransposeA 

func MatMulTransposeA(value bool) MatMulAttr

MatMulTransposeA sets the optional transpose_a attribute to value.

value: If true, "a" is transposed before multiplication. If not specified, defaults to false

func MatMulTransposeB 

func MatMulTransposeB(value bool) MatMulAttr

MatMulTransposeB sets the optional transpose_b attribute to value.

value: If true, "b" is transposed before multiplication. If not specified, defaults to false

type MatrixDiagPartV3Attr 

type MatrixDiagPartV3Attr func(optionalAttr)

MatrixDiagPartV3Attr is an optional argument to MatrixDiagPartV3.

func MatrixDiagPartV3Align 

func MatrixDiagPartV3Align(value string) MatrixDiagPartV3Attr

MatrixDiagPartV3Align sets the optional align attribute to value.

value: Some diagonals are shorter than `max_diag_len` and need to be padded. `align` is a string specifying how superdiagonals and subdiagonals should be aligned, respectively. There are four possible alignments: "RIGHT_LEFT" (default), "LEFT_RIGHT", "LEFT_LEFT", and "RIGHT_RIGHT". "RIGHT_LEFT" aligns superdiagonals to the right (left-pads the row) and subdiagonals to the left (right-pads the row). It is the packing format LAPACK uses. cuSPARSE uses "LEFT_RIGHT", which is the opposite alignment. If not specified, defaults to "RIGHT_LEFT"

type MatrixDiagV3Attr 

type MatrixDiagV3Attr func(optionalAttr)

MatrixDiagV3Attr is an optional argument to MatrixDiagV3.

func MatrixDiagV3Align 

func MatrixDiagV3Align(value string) MatrixDiagV3Attr

MatrixDiagV3Align sets the optional align attribute to value.

value: Some diagonals are shorter than `max_diag_len` and need to be padded. `align` is a string specifying how superdiagonals and subdiagonals should be aligned, respectively. There are four possible alignments: "RIGHT_LEFT" (default), "LEFT_RIGHT", "LEFT_LEFT", and "RIGHT_RIGHT". "RIGHT_LEFT" aligns superdiagonals to the right (left-pads the row) and subdiagonals to the left (right-pads the row). It is the packing format LAPACK uses. cuSPARSE uses "LEFT_RIGHT", which is the opposite alignment. If not specified, defaults to "RIGHT_LEFT"

type MatrixInverseAttr 

type MatrixInverseAttr func(optionalAttr)

MatrixInverseAttr is an optional argument to MatrixInverse.

func MatrixInverseAdjoint 

func MatrixInverseAdjoint(value bool) MatrixInverseAttr

MatrixInverseAdjoint sets the optional adjoint attribute to value. If not specified, defaults to false

type MatrixSetDiagV3Attr 

type MatrixSetDiagV3Attr func(optionalAttr)

MatrixSetDiagV3Attr is an optional argument to MatrixSetDiagV3.

func MatrixSetDiagV3Align 

func MatrixSetDiagV3Align(value string) MatrixSetDiagV3Attr

MatrixSetDiagV3Align sets the optional align attribute to value.

value: Some diagonals are shorter than `max_diag_len` and need to be padded. `align` is a string specifying how superdiagonals and subdiagonals should be aligned, respectively. There are four possible alignments: "RIGHT_LEFT" (default), "LEFT_RIGHT", "LEFT_LEFT", and "RIGHT_RIGHT". "RIGHT_LEFT" aligns superdiagonals to the right (left-pads the row) and subdiagonals to the left (right-pads the row). It is the packing format LAPACK uses. cuSPARSE uses "LEFT_RIGHT", which is the opposite alignment. If not specified, defaults to "RIGHT_LEFT"

type MatrixSolveAttr 

type MatrixSolveAttr func(optionalAttr)

MatrixSolveAttr is an optional argument to MatrixSolve.

func MatrixSolveAdjoint 

func MatrixSolveAdjoint(value bool) MatrixSolveAttr

MatrixSolveAdjoint sets the optional adjoint attribute to value.

value: Boolean indicating whether to solve with `matrix` or its (block-wise) adjoint. If not specified, defaults to false

type MatrixSolveLsAttr 

type MatrixSolveLsAttr func(optionalAttr)

MatrixSolveLsAttr is an optional argument to MatrixSolveLs.

func MatrixSolveLsFast 

func MatrixSolveLsFast(value bool) MatrixSolveLsAttr

MatrixSolveLsFast sets the optional fast attribute to value. If not specified, defaults to true

type MatrixTriangularSolveAttr 

type MatrixTriangularSolveAttr func(optionalAttr)

MatrixTriangularSolveAttr is an optional argument to MatrixTriangularSolve.

func MatrixTriangularSolveAdjoint 

func MatrixTriangularSolveAdjoint(value bool) MatrixTriangularSolveAttr

MatrixTriangularSolveAdjoint sets the optional adjoint attribute to value.

value: Boolean indicating whether to solve with `matrix` or its (block-wise) 

adjoint.

@compatibility(numpy) Equivalent to scipy.linalg.solve_triangular @end_compatibility If not specified, defaults to false

func MatrixTriangularSolveLower 

func MatrixTriangularSolveLower(value bool) MatrixTriangularSolveAttr

MatrixTriangularSolveLower sets the optional lower attribute to value.

value: Boolean indicating whether the innermost matrices in `matrix` are lower or upper triangular. If not specified, defaults to true

type MaxAttr 

type MaxAttr func(optionalAttr)

MaxAttr is an optional argument to Max.

func MaxKeepDims 

func MaxKeepDims(value bool) MaxAttr

MaxKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type MaxPool3DAttr 

type MaxPool3DAttr func(optionalAttr)

MaxPool3DAttr is an optional argument to MaxPool3D.

func MaxPool3DDataFormat 

func MaxPool3DDataFormat(value string) MaxPool3DAttr

MaxPool3DDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

type MaxPool3DGradAttr 

type MaxPool3DGradAttr func(optionalAttr)

MaxPool3DGradAttr is an optional argument to MaxPool3DGrad.

func MaxPool3DGradDataFormat 

func MaxPool3DGradDataFormat(value string) MaxPool3DGradAttr

MaxPool3DGradDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

type MaxPool3DGradGradAttr 

type MaxPool3DGradGradAttr func(optionalAttr)

MaxPool3DGradGradAttr is an optional argument to MaxPool3DGradGrad.

func MaxPool3DGradGradDataFormat 

func MaxPool3DGradGradDataFormat(value string) MaxPool3DGradGradAttr

MaxPool3DGradGradDataFormat sets the optional data_format attribute to value.

value: The data format of the input and output data. With the default format "NDHWC", the data is stored in the order of: 

[batch, in_depth, in_height, in_width, in_channels].

Alternatively, the format could be "NCDHW", the data storage order is: 

[batch, in_channels, in_depth, in_height, in_width].

If not specified, defaults to "NDHWC"

type MaxPoolAttr 

type MaxPoolAttr func(optionalAttr)

MaxPoolAttr is an optional argument to MaxPool.

func MaxPoolDataFormat 

func MaxPoolDataFormat(value string) MaxPoolAttr

MaxPoolDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func MaxPoolExplicitPaddings 

func MaxPoolExplicitPaddings(value []int64) MaxPoolAttr

MaxPoolExplicitPaddings sets the optional explicit_paddings attribute to value. If not specified, defaults to {}

type MaxPoolGradAttr 

type MaxPoolGradAttr func(optionalAttr)

MaxPoolGradAttr is an optional argument to MaxPoolGrad.

func MaxPoolGradDataFormat 

func MaxPoolGradDataFormat(value string) MaxPoolGradAttr

MaxPoolGradDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

func MaxPoolGradExplicitPaddings 

func MaxPoolGradExplicitPaddings(value []int64) MaxPoolGradAttr

MaxPoolGradExplicitPaddings sets the optional explicit_paddings attribute to value. If not specified, defaults to {}

type MaxPoolGradGradAttr 

type MaxPoolGradGradAttr func(optionalAttr)

MaxPoolGradGradAttr is an optional argument to MaxPoolGradGrad.

func MaxPoolGradGradDataFormat 

func MaxPoolGradGradDataFormat(value string) MaxPoolGradGradAttr

MaxPoolGradGradDataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type MaxPoolGradGradV2Attr 

type MaxPoolGradGradV2Attr func(optionalAttr)

MaxPoolGradGradV2Attr is an optional argument to MaxPoolGradGradV2.

func MaxPoolGradGradV2DataFormat 

func MaxPoolGradGradV2DataFormat(value string) MaxPoolGradGradV2Attr

MaxPoolGradGradV2DataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type MaxPoolGradGradWithArgmaxAttr 

type MaxPoolGradGradWithArgmaxAttr func(optionalAttr)

MaxPoolGradGradWithArgmaxAttr is an optional argument to MaxPoolGradGradWithArgmax.

func MaxPoolGradGradWithArgmaxIncludeBatchInIndex 

func MaxPoolGradGradWithArgmaxIncludeBatchInIndex(value bool) MaxPoolGradGradWithArgmaxAttr

MaxPoolGradGradWithArgmaxIncludeBatchInIndex sets the optional include_batch_in_index attribute to value.

value: Whether to include batch dimension in flattened index of `argmax`. If not specified, defaults to false

type MaxPoolGradV2Attr 

type MaxPoolGradV2Attr func(optionalAttr)

MaxPoolGradV2Attr is an optional argument to MaxPoolGradV2.

func MaxPoolGradV2DataFormat 

func MaxPoolGradV2DataFormat(value string) MaxPoolGradV2Attr

MaxPoolGradV2DataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type MaxPoolGradWithArgmaxAttr 

type MaxPoolGradWithArgmaxAttr func(optionalAttr)

MaxPoolGradWithArgmaxAttr is an optional argument to MaxPoolGradWithArgmax.

func MaxPoolGradWithArgmaxIncludeBatchInIndex 

func MaxPoolGradWithArgmaxIncludeBatchInIndex(value bool) MaxPoolGradWithArgmaxAttr

MaxPoolGradWithArgmaxIncludeBatchInIndex sets the optional include_batch_in_index attribute to value.

value: Whether to include batch dimension in flattened index of `argmax`. If not specified, defaults to false

type MaxPoolV2Attr 

type MaxPoolV2Attr func(optionalAttr)

MaxPoolV2Attr is an optional argument to MaxPoolV2.

func MaxPoolV2DataFormat 

func MaxPoolV2DataFormat(value string) MaxPoolV2Attr

MaxPoolV2DataFormat sets the optional data_format attribute to value.

value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of: 

[batch, in_height, in_width, in_channels].

Alternatively, the format could be "NCHW", the data storage order of: 

[batch, in_channels, in_height, in_width].

If not specified, defaults to "NHWC"

type MaxPoolWithArgmaxAttr 

type MaxPoolWithArgmaxAttr func(optionalAttr)

MaxPoolWithArgmaxAttr is an optional argument to MaxPoolWithArgmax.

func MaxPoolWithArgmaxIncludeBatchInIndex 

func MaxPoolWithArgmaxIncludeBatchInIndex(value bool) MaxPoolWithArgmaxAttr

MaxPoolWithArgmaxIncludeBatchInIndex sets the optional include_batch_in_index attribute to value.

value: Whether to include batch dimension in flattened index of `argmax`. If not specified, defaults to false

func MaxPoolWithArgmaxTargmax 

func MaxPoolWithArgmaxTargmax(value tf.DataType) MaxPoolWithArgmaxAttr

MaxPoolWithArgmaxTargmax sets the optional Targmax attribute to value. If not specified, defaults to DT_INT64

type MeanAttr 

type MeanAttr func(optionalAttr)

MeanAttr is an optional argument to Mean.

func MeanKeepDims 

func MeanKeepDims(value bool) MeanAttr

MeanKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type MergeDedupDataAttr added in v0.5.0 

type MergeDedupDataAttr func(optionalAttr)

MergeDedupDataAttr is an optional argument to MergeDedupData.

func MergeDedupDataConfig added in v0.5.0 

func MergeDedupDataConfig(value string) MergeDedupDataAttr

MergeDedupDataConfig sets the optional config attribute to value. If not specified, defaults to ""

type MergeV2CheckpointsAttr 

type MergeV2CheckpointsAttr func(optionalAttr)

MergeV2CheckpointsAttr is an optional argument to MergeV2Checkpoints.

func MergeV2CheckpointsAllowMissingFiles added in v0.2.0 

func MergeV2CheckpointsAllowMissingFiles(value bool) MergeV2CheckpointsAttr

MergeV2CheckpointsAllowMissingFiles sets the optional allow_missing_files attribute to value.

value: see above. If not specified, defaults to false

func MergeV2CheckpointsDeleteOldDirs 

func MergeV2CheckpointsDeleteOldDirs(value bool) MergeV2CheckpointsAttr

MergeV2CheckpointsDeleteOldDirs sets the optional delete_old_dirs attribute to value.

value: see above. If not specified, defaults to true

type MfccAttr 

type MfccAttr func(optionalAttr)

MfccAttr is an optional argument to Mfcc.

func MfccDctCoefficientCount 

func MfccDctCoefficientCount(value int64) MfccAttr

MfccDctCoefficientCount sets the optional dct_coefficient_count attribute to value.

value: How many output channels to produce per time slice. If not specified, defaults to 13

func MfccFilterbankChannelCount 

func MfccFilterbankChannelCount(value int64) MfccAttr

MfccFilterbankChannelCount sets the optional filterbank_channel_count attribute to value.

value: Resolution of the Mel bank used internally. If not specified, defaults to 40

func MfccLowerFrequencyLimit 

func MfccLowerFrequencyLimit(value float32) MfccAttr

MfccLowerFrequencyLimit sets the optional lower_frequency_limit attribute to value.

value: The lowest frequency to use when calculating the ceptstrum. If not specified, defaults to 20

func MfccUpperFrequencyLimit 

func MfccUpperFrequencyLimit(value float32) MfccAttr

MfccUpperFrequencyLimit sets the optional upper_frequency_limit attribute to value.

value: The highest frequency to use when calculating the ceptstrum. If not specified, defaults to 4000

type MinAttr 

type MinAttr func(optionalAttr)

MinAttr is an optional argument to Min.

func MinKeepDims 

func MinKeepDims(value bool) MinAttr

MinKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type ModelDatasetAttr 

type ModelDatasetAttr func(optionalAttr)

ModelDatasetAttr is an optional argument to ModelDataset.

func ModelDatasetAlgorithm 

func ModelDatasetAlgorithm(value int64) ModelDatasetAttr

ModelDatasetAlgorithm sets the optional algorithm attribute to value. If not specified, defaults to 0

func ModelDatasetCpuBudget 

func ModelDatasetCpuBudget(value int64) ModelDatasetAttr

ModelDatasetCpuBudget sets the optional cpu_budget attribute to value. If not specified, defaults to 0

func ModelDatasetRamBudget 

func ModelDatasetRamBudget(value int64) ModelDatasetAttr

ModelDatasetRamBudget sets the optional ram_budget attribute to value. If not specified, defaults to 0

type MultiDeviceIteratorFromStringHandleAttr 

type MultiDeviceIteratorFromStringHandleAttr func(optionalAttr)

MultiDeviceIteratorFromStringHandleAttr is an optional argument to MultiDeviceIteratorFromStringHandle.

func MultiDeviceIteratorFromStringHandleOutputShapes 

func MultiDeviceIteratorFromStringHandleOutputShapes(value []tf.Shape) MultiDeviceIteratorFromStringHandleAttr

MultiDeviceIteratorFromStringHandleOutputShapes sets the optional output_shapes attribute to value.

value: The list of shapes being produced. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func MultiDeviceIteratorFromStringHandleOutputTypes 

func MultiDeviceIteratorFromStringHandleOutputTypes(value []tf.DataType) MultiDeviceIteratorFromStringHandleAttr

MultiDeviceIteratorFromStringHandleOutputTypes sets the optional output_types attribute to value.

value: The type list for the return values. If not specified, defaults to {}

REQUIRES: len(value) >= 0

type MultinomialAttr 

type MultinomialAttr func(optionalAttr)

MultinomialAttr is an optional argument to Multinomial.

func MultinomialOutputDtype 

func MultinomialOutputDtype(value tf.DataType) MultinomialAttr

MultinomialOutputDtype sets the optional output_dtype attribute to value. If not specified, defaults to DT_INT64

func MultinomialSeed 

func MultinomialSeed(value int64) MultinomialAttr

MultinomialSeed sets the optional seed attribute to value.

value: If either seed or seed2 is set to be non-zero, the internal random number generator is seeded by the given seed. Otherwise, a random seed is used. If not specified, defaults to 0

func MultinomialSeed2 

func MultinomialSeed2(value int64) MultinomialAttr

MultinomialSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type MutableDenseHashTableV2Attr 

type MutableDenseHashTableV2Attr func(optionalAttr)

MutableDenseHashTableV2Attr is an optional argument to MutableDenseHashTableV2.

func MutableDenseHashTableV2Container 

func MutableDenseHashTableV2Container(value string) MutableDenseHashTableV2Attr

MutableDenseHashTableV2Container sets the optional container attribute to value.

value: If non-empty, this table is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func MutableDenseHashTableV2InitialNumBuckets 

func MutableDenseHashTableV2InitialNumBuckets(value int64) MutableDenseHashTableV2Attr

MutableDenseHashTableV2InitialNumBuckets sets the optional initial_num_buckets attribute to value.

value: The initial number of hash table buckets. Must be a power to 2. If not specified, defaults to 131072

func MutableDenseHashTableV2MaxLoadFactor 

func MutableDenseHashTableV2MaxLoadFactor(value float32) MutableDenseHashTableV2Attr

MutableDenseHashTableV2MaxLoadFactor sets the optional max_load_factor attribute to value.

value: The maximum ratio between number of entries and number of buckets before growing the table. Must be between 0 and 1. If not specified, defaults to 0.8

func MutableDenseHashTableV2SharedName 

func MutableDenseHashTableV2SharedName(value string) MutableDenseHashTableV2Attr

MutableDenseHashTableV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this table is shared under the given name across multiple sessions. If not specified, defaults to ""

func MutableDenseHashTableV2UseNodeNameSharing 

func MutableDenseHashTableV2UseNodeNameSharing(value bool) MutableDenseHashTableV2Attr

MutableDenseHashTableV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value. If not specified, defaults to false

func MutableDenseHashTableV2ValueShape 

func MutableDenseHashTableV2ValueShape(value tf.Shape) MutableDenseHashTableV2Attr

MutableDenseHashTableV2ValueShape sets the optional value_shape attribute to value.

value: The shape of each value. If not specified, defaults to {}

type MutableHashTableOfTensorsV2Attr 

type MutableHashTableOfTensorsV2Attr func(optionalAttr)

MutableHashTableOfTensorsV2Attr is an optional argument to MutableHashTableOfTensorsV2.

func MutableHashTableOfTensorsV2Container 

func MutableHashTableOfTensorsV2Container(value string) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2Container sets the optional container attribute to value.

value: If non-empty, this table is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func MutableHashTableOfTensorsV2SharedName 

func MutableHashTableOfTensorsV2SharedName(value string) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this table is shared under the given name across multiple sessions. If not specified, defaults to ""

func MutableHashTableOfTensorsV2UseNodeNameSharing 

func MutableHashTableOfTensorsV2UseNodeNameSharing(value bool) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value. If not specified, defaults to false

func MutableHashTableOfTensorsV2ValueShape 

func MutableHashTableOfTensorsV2ValueShape(value tf.Shape) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2ValueShape sets the optional value_shape attribute to value. If not specified, defaults to {}

type MutableHashTableV2Attr 

type MutableHashTableV2Attr func(optionalAttr)

MutableHashTableV2Attr is an optional argument to MutableHashTableV2.

func MutableHashTableV2Container 

func MutableHashTableV2Container(value string) MutableHashTableV2Attr

MutableHashTableV2Container sets the optional container attribute to value.

value: If non-empty, this table is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func MutableHashTableV2SharedName 

func MutableHashTableV2SharedName(value string) MutableHashTableV2Attr

MutableHashTableV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this table is shared under the given name across multiple sessions. If not specified, defaults to ""

func MutableHashTableV2UseNodeNameSharing 

func MutableHashTableV2UseNodeNameSharing(value bool) MutableHashTableV2Attr

MutableHashTableV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value.

value: If true and shared_name is empty, the table is shared using the node name. If not specified, defaults to false

type MutexV2Attr 

type MutexV2Attr func(optionalAttr)

MutexV2Attr is an optional argument to MutexV2.

func MutexV2Container 

func MutexV2Container(value string) MutexV2Attr

MutexV2Container sets the optional container attribute to value.

value: If non-empty, this variable is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func MutexV2SharedName 

func MutexV2SharedName(value string) MutexV2Attr

MutexV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this variable is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

type NonDeterministicIntsAttr 

type NonDeterministicIntsAttr func(optionalAttr)

NonDeterministicIntsAttr is an optional argument to NonDeterministicInts.

func NonDeterministicIntsDtype 

func NonDeterministicIntsDtype(value tf.DataType) NonDeterministicIntsAttr

NonDeterministicIntsDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_INT64

type NonMaxSuppressionAttr 

type NonMaxSuppressionAttr func(optionalAttr)

NonMaxSuppressionAttr is an optional argument to NonMaxSuppression.

func NonMaxSuppressionIouThreshold 

func NonMaxSuppressionIouThreshold(value float32) NonMaxSuppressionAttr

NonMaxSuppressionIouThreshold sets the optional iou_threshold attribute to value.

value: A float representing the threshold for deciding whether boxes overlap too much with respect to IOU. If not specified, defaults to 0.5

type NonMaxSuppressionV4Attr 

type NonMaxSuppressionV4Attr func(optionalAttr)

NonMaxSuppressionV4Attr is an optional argument to NonMaxSuppressionV4.

func NonMaxSuppressionV4PadToMaxOutputSize 

func NonMaxSuppressionV4PadToMaxOutputSize(value bool) NonMaxSuppressionV4Attr

NonMaxSuppressionV4PadToMaxOutputSize sets the optional pad_to_max_output_size attribute to value.

value: If true, the output `selected_indices` is padded to be of length `max_output_size`. Defaults to false. If not specified, defaults to false

type NonMaxSuppressionV5Attr 

type NonMaxSuppressionV5Attr func(optionalAttr)

NonMaxSuppressionV5Attr is an optional argument to NonMaxSuppressionV5.

func NonMaxSuppressionV5PadToMaxOutputSize 

func NonMaxSuppressionV5PadToMaxOutputSize(value bool) NonMaxSuppressionV5Attr

NonMaxSuppressionV5PadToMaxOutputSize sets the optional pad_to_max_output_size attribute to value.

value: If true, the output `selected_indices` is padded to be of length `max_output_size`. Defaults to false. If not specified, defaults to false

type NotEqualAttr 

type NotEqualAttr func(optionalAttr)

NotEqualAttr is an optional argument to NotEqual.

func NotEqualIncompatibleShapeError 

func NotEqualIncompatibleShapeError(value bool) NotEqualAttr

NotEqualIncompatibleShapeError sets the optional incompatible_shape_error attribute to value. If not specified, defaults to true

type NthElementAttr 

type NthElementAttr func(optionalAttr)

NthElementAttr is an optional argument to NthElement.

func NthElementReverse 

func NthElementReverse(value bool) NthElementAttr

NthElementReverse sets the optional reverse attribute to value.

value: When set to True, find the nth-largest value in the vector and vice versa. If not specified, defaults to false

type OneHotAttr 

type OneHotAttr func(optionalAttr)

OneHotAttr is an optional argument to OneHot.

func OneHotAxis 

func OneHotAxis(value int64) OneHotAttr

OneHotAxis sets the optional axis attribute to value.

value: The axis to fill (default: -1, a new inner-most axis). If not specified, defaults to -1

type OptimizeDatasetAttr 

type OptimizeDatasetAttr func(optionalAttr)

OptimizeDatasetAttr is an optional argument to OptimizeDataset.

func OptimizeDatasetOptimizationConfigs 

func OptimizeDatasetOptimizationConfigs(value []string) OptimizeDatasetAttr

OptimizeDatasetOptimizationConfigs sets the optional optimization_configs attribute to value. If not specified, defaults to {}

type OptimizeDatasetV2Attr 

type OptimizeDatasetV2Attr func(optionalAttr)

OptimizeDatasetV2Attr is an optional argument to OptimizeDatasetV2.

func OptimizeDatasetV2OptimizationConfigs 

func OptimizeDatasetV2OptimizationConfigs(value []string) OptimizeDatasetV2Attr

OptimizeDatasetV2OptimizationConfigs sets the optional optimization_configs attribute to value. If not specified, defaults to {}

type OptionsDatasetAttr 

type OptionsDatasetAttr func(optionalAttr)

OptionsDatasetAttr is an optional argument to OptionsDataset.

func OptionsDatasetMetadata 

func OptionsDatasetMetadata(value string) OptionsDatasetAttr

OptionsDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type OrderedMapClearAttr 

type OrderedMapClearAttr func(optionalAttr)

OrderedMapClearAttr is an optional argument to OrderedMapClear.

func OrderedMapClearCapacity 

func OrderedMapClearCapacity(value int64) OrderedMapClearAttr

OrderedMapClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapClearContainer 

func OrderedMapClearContainer(value string) OrderedMapClearAttr

OrderedMapClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapClearMemoryLimit 

func OrderedMapClearMemoryLimit(value int64) OrderedMapClearAttr

OrderedMapClearMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapClearSharedName 

func OrderedMapClearSharedName(value string) OrderedMapClearAttr

OrderedMapClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapIncompleteSizeAttr 

type OrderedMapIncompleteSizeAttr func(optionalAttr)

OrderedMapIncompleteSizeAttr is an optional argument to OrderedMapIncompleteSize.

func OrderedMapIncompleteSizeCapacity 

func OrderedMapIncompleteSizeCapacity(value int64) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapIncompleteSizeContainer 

func OrderedMapIncompleteSizeContainer(value string) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapIncompleteSizeMemoryLimit 

func OrderedMapIncompleteSizeMemoryLimit(value int64) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapIncompleteSizeSharedName 

func OrderedMapIncompleteSizeSharedName(value string) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapPeekAttr 

type OrderedMapPeekAttr func(optionalAttr)

OrderedMapPeekAttr is an optional argument to OrderedMapPeek.

func OrderedMapPeekCapacity 

func OrderedMapPeekCapacity(value int64) OrderedMapPeekAttr

OrderedMapPeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapPeekContainer 

func OrderedMapPeekContainer(value string) OrderedMapPeekAttr

OrderedMapPeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapPeekMemoryLimit 

func OrderedMapPeekMemoryLimit(value int64) OrderedMapPeekAttr

OrderedMapPeekMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapPeekSharedName 

func OrderedMapPeekSharedName(value string) OrderedMapPeekAttr

OrderedMapPeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapSizeAttr 

type OrderedMapSizeAttr func(optionalAttr)

OrderedMapSizeAttr is an optional argument to OrderedMapSize.

func OrderedMapSizeCapacity 

func OrderedMapSizeCapacity(value int64) OrderedMapSizeAttr

OrderedMapSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapSizeContainer 

func OrderedMapSizeContainer(value string) OrderedMapSizeAttr

OrderedMapSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapSizeMemoryLimit 

func OrderedMapSizeMemoryLimit(value int64) OrderedMapSizeAttr

OrderedMapSizeMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapSizeSharedName 

func OrderedMapSizeSharedName(value string) OrderedMapSizeAttr

OrderedMapSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapStageAttr 

type OrderedMapStageAttr func(optionalAttr)

OrderedMapStageAttr is an optional argument to OrderedMapStage.

func OrderedMapStageCapacity 

func OrderedMapStageCapacity(value int64) OrderedMapStageAttr

OrderedMapStageCapacity sets the optional capacity attribute to value.

value: Maximum number of elements in the Staging Area. If > 0, inserts on the container will block when the capacity is reached. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapStageContainer 

func OrderedMapStageContainer(value string) OrderedMapStageAttr

OrderedMapStageContainer sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func OrderedMapStageMemoryLimit 

func OrderedMapStageMemoryLimit(value int64) OrderedMapStageAttr

OrderedMapStageMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapStageSharedName 

func OrderedMapStageSharedName(value string) OrderedMapStageAttr

OrderedMapStageSharedName sets the optional shared_name attribute to value.

value: It is necessary to match this name to the matching Unstage Op. If not specified, defaults to ""

type OrderedMapUnstageAttr 

type OrderedMapUnstageAttr func(optionalAttr)

OrderedMapUnstageAttr is an optional argument to OrderedMapUnstage.

func OrderedMapUnstageCapacity 

func OrderedMapUnstageCapacity(value int64) OrderedMapUnstageAttr

OrderedMapUnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageContainer 

func OrderedMapUnstageContainer(value string) OrderedMapUnstageAttr

OrderedMapUnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapUnstageMemoryLimit 

func OrderedMapUnstageMemoryLimit(value int64) OrderedMapUnstageAttr

OrderedMapUnstageMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageSharedName 

func OrderedMapUnstageSharedName(value string) OrderedMapUnstageAttr

OrderedMapUnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapUnstageNoKeyAttr 

type OrderedMapUnstageNoKeyAttr func(optionalAttr)

OrderedMapUnstageNoKeyAttr is an optional argument to OrderedMapUnstageNoKey.

func OrderedMapUnstageNoKeyCapacity 

func OrderedMapUnstageNoKeyCapacity(value int64) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeyCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageNoKeyContainer 

func OrderedMapUnstageNoKeyContainer(value string) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeyContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapUnstageNoKeyMemoryLimit 

func OrderedMapUnstageNoKeyMemoryLimit(value int64) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeyMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageNoKeySharedName 

func OrderedMapUnstageNoKeySharedName(value string) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeySharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OutfeedDequeueAttr 

type OutfeedDequeueAttr func(optionalAttr)

OutfeedDequeueAttr is an optional argument to OutfeedDequeue.

func OutfeedDequeueDeviceOrdinal 

func OutfeedDequeueDeviceOrdinal(value int64) OutfeedDequeueAttr

OutfeedDequeueDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. If not specified, defaults to -1

type OutfeedDequeueTupleAttr 

type OutfeedDequeueTupleAttr func(optionalAttr)

OutfeedDequeueTupleAttr is an optional argument to OutfeedDequeueTuple.

func OutfeedDequeueTupleDeviceOrdinal 

func OutfeedDequeueTupleDeviceOrdinal(value int64) OutfeedDequeueTupleAttr

OutfeedDequeueTupleDeviceOrdinal sets the optional device_ordinal attribute to value.

value: The TPU device to use. This should be -1 when the Op is running on a TPU device, and >= 0 when the Op is running on the CPU device. If not specified, defaults to -1

type PackAttr 

type PackAttr func(optionalAttr)

PackAttr is an optional argument to Pack.

func PackAxis 

func PackAxis(value int64) PackAttr

PackAxis sets the optional axis attribute to value.

value: Dimension along which to pack. Negative values wrap around, so the valid range is `[-(R+1), R+1)`. If not specified, defaults to 0

type PaddedBatchDatasetAttr 

type PaddedBatchDatasetAttr func(optionalAttr)

PaddedBatchDatasetAttr is an optional argument to PaddedBatchDataset.

func PaddedBatchDatasetMetadata 

func PaddedBatchDatasetMetadata(value string) PaddedBatchDatasetAttr

PaddedBatchDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type PaddedBatchDatasetV2Attr 

type PaddedBatchDatasetV2Attr func(optionalAttr)

PaddedBatchDatasetV2Attr is an optional argument to PaddedBatchDatasetV2.

func PaddedBatchDatasetV2Metadata 

func PaddedBatchDatasetV2Metadata(value string) PaddedBatchDatasetV2Attr

PaddedBatchDatasetV2Metadata sets the optional metadata attribute to value. If not specified, defaults to ""

func PaddedBatchDatasetV2ParallelCopy 

func PaddedBatchDatasetV2ParallelCopy(value bool) PaddedBatchDatasetV2Attr

PaddedBatchDatasetV2ParallelCopy sets the optional parallel_copy attribute to value. If not specified, defaults to false

type PaddingFIFOQueueV2Attr 

type PaddingFIFOQueueV2Attr func(optionalAttr)

PaddingFIFOQueueV2Attr is an optional argument to PaddingFIFOQueueV2.

func PaddingFIFOQueueV2Capacity 

func PaddingFIFOQueueV2Capacity(value int64) PaddingFIFOQueueV2Attr

PaddingFIFOQueueV2Capacity sets the optional capacity attribute to value.

value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1

func PaddingFIFOQueueV2Container 

func PaddingFIFOQueueV2Container(value string) PaddingFIFOQueueV2Attr

PaddingFIFOQueueV2Container sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func PaddingFIFOQueueV2Shapes 

func PaddingFIFOQueueV2Shapes(value []tf.Shape) PaddingFIFOQueueV2Attr

PaddingFIFOQueueV2Shapes sets the optional shapes attribute to value.

value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. Shapes of fixed rank but variable size are allowed by setting any shape dimension to -1. In this case, the inputs' shape may vary along the given dimension, and DequeueMany will pad the given dimension with zeros up to the maximum shape of all elements in the given batch. If the length of this attr is 0, different queue elements may have different ranks and shapes, but only one element may be dequeued at a time. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func PaddingFIFOQueueV2SharedName 

func PaddingFIFOQueueV2SharedName(value string) PaddingFIFOQueueV2Attr

PaddingFIFOQueueV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""

type ParameterizedTruncatedNormalAttr 

type ParameterizedTruncatedNormalAttr func(optionalAttr)

ParameterizedTruncatedNormalAttr is an optional argument to ParameterizedTruncatedNormal.

func ParameterizedTruncatedNormalSeed 

func ParameterizedTruncatedNormalSeed(value int64) ParameterizedTruncatedNormalAttr

ParameterizedTruncatedNormalSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func ParameterizedTruncatedNormalSeed2 

func ParameterizedTruncatedNormalSeed2(value int64) ParameterizedTruncatedNormalAttr

ParameterizedTruncatedNormalSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type ParseExampleDatasetAttr 

type ParseExampleDatasetAttr func(optionalAttr)

ParseExampleDatasetAttr is an optional argument to ParseExampleDataset.

func ParseExampleDatasetRaggedKeys 

func ParseExampleDatasetRaggedKeys(value []string) ParseExampleDatasetAttr

ParseExampleDatasetRaggedKeys sets the optional ragged_keys attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseExampleDatasetRaggedSplitTypes 

func ParseExampleDatasetRaggedSplitTypes(value []tf.DataType) ParseExampleDatasetAttr

ParseExampleDatasetRaggedSplitTypes sets the optional ragged_split_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseExampleDatasetRaggedValueTypes 

func ParseExampleDatasetRaggedValueTypes(value []tf.DataType) ParseExampleDatasetAttr

ParseExampleDatasetRaggedValueTypes sets the optional ragged_value_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseExampleDatasetSloppy 

func ParseExampleDatasetSloppy(value bool) ParseExampleDatasetAttr

ParseExampleDatasetSloppy sets the optional sloppy attribute to value. If not specified, defaults to false

type ParseExampleDatasetV2Attr 

type ParseExampleDatasetV2Attr func(optionalAttr)

ParseExampleDatasetV2Attr is an optional argument to ParseExampleDatasetV2.

func ParseExampleDatasetV2Deterministic 

func ParseExampleDatasetV2Deterministic(value string) ParseExampleDatasetV2Attr

ParseExampleDatasetV2Deterministic sets the optional deterministic attribute to value.

value: A string indicating the op-level determinism to use. Deterministic controls whether the dataset is allowed to return elements out of order if the next element to be returned isn't available, but a later element is. Options are "true", "false", and "default". "default" indicates that determinism should be decided by the `experimental_deterministic` parameter of `tf.data.Options`. If not specified, defaults to "default"

func ParseExampleDatasetV2RaggedKeys 

func ParseExampleDatasetV2RaggedKeys(value []string) ParseExampleDatasetV2Attr

ParseExampleDatasetV2RaggedKeys sets the optional ragged_keys attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseExampleDatasetV2RaggedSplitTypes 

func ParseExampleDatasetV2RaggedSplitTypes(value []tf.DataType) ParseExampleDatasetV2Attr

ParseExampleDatasetV2RaggedSplitTypes sets the optional ragged_split_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseExampleDatasetV2RaggedValueTypes 

func ParseExampleDatasetV2RaggedValueTypes(value []tf.DataType) ParseExampleDatasetV2Attr

ParseExampleDatasetV2RaggedValueTypes sets the optional ragged_value_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

type ParseSequenceExampleAttr 

type ParseSequenceExampleAttr func(optionalAttr)

ParseSequenceExampleAttr is an optional argument to ParseSequenceExample.

func ParseSequenceExampleContextDenseShapes 

func ParseSequenceExampleContextDenseShapes(value []tf.Shape) ParseSequenceExampleAttr

ParseSequenceExampleContextDenseShapes sets the optional context_dense_shapes attribute to value.

value: A list of Ncontext_dense shapes; the shapes of data in each context Feature given in context_dense_keys. The number of elements in the Feature corresponding to context_dense_key[j] must always equal context_dense_shapes[j].NumEntries(). The shape of context_dense_values[j] will match context_dense_shapes[j]. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleContextSparseTypes 

func ParseSequenceExampleContextSparseTypes(value []tf.DataType) ParseSequenceExampleAttr

ParseSequenceExampleContextSparseTypes sets the optional context_sparse_types attribute to value.

value: A list of Ncontext_sparse types; the data types of data in each context Feature given in context_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleFeatureListDenseShapes 

func ParseSequenceExampleFeatureListDenseShapes(value []tf.Shape) ParseSequenceExampleAttr

ParseSequenceExampleFeatureListDenseShapes sets the optional feature_list_dense_shapes attribute to value.

value: A list of Nfeature_list_dense shapes; the shapes of data in each FeatureList given in feature_list_dense_keys. The shape of each Feature in the FeatureList corresponding to feature_list_dense_key[j] must always equal feature_list_dense_shapes[j].NumEntries(). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleFeatureListDenseTypes 

func ParseSequenceExampleFeatureListDenseTypes(value []tf.DataType) ParseSequenceExampleAttr

ParseSequenceExampleFeatureListDenseTypes sets the optional feature_list_dense_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleFeatureListSparseTypes 

func ParseSequenceExampleFeatureListSparseTypes(value []tf.DataType) ParseSequenceExampleAttr

ParseSequenceExampleFeatureListSparseTypes sets the optional feature_list_sparse_types attribute to value.

value: A list of Nfeature_list_sparse types; the data types of data in each FeatureList given in feature_list_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleNcontextDense 

func ParseSequenceExampleNcontextDense(value int64) ParseSequenceExampleAttr

ParseSequenceExampleNcontextDense sets the optional Ncontext_dense attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func ParseSequenceExampleNcontextSparse 

func ParseSequenceExampleNcontextSparse(value int64) ParseSequenceExampleAttr

ParseSequenceExampleNcontextSparse sets the optional Ncontext_sparse attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func ParseSequenceExampleNfeatureListDense 

func ParseSequenceExampleNfeatureListDense(value int64) ParseSequenceExampleAttr

ParseSequenceExampleNfeatureListDense sets the optional Nfeature_list_dense attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func ParseSequenceExampleNfeatureListSparse 

func ParseSequenceExampleNfeatureListSparse(value int64) ParseSequenceExampleAttr

ParseSequenceExampleNfeatureListSparse sets the optional Nfeature_list_sparse attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

type ParseSequenceExampleV2Attr 

type ParseSequenceExampleV2Attr func(optionalAttr)

ParseSequenceExampleV2Attr is an optional argument to ParseSequenceExampleV2.

func ParseSequenceExampleV2ContextDenseShapes 

func ParseSequenceExampleV2ContextDenseShapes(value []tf.Shape) ParseSequenceExampleV2Attr

ParseSequenceExampleV2ContextDenseShapes sets the optional context_dense_shapes attribute to value.

value: A list of Ncontext_dense shapes; the shapes of data in each context Feature given in context_dense_keys. The number of elements in the Feature corresponding to context_dense_key[j] must always equal context_dense_shapes[j].NumEntries(). The shape of context_dense_values[j] will match context_dense_shapes[j]. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2ContextRaggedSplitTypes 

func ParseSequenceExampleV2ContextRaggedSplitTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2ContextRaggedSplitTypes sets the optional context_ragged_split_types attribute to value.

value: RaggedTensor.row_split dtypes for the ragged context features. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2ContextRaggedValueTypes 

func ParseSequenceExampleV2ContextRaggedValueTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2ContextRaggedValueTypes sets the optional context_ragged_value_types attribute to value.

value: RaggedTensor.value dtypes for the ragged context features. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2ContextSparseTypes 

func ParseSequenceExampleV2ContextSparseTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2ContextSparseTypes sets the optional context_sparse_types attribute to value.

value: A list of Ncontext_sparse types; the data types of data in each context Feature given in context_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2FeatureListDenseShapes 

func ParseSequenceExampleV2FeatureListDenseShapes(value []tf.Shape) ParseSequenceExampleV2Attr

ParseSequenceExampleV2FeatureListDenseShapes sets the optional feature_list_dense_shapes attribute to value.

value: A list of Nfeature_list_dense shapes; the shapes of data in each FeatureList given in feature_list_dense_keys. The shape of each Feature in the FeatureList corresponding to feature_list_dense_key[j] must always equal feature_list_dense_shapes[j].NumEntries(). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2FeatureListDenseTypes 

func ParseSequenceExampleV2FeatureListDenseTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2FeatureListDenseTypes sets the optional feature_list_dense_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2FeatureListRaggedSplitTypes 

func ParseSequenceExampleV2FeatureListRaggedSplitTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2FeatureListRaggedSplitTypes sets the optional feature_list_ragged_split_types attribute to value.

value: RaggedTensor.row_split dtypes for the ragged FeatureList features. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2FeatureListRaggedValueTypes 

func ParseSequenceExampleV2FeatureListRaggedValueTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2FeatureListRaggedValueTypes sets the optional feature_list_ragged_value_types attribute to value.

value: RaggedTensor.value dtypes for the ragged FeatureList features. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2FeatureListSparseTypes 

func ParseSequenceExampleV2FeatureListSparseTypes(value []tf.DataType) ParseSequenceExampleV2Attr

ParseSequenceExampleV2FeatureListSparseTypes sets the optional feature_list_sparse_types attribute to value.

value: A list of Nfeature_list_sparse types; the data types of data in each FeatureList given in feature_list_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSequenceExampleV2NcontextSparse 

func ParseSequenceExampleV2NcontextSparse(value int64) ParseSequenceExampleV2Attr

ParseSequenceExampleV2NcontextSparse sets the optional Ncontext_sparse attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func ParseSequenceExampleV2NfeatureListDense 

func ParseSequenceExampleV2NfeatureListDense(value int64) ParseSequenceExampleV2Attr

ParseSequenceExampleV2NfeatureListDense sets the optional Nfeature_list_dense attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func ParseSequenceExampleV2NfeatureListSparse 

func ParseSequenceExampleV2NfeatureListSparse(value int64) ParseSequenceExampleV2Attr

ParseSequenceExampleV2NfeatureListSparse sets the optional Nfeature_list_sparse attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

type ParseSingleSequenceExampleAttr 

type ParseSingleSequenceExampleAttr func(optionalAttr)

ParseSingleSequenceExampleAttr is an optional argument to ParseSingleSequenceExample.

func ParseSingleSequenceExampleContextDenseShapes 

func ParseSingleSequenceExampleContextDenseShapes(value []tf.Shape) ParseSingleSequenceExampleAttr

ParseSingleSequenceExampleContextDenseShapes sets the optional context_dense_shapes attribute to value.

value: A list of Ncontext_dense shapes; the shapes of data in each context Feature given in context_dense_keys. The number of elements in the Feature corresponding to context_dense_key[j] must always equal context_dense_shapes[j].NumEntries(). The shape of context_dense_values[j] will match context_dense_shapes[j]. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSingleSequenceExampleContextSparseTypes 

func ParseSingleSequenceExampleContextSparseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr

ParseSingleSequenceExampleContextSparseTypes sets the optional context_sparse_types attribute to value.

value: A list of Ncontext_sparse types; the data types of data in each context Feature given in context_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSingleSequenceExampleFeatureListDenseShapes 

func ParseSingleSequenceExampleFeatureListDenseShapes(value []tf.Shape) ParseSingleSequenceExampleAttr

ParseSingleSequenceExampleFeatureListDenseShapes sets the optional feature_list_dense_shapes attribute to value.

value: A list of Nfeature_list_dense shapes; the shapes of data in each FeatureList given in feature_list_dense_keys. The shape of each Feature in the FeatureList corresponding to feature_list_dense_key[j] must always equal feature_list_dense_shapes[j].NumEntries(). If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSingleSequenceExampleFeatureListDenseTypes 

func ParseSingleSequenceExampleFeatureListDenseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr

ParseSingleSequenceExampleFeatureListDenseTypes sets the optional feature_list_dense_types attribute to value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func ParseSingleSequenceExampleFeatureListSparseTypes 

func ParseSingleSequenceExampleFeatureListSparseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr

ParseSingleSequenceExampleFeatureListSparseTypes sets the optional feature_list_sparse_types attribute to value.

value: A list of Nfeature_list_sparse types; the data types of data in each FeatureList given in feature_list_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to {}

REQUIRES: len(value) >= 0

type PlaceholderAttr 

type PlaceholderAttr func(optionalAttr)

PlaceholderAttr is an optional argument to Placeholder.

func PlaceholderShape 

func PlaceholderShape(value tf.Shape) PlaceholderAttr

PlaceholderShape sets the optional shape attribute to value.

value: (Optional) The shape of the tensor. If the shape has 0 dimensions, the shape is unconstrained. If not specified, defaults to {unknown_rank:true}

type PrefetchDatasetAttr 

type PrefetchDatasetAttr func(optionalAttr)

PrefetchDatasetAttr is an optional argument to PrefetchDataset.

func PrefetchDatasetBufferSizeMin 

func PrefetchDatasetBufferSizeMin(value int64) PrefetchDatasetAttr

PrefetchDatasetBufferSizeMin sets the optional buffer_size_min attribute to value. If not specified, defaults to 0

func PrefetchDatasetLegacyAutotune 

func PrefetchDatasetLegacyAutotune(value bool) PrefetchDatasetAttr

PrefetchDatasetLegacyAutotune sets the optional legacy_autotune attribute to value. If not specified, defaults to true

func PrefetchDatasetMetadata 

func PrefetchDatasetMetadata(value string) PrefetchDatasetAttr

PrefetchDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func PrefetchDatasetSlackPeriod 

func PrefetchDatasetSlackPeriod(value int64) PrefetchDatasetAttr

PrefetchDatasetSlackPeriod sets the optional slack_period attribute to value. If not specified, defaults to 0

type PrelinearizeAttr 

type PrelinearizeAttr func(optionalAttr)

PrelinearizeAttr is an optional argument to Prelinearize.

func PrelinearizeLayout 

func PrelinearizeLayout(value []int64) PrelinearizeAttr

PrelinearizeLayout sets the optional layout attribute to value.

value: A vector holding the requested layout in minor-to-major sequence. If a layout attribute is passed but its values are all -1 the layout will be computed by the infeed operation. If not specified, defaults to {}

func PrelinearizeShape 

func PrelinearizeShape(value tf.Shape) PrelinearizeAttr

PrelinearizeShape sets the optional shape attribute to value.

value: The shape of the tensor. If not specified, defaults to {}

type PrelinearizeTupleAttr 

type PrelinearizeTupleAttr func(optionalAttr)

PrelinearizeTupleAttr is an optional argument to PrelinearizeTuple.

func PrelinearizeTupleLayouts 

func PrelinearizeTupleLayouts(value []int64) PrelinearizeTupleAttr

PrelinearizeTupleLayouts sets the optional layouts attribute to value.

value: A vector holding the requested layout in minor-to-major sequence for all the tuple shapes in the order the shapes appear in the "shapes" input. The layout elements for a sub-shape can be set to -1 in which case the corresponding layout will be computed by the infeed operation. If not specified, defaults to {}

type PreventGradientAttr 

type PreventGradientAttr func(optionalAttr)

PreventGradientAttr is an optional argument to PreventGradient.

func PreventGradientMessage 

func PreventGradientMessage(value string) PreventGradientAttr

PreventGradientMessage sets the optional message attribute to value.

value: Will be printed in the error when anyone tries to differentiate this operation. If not specified, defaults to ""

type PrintAttr 

type PrintAttr func(optionalAttr)

PrintAttr is an optional argument to Print.

func PrintFirstN 

func PrintFirstN(value int64) PrintAttr

PrintFirstN sets the optional first_n attribute to value.

value: Only log `first_n` number of times. -1 disables logging. If not specified, defaults to -1

func PrintMessage 

func PrintMessage(value string) PrintAttr

PrintMessage sets the optional message attribute to value.

value: A string, prefix of the error message. If not specified, defaults to ""

func PrintSummarize 

func PrintSummarize(value int64) PrintAttr

PrintSummarize sets the optional summarize attribute to value.

value: Only print this many entries of each tensor. If not specified, defaults to 3

type PrintV2Attr 

type PrintV2Attr func(optionalAttr)

PrintV2Attr is an optional argument to PrintV2.

func PrintV2End 

func PrintV2End(value string) PrintV2Attr

PrintV2End sets the optional end attribute to value. If not specified, defaults to "\n"

func PrintV2OutputStream 

func PrintV2OutputStream(value string) PrintV2Attr

PrintV2OutputStream sets the optional output_stream attribute to value.

value: A string specifying the output stream or logging level to print to. If not specified, defaults to "stderr"

type PriorityQueueV2Attr 

type PriorityQueueV2Attr func(optionalAttr)

PriorityQueueV2Attr is an optional argument to PriorityQueueV2.

func PriorityQueueV2Capacity 

func PriorityQueueV2Capacity(value int64) PriorityQueueV2Attr

PriorityQueueV2Capacity sets the optional capacity attribute to value.

value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1

func PriorityQueueV2ComponentTypes 

func PriorityQueueV2ComponentTypes(value []tf.DataType) PriorityQueueV2Attr

PriorityQueueV2ComponentTypes sets the optional component_types attribute to value.

value: The type of each component in a value. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func PriorityQueueV2Container 

func PriorityQueueV2Container(value string) PriorityQueueV2Attr

PriorityQueueV2Container sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func PriorityQueueV2SharedName 

func PriorityQueueV2SharedName(value string) PriorityQueueV2Attr

PriorityQueueV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""

type ProdAttr 

type ProdAttr func(optionalAttr)

ProdAttr is an optional argument to Prod.

func ProdKeepDims 

func ProdKeepDims(value bool) ProdAttr

ProdKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type QrAttr 

type QrAttr func(optionalAttr)

QrAttr is an optional argument to Qr.

func QrFullMatrices 

func QrFullMatrices(value bool) QrAttr

QrFullMatrices sets the optional full_matrices attribute to value.

value: If true, compute full-sized `q` and `r`. If false (the default), compute only the leading `P` columns of `q`. If not specified, defaults to false

type QuantizeAndDequantizeAttr 

type QuantizeAndDequantizeAttr func(optionalAttr)

QuantizeAndDequantizeAttr is an optional argument to QuantizeAndDequantize.

func QuantizeAndDequantizeInputMax 

func QuantizeAndDequantizeInputMax(value float32) QuantizeAndDequantizeAttr

QuantizeAndDequantizeInputMax sets the optional input_max attribute to value. If not specified, defaults to 0

func QuantizeAndDequantizeInputMin 

func QuantizeAndDequantizeInputMin(value float32) QuantizeAndDequantizeAttr

QuantizeAndDequantizeInputMin sets the optional input_min attribute to value. If not specified, defaults to 0

func QuantizeAndDequantizeNumBits 

func QuantizeAndDequantizeNumBits(value int64) QuantizeAndDequantizeAttr

QuantizeAndDequantizeNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

func QuantizeAndDequantizeRangeGiven 

func QuantizeAndDequantizeRangeGiven(value bool) QuantizeAndDequantizeAttr

QuantizeAndDequantizeRangeGiven sets the optional range_given attribute to value. If not specified, defaults to false

func QuantizeAndDequantizeSignedInput 

func QuantizeAndDequantizeSignedInput(value bool) QuantizeAndDequantizeAttr

QuantizeAndDequantizeSignedInput sets the optional signed_input attribute to value. If not specified, defaults to true

type QuantizeAndDequantizeV2Attr 

type QuantizeAndDequantizeV2Attr func(optionalAttr)

QuantizeAndDequantizeV2Attr is an optional argument to QuantizeAndDequantizeV2.

func QuantizeAndDequantizeV2Axis 

func QuantizeAndDequantizeV2Axis(value int64) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2Axis sets the optional axis attribute to value.

value: If specified, this axis is treated as a channel or slice axis, and a separate quantization range is used for each channel or slice along this axis. If not specified, defaults to -1

func QuantizeAndDequantizeV2NarrowRange 

func QuantizeAndDequantizeV2NarrowRange(value bool) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2NarrowRange sets the optional narrow_range attribute to value.

value: If True, then the absolute value of the quantized minimum value is the same as the quantized maximum value, instead of 1 greater. i.e. for 8 bit quantization, the minimum value is -127 instead of -128. If not specified, defaults to false

func QuantizeAndDequantizeV2NumBits 

func QuantizeAndDequantizeV2NumBits(value int64) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2NumBits sets the optional num_bits attribute to value.

value: The bitwidth of the quantization. If not specified, defaults to 8

func QuantizeAndDequantizeV2RangeGiven 

func QuantizeAndDequantizeV2RangeGiven(value bool) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2RangeGiven sets the optional range_given attribute to value.

value: Whether the range is given or should be determined from the `input` tensor. If not specified, defaults to false

func QuantizeAndDequantizeV2RoundMode 

func QuantizeAndDequantizeV2RoundMode(value string) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2RoundMode sets the optional round_mode attribute to value.

value: The 'round_mode' attribute controls which rounding tie-breaking algorithm is used when rounding float values to their quantized equivalents. The following rounding modes are currently supported:

  • HALF_TO_EVEN: this is the default round_mode.
  • HALF_UP: round towards positive. In this mode 7.5 rounds up to 8 and -7.5 rounds up to -7.

If not specified, defaults to "HALF_TO_EVEN"

func QuantizeAndDequantizeV2SignedInput 

func QuantizeAndDequantizeV2SignedInput(value bool) QuantizeAndDequantizeV2Attr

QuantizeAndDequantizeV2SignedInput sets the optional signed_input attribute to value.

value: Whether the quantization is signed or unsigned. (actually this parameter should have been called <b>`signed_output`</b>) If not specified, defaults to true

type QuantizeAndDequantizeV3Attr 

type QuantizeAndDequantizeV3Attr func(optionalAttr)

QuantizeAndDequantizeV3Attr is an optional argument to QuantizeAndDequantizeV3.

func QuantizeAndDequantizeV3Axis 

func QuantizeAndDequantizeV3Axis(value int64) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3Axis sets the optional axis attribute to value. If not specified, defaults to -1

func QuantizeAndDequantizeV3NarrowRange 

func QuantizeAndDequantizeV3NarrowRange(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3NarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func QuantizeAndDequantizeV3RangeGiven 

func QuantizeAndDequantizeV3RangeGiven(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3RangeGiven sets the optional range_given attribute to value. If not specified, defaults to true

func QuantizeAndDequantizeV3SignedInput 

func QuantizeAndDequantizeV3SignedInput(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3SignedInput sets the optional signed_input attribute to value. If not specified, defaults to true

type QuantizeAndDequantizeV4Attr 

type QuantizeAndDequantizeV4Attr func(optionalAttr)

QuantizeAndDequantizeV4Attr is an optional argument to QuantizeAndDequantizeV4.

func QuantizeAndDequantizeV4Axis 

func QuantizeAndDequantizeV4Axis(value int64) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4Axis sets the optional axis attribute to value.

value: If specified, this axis is treated as a channel or slice axis, and a separate quantization range is used for each channel or slice along this axis. If not specified, defaults to -1

func QuantizeAndDequantizeV4NarrowRange 

func QuantizeAndDequantizeV4NarrowRange(value bool) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4NarrowRange sets the optional narrow_range attribute to value.

value: If True, then the absolute value of the quantized minimum value is the same as the quantized maximum value, instead of 1 greater. i.e. for 8 bit quantization, the minimum value is -127 instead of -128. If not specified, defaults to false

func QuantizeAndDequantizeV4NumBits 

func QuantizeAndDequantizeV4NumBits(value int64) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4NumBits sets the optional num_bits attribute to value.

value: The bitwidth of the quantization. If not specified, defaults to 8

func QuantizeAndDequantizeV4RangeGiven 

func QuantizeAndDequantizeV4RangeGiven(value bool) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4RangeGiven sets the optional range_given attribute to value.

value: Whether the range is given or should be determined from the `input` tensor. If not specified, defaults to false

func QuantizeAndDequantizeV4RoundMode 

func QuantizeAndDequantizeV4RoundMode(value string) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4RoundMode sets the optional round_mode attribute to value.

value: The 'round_mode' attribute controls which rounding tie-breaking algorithm is used when rounding float values to their quantized equivalents. The following rounding modes are currently supported:

  • HALF_TO_EVEN: this is the default round_mode.
  • HALF_UP: round towards positive. In this mode 7.5 rounds up to 8 and -7.5 rounds up to -7.

If not specified, defaults to "HALF_TO_EVEN"

func QuantizeAndDequantizeV4SignedInput 

func QuantizeAndDequantizeV4SignedInput(value bool) QuantizeAndDequantizeV4Attr

QuantizeAndDequantizeV4SignedInput sets the optional signed_input attribute to value.

value: Whether the quantization is signed or unsigned. (actually this parameter should have been called <b>`signed_output`</b>) If not specified, defaults to true

type QuantizeAndDequantizeV4GradAttr 

type QuantizeAndDequantizeV4GradAttr func(optionalAttr)

QuantizeAndDequantizeV4GradAttr is an optional argument to QuantizeAndDequantizeV4Grad.

func QuantizeAndDequantizeV4GradAxis 

func QuantizeAndDequantizeV4GradAxis(value int64) QuantizeAndDequantizeV4GradAttr

QuantizeAndDequantizeV4GradAxis sets the optional axis attribute to value. If not specified, defaults to -1

type QuantizeV2Attr 

type QuantizeV2Attr func(optionalAttr)

QuantizeV2Attr is an optional argument to QuantizeV2.

func QuantizeV2Axis 

func QuantizeV2Axis(value int64) QuantizeV2Attr

QuantizeV2Axis sets the optional axis attribute to value. If not specified, defaults to -1

func QuantizeV2EnsureMinimumRange 

func QuantizeV2EnsureMinimumRange(value float32) QuantizeV2Attr

QuantizeV2EnsureMinimumRange sets the optional ensure_minimum_range attribute to value. If not specified, defaults to 0.01

func QuantizeV2Mode 

func QuantizeV2Mode(value string) QuantizeV2Attr

QuantizeV2Mode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"

func QuantizeV2NarrowRange 

func QuantizeV2NarrowRange(value bool) QuantizeV2Attr

QuantizeV2NarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func QuantizeV2RoundMode 

func QuantizeV2RoundMode(value string) QuantizeV2Attr

QuantizeV2RoundMode sets the optional round_mode attribute to value. If not specified, defaults to "HALF_AWAY_FROM_ZERO"

type QuantizedAddAttr 

type QuantizedAddAttr func(optionalAttr)

QuantizedAddAttr is an optional argument to QuantizedAdd.

func QuantizedAddToutput 

func QuantizedAddToutput(value tf.DataType) QuantizedAddAttr

QuantizedAddToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

type QuantizedConv2DAttr 

type QuantizedConv2DAttr func(optionalAttr)

QuantizedConv2DAttr is an optional argument to QuantizedConv2D.

func QuantizedConv2DDilations 

func QuantizedConv2DDilations(value []int64) QuantizedConv2DAttr

QuantizedConv2DDilations sets the optional dilations attribute to value.

value: 1-D tensor of length 4. The dilation factor for each dimension of `input`. If set to k > 1, there will be k-1 skipped cells between each filter element on that dimension. The dimension order is determined by the value of `data_format`, see above for details. Dilations in the batch and depth dimensions must be 1. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedConv2DOutType 

func QuantizedConv2DOutType(value tf.DataType) QuantizedConv2DAttr

QuantizedConv2DOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QINT32

type QuantizedConv2DPerChannelAttr 

type QuantizedConv2DPerChannelAttr func(optionalAttr)

QuantizedConv2DPerChannelAttr is an optional argument to QuantizedConv2DPerChannel.

func QuantizedConv2DPerChannelDilations 

func QuantizedConv2DPerChannelDilations(value []int64) QuantizedConv2DPerChannelAttr

QuantizedConv2DPerChannelDilations sets the optional dilations attribute to value.

value: list of dilation values. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedConv2DPerChannelOutType 

func QuantizedConv2DPerChannelOutType(value tf.DataType) QuantizedConv2DPerChannelAttr

QuantizedConv2DPerChannelOutType sets the optional out_type attribute to value.

value: The quantized type of output tensor that needs to be converted. If not specified, defaults to DT_QINT32

type QuantizedDepthwiseConv2DAttr 

type QuantizedDepthwiseConv2DAttr func(optionalAttr)

QuantizedDepthwiseConv2DAttr is an optional argument to QuantizedDepthwiseConv2D.

func QuantizedDepthwiseConv2DDilations 

func QuantizedDepthwiseConv2DDilations(value []int64) QuantizedDepthwiseConv2DAttr

QuantizedDepthwiseConv2DDilations sets the optional dilations attribute to value.

value: List of dilation values. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedDepthwiseConv2DOutType 

func QuantizedDepthwiseConv2DOutType(value tf.DataType) QuantizedDepthwiseConv2DAttr

QuantizedDepthwiseConv2DOutType sets the optional out_type attribute to value.

value: The type of the output. If not specified, defaults to DT_QINT32

type QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr 

type QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr is an optional argument to QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize.

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeDilations 

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeDilations(value []int64) QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr

QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeDilations sets the optional dilations attribute to value.

value: List of dilation values. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeOutType 

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeOutType(value tf.DataType) QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr

QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeOutType sets the optional out_type attribute to value.

value: The type of the output. If not specified, defaults to DT_QUINT8

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizePaddingList 

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizePaddingList(value []int64) QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr

QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizePaddingList sets the optional padding_list attribute to value. If not specified, defaults to {}

type QuantizedDepthwiseConv2DWithBiasAndReluAttr 

type QuantizedDepthwiseConv2DWithBiasAndReluAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAndReluAttr is an optional argument to QuantizedDepthwiseConv2DWithBiasAndRelu.

func QuantizedDepthwiseConv2DWithBiasAndReluDilations 

func QuantizedDepthwiseConv2DWithBiasAndReluDilations(value []int64) QuantizedDepthwiseConv2DWithBiasAndReluAttr

QuantizedDepthwiseConv2DWithBiasAndReluDilations sets the optional dilations attribute to value.

value: List of dilation values. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedDepthwiseConv2DWithBiasAndReluOutType 

func QuantizedDepthwiseConv2DWithBiasAndReluOutType(value tf.DataType) QuantizedDepthwiseConv2DWithBiasAndReluAttr

QuantizedDepthwiseConv2DWithBiasAndReluOutType sets the optional out_type attribute to value.

value: The type of the output. If not specified, defaults to DT_QINT32

func QuantizedDepthwiseConv2DWithBiasAndReluPaddingList 

func QuantizedDepthwiseConv2DWithBiasAndReluPaddingList(value []int64) QuantizedDepthwiseConv2DWithBiasAndReluAttr

QuantizedDepthwiseConv2DWithBiasAndReluPaddingList sets the optional padding_list attribute to value. If not specified, defaults to {}

type QuantizedDepthwiseConv2DWithBiasAttr 

type QuantizedDepthwiseConv2DWithBiasAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAttr is an optional argument to QuantizedDepthwiseConv2DWithBias.

func QuantizedDepthwiseConv2DWithBiasDilations 

func QuantizedDepthwiseConv2DWithBiasDilations(value []int64) QuantizedDepthwiseConv2DWithBiasAttr

QuantizedDepthwiseConv2DWithBiasDilations sets the optional dilations attribute to value.

value: List of dilation values. If not specified, defaults to {i:1 i:1 i:1 i:1}

func QuantizedDepthwiseConv2DWithBiasOutType 

func QuantizedDepthwiseConv2DWithBiasOutType(value tf.DataType) QuantizedDepthwiseConv2DWithBiasAttr

QuantizedDepthwiseConv2DWithBiasOutType sets the optional out_type attribute to value.

value: The type of the output. If not specified, defaults to DT_QINT32

type QuantizedInstanceNormAttr 

type QuantizedInstanceNormAttr func(optionalAttr)

QuantizedInstanceNormAttr is an optional argument to QuantizedInstanceNorm.

func QuantizedInstanceNormGivenYMax 

func QuantizedInstanceNormGivenYMax(value float32) QuantizedInstanceNormAttr

QuantizedInstanceNormGivenYMax sets the optional given_y_max attribute to value.

value: Output in `y_max` if `output_range_given` is True. If not specified, defaults to 0

func QuantizedInstanceNormGivenYMin 

func QuantizedInstanceNormGivenYMin(value float32) QuantizedInstanceNormAttr

QuantizedInstanceNormGivenYMin sets the optional given_y_min attribute to value.

value: Output in `y_min` if `output_range_given` is True. If not specified, defaults to 0

func QuantizedInstanceNormMinSeparation 

func QuantizedInstanceNormMinSeparation(value float32) QuantizedInstanceNormAttr

QuantizedInstanceNormMinSeparation sets the optional min_separation attribute to value.

value: Minimum value of `y_max - y_min` If not specified, defaults to 0.001

func QuantizedInstanceNormOutputRangeGiven 

func QuantizedInstanceNormOutputRangeGiven(value bool) QuantizedInstanceNormAttr

QuantizedInstanceNormOutputRangeGiven sets the optional output_range_given attribute to value.

value: If True, `given_y_min` and `given_y_min` and `given_y_max` are used as the output range. Otherwise, the implementation computes the output range. If not specified, defaults to false

func QuantizedInstanceNormVarianceEpsilon 

func QuantizedInstanceNormVarianceEpsilon(value float32) QuantizedInstanceNormAttr

QuantizedInstanceNormVarianceEpsilon sets the optional variance_epsilon attribute to value.

value: A small float number to avoid dividing by 0. If not specified, defaults to 1e-05

type QuantizedMatMulAttr 

type QuantizedMatMulAttr func(optionalAttr)

QuantizedMatMulAttr is an optional argument to QuantizedMatMul.

func QuantizedMatMulTactivation 

func QuantizedMatMulTactivation(value tf.DataType) QuantizedMatMulAttr

QuantizedMatMulTactivation sets the optional Tactivation attribute to value.

value: The type of output produced by activation function following this operation. If not specified, defaults to DT_QUINT8

func QuantizedMatMulToutput 

func QuantizedMatMulToutput(value tf.DataType) QuantizedMatMulAttr

QuantizedMatMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulTransposeA 

func QuantizedMatMulTransposeA(value bool) QuantizedMatMulAttr

QuantizedMatMulTransposeA sets the optional transpose_a attribute to value.

value: If true, `a` is transposed before multiplication. If not specified, defaults to false

func QuantizedMatMulTransposeB 

func QuantizedMatMulTransposeB(value bool) QuantizedMatMulAttr

QuantizedMatMulTransposeB sets the optional transpose_b attribute to value.

value: If true, `b` is transposed before multiplication. If not specified, defaults to false

type QuantizedMatMulWithBiasAndReluAndRequantizeAttr 

type QuantizedMatMulWithBiasAndReluAndRequantizeAttr func(optionalAttr)

QuantizedMatMulWithBiasAndReluAndRequantizeAttr is an optional argument to QuantizedMatMulWithBiasAndReluAndRequantize.

func QuantizedMatMulWithBiasAndReluAndRequantizeInputQuantMode 

func QuantizedMatMulWithBiasAndReluAndRequantizeInputQuantMode(value string) QuantizedMatMulWithBiasAndReluAndRequantizeAttr

QuantizedMatMulWithBiasAndReluAndRequantizeInputQuantMode sets the optional input_quant_mode attribute to value.

value: Input data quantization mode. Either MIN_FIRST(default) or SCALED. If not specified, defaults to "MIN_FIRST"

func QuantizedMatMulWithBiasAndReluAndRequantizeToutput 

func QuantizedMatMulWithBiasAndReluAndRequantizeToutput(value tf.DataType) QuantizedMatMulWithBiasAndReluAndRequantizeAttr

QuantizedMatMulWithBiasAndReluAndRequantizeToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QUINT8

func QuantizedMatMulWithBiasAndReluAndRequantizeTransposeA 

func QuantizedMatMulWithBiasAndReluAndRequantizeTransposeA(value bool) QuantizedMatMulWithBiasAndReluAndRequantizeAttr

QuantizedMatMulWithBiasAndReluAndRequantizeTransposeA sets the optional transpose_a attribute to value.

value: If true, `a` is transposed before multiplication. If not specified, defaults to false

func QuantizedMatMulWithBiasAndReluAndRequantizeTransposeB 

func QuantizedMatMulWithBiasAndReluAndRequantizeTransposeB(value bool) QuantizedMatMulWithBiasAndReluAndRequantizeAttr

QuantizedMatMulWithBiasAndReluAndRequantizeTransposeB sets the optional transpose_b attribute to value.

value: If true, `b` is transposed before multiplication. If not specified, defaults to false

type QuantizedMatMulWithBiasAndReluAttr 

type QuantizedMatMulWithBiasAndReluAttr func(optionalAttr)

QuantizedMatMulWithBiasAndReluAttr is an optional argument to QuantizedMatMulWithBiasAndRelu.

func QuantizedMatMulWithBiasAndReluInputQuantMode 

func QuantizedMatMulWithBiasAndReluInputQuantMode(value string) QuantizedMatMulWithBiasAndReluAttr

QuantizedMatMulWithBiasAndReluInputQuantMode sets the optional input_quant_mode attribute to value.

value: Input data quantization mode. Either MIN_FIRST(default) or SCALED. If not specified, defaults to "MIN_FIRST"

func QuantizedMatMulWithBiasAndReluToutput 

func QuantizedMatMulWithBiasAndReluToutput(value tf.DataType) QuantizedMatMulWithBiasAndReluAttr

QuantizedMatMulWithBiasAndReluToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulWithBiasAndReluTransposeA 

func QuantizedMatMulWithBiasAndReluTransposeA(value bool) QuantizedMatMulWithBiasAndReluAttr

QuantizedMatMulWithBiasAndReluTransposeA sets the optional transpose_a attribute to value.

value: If true, `a` is transposed before multiplication. If not specified, defaults to false

func QuantizedMatMulWithBiasAndReluTransposeB 

func QuantizedMatMulWithBiasAndReluTransposeB(value bool) QuantizedMatMulWithBiasAndReluAttr

QuantizedMatMulWithBiasAndReluTransposeB sets the optional transpose_b attribute to value.

value: If true, `b` is transposed before multiplication. If not specified, defaults to false

type QuantizedMatMulWithBiasAttr 

type QuantizedMatMulWithBiasAttr func(optionalAttr)

QuantizedMatMulWithBiasAttr is an optional argument to QuantizedMatMulWithBias.

func QuantizedMatMulWithBiasInputQuantMode 

func QuantizedMatMulWithBiasInputQuantMode(value string) QuantizedMatMulWithBiasAttr

QuantizedMatMulWithBiasInputQuantMode sets the optional input_quant_mode attribute to value.

value: Input data quantization mode. Either MIN_FIRST(default) or SCALED. If not specified, defaults to "MIN_FIRST"

func QuantizedMatMulWithBiasToutput 

func QuantizedMatMulWithBiasToutput(value tf.DataType) QuantizedMatMulWithBiasAttr

QuantizedMatMulWithBiasToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulWithBiasTransposeA 

func QuantizedMatMulWithBiasTransposeA(value bool) QuantizedMatMulWithBiasAttr

QuantizedMatMulWithBiasTransposeA sets the optional transpose_a attribute to value.

value: If true, `a` is transposed before multiplication. If not specified, defaults to false

func QuantizedMatMulWithBiasTransposeB 

func QuantizedMatMulWithBiasTransposeB(value bool) QuantizedMatMulWithBiasAttr

QuantizedMatMulWithBiasTransposeB sets the optional transpose_b attribute to value.

value: If true, `b` is transposed before multiplication. If not specified, defaults to false

type QuantizedMulAttr 

type QuantizedMulAttr func(optionalAttr)

QuantizedMulAttr is an optional argument to QuantizedMul.

func QuantizedMulToutput 

func QuantizedMulToutput(value tf.DataType) QuantizedMulAttr

QuantizedMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

type QuantizedRelu6Attr 

type QuantizedRelu6Attr func(optionalAttr)

QuantizedRelu6Attr is an optional argument to QuantizedRelu6.

func QuantizedRelu6OutType 

func QuantizedRelu6OutType(value tf.DataType) QuantizedRelu6Attr

QuantizedRelu6OutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedReluAttr 

type QuantizedReluAttr func(optionalAttr)

QuantizedReluAttr is an optional argument to QuantizedRelu.

func QuantizedReluOutType 

func QuantizedReluOutType(value tf.DataType) QuantizedReluAttr

QuantizedReluOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedReluXAttr 

type QuantizedReluXAttr func(optionalAttr)

QuantizedReluXAttr is an optional argument to QuantizedReluX.

func QuantizedReluXOutType 

func QuantizedReluXOutType(value tf.DataType) QuantizedReluXAttr

QuantizedReluXOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedResizeBilinearAttr 

type QuantizedResizeBilinearAttr func(optionalAttr)

QuantizedResizeBilinearAttr is an optional argument to QuantizedResizeBilinear.

func QuantizedResizeBilinearAlignCorners 

func QuantizedResizeBilinearAlignCorners(value bool) QuantizedResizeBilinearAttr

QuantizedResizeBilinearAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

func QuantizedResizeBilinearHalfPixelCenters 

func QuantizedResizeBilinearHalfPixelCenters(value bool) QuantizedResizeBilinearAttr

QuantizedResizeBilinearHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type QueueCloseV2Attr 

type QueueCloseV2Attr func(optionalAttr)

QueueCloseV2Attr is an optional argument to QueueCloseV2.

func QueueCloseV2CancelPendingEnqueues 

func QueueCloseV2CancelPendingEnqueues(value bool) QueueCloseV2Attr

QueueCloseV2CancelPendingEnqueues sets the optional cancel_pending_enqueues attribute to value.

value: If true, all pending enqueue requests that are blocked on the given queue will be canceled. If not specified, defaults to false

type QueueDequeueManyV2Attr 

type QueueDequeueManyV2Attr func(optionalAttr)

QueueDequeueManyV2Attr is an optional argument to QueueDequeueManyV2.

func QueueDequeueManyV2TimeoutMs 

func QueueDequeueManyV2TimeoutMs(value int64) QueueDequeueManyV2Attr

QueueDequeueManyV2TimeoutMs sets the optional timeout_ms attribute to value.

value: If the queue has fewer than n elements, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1

type QueueDequeueUpToV2Attr 

type QueueDequeueUpToV2Attr func(optionalAttr)

QueueDequeueUpToV2Attr is an optional argument to QueueDequeueUpToV2.

func QueueDequeueUpToV2TimeoutMs 

func QueueDequeueUpToV2TimeoutMs(value int64) QueueDequeueUpToV2Attr

QueueDequeueUpToV2TimeoutMs sets the optional timeout_ms attribute to value.

value: If the queue has fewer than n elements, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1

type QueueDequeueV2Attr 

type QueueDequeueV2Attr func(optionalAttr)

QueueDequeueV2Attr is an optional argument to QueueDequeueV2.

func QueueDequeueV2TimeoutMs 

func QueueDequeueV2TimeoutMs(value int64) QueueDequeueV2Attr

QueueDequeueV2TimeoutMs sets the optional timeout_ms attribute to value.

value: If the queue is empty, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1

type QueueEnqueueManyV2Attr 

type QueueEnqueueManyV2Attr func(optionalAttr)

QueueEnqueueManyV2Attr is an optional argument to QueueEnqueueManyV2.

func QueueEnqueueManyV2TimeoutMs 

func QueueEnqueueManyV2TimeoutMs(value int64) QueueEnqueueManyV2Attr

QueueEnqueueManyV2TimeoutMs sets the optional timeout_ms attribute to value.

value: If the queue is too full, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1

type QueueEnqueueV2Attr 

type QueueEnqueueV2Attr func(optionalAttr)

QueueEnqueueV2Attr is an optional argument to QueueEnqueueV2.

func QueueEnqueueV2TimeoutMs 

func QueueEnqueueV2TimeoutMs(value int64) QueueEnqueueV2Attr

QueueEnqueueV2TimeoutMs sets the optional timeout_ms attribute to value.

value: If the queue is full, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1

type RFFT2DAttr 

type RFFT2DAttr func(optionalAttr)

RFFT2DAttr is an optional argument to RFFT2D.

func RFFT2DTcomplex 

func RFFT2DTcomplex(value tf.DataType) RFFT2DAttr

RFFT2DTcomplex sets the optional Tcomplex attribute to value. If not specified, defaults to DT_COMPLEX64

type RFFT3DAttr 

type RFFT3DAttr func(optionalAttr)

RFFT3DAttr is an optional argument to RFFT3D.

func RFFT3DTcomplex 

func RFFT3DTcomplex(value tf.DataType) RFFT3DAttr

RFFT3DTcomplex sets the optional Tcomplex attribute to value. If not specified, defaults to DT_COMPLEX64

type RFFTAttr 

type RFFTAttr func(optionalAttr)

RFFTAttr is an optional argument to RFFT.

func RFFTTcomplex 

func RFFTTcomplex(value tf.DataType) RFFTAttr

RFFTTcomplex sets the optional Tcomplex attribute to value. If not specified, defaults to DT_COMPLEX64

type RFFTNDAttr added in v0.7.0 

type RFFTNDAttr func(optionalAttr)

RFFTNDAttr is an optional argument to RFFTND.

func RFFTNDTcomplex added in v0.7.0 

func RFFTNDTcomplex(value tf.DataType) RFFTNDAttr

RFFTNDTcomplex sets the optional Tcomplex attribute to value. If not specified, defaults to DT_COMPLEX64

type RaggedBincountAttr 

type RaggedBincountAttr func(optionalAttr)

RaggedBincountAttr is an optional argument to RaggedBincount.

func RaggedBincountBinaryOutput 

func RaggedBincountBinaryOutput(value bool) RaggedBincountAttr

RaggedBincountBinaryOutput sets the optional binary_output attribute to value.

value: bool; Whether the kernel should count the appearance or number of occurrences. If not specified, defaults to false

type RaggedCountSparseOutputAttr 

type RaggedCountSparseOutputAttr func(optionalAttr)

RaggedCountSparseOutputAttr is an optional argument to RaggedCountSparseOutput.

func RaggedCountSparseOutputMaxlength 

func RaggedCountSparseOutputMaxlength(value int64) RaggedCountSparseOutputAttr

RaggedCountSparseOutputMaxlength sets the optional maxlength attribute to value.

value: Maximum value to count. Can be set to -1 for no maximum. If not specified, defaults to -1

REQUIRES: value >= -1

func RaggedCountSparseOutputMinlength 

func RaggedCountSparseOutputMinlength(value int64) RaggedCountSparseOutputAttr

RaggedCountSparseOutputMinlength sets the optional minlength attribute to value.

value: Minimum value to count. Can be set to -1 for no minimum. If not specified, defaults to -1

REQUIRES: value >= -1

type RaggedRangeAttr 

type RaggedRangeAttr func(optionalAttr)

RaggedRangeAttr is an optional argument to RaggedRange.

func RaggedRangeTsplits 

func RaggedRangeTsplits(value tf.DataType) RaggedRangeAttr

RaggedRangeTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type RaggedTensorFromVariantAttr 

type RaggedTensorFromVariantAttr func(optionalAttr)

RaggedTensorFromVariantAttr is an optional argument to RaggedTensorFromVariant.

func RaggedTensorFromVariantTsplits 

func RaggedTensorFromVariantTsplits(value tf.DataType) RaggedTensorFromVariantAttr

RaggedTensorFromVariantTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type RandomCropAttr 

type RandomCropAttr func(optionalAttr)

RandomCropAttr is an optional argument to RandomCrop.

func RandomCropSeed 

func RandomCropSeed(value int64) RandomCropAttr

RandomCropSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomCropSeed2 

func RandomCropSeed2(value int64) RandomCropAttr

RandomCropSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type RandomDatasetAttr 

type RandomDatasetAttr func(optionalAttr)

RandomDatasetAttr is an optional argument to RandomDataset.

func RandomDatasetMetadata 

func RandomDatasetMetadata(value string) RandomDatasetAttr

RandomDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type RandomDatasetV2Attr added in v0.4.0 

type RandomDatasetV2Attr func(optionalAttr)

RandomDatasetV2Attr is an optional argument to RandomDatasetV2.

func RandomDatasetV2Metadata added in v0.4.0 

func RandomDatasetV2Metadata(value string) RandomDatasetV2Attr

RandomDatasetV2Metadata sets the optional metadata attribute to value. If not specified, defaults to ""

func RandomDatasetV2RerandomizeEachIteration added in v0.4.0 

func RandomDatasetV2RerandomizeEachIteration(value bool) RandomDatasetV2Attr

RandomDatasetV2RerandomizeEachIteration sets the optional rerandomize_each_iteration attribute to value.

value: A boolean attribute to rerandomize the sequence of random numbers generated at each epoch. If not specified, defaults to false

type RandomGammaAttr 

type RandomGammaAttr func(optionalAttr)

RandomGammaAttr is an optional argument to RandomGamma.

func RandomGammaSeed 

func RandomGammaSeed(value int64) RandomGammaAttr

RandomGammaSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomGammaSeed2 

func RandomGammaSeed2(value int64) RandomGammaAttr

RandomGammaSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RandomIndexShuffleAttr added in v0.3.0 

type RandomIndexShuffleAttr func(optionalAttr)

RandomIndexShuffleAttr is an optional argument to RandomIndexShuffle.

func RandomIndexShuffleRounds added in v0.3.0 

func RandomIndexShuffleRounds(value int64) RandomIndexShuffleAttr

RandomIndexShuffleRounds sets the optional rounds attribute to value.

value: The number of rounds to use the in block cipher. If not specified, defaults to 4

type RandomPoissonAttr 

type RandomPoissonAttr func(optionalAttr)

RandomPoissonAttr is an optional argument to RandomPoisson.

func RandomPoissonSeed 

func RandomPoissonSeed(value int64) RandomPoissonAttr

RandomPoissonSeed sets the optional seed attribute to value. If not specified, defaults to 0

func RandomPoissonSeed2 

func RandomPoissonSeed2(value int64) RandomPoissonAttr

RandomPoissonSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type RandomPoissonV2Attr 

type RandomPoissonV2Attr func(optionalAttr)

RandomPoissonV2Attr is an optional argument to RandomPoissonV2.

func RandomPoissonV2Dtype 

func RandomPoissonV2Dtype(value tf.DataType) RandomPoissonV2Attr

RandomPoissonV2Dtype sets the optional dtype attribute to value. If not specified, defaults to DT_INT64

func RandomPoissonV2Seed 

func RandomPoissonV2Seed(value int64) RandomPoissonV2Attr

RandomPoissonV2Seed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomPoissonV2Seed2 

func RandomPoissonV2Seed2(value int64) RandomPoissonV2Attr

RandomPoissonV2Seed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RandomShuffleAttr 

type RandomShuffleAttr func(optionalAttr)

RandomShuffleAttr is an optional argument to RandomShuffle.

func RandomShuffleSeed 

func RandomShuffleSeed(value int64) RandomShuffleAttr

RandomShuffleSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomShuffleSeed2 

func RandomShuffleSeed2(value int64) RandomShuffleAttr

RandomShuffleSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RandomShuffleQueueV2Attr 

type RandomShuffleQueueV2Attr func(optionalAttr)

RandomShuffleQueueV2Attr is an optional argument to RandomShuffleQueueV2.

func RandomShuffleQueueV2Capacity 

func RandomShuffleQueueV2Capacity(value int64) RandomShuffleQueueV2Attr

RandomShuffleQueueV2Capacity sets the optional capacity attribute to value.

value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1

func RandomShuffleQueueV2Container 

func RandomShuffleQueueV2Container(value string) RandomShuffleQueueV2Attr

RandomShuffleQueueV2Container sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func RandomShuffleQueueV2MinAfterDequeue 

func RandomShuffleQueueV2MinAfterDequeue(value int64) RandomShuffleQueueV2Attr

RandomShuffleQueueV2MinAfterDequeue sets the optional min_after_dequeue attribute to value.

value: Dequeue will block unless there would be this many elements after the dequeue or the queue is closed. This ensures a minimum level of mixing of elements. If not specified, defaults to 0

func RandomShuffleQueueV2Seed 

func RandomShuffleQueueV2Seed(value int64) RandomShuffleQueueV2Attr

RandomShuffleQueueV2Seed sets the optional seed attribute to value.

value: If either seed or seed2 is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. If not specified, defaults to 0

func RandomShuffleQueueV2Seed2 

func RandomShuffleQueueV2Seed2(value int64) RandomShuffleQueueV2Attr

RandomShuffleQueueV2Seed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

func RandomShuffleQueueV2Shapes 

func RandomShuffleQueueV2Shapes(value []tf.Shape) RandomShuffleQueueV2Attr

RandomShuffleQueueV2Shapes sets the optional shapes attribute to value.

value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time. If not specified, defaults to {}

REQUIRES: len(value) >= 0

func RandomShuffleQueueV2SharedName 

func RandomShuffleQueueV2SharedName(value string) RandomShuffleQueueV2Attr

RandomShuffleQueueV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""

type RandomStandardNormalAttr 

type RandomStandardNormalAttr func(optionalAttr)

RandomStandardNormalAttr is an optional argument to RandomStandardNormal.

func RandomStandardNormalSeed 

func RandomStandardNormalSeed(value int64) RandomStandardNormalAttr

RandomStandardNormalSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomStandardNormalSeed2 

func RandomStandardNormalSeed2(value int64) RandomStandardNormalAttr

RandomStandardNormalSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RandomUniformAttr 

type RandomUniformAttr func(optionalAttr)

RandomUniformAttr is an optional argument to RandomUniform.

func RandomUniformSeed 

func RandomUniformSeed(value int64) RandomUniformAttr

RandomUniformSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomUniformSeed2 

func RandomUniformSeed2(value int64) RandomUniformAttr

RandomUniformSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RandomUniformIntAttr 

type RandomUniformIntAttr func(optionalAttr)

RandomUniformIntAttr is an optional argument to RandomUniformInt.

func RandomUniformIntSeed 

func RandomUniformIntSeed(value int64) RandomUniformIntAttr

RandomUniformIntSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func RandomUniformIntSeed2 

func RandomUniformIntSeed2(value int64) RandomUniformIntAttr

RandomUniformIntSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type RangeDatasetAttr 

type RangeDatasetAttr func(optionalAttr)

RangeDatasetAttr is an optional argument to RangeDataset.

func RangeDatasetMetadata 

func RangeDatasetMetadata(value string) RangeDatasetAttr

RangeDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func RangeDatasetReplicateOnSplit added in v0.2.0 

func RangeDatasetReplicateOnSplit(value bool) RangeDatasetAttr

RangeDatasetReplicateOnSplit sets the optional replicate_on_split attribute to value. If not specified, defaults to false

type ReadVariableXlaSplitNDAttr 

type ReadVariableXlaSplitNDAttr func(optionalAttr)

ReadVariableXlaSplitNDAttr is an optional argument to ReadVariableXlaSplitND.

func ReadVariableXlaSplitNDPaddings 

func ReadVariableXlaSplitNDPaddings(value []int64) ReadVariableXlaSplitNDAttr

ReadVariableXlaSplitNDPaddings sets the optional paddings attribute to value.

value: Optional list of right paddings per dimension of input tensor to apply before splitting. This can be used to make a dimension evenly divisible. If not specified, defaults to {}

type RealAttr 

type RealAttr func(optionalAttr)

RealAttr is an optional argument to Real.

func RealTout 

func RealTout(value tf.DataType) RealAttr

RealTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type RebatchDatasetAttr 

type RebatchDatasetAttr func(optionalAttr)

RebatchDatasetAttr is an optional argument to RebatchDataset.

func RebatchDatasetUseFallback 

func RebatchDatasetUseFallback(value bool) RebatchDatasetAttr

RebatchDatasetUseFallback sets the optional use_fallback attribute to value. If not specified, defaults to true

type RecordInputAttr 

type RecordInputAttr func(optionalAttr)

RecordInputAttr is an optional argument to RecordInput.

func RecordInputBatchSize 

func RecordInputBatchSize(value int64) RecordInputAttr

RecordInputBatchSize sets the optional batch_size attribute to value.

value: The batch size. If not specified, defaults to 32

func RecordInputCompressionType 

func RecordInputCompressionType(value string) RecordInputAttr

RecordInputCompressionType sets the optional compression_type attribute to value.

value: The type of compression for the file. Currently ZLIB and GZIP are supported. Defaults to none. If not specified, defaults to ""

func RecordInputFileBufferSize 

func RecordInputFileBufferSize(value int64) RecordInputAttr

RecordInputFileBufferSize sets the optional file_buffer_size attribute to value.

value: The randomization shuffling buffer. If not specified, defaults to 10000

func RecordInputFileParallelism 

func RecordInputFileParallelism(value int64) RecordInputAttr

RecordInputFileParallelism sets the optional file_parallelism attribute to value.

value: How many sstables are opened and concurrently iterated over. If not specified, defaults to 16

func RecordInputFileRandomSeed 

func RecordInputFileRandomSeed(value int64) RecordInputAttr

RecordInputFileRandomSeed sets the optional file_random_seed attribute to value.

value: Random seeds used to produce randomized records. If not specified, defaults to 301

func RecordInputFileShuffleShiftRatio 

func RecordInputFileShuffleShiftRatio(value float32) RecordInputAttr

RecordInputFileShuffleShiftRatio sets the optional file_shuffle_shift_ratio attribute to value.

value: Shifts the list of files after the list is randomly shuffled. If not specified, defaults to 0

type RecvAttr 

type RecvAttr func(optionalAttr)

RecvAttr is an optional argument to Recv.

func RecvClientTerminated 

func RecvClientTerminated(value bool) RecvAttr

RecvClientTerminated sets the optional client_terminated attribute to value.

value: If set to true, this indicates that the node was added to the graph as a result of a client-side feed or fetch of Tensor data, in which case the corresponding send or recv is expected to be managed locally by the caller. If not specified, defaults to false

type ReduceJoinAttr 

type ReduceJoinAttr func(optionalAttr)

ReduceJoinAttr is an optional argument to ReduceJoin.

func ReduceJoinKeepDims 

func ReduceJoinKeepDims(value bool) ReduceJoinAttr

ReduceJoinKeepDims sets the optional keep_dims attribute to value.

value: If `True`, retain reduced dimensions with length `1`. If not specified, defaults to false

func ReduceJoinSeparator 

func ReduceJoinSeparator(value string) ReduceJoinAttr

ReduceJoinSeparator sets the optional separator attribute to value.

value: The separator to use when joining. If not specified, defaults to ""

type RegexReplaceAttr 

type RegexReplaceAttr func(optionalAttr)

RegexReplaceAttr is an optional argument to RegexReplace.

func RegexReplaceReplaceGlobal 

func RegexReplaceReplaceGlobal(value bool) RegexReplaceAttr

RegexReplaceReplaceGlobal sets the optional replace_global attribute to value.

value: If True, the replacement is global (that is, all matches of the `pattern` regular expression in each input string are rewritten), otherwise the `rewrite` substitution is only made for the first `pattern` match. If not specified, defaults to true

type RegisterDatasetAttr 

type RegisterDatasetAttr func(optionalAttr)

RegisterDatasetAttr is an optional argument to RegisterDataset.

func RegisterDatasetElementSpec 

func RegisterDatasetElementSpec(value string) RegisterDatasetAttr

RegisterDatasetElementSpec sets the optional element_spec attribute to value. If not specified, defaults to ""

func RegisterDatasetMetadata 

func RegisterDatasetMetadata(value string) RegisterDatasetAttr

RegisterDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type RegisterDatasetV2Attr added in v0.2.0 

type RegisterDatasetV2Attr func(optionalAttr)

RegisterDatasetV2Attr is an optional argument to RegisterDatasetV2.

func RegisterDatasetV2ElementSpec added in v0.2.0 

func RegisterDatasetV2ElementSpec(value string) RegisterDatasetV2Attr

RegisterDatasetV2ElementSpec sets the optional element_spec attribute to value. If not specified, defaults to ""

func RegisterDatasetV2Metadata added in v0.2.0 

func RegisterDatasetV2Metadata(value string) RegisterDatasetV2Attr

RegisterDatasetV2Metadata sets the optional metadata attribute to value. If not specified, defaults to ""

func RegisterDatasetV2RequestedDatasetId added in v0.2.0 

func RegisterDatasetV2RequestedDatasetId(value string) RegisterDatasetV2Attr

RegisterDatasetV2RequestedDatasetId sets the optional requested_dataset_id attribute to value. If not specified, defaults to ""

type RepeatDatasetAttr 

type RepeatDatasetAttr func(optionalAttr)

RepeatDatasetAttr is an optional argument to RepeatDataset.

func RepeatDatasetMetadata 

func RepeatDatasetMetadata(value string) RepeatDatasetAttr

RepeatDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type RequantizePerChannelAttr 

type RequantizePerChannelAttr func(optionalAttr)

RequantizePerChannelAttr is an optional argument to RequantizePerChannel.

func RequantizePerChannelOutType 

func RequantizePerChannelOutType(value tf.DataType) RequantizePerChannelAttr

RequantizePerChannelOutType sets the optional out_type attribute to value.

value: The quantized type of output tensor that needs to be converted. If not specified, defaults to DT_QUINT8

type ResizeAreaAttr 

type ResizeAreaAttr func(optionalAttr)

ResizeAreaAttr is an optional argument to ResizeArea.

func ResizeAreaAlignCorners 

func ResizeAreaAlignCorners(value bool) ResizeAreaAttr

ResizeAreaAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

type ResizeBicubicAttr 

type ResizeBicubicAttr func(optionalAttr)

ResizeBicubicAttr is an optional argument to ResizeBicubic.

func ResizeBicubicAlignCorners 

func ResizeBicubicAlignCorners(value bool) ResizeBicubicAttr

ResizeBicubicAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

func ResizeBicubicHalfPixelCenters 

func ResizeBicubicHalfPixelCenters(value bool) ResizeBicubicAttr

ResizeBicubicHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBicubicGradAttr 

type ResizeBicubicGradAttr func(optionalAttr)

ResizeBicubicGradAttr is an optional argument to ResizeBicubicGrad.

func ResizeBicubicGradAlignCorners 

func ResizeBicubicGradAlignCorners(value bool) ResizeBicubicGradAttr

ResizeBicubicGradAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and grad tensors are aligned. Defaults to false. If not specified, defaults to false

func ResizeBicubicGradHalfPixelCenters 

func ResizeBicubicGradHalfPixelCenters(value bool) ResizeBicubicGradAttr

ResizeBicubicGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBilinearAttr 

type ResizeBilinearAttr func(optionalAttr)

ResizeBilinearAttr is an optional argument to ResizeBilinear.

func ResizeBilinearAlignCorners 

func ResizeBilinearAlignCorners(value bool) ResizeBilinearAttr

ResizeBilinearAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

func ResizeBilinearHalfPixelCenters 

func ResizeBilinearHalfPixelCenters(value bool) ResizeBilinearAttr

ResizeBilinearHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBilinearGradAttr 

type ResizeBilinearGradAttr func(optionalAttr)

ResizeBilinearGradAttr is an optional argument to ResizeBilinearGrad.

func ResizeBilinearGradAlignCorners 

func ResizeBilinearGradAlignCorners(value bool) ResizeBilinearGradAttr

ResizeBilinearGradAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and grad tensors are aligned. Defaults to false. If not specified, defaults to false

func ResizeBilinearGradHalfPixelCenters 

func ResizeBilinearGradHalfPixelCenters(value bool) ResizeBilinearGradAttr

ResizeBilinearGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeNearestNeighborAttr 

type ResizeNearestNeighborAttr func(optionalAttr)

ResizeNearestNeighborAttr is an optional argument to ResizeNearestNeighbor.

func ResizeNearestNeighborAlignCorners 

func ResizeNearestNeighborAlignCorners(value bool) ResizeNearestNeighborAttr

ResizeNearestNeighborAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and output tensors are aligned, preserving the values at the corner pixels. Defaults to false. If not specified, defaults to false

func ResizeNearestNeighborHalfPixelCenters 

func ResizeNearestNeighborHalfPixelCenters(value bool) ResizeNearestNeighborAttr

ResizeNearestNeighborHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeNearestNeighborGradAttr 

type ResizeNearestNeighborGradAttr func(optionalAttr)

ResizeNearestNeighborGradAttr is an optional argument to ResizeNearestNeighborGrad.

func ResizeNearestNeighborGradAlignCorners 

func ResizeNearestNeighborGradAlignCorners(value bool) ResizeNearestNeighborGradAttr

ResizeNearestNeighborGradAlignCorners sets the optional align_corners attribute to value.

value: If true, the centers of the 4 corner pixels of the input and grad tensors are aligned. Defaults to false. If not specified, defaults to false

func ResizeNearestNeighborGradHalfPixelCenters 

func ResizeNearestNeighborGradHalfPixelCenters(value bool) ResizeNearestNeighborGradAttr

ResizeNearestNeighborGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResourceApplyAdaMaxAttr 

type ResourceApplyAdaMaxAttr func(optionalAttr)

ResourceApplyAdaMaxAttr is an optional argument to ResourceApplyAdaMax.

func ResourceApplyAdaMaxUseLocking 

func ResourceApplyAdaMaxUseLocking(value bool) ResourceApplyAdaMaxAttr

ResourceApplyAdaMaxUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, m, and v tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAdadeltaAttr 

type ResourceApplyAdadeltaAttr func(optionalAttr)

ResourceApplyAdadeltaAttr is an optional argument to ResourceApplyAdadelta.

func ResourceApplyAdadeltaUseLocking 

func ResourceApplyAdadeltaUseLocking(value bool) ResourceApplyAdadeltaAttr

ResourceApplyAdadeltaUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var, accum and update_accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAdagradAttr 

type ResourceApplyAdagradAttr func(optionalAttr)

ResourceApplyAdagradAttr is an optional argument to ResourceApplyAdagrad.

func ResourceApplyAdagradUpdateSlots 

func ResourceApplyAdagradUpdateSlots(value bool) ResourceApplyAdagradAttr

ResourceApplyAdagradUpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceApplyAdagradUseLocking 

func ResourceApplyAdagradUseLocking(value bool) ResourceApplyAdagradAttr

ResourceApplyAdagradUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAdagradDAAttr 

type ResourceApplyAdagradDAAttr func(optionalAttr)

ResourceApplyAdagradDAAttr is an optional argument to ResourceApplyAdagradDA.

func ResourceApplyAdagradDAUseLocking 

func ResourceApplyAdagradDAUseLocking(value bool) ResourceApplyAdagradDAAttr

ResourceApplyAdagradDAUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAdagradV2Attr 

type ResourceApplyAdagradV2Attr func(optionalAttr)

ResourceApplyAdagradV2Attr is an optional argument to ResourceApplyAdagradV2.

func ResourceApplyAdagradV2UpdateSlots 

func ResourceApplyAdagradV2UpdateSlots(value bool) ResourceApplyAdagradV2Attr

ResourceApplyAdagradV2UpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceApplyAdagradV2UseLocking 

func ResourceApplyAdagradV2UseLocking(value bool) ResourceApplyAdagradV2Attr

ResourceApplyAdagradV2UseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAdamAttr 

type ResourceApplyAdamAttr func(optionalAttr)

ResourceApplyAdamAttr is an optional argument to ResourceApplyAdam.

func ResourceApplyAdamUseLocking 

func ResourceApplyAdamUseLocking(value bool) ResourceApplyAdamAttr

ResourceApplyAdamUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, m, and v tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

func ResourceApplyAdamUseNesterov 

func ResourceApplyAdamUseNesterov(value bool) ResourceApplyAdamAttr

ResourceApplyAdamUseNesterov sets the optional use_nesterov attribute to value.

value: If `True`, uses the nesterov update. If not specified, defaults to false

type ResourceApplyAdamWithAmsgradAttr 

type ResourceApplyAdamWithAmsgradAttr func(optionalAttr)

ResourceApplyAdamWithAmsgradAttr is an optional argument to ResourceApplyAdamWithAmsgrad.

func ResourceApplyAdamWithAmsgradUseLocking 

func ResourceApplyAdamWithAmsgradUseLocking(value bool) ResourceApplyAdamWithAmsgradAttr

ResourceApplyAdamWithAmsgradUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, m, and v tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyAddSignAttr 

type ResourceApplyAddSignAttr func(optionalAttr)

ResourceApplyAddSignAttr is an optional argument to ResourceApplyAddSign.

func ResourceApplyAddSignUseLocking 

func ResourceApplyAddSignUseLocking(value bool) ResourceApplyAddSignAttr

ResourceApplyAddSignUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and m tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyCenteredRMSPropAttr 

type ResourceApplyCenteredRMSPropAttr func(optionalAttr)

ResourceApplyCenteredRMSPropAttr is an optional argument to ResourceApplyCenteredRMSProp.

func ResourceApplyCenteredRMSPropUseLocking 

func ResourceApplyCenteredRMSPropUseLocking(value bool) ResourceApplyCenteredRMSPropAttr

ResourceApplyCenteredRMSPropUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, mg, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyFtrlAttr 

type ResourceApplyFtrlAttr func(optionalAttr)

ResourceApplyFtrlAttr is an optional argument to ResourceApplyFtrl.

func ResourceApplyFtrlMultiplyLinearByLr 

func ResourceApplyFtrlMultiplyLinearByLr(value bool) ResourceApplyFtrlAttr

ResourceApplyFtrlMultiplyLinearByLr sets the optional multiply_linear_by_lr attribute to value. If not specified, defaults to false

func ResourceApplyFtrlUseLocking 

func ResourceApplyFtrlUseLocking(value bool) ResourceApplyFtrlAttr

ResourceApplyFtrlUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyFtrlV2Attr 

type ResourceApplyFtrlV2Attr func(optionalAttr)

ResourceApplyFtrlV2Attr is an optional argument to ResourceApplyFtrlV2.

func ResourceApplyFtrlV2MultiplyLinearByLr 

func ResourceApplyFtrlV2MultiplyLinearByLr(value bool) ResourceApplyFtrlV2Attr

ResourceApplyFtrlV2MultiplyLinearByLr sets the optional multiply_linear_by_lr attribute to value. If not specified, defaults to false

func ResourceApplyFtrlV2UseLocking 

func ResourceApplyFtrlV2UseLocking(value bool) ResourceApplyFtrlV2Attr

ResourceApplyFtrlV2UseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyGradientDescentAttr 

type ResourceApplyGradientDescentAttr func(optionalAttr)

ResourceApplyGradientDescentAttr is an optional argument to ResourceApplyGradientDescent.

func ResourceApplyGradientDescentUseLocking 

func ResourceApplyGradientDescentUseLocking(value bool) ResourceApplyGradientDescentAttr

ResourceApplyGradientDescentUseLocking sets the optional use_locking attribute to value.

value: If `True`, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyKerasMomentumAttr 

type ResourceApplyKerasMomentumAttr func(optionalAttr)

ResourceApplyKerasMomentumAttr is an optional argument to ResourceApplyKerasMomentum.

func ResourceApplyKerasMomentumUseLocking 

func ResourceApplyKerasMomentumUseLocking(value bool) ResourceApplyKerasMomentumAttr

ResourceApplyKerasMomentumUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

func ResourceApplyKerasMomentumUseNesterov 

func ResourceApplyKerasMomentumUseNesterov(value bool) ResourceApplyKerasMomentumAttr

ResourceApplyKerasMomentumUseNesterov sets the optional use_nesterov attribute to value.

value: If `True`, the tensor passed to compute grad will be var + momentum * accum, so in the end, the var you get is actually var + momentum * accum. If not specified, defaults to false

type ResourceApplyMomentumAttr 

type ResourceApplyMomentumAttr func(optionalAttr)

ResourceApplyMomentumAttr is an optional argument to ResourceApplyMomentum.

func ResourceApplyMomentumUseLocking 

func ResourceApplyMomentumUseLocking(value bool) ResourceApplyMomentumAttr

ResourceApplyMomentumUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

func ResourceApplyMomentumUseNesterov 

func ResourceApplyMomentumUseNesterov(value bool) ResourceApplyMomentumAttr

ResourceApplyMomentumUseNesterov sets the optional use_nesterov attribute to value.

value: If `True`, the tensor passed to compute grad will be var - lr * momentum * accum, so in the end, the var you get is actually var - lr * momentum * accum. If not specified, defaults to false

type ResourceApplyPowerSignAttr 

type ResourceApplyPowerSignAttr func(optionalAttr)

ResourceApplyPowerSignAttr is an optional argument to ResourceApplyPowerSign.

func ResourceApplyPowerSignUseLocking 

func ResourceApplyPowerSignUseLocking(value bool) ResourceApplyPowerSignAttr

ResourceApplyPowerSignUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and m tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyProximalAdagradAttr 

type ResourceApplyProximalAdagradAttr func(optionalAttr)

ResourceApplyProximalAdagradAttr is an optional argument to ResourceApplyProximalAdagrad.

func ResourceApplyProximalAdagradUseLocking 

func ResourceApplyProximalAdagradUseLocking(value bool) ResourceApplyProximalAdagradAttr

ResourceApplyProximalAdagradUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyProximalGradientDescentAttr 

type ResourceApplyProximalGradientDescentAttr func(optionalAttr)

ResourceApplyProximalGradientDescentAttr is an optional argument to ResourceApplyProximalGradientDescent.

func ResourceApplyProximalGradientDescentUseLocking 

func ResourceApplyProximalGradientDescentUseLocking(value bool) ResourceApplyProximalGradientDescentAttr

ResourceApplyProximalGradientDescentUseLocking sets the optional use_locking attribute to value.

value: If True, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceApplyRMSPropAttr 

type ResourceApplyRMSPropAttr func(optionalAttr)

ResourceApplyRMSPropAttr is an optional argument to ResourceApplyRMSProp.

func ResourceApplyRMSPropUseLocking 

func ResourceApplyRMSPropUseLocking(value bool) ResourceApplyRMSPropAttr

ResourceApplyRMSPropUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceConditionalAccumulatorAttr 

type ResourceConditionalAccumulatorAttr func(optionalAttr)

ResourceConditionalAccumulatorAttr is an optional argument to ResourceConditionalAccumulator.

func ResourceConditionalAccumulatorContainer 

func ResourceConditionalAccumulatorContainer(value string) ResourceConditionalAccumulatorAttr

ResourceConditionalAccumulatorContainer sets the optional container attribute to value.

value: If non-empty, this accumulator is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func ResourceConditionalAccumulatorReductionType 

func ResourceConditionalAccumulatorReductionType(value string) ResourceConditionalAccumulatorAttr

ResourceConditionalAccumulatorReductionType sets the optional reduction_type attribute to value. If not specified, defaults to "MEAN"

func ResourceConditionalAccumulatorSharedName 

func ResourceConditionalAccumulatorSharedName(value string) ResourceConditionalAccumulatorAttr

ResourceConditionalAccumulatorSharedName sets the optional shared_name attribute to value.

value: If non-empty, this accumulator will be shared under the given name across multiple sessions. If not specified, defaults to ""

type ResourceGatherAttr 

type ResourceGatherAttr func(optionalAttr)

ResourceGatherAttr is an optional argument to ResourceGather.

func ResourceGatherBatchDims 

func ResourceGatherBatchDims(value int64) ResourceGatherAttr

ResourceGatherBatchDims sets the optional batch_dims attribute to value. If not specified, defaults to 0

func ResourceGatherValidateIndices 

func ResourceGatherValidateIndices(value bool) ResourceGatherAttr

ResourceGatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type ResourceScatterNdAddAttr 

type ResourceScatterNdAddAttr func(optionalAttr)

ResourceScatterNdAddAttr is an optional argument to ResourceScatterNdAdd.

func ResourceScatterNdAddBadIndicesPolicy added in v0.8.2 

func ResourceScatterNdAddBadIndicesPolicy(value string) ResourceScatterNdAddAttr

ResourceScatterNdAddBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

func ResourceScatterNdAddUseLocking 

func ResourceScatterNdAddUseLocking(value bool) ResourceScatterNdAddAttr

ResourceScatterNdAddUseLocking sets the optional use_locking attribute to value.

value: An optional bool. Defaults to True. If True, the assignment will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to true

type ResourceScatterNdSubAttr 

type ResourceScatterNdSubAttr func(optionalAttr)

ResourceScatterNdSubAttr is an optional argument to ResourceScatterNdSub.

func ResourceScatterNdSubBadIndicesPolicy added in v0.8.2 

func ResourceScatterNdSubBadIndicesPolicy(value string) ResourceScatterNdSubAttr

ResourceScatterNdSubBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

func ResourceScatterNdSubUseLocking 

func ResourceScatterNdSubUseLocking(value bool) ResourceScatterNdSubAttr

ResourceScatterNdSubUseLocking sets the optional use_locking attribute to value.

value: An optional bool. Defaults to True. If True, the assignment will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to true

type ResourceScatterNdUpdateAttr 

type ResourceScatterNdUpdateAttr func(optionalAttr)

ResourceScatterNdUpdateAttr is an optional argument to ResourceScatterNdUpdate.

func ResourceScatterNdUpdateBadIndicesPolicy added in v0.8.2 

func ResourceScatterNdUpdateBadIndicesPolicy(value string) ResourceScatterNdUpdateAttr

ResourceScatterNdUpdateBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

func ResourceScatterNdUpdateUseLocking 

func ResourceScatterNdUpdateUseLocking(value bool) ResourceScatterNdUpdateAttr

ResourceScatterNdUpdateUseLocking sets the optional use_locking attribute to value.

value: An optional bool. Defaults to True. If True, the assignment will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to true

type ResourceSparseApplyAdadeltaAttr 

type ResourceSparseApplyAdadeltaAttr func(optionalAttr)

ResourceSparseApplyAdadeltaAttr is an optional argument to ResourceSparseApplyAdadelta.

func ResourceSparseApplyAdadeltaUseLocking 

func ResourceSparseApplyAdadeltaUseLocking(value bool) ResourceSparseApplyAdadeltaAttr

ResourceSparseApplyAdadeltaUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyAdagradAttr 

type ResourceSparseApplyAdagradAttr func(optionalAttr)

ResourceSparseApplyAdagradAttr is an optional argument to ResourceSparseApplyAdagrad.

func ResourceSparseApplyAdagradUpdateSlots 

func ResourceSparseApplyAdagradUpdateSlots(value bool) ResourceSparseApplyAdagradAttr

ResourceSparseApplyAdagradUpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceSparseApplyAdagradUseLocking 

func ResourceSparseApplyAdagradUseLocking(value bool) ResourceSparseApplyAdagradAttr

ResourceSparseApplyAdagradUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyAdagradDAAttr 

type ResourceSparseApplyAdagradDAAttr func(optionalAttr)

ResourceSparseApplyAdagradDAAttr is an optional argument to ResourceSparseApplyAdagradDA.

func ResourceSparseApplyAdagradDAUseLocking 

func ResourceSparseApplyAdagradDAUseLocking(value bool) ResourceSparseApplyAdagradDAAttr

ResourceSparseApplyAdagradDAUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyAdagradV2Attr 

type ResourceSparseApplyAdagradV2Attr func(optionalAttr)

ResourceSparseApplyAdagradV2Attr is an optional argument to ResourceSparseApplyAdagradV2.

func ResourceSparseApplyAdagradV2UpdateSlots 

func ResourceSparseApplyAdagradV2UpdateSlots(value bool) ResourceSparseApplyAdagradV2Attr

ResourceSparseApplyAdagradV2UpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceSparseApplyAdagradV2UseLocking 

func ResourceSparseApplyAdagradV2UseLocking(value bool) ResourceSparseApplyAdagradV2Attr

ResourceSparseApplyAdagradV2UseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyCenteredRMSPropAttr 

type ResourceSparseApplyCenteredRMSPropAttr func(optionalAttr)

ResourceSparseApplyCenteredRMSPropAttr is an optional argument to ResourceSparseApplyCenteredRMSProp.

func ResourceSparseApplyCenteredRMSPropUseLocking 

func ResourceSparseApplyCenteredRMSPropUseLocking(value bool) ResourceSparseApplyCenteredRMSPropAttr

ResourceSparseApplyCenteredRMSPropUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, mg, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyFtrlAttr 

type ResourceSparseApplyFtrlAttr func(optionalAttr)

ResourceSparseApplyFtrlAttr is an optional argument to ResourceSparseApplyFtrl.

func ResourceSparseApplyFtrlMultiplyLinearByLr 

func ResourceSparseApplyFtrlMultiplyLinearByLr(value bool) ResourceSparseApplyFtrlAttr

ResourceSparseApplyFtrlMultiplyLinearByLr sets the optional multiply_linear_by_lr attribute to value. If not specified, defaults to false

func ResourceSparseApplyFtrlUseLocking 

func ResourceSparseApplyFtrlUseLocking(value bool) ResourceSparseApplyFtrlAttr

ResourceSparseApplyFtrlUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyFtrlV2Attr 

type ResourceSparseApplyFtrlV2Attr func(optionalAttr)

ResourceSparseApplyFtrlV2Attr is an optional argument to ResourceSparseApplyFtrlV2.

func ResourceSparseApplyFtrlV2MultiplyLinearByLr 

func ResourceSparseApplyFtrlV2MultiplyLinearByLr(value bool) ResourceSparseApplyFtrlV2Attr

ResourceSparseApplyFtrlV2MultiplyLinearByLr sets the optional multiply_linear_by_lr attribute to value. If not specified, defaults to false

func ResourceSparseApplyFtrlV2UseLocking 

func ResourceSparseApplyFtrlV2UseLocking(value bool) ResourceSparseApplyFtrlV2Attr

ResourceSparseApplyFtrlV2UseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyKerasMomentumAttr 

type ResourceSparseApplyKerasMomentumAttr func(optionalAttr)

ResourceSparseApplyKerasMomentumAttr is an optional argument to ResourceSparseApplyKerasMomentum.

func ResourceSparseApplyKerasMomentumUseLocking 

func ResourceSparseApplyKerasMomentumUseLocking(value bool) ResourceSparseApplyKerasMomentumAttr

ResourceSparseApplyKerasMomentumUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

func ResourceSparseApplyKerasMomentumUseNesterov 

func ResourceSparseApplyKerasMomentumUseNesterov(value bool) ResourceSparseApplyKerasMomentumAttr

ResourceSparseApplyKerasMomentumUseNesterov sets the optional use_nesterov attribute to value.

value: If `True`, the tensor passed to compute grad will be var + momentum * accum, so in the end, the var you get is actually var + momentum * accum. If not specified, defaults to false

type ResourceSparseApplyMomentumAttr 

type ResourceSparseApplyMomentumAttr func(optionalAttr)

ResourceSparseApplyMomentumAttr is an optional argument to ResourceSparseApplyMomentum.

func ResourceSparseApplyMomentumUseLocking 

func ResourceSparseApplyMomentumUseLocking(value bool) ResourceSparseApplyMomentumAttr

ResourceSparseApplyMomentumUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

func ResourceSparseApplyMomentumUseNesterov 

func ResourceSparseApplyMomentumUseNesterov(value bool) ResourceSparseApplyMomentumAttr

ResourceSparseApplyMomentumUseNesterov sets the optional use_nesterov attribute to value.

value: If `True`, the tensor passed to compute grad will be var - lr * momentum * accum, so in the end, the var you get is actually var - lr * momentum * accum. If not specified, defaults to false

type ResourceSparseApplyProximalAdagradAttr 

type ResourceSparseApplyProximalAdagradAttr func(optionalAttr)

ResourceSparseApplyProximalAdagradAttr is an optional argument to ResourceSparseApplyProximalAdagrad.

func ResourceSparseApplyProximalAdagradUseLocking 

func ResourceSparseApplyProximalAdagradUseLocking(value bool) ResourceSparseApplyProximalAdagradAttr

ResourceSparseApplyProximalAdagradUseLocking sets the optional use_locking attribute to value.

value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyProximalGradientDescentAttr 

type ResourceSparseApplyProximalGradientDescentAttr func(optionalAttr)

ResourceSparseApplyProximalGradientDescentAttr is an optional argument to ResourceSparseApplyProximalGradientDescent.

func ResourceSparseApplyProximalGradientDescentUseLocking 

func ResourceSparseApplyProximalGradientDescentUseLocking(value bool) ResourceSparseApplyProximalGradientDescentAttr

ResourceSparseApplyProximalGradientDescentUseLocking sets the optional use_locking attribute to value.

value: If True, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceSparseApplyRMSPropAttr 

type ResourceSparseApplyRMSPropAttr func(optionalAttr)

ResourceSparseApplyRMSPropAttr is an optional argument to ResourceSparseApplyRMSProp.

func ResourceSparseApplyRMSPropUseLocking 

func ResourceSparseApplyRMSPropUseLocking(value bool) ResourceSparseApplyRMSPropAttr

ResourceSparseApplyRMSPropUseLocking sets the optional use_locking attribute to value.

value: If `True`, updating of the var, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false

type ResourceStridedSliceAssignAttr 

type ResourceStridedSliceAssignAttr func(optionalAttr)

ResourceStridedSliceAssignAttr is an optional argument to ResourceStridedSliceAssign.

func ResourceStridedSliceAssignBeginMask 

func ResourceStridedSliceAssignBeginMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignEllipsisMask 

func ResourceStridedSliceAssignEllipsisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignEndMask 

func ResourceStridedSliceAssignEndMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignNewAxisMask 

func ResourceStridedSliceAssignNewAxisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignShrinkAxisMask 

func ResourceStridedSliceAssignShrinkAxisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type RestoreAttr 

type RestoreAttr func(optionalAttr)

RestoreAttr is an optional argument to Restore.

func RestorePreferredShard 

func RestorePreferredShard(value int64) RestoreAttr

RestorePreferredShard sets the optional preferred_shard attribute to value.

value: Index of file to open first if multiple files match `file_pattern`. If not specified, defaults to -1

type RestoreSliceAttr 

type RestoreSliceAttr func(optionalAttr)

RestoreSliceAttr is an optional argument to RestoreSlice.

func RestoreSlicePreferredShard 

func RestoreSlicePreferredShard(value int64) RestoreSliceAttr

RestoreSlicePreferredShard sets the optional preferred_shard attribute to value.

value: Index of file to open first if multiple files match `file_pattern`. See the documentation for `Restore`. If not specified, defaults to -1

type RetrieveTPUEmbeddingADAMParametersAttr 

type RetrieveTPUEmbeddingADAMParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingADAMParametersAttr is an optional argument to RetrieveTPUEmbeddingADAMParameters.

func RetrieveTPUEmbeddingADAMParametersConfig 

func RetrieveTPUEmbeddingADAMParametersConfig(value string) RetrieveTPUEmbeddingADAMParametersAttr

RetrieveTPUEmbeddingADAMParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingADAMParametersTableId 

func RetrieveTPUEmbeddingADAMParametersTableId(value int64) RetrieveTPUEmbeddingADAMParametersAttr

RetrieveTPUEmbeddingADAMParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingADAMParametersTableName 

func RetrieveTPUEmbeddingADAMParametersTableName(value string) RetrieveTPUEmbeddingADAMParametersAttr

RetrieveTPUEmbeddingADAMParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdadeltaParametersAttr 

type RetrieveTPUEmbeddingAdadeltaParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingAdadeltaParametersAttr is an optional argument to RetrieveTPUEmbeddingAdadeltaParameters.

func RetrieveTPUEmbeddingAdadeltaParametersConfig 

func RetrieveTPUEmbeddingAdadeltaParametersConfig(value string) RetrieveTPUEmbeddingAdadeltaParametersAttr

RetrieveTPUEmbeddingAdadeltaParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingAdadeltaParametersTableId 

func RetrieveTPUEmbeddingAdadeltaParametersTableId(value int64) RetrieveTPUEmbeddingAdadeltaParametersAttr

RetrieveTPUEmbeddingAdadeltaParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingAdadeltaParametersTableName 

func RetrieveTPUEmbeddingAdadeltaParametersTableName(value string) RetrieveTPUEmbeddingAdadeltaParametersAttr

RetrieveTPUEmbeddingAdadeltaParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdagradMomentumParametersAttr 

type RetrieveTPUEmbeddingAdagradMomentumParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingAdagradMomentumParametersAttr is an optional argument to RetrieveTPUEmbeddingAdagradMomentumParameters.

func RetrieveTPUEmbeddingAdagradMomentumParametersConfig 

func RetrieveTPUEmbeddingAdagradMomentumParametersConfig(value string) RetrieveTPUEmbeddingAdagradMomentumParametersAttr

RetrieveTPUEmbeddingAdagradMomentumParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingAdagradMomentumParametersTableId 

func RetrieveTPUEmbeddingAdagradMomentumParametersTableId(value int64) RetrieveTPUEmbeddingAdagradMomentumParametersAttr

RetrieveTPUEmbeddingAdagradMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingAdagradMomentumParametersTableName 

func RetrieveTPUEmbeddingAdagradMomentumParametersTableName(value string) RetrieveTPUEmbeddingAdagradMomentumParametersAttr

RetrieveTPUEmbeddingAdagradMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdagradParametersAttr 

type RetrieveTPUEmbeddingAdagradParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingAdagradParametersAttr is an optional argument to RetrieveTPUEmbeddingAdagradParameters.

func RetrieveTPUEmbeddingAdagradParametersConfig 

func RetrieveTPUEmbeddingAdagradParametersConfig(value string) RetrieveTPUEmbeddingAdagradParametersAttr

RetrieveTPUEmbeddingAdagradParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingAdagradParametersTableId 

func RetrieveTPUEmbeddingAdagradParametersTableId(value int64) RetrieveTPUEmbeddingAdagradParametersAttr

RetrieveTPUEmbeddingAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingAdagradParametersTableName 

func RetrieveTPUEmbeddingAdagradParametersTableName(value string) RetrieveTPUEmbeddingAdagradParametersAttr

RetrieveTPUEmbeddingAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingCenteredRMSPropParametersAttr 

type RetrieveTPUEmbeddingCenteredRMSPropParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingCenteredRMSPropParametersAttr is an optional argument to RetrieveTPUEmbeddingCenteredRMSPropParameters.

func RetrieveTPUEmbeddingCenteredRMSPropParametersConfig 

func RetrieveTPUEmbeddingCenteredRMSPropParametersConfig(value string) RetrieveTPUEmbeddingCenteredRMSPropParametersAttr

RetrieveTPUEmbeddingCenteredRMSPropParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableId 

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableId(value int64) RetrieveTPUEmbeddingCenteredRMSPropParametersAttr

RetrieveTPUEmbeddingCenteredRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableName 

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableName(value string) RetrieveTPUEmbeddingCenteredRMSPropParametersAttr

RetrieveTPUEmbeddingCenteredRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingFTRLParametersAttr 

type RetrieveTPUEmbeddingFTRLParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingFTRLParametersAttr is an optional argument to RetrieveTPUEmbeddingFTRLParameters.

func RetrieveTPUEmbeddingFTRLParametersConfig 

func RetrieveTPUEmbeddingFTRLParametersConfig(value string) RetrieveTPUEmbeddingFTRLParametersAttr

RetrieveTPUEmbeddingFTRLParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingFTRLParametersTableId 

func RetrieveTPUEmbeddingFTRLParametersTableId(value int64) RetrieveTPUEmbeddingFTRLParametersAttr

RetrieveTPUEmbeddingFTRLParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingFTRLParametersTableName 

func RetrieveTPUEmbeddingFTRLParametersTableName(value string) RetrieveTPUEmbeddingFTRLParametersAttr

RetrieveTPUEmbeddingFTRLParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr 

type RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr is an optional argument to RetrieveTPUEmbeddingFrequencyEstimatorParameters.

func RetrieveTPUEmbeddingFrequencyEstimatorParametersConfig 

func RetrieveTPUEmbeddingFrequencyEstimatorParametersConfig(value string) RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr

RetrieveTPUEmbeddingFrequencyEstimatorParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingFrequencyEstimatorParametersTableId 

func RetrieveTPUEmbeddingFrequencyEstimatorParametersTableId(value int64) RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr

RetrieveTPUEmbeddingFrequencyEstimatorParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingFrequencyEstimatorParametersTableName 

func RetrieveTPUEmbeddingFrequencyEstimatorParametersTableName(value string) RetrieveTPUEmbeddingFrequencyEstimatorParametersAttr

RetrieveTPUEmbeddingFrequencyEstimatorParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingMDLAdagradLightParametersAttr 

type RetrieveTPUEmbeddingMDLAdagradLightParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingMDLAdagradLightParametersAttr is an optional argument to RetrieveTPUEmbeddingMDLAdagradLightParameters.

func RetrieveTPUEmbeddingMDLAdagradLightParametersConfig 

func RetrieveTPUEmbeddingMDLAdagradLightParametersConfig(value string) RetrieveTPUEmbeddingMDLAdagradLightParametersAttr

RetrieveTPUEmbeddingMDLAdagradLightParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableId 

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableId(value int64) RetrieveTPUEmbeddingMDLAdagradLightParametersAttr

RetrieveTPUEmbeddingMDLAdagradLightParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableName 

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableName(value string) RetrieveTPUEmbeddingMDLAdagradLightParametersAttr

RetrieveTPUEmbeddingMDLAdagradLightParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingMomentumParametersAttr 

type RetrieveTPUEmbeddingMomentumParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingMomentumParametersAttr is an optional argument to RetrieveTPUEmbeddingMomentumParameters.

func RetrieveTPUEmbeddingMomentumParametersConfig 

func RetrieveTPUEmbeddingMomentumParametersConfig(value string) RetrieveTPUEmbeddingMomentumParametersAttr

RetrieveTPUEmbeddingMomentumParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingMomentumParametersTableId 

func RetrieveTPUEmbeddingMomentumParametersTableId(value int64) RetrieveTPUEmbeddingMomentumParametersAttr

RetrieveTPUEmbeddingMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingMomentumParametersTableName 

func RetrieveTPUEmbeddingMomentumParametersTableName(value string) RetrieveTPUEmbeddingMomentumParametersAttr

RetrieveTPUEmbeddingMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingProximalAdagradParametersAttr 

type RetrieveTPUEmbeddingProximalAdagradParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingProximalAdagradParametersAttr is an optional argument to RetrieveTPUEmbeddingProximalAdagradParameters.

func RetrieveTPUEmbeddingProximalAdagradParametersConfig 

func RetrieveTPUEmbeddingProximalAdagradParametersConfig(value string) RetrieveTPUEmbeddingProximalAdagradParametersAttr

RetrieveTPUEmbeddingProximalAdagradParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingProximalAdagradParametersTableId 

func RetrieveTPUEmbeddingProximalAdagradParametersTableId(value int64) RetrieveTPUEmbeddingProximalAdagradParametersAttr

RetrieveTPUEmbeddingProximalAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingProximalAdagradParametersTableName 

func RetrieveTPUEmbeddingProximalAdagradParametersTableName(value string) RetrieveTPUEmbeddingProximalAdagradParametersAttr

RetrieveTPUEmbeddingProximalAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingRMSPropParametersAttr 

type RetrieveTPUEmbeddingRMSPropParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingRMSPropParametersAttr is an optional argument to RetrieveTPUEmbeddingRMSPropParameters.

func RetrieveTPUEmbeddingRMSPropParametersConfig 

func RetrieveTPUEmbeddingRMSPropParametersConfig(value string) RetrieveTPUEmbeddingRMSPropParametersAttr

RetrieveTPUEmbeddingRMSPropParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingRMSPropParametersTableId 

func RetrieveTPUEmbeddingRMSPropParametersTableId(value int64) RetrieveTPUEmbeddingRMSPropParametersAttr

RetrieveTPUEmbeddingRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingRMSPropParametersTableName 

func RetrieveTPUEmbeddingRMSPropParametersTableName(value string) RetrieveTPUEmbeddingRMSPropParametersAttr

RetrieveTPUEmbeddingRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr 

type RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr is an optional argument to RetrieveTPUEmbeddingStochasticGradientDescentParameters.

func RetrieveTPUEmbeddingStochasticGradientDescentParametersConfig 

func RetrieveTPUEmbeddingStochasticGradientDescentParametersConfig(value string) RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr

RetrieveTPUEmbeddingStochasticGradientDescentParametersConfig sets the optional config attribute to value. If not specified, defaults to ""

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId 

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId(value int64) RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr

RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName 

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName(value string) RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr

RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type ReverseSequenceAttr 

type ReverseSequenceAttr func(optionalAttr)

ReverseSequenceAttr is an optional argument to ReverseSequence.

func ReverseSequenceBatchDim 

func ReverseSequenceBatchDim(value int64) ReverseSequenceAttr

ReverseSequenceBatchDim sets the optional batch_dim attribute to value.

value: The dimension along which reversal is performed. If not specified, defaults to 0

type SampleDistortedBoundingBoxAttr 

type SampleDistortedBoundingBoxAttr func(optionalAttr)

SampleDistortedBoundingBoxAttr is an optional argument to SampleDistortedBoundingBox.

func SampleDistortedBoundingBoxAreaRange 

func SampleDistortedBoundingBoxAreaRange(value []float32) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxAreaRange sets the optional area_range attribute to value.

value: The cropped area of the image must contain a fraction of the supplied image within this range. If not specified, defaults to {f:0.05 f:1}

func SampleDistortedBoundingBoxAspectRatioRange 

func SampleDistortedBoundingBoxAspectRatioRange(value []float32) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxAspectRatioRange sets the optional aspect_ratio_range attribute to value.

value: The cropped area of the image must have an aspect ratio = width / height within this range. If not specified, defaults to {f:0.75 f:1.33}

func SampleDistortedBoundingBoxMaxAttempts 

func SampleDistortedBoundingBoxMaxAttempts(value int64) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxMaxAttempts sets the optional max_attempts attribute to value.

value: Number of attempts at generating a cropped region of the image of the specified constraints. After `max_attempts` failures, return the entire image. If not specified, defaults to 100

func SampleDistortedBoundingBoxMinObjectCovered 

func SampleDistortedBoundingBoxMinObjectCovered(value float32) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxMinObjectCovered sets the optional min_object_covered attribute to value.

value: The cropped area of the image must contain at least this fraction of any bounding box supplied. The value of this parameter should be non-negative. In the case of 0, the cropped area does not need to overlap any of the bounding boxes supplied. If not specified, defaults to 0.1

func SampleDistortedBoundingBoxSeed 

func SampleDistortedBoundingBoxSeed(value int64) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to non-zero, the random number generator is seeded by the given `seed`. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func SampleDistortedBoundingBoxSeed2 

func SampleDistortedBoundingBoxSeed2(value int64) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

func SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes 

func SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes(value bool) SampleDistortedBoundingBoxAttr

SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes sets the optional use_image_if_no_bounding_boxes attribute to value.

value: Controls behavior if no bounding boxes supplied. If true, assume an implicit bounding box covering the whole input. If false, raise an error. If not specified, defaults to false

type SampleDistortedBoundingBoxV2Attr 

type SampleDistortedBoundingBoxV2Attr func(optionalAttr)

SampleDistortedBoundingBoxV2Attr is an optional argument to SampleDistortedBoundingBoxV2.

func SampleDistortedBoundingBoxV2AreaRange 

func SampleDistortedBoundingBoxV2AreaRange(value []float32) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2AreaRange sets the optional area_range attribute to value.

value: The cropped area of the image must contain a fraction of the supplied image within this range. If not specified, defaults to {f:0.05 f:1}

func SampleDistortedBoundingBoxV2AspectRatioRange 

func SampleDistortedBoundingBoxV2AspectRatioRange(value []float32) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2AspectRatioRange sets the optional aspect_ratio_range attribute to value.

value: The cropped area of the image must have an aspect ratio = width / height within this range. If not specified, defaults to {f:0.75 f:1.33}

func SampleDistortedBoundingBoxV2MaxAttempts 

func SampleDistortedBoundingBoxV2MaxAttempts(value int64) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2MaxAttempts sets the optional max_attempts attribute to value.

value: Number of attempts at generating a cropped region of the image of the specified constraints. After `max_attempts` failures, return the entire image. If not specified, defaults to 100

func SampleDistortedBoundingBoxV2Seed 

func SampleDistortedBoundingBoxV2Seed(value int64) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2Seed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to non-zero, the random number generator is seeded by the given `seed`. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func SampleDistortedBoundingBoxV2Seed2 

func SampleDistortedBoundingBoxV2Seed2(value int64) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2Seed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

func SampleDistortedBoundingBoxV2UseImageIfNoBoundingBoxes 

func SampleDistortedBoundingBoxV2UseImageIfNoBoundingBoxes(value bool) SampleDistortedBoundingBoxV2Attr

SampleDistortedBoundingBoxV2UseImageIfNoBoundingBoxes sets the optional use_image_if_no_bounding_boxes attribute to value.

value: Controls behavior if no bounding boxes supplied. If true, assume an implicit bounding box covering the whole input. If false, raise an error. If not specified, defaults to false

type ScatterNdAttr added in v0.8.2 

type ScatterNdAttr func(optionalAttr)

ScatterNdAttr is an optional argument to ScatterNd.

func ScatterNdBadIndicesPolicy added in v0.8.2 

func ScatterNdBadIndicesPolicy(value string) ScatterNdAttr

ScatterNdBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type ScatterNdNonAliasingAddAttr added in v0.8.2 

type ScatterNdNonAliasingAddAttr func(optionalAttr)

ScatterNdNonAliasingAddAttr is an optional argument to ScatterNdNonAliasingAdd.

func ScatterNdNonAliasingAddBadIndicesPolicy added in v0.8.2 

func ScatterNdNonAliasingAddBadIndicesPolicy(value string) ScatterNdNonAliasingAddAttr

ScatterNdNonAliasingAddBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type Scope 

type Scope struct {
	// contains filtered or unexported fields
}

Scope encapsulates common operation properties when building a Graph.

A Scope object (and its derivatives, e.g., obtained from Scope.SubScope) act as a builder for graphs. They allow common properties (such as a name prefix) to be specified for multiple operations being added to the graph.

A Scope object and all its derivatives (e.g., obtained from Scope.SubScope) are not safe for concurrent use by multiple goroutines.

func NewScope 

func NewScope() *Scope

NewScope creates a Scope initialized with an empty Graph.

func NewScopeWithGraph 

func NewScopeWithGraph(g *tf.Graph) *Scope

NewScopeWithGraph creates a Scope initialized with the Graph thats passed in

func (*Scope) AddOperation 

func (s *Scope) AddOperation(args tf.OpSpec) *tf.Operation

AddOperation adds the operation to the Graph managed by s.

If there is a name prefix associated with s (such as if s was created by a call to SubScope), then this prefix will be applied to the name of the operation being added. See also Graph.AddOperation.

func (*Scope) Err 

func (s *Scope) Err() error

Err returns the error, if any, encountered during the construction of the Graph managed by s.

Once Err returns a non-nil error, all future calls will do the same, indicating that the scope should be discarded as the graph could not be constructed.

func (*Scope) Finalize 

func (s *Scope) Finalize() (*tf.Graph, error)

Finalize returns the Graph on which this scope operates on and renders s unusable. If there was an error during graph construction, that error is returned instead.

func (*Scope) SubScope 

func (s *Scope) SubScope(namespace string) *Scope

SubScope returns a new Scope which will cause all operations added to the graph to be namespaced with 'namespace'. If namespace collides with an existing namespace within the scope, then a suffix will be added.

Example 
var (
	s  = NewScope()
	c1 = Const(s.SubScope("x"), int64(1))
	c2 = Const(s.SubScope("x"), int64(1))
)
if s.Err() != nil {
	panic(s.Err())
}
fmt.Println(c1.Op.Name(), c2.Op.Name())

Output:

x/Const x_1/Const

func (*Scope) UpdateErr 

func (s *Scope) UpdateErr(op string, err error)

UpdateErr is used to notify Scope of any graph construction errors while creating the operation op.

func (*Scope) WithControlDependencies 

func (s *Scope) WithControlDependencies(ops ...*tf.Operation) *Scope

WithControlDependencies returns a new Scope which will cause all operations added to the graph to execute only after all the provided operations have executed first (in addition to any other control dependencies in s).

func (*Scope) WithDevice 

func (s *Scope) WithDevice(device string) *Scope

WithDevice returns a new Scope which will cause all operations added to the graph to execute on devices that match the provided device specification.

For example, WithDevice("/device:GPU:0") will cause operations added to the graph to execute on GPU #0.

An empty string removes any device restrictions.

type SdcaOptimizerAttr 

type SdcaOptimizerAttr func(optionalAttr)

SdcaOptimizerAttr is an optional argument to SdcaOptimizer.

func SdcaOptimizerAdaptative 

func SdcaOptimizerAdaptative(value bool) SdcaOptimizerAttr

SdcaOptimizerAdaptative sets the optional adaptative attribute to value.

value: Whether to use Adaptive SDCA for the inner loop. If not specified, defaults to true

type SdcaOptimizerV2Attr 

type SdcaOptimizerV2Attr func(optionalAttr)

SdcaOptimizerV2Attr is an optional argument to SdcaOptimizerV2.

func SdcaOptimizerV2Adaptive 

func SdcaOptimizerV2Adaptive(value bool) SdcaOptimizerV2Attr

SdcaOptimizerV2Adaptive sets the optional adaptive attribute to value.

value: Whether to use Adaptive SDCA for the inner loop. If not specified, defaults to true

type SelfAdjointEigV2Attr 

type SelfAdjointEigV2Attr func(optionalAttr)

SelfAdjointEigV2Attr is an optional argument to SelfAdjointEigV2.

func SelfAdjointEigV2ComputeV 

func SelfAdjointEigV2ComputeV(value bool) SelfAdjointEigV2Attr

SelfAdjointEigV2ComputeV sets the optional compute_v attribute to value.

value: If `True` then eigenvectors will be computed and returned in `v`. Otherwise, only the eigenvalues will be computed. If not specified, defaults to true

type SendAttr 

type SendAttr func(optionalAttr)

SendAttr is an optional argument to Send.

func SendClientTerminated 

func SendClientTerminated(value bool) SendAttr

SendClientTerminated sets the optional client_terminated attribute to value.

value: If set to true, this indicates that the node was added to the graph as a result of a client-side feed or fetch of Tensor data, in which case the corresponding send or recv is expected to be managed locally by the caller. If not specified, defaults to false

type SerializeIteratorAttr 

type SerializeIteratorAttr func(optionalAttr)

SerializeIteratorAttr is an optional argument to SerializeIterator.

func SerializeIteratorExternalStatePolicy 

func SerializeIteratorExternalStatePolicy(value int64) SerializeIteratorAttr

SerializeIteratorExternalStatePolicy sets the optional external_state_policy attribute to value. If not specified, defaults to 0

type SerializeManySparseAttr 

type SerializeManySparseAttr func(optionalAttr)

SerializeManySparseAttr is an optional argument to SerializeManySparse.

func SerializeManySparseOutType 

func SerializeManySparseOutType(value tf.DataType) SerializeManySparseAttr

SerializeManySparseOutType sets the optional out_type attribute to value.

value: The `dtype` to use for serialization; the supported types are `string` (default) and `variant`. If not specified, defaults to DT_STRING

type SerializeSparseAttr 

type SerializeSparseAttr func(optionalAttr)

SerializeSparseAttr is an optional argument to SerializeSparse.

func SerializeSparseOutType 

func SerializeSparseOutType(value tf.DataType) SerializeSparseAttr

SerializeSparseOutType sets the optional out_type attribute to value.

value: The `dtype` to use for serialization; the supported types are `string` (default) and `variant`. If not specified, defaults to DT_STRING

type SetSizeAttr 

type SetSizeAttr func(optionalAttr)

SetSizeAttr is an optional argument to SetSize.

func SetSizeValidateIndices 

func SetSizeValidateIndices(value bool) SetSizeAttr

SetSizeValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type ShapeAttr 

type ShapeAttr func(optionalAttr)

ShapeAttr is an optional argument to Shape.

func ShapeOutType 

func ShapeOutType(value tf.DataType) ShapeAttr

ShapeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type ShapeNAttr 

type ShapeNAttr func(optionalAttr)

ShapeNAttr is an optional argument to ShapeN.

func ShapeNOutType 

func ShapeNOutType(value tf.DataType) ShapeNAttr

ShapeNOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type ShardDatasetAttr 

type ShardDatasetAttr func(optionalAttr)

ShardDatasetAttr is an optional argument to ShardDataset.

func ShardDatasetMetadata 

func ShardDatasetMetadata(value string) ShardDatasetAttr

ShardDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func ShardDatasetRequireNonEmpty 

func ShardDatasetRequireNonEmpty(value bool) ShardDatasetAttr

ShardDatasetRequireNonEmpty sets the optional require_non_empty attribute to value. If not specified, defaults to false

type ShuffleAndRepeatDatasetAttr 

type ShuffleAndRepeatDatasetAttr func(optionalAttr)

ShuffleAndRepeatDatasetAttr is an optional argument to ShuffleAndRepeatDataset.

func ShuffleAndRepeatDatasetMetadata 

func ShuffleAndRepeatDatasetMetadata(value string) ShuffleAndRepeatDatasetAttr

ShuffleAndRepeatDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func ShuffleAndRepeatDatasetReshuffleEachIteration 

func ShuffleAndRepeatDatasetReshuffleEachIteration(value bool) ShuffleAndRepeatDatasetAttr

ShuffleAndRepeatDatasetReshuffleEachIteration sets the optional reshuffle_each_iteration attribute to value. If not specified, defaults to true

type ShuffleDatasetAttr 

type ShuffleDatasetAttr func(optionalAttr)

ShuffleDatasetAttr is an optional argument to ShuffleDataset.

func ShuffleDatasetMetadata 

func ShuffleDatasetMetadata(value string) ShuffleDatasetAttr

ShuffleDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func ShuffleDatasetReshuffleEachIteration 

func ShuffleDatasetReshuffleEachIteration(value bool) ShuffleDatasetAttr

ShuffleDatasetReshuffleEachIteration sets the optional reshuffle_each_iteration attribute to value.

value: If true, each iterator over this dataset will be given a different pseudorandomly generated seed, based on a sequence seeded by the `seed` and `seed2` inputs. If false, each iterator will be given the same seed, and repeated iteration over this dataset will yield the exact same sequence of results. If not specified, defaults to true

type SizeAttr 

type SizeAttr func(optionalAttr)

SizeAttr is an optional argument to Size.

func SizeOutType 

func SizeOutType(value tf.DataType) SizeAttr

SizeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type SkipDatasetAttr 

type SkipDatasetAttr func(optionalAttr)

SkipDatasetAttr is an optional argument to SkipDataset.

func SkipDatasetMetadata 

func SkipDatasetMetadata(value string) SkipDatasetAttr

SkipDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type SkipgramAttr 

type SkipgramAttr func(optionalAttr)

SkipgramAttr is an optional argument to Skipgram.

func SkipgramMinCount 

func SkipgramMinCount(value int64) SkipgramAttr

SkipgramMinCount sets the optional min_count attribute to value.

value: The minimum number of word occurrences for it to be included in the vocabulary. If not specified, defaults to 5

func SkipgramSubsample 

func SkipgramSubsample(value float32) SkipgramAttr

SkipgramSubsample sets the optional subsample attribute to value.

value: Threshold for word occurrence. Words that appear with higher frequency will be randomly down-sampled. Set to 0 to disable. If not specified, defaults to 0.001

func SkipgramWindowSize 

func SkipgramWindowSize(value int64) SkipgramAttr

SkipgramWindowSize sets the optional window_size attribute to value.

value: The number of words to predict to the left and right of the target. If not specified, defaults to 5

type SlidingWindowDatasetAttr 

type SlidingWindowDatasetAttr func(optionalAttr)

SlidingWindowDatasetAttr is an optional argument to SlidingWindowDataset.

func SlidingWindowDatasetDropRemainder 

func SlidingWindowDatasetDropRemainder(value bool) SlidingWindowDatasetAttr

SlidingWindowDatasetDropRemainder sets the optional drop_remainder attribute to value. If not specified, defaults to true

type SnapshotDatasetAttr 

type SnapshotDatasetAttr func(optionalAttr)

SnapshotDatasetAttr is an optional argument to SnapshotDataset.

func SnapshotDatasetCompression 

func SnapshotDatasetCompression(value string) SnapshotDatasetAttr

SnapshotDatasetCompression sets the optional compression attribute to value. If not specified, defaults to ""

func SnapshotDatasetMode 

func SnapshotDatasetMode(value string) SnapshotDatasetAttr

SnapshotDatasetMode sets the optional mode attribute to value. If not specified, defaults to "auto"

func SnapshotDatasetNumReaderThreads 

func SnapshotDatasetNumReaderThreads(value int64) SnapshotDatasetAttr

SnapshotDatasetNumReaderThreads sets the optional num_reader_threads attribute to value. If not specified, defaults to 1

func SnapshotDatasetNumWriterThreads 

func SnapshotDatasetNumWriterThreads(value int64) SnapshotDatasetAttr

SnapshotDatasetNumWriterThreads sets the optional num_writer_threads attribute to value. If not specified, defaults to 1

func SnapshotDatasetPendingSnapshotExpirySeconds 

func SnapshotDatasetPendingSnapshotExpirySeconds(value int64) SnapshotDatasetAttr

SnapshotDatasetPendingSnapshotExpirySeconds sets the optional pending_snapshot_expiry_seconds attribute to value. If not specified, defaults to 86400

func SnapshotDatasetReaderBufferSize 

func SnapshotDatasetReaderBufferSize(value int64) SnapshotDatasetAttr

SnapshotDatasetReaderBufferSize sets the optional reader_buffer_size attribute to value. If not specified, defaults to 1

func SnapshotDatasetReaderPathPrefix 

func SnapshotDatasetReaderPathPrefix(value string) SnapshotDatasetAttr

SnapshotDatasetReaderPathPrefix sets the optional reader_path_prefix attribute to value. If not specified, defaults to ""

func SnapshotDatasetSeed 

func SnapshotDatasetSeed(value int64) SnapshotDatasetAttr

SnapshotDatasetSeed sets the optional seed attribute to value. If not specified, defaults to 0

func SnapshotDatasetSeed2 

func SnapshotDatasetSeed2(value int64) SnapshotDatasetAttr

SnapshotDatasetSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

func SnapshotDatasetShardSizeBytes 

func SnapshotDatasetShardSizeBytes(value int64) SnapshotDatasetAttr

SnapshotDatasetShardSizeBytes sets the optional shard_size_bytes attribute to value. If not specified, defaults to 10737418240

func SnapshotDatasetShuffleOnRead 

func SnapshotDatasetShuffleOnRead(value bool) SnapshotDatasetAttr

SnapshotDatasetShuffleOnRead sets the optional shuffle_on_read attribute to value. If not specified, defaults to false

func SnapshotDatasetSnapshotName 

func SnapshotDatasetSnapshotName(value string) SnapshotDatasetAttr

SnapshotDatasetSnapshotName sets the optional snapshot_name attribute to value. If not specified, defaults to ""

func SnapshotDatasetWriterBufferSize 

func SnapshotDatasetWriterBufferSize(value int64) SnapshotDatasetAttr

SnapshotDatasetWriterBufferSize sets the optional writer_buffer_size attribute to value. If not specified, defaults to 1

func SnapshotDatasetWriterPathPrefix 

func SnapshotDatasetWriterPathPrefix(value string) SnapshotDatasetAttr

SnapshotDatasetWriterPathPrefix sets the optional writer_path_prefix attribute to value. If not specified, defaults to ""

type SobolSampleAttr 

type SobolSampleAttr func(optionalAttr)

SobolSampleAttr is an optional argument to SobolSample.

func SobolSampleDtype 

func SobolSampleDtype(value tf.DataType) SobolSampleAttr

SobolSampleDtype sets the optional dtype attribute to value.

value: The type of the sample. One of: `float32` or `float64`. If not specified, defaults to DT_FLOAT

type SpaceToDepthAttr 

type SpaceToDepthAttr func(optionalAttr)

SpaceToDepthAttr is an optional argument to SpaceToDepth.

func SpaceToDepthDataFormat 

func SpaceToDepthDataFormat(value string) SpaceToDepthAttr

SpaceToDepthDataFormat sets the optional data_format attribute to value. If not specified, defaults to "NHWC"

type SparseBincountAttr 

type SparseBincountAttr func(optionalAttr)

SparseBincountAttr is an optional argument to SparseBincount.

func SparseBincountBinaryOutput 

func SparseBincountBinaryOutput(value bool) SparseBincountAttr

SparseBincountBinaryOutput sets the optional binary_output attribute to value.

value: bool; Whether the kernel should count the appearance or number of occurrences. If not specified, defaults to false

type SparseCountSparseOutputAttr 

type SparseCountSparseOutputAttr func(optionalAttr)

SparseCountSparseOutputAttr is an optional argument to SparseCountSparseOutput.

func SparseCountSparseOutputMaxlength 

func SparseCountSparseOutputMaxlength(value int64) SparseCountSparseOutputAttr

SparseCountSparseOutputMaxlength sets the optional maxlength attribute to value.

value: Maximum value to count. Can be set to -1 for no maximum. If not specified, defaults to -1

REQUIRES: value >= -1

func SparseCountSparseOutputMinlength 

func SparseCountSparseOutputMinlength(value int64) SparseCountSparseOutputAttr

SparseCountSparseOutputMinlength sets the optional minlength attribute to value.

value: Minimum value to count. Can be set to -1 for no minimum. If not specified, defaults to -1

REQUIRES: value >= -1

type SparseMatMulAttr 

type SparseMatMulAttr func(optionalAttr)

SparseMatMulAttr is an optional argument to SparseMatMul.

func SparseMatMulAIsSparse 

func SparseMatMulAIsSparse(value bool) SparseMatMulAttr

SparseMatMulAIsSparse sets the optional a_is_sparse attribute to value. If not specified, defaults to false

func SparseMatMulBIsSparse 

func SparseMatMulBIsSparse(value bool) SparseMatMulAttr

SparseMatMulBIsSparse sets the optional b_is_sparse attribute to value. If not specified, defaults to false

func SparseMatMulTransposeA 

func SparseMatMulTransposeA(value bool) SparseMatMulAttr

SparseMatMulTransposeA sets the optional transpose_a attribute to value. If not specified, defaults to false

func SparseMatMulTransposeB 

func SparseMatMulTransposeB(value bool) SparseMatMulAttr

SparseMatMulTransposeB sets the optional transpose_b attribute to value. If not specified, defaults to false

type SparseMatrixMatMulAttr 

type SparseMatrixMatMulAttr func(optionalAttr)

SparseMatrixMatMulAttr is an optional argument to SparseMatrixMatMul.

func SparseMatrixMatMulAdjointA 

func SparseMatrixMatMulAdjointA(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulAdjointA sets the optional adjoint_a attribute to value.

value: Indicates whether `a` should be conjugate-transposed. If not specified, defaults to false

func SparseMatrixMatMulAdjointB 

func SparseMatrixMatMulAdjointB(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulAdjointB sets the optional adjoint_b attribute to value.

value: Indicates whether `b` should be conjugate-transposed. If not specified, defaults to false

func SparseMatrixMatMulConjugateOutput 

func SparseMatrixMatMulConjugateOutput(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulConjugateOutput sets the optional conjugate_output attribute to value.

value: Conjugates the product of `a` and `b`. If not specified, defaults to false

func SparseMatrixMatMulTransposeA 

func SparseMatrixMatMulTransposeA(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulTransposeA sets the optional transpose_a attribute to value.

value: Indicates whether `a` should be transposed. If not specified, defaults to false

func SparseMatrixMatMulTransposeB 

func SparseMatrixMatMulTransposeB(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulTransposeB sets the optional transpose_b attribute to value.

value: Indicates whether `b` should be transposed. If not specified, defaults to false

func SparseMatrixMatMulTransposeOutput 

func SparseMatrixMatMulTransposeOutput(value bool) SparseMatrixMatMulAttr

SparseMatrixMatMulTransposeOutput sets the optional transpose_output attribute to value.

value: Transposes the product of `a` and `b`. If not specified, defaults to false

type SparseMatrixSparseMatMulAttr 

type SparseMatrixSparseMatMulAttr func(optionalAttr)

SparseMatrixSparseMatMulAttr is an optional argument to SparseMatrixSparseMatMul.

func SparseMatrixSparseMatMulAdjointA 

func SparseMatrixSparseMatMulAdjointA(value bool) SparseMatrixSparseMatMulAttr

SparseMatrixSparseMatMulAdjointA sets the optional adjoint_a attribute to value.

value: Indicates whether `a` should be conjugate-transposed. If not specified, defaults to false

func SparseMatrixSparseMatMulAdjointB 

func SparseMatrixSparseMatMulAdjointB(value bool) SparseMatrixSparseMatMulAttr

SparseMatrixSparseMatMulAdjointB sets the optional adjoint_b attribute to value.

value: Indicates whether `b` should be conjugate-transposed. If not specified, defaults to false

func SparseMatrixSparseMatMulTransposeA 

func SparseMatrixSparseMatMulTransposeA(value bool) SparseMatrixSparseMatMulAttr

SparseMatrixSparseMatMulTransposeA sets the optional transpose_a attribute to value.

value: Indicates whether `a` should be transposed. If not specified, defaults to false

func SparseMatrixSparseMatMulTransposeB 

func SparseMatrixSparseMatMulTransposeB(value bool) SparseMatrixSparseMatMulAttr

SparseMatrixSparseMatMulTransposeB sets the optional transpose_b attribute to value.

value: Indicates whether `b` should be transposed. If not specified, defaults to false

type SparseMatrixTransposeAttr 

type SparseMatrixTransposeAttr func(optionalAttr)

SparseMatrixTransposeAttr is an optional argument to SparseMatrixTranspose.

func SparseMatrixTransposeConjugate 

func SparseMatrixTransposeConjugate(value bool) SparseMatrixTransposeAttr

SparseMatrixTransposeConjugate sets the optional conjugate attribute to value.

value: Indicates whether `input` should be conjugated. If not specified, defaults to false

type SparseReduceMaxAttr 

type SparseReduceMaxAttr func(optionalAttr)

SparseReduceMaxAttr is an optional argument to SparseReduceMax.

func SparseReduceMaxKeepDims 

func SparseReduceMaxKeepDims(value bool) SparseReduceMaxAttr

SparseReduceMaxKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type SparseReduceMaxSparseAttr 

type SparseReduceMaxSparseAttr func(optionalAttr)

SparseReduceMaxSparseAttr is an optional argument to SparseReduceMaxSparse.

func SparseReduceMaxSparseKeepDims 

func SparseReduceMaxSparseKeepDims(value bool) SparseReduceMaxSparseAttr

SparseReduceMaxSparseKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type SparseReduceSumAttr 

type SparseReduceSumAttr func(optionalAttr)

SparseReduceSumAttr is an optional argument to SparseReduceSum.

func SparseReduceSumKeepDims 

func SparseReduceSumKeepDims(value bool) SparseReduceSumAttr

SparseReduceSumKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type SparseReduceSumSparseAttr 

type SparseReduceSumSparseAttr func(optionalAttr)

SparseReduceSumSparseAttr is an optional argument to SparseReduceSumSparse.

func SparseReduceSumSparseKeepDims 

func SparseReduceSumSparseKeepDims(value bool) SparseReduceSumSparseAttr

SparseReduceSumSparseKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type SparseSegmentMeanAttr added in v0.7.0 

type SparseSegmentMeanAttr func(optionalAttr)

SparseSegmentMeanAttr is an optional argument to SparseSegmentMean.

func SparseSegmentMeanSparseGradient added in v0.7.0 

func SparseSegmentMeanSparseGradient(value bool) SparseSegmentMeanAttr

SparseSegmentMeanSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseSegmentMeanWithNumSegmentsAttr added in v0.7.0 

type SparseSegmentMeanWithNumSegmentsAttr func(optionalAttr)

SparseSegmentMeanWithNumSegmentsAttr is an optional argument to SparseSegmentMeanWithNumSegments.

func SparseSegmentMeanWithNumSegmentsSparseGradient added in v0.7.0 

func SparseSegmentMeanWithNumSegmentsSparseGradient(value bool) SparseSegmentMeanWithNumSegmentsAttr

SparseSegmentMeanWithNumSegmentsSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseSegmentSqrtNAttr added in v0.7.0 

type SparseSegmentSqrtNAttr func(optionalAttr)

SparseSegmentSqrtNAttr is an optional argument to SparseSegmentSqrtN.

func SparseSegmentSqrtNSparseGradient added in v0.7.0 

func SparseSegmentSqrtNSparseGradient(value bool) SparseSegmentSqrtNAttr

SparseSegmentSqrtNSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseSegmentSqrtNWithNumSegmentsAttr added in v0.7.0 

type SparseSegmentSqrtNWithNumSegmentsAttr func(optionalAttr)

SparseSegmentSqrtNWithNumSegmentsAttr is an optional argument to SparseSegmentSqrtNWithNumSegments.

func SparseSegmentSqrtNWithNumSegmentsSparseGradient added in v0.7.0 

func SparseSegmentSqrtNWithNumSegmentsSparseGradient(value bool) SparseSegmentSqrtNWithNumSegmentsAttr

SparseSegmentSqrtNWithNumSegmentsSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseSegmentSumAttr added in v0.7.0 

type SparseSegmentSumAttr func(optionalAttr)

SparseSegmentSumAttr is an optional argument to SparseSegmentSum.

func SparseSegmentSumSparseGradient added in v0.7.0 

func SparseSegmentSumSparseGradient(value bool) SparseSegmentSumAttr

SparseSegmentSumSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseSegmentSumWithNumSegmentsAttr added in v0.7.0 

type SparseSegmentSumWithNumSegmentsAttr func(optionalAttr)

SparseSegmentSumWithNumSegmentsAttr is an optional argument to SparseSegmentSumWithNumSegments.

func SparseSegmentSumWithNumSegmentsSparseGradient added in v0.7.0 

func SparseSegmentSumWithNumSegmentsSparseGradient(value bool) SparseSegmentSumWithNumSegmentsAttr

SparseSegmentSumWithNumSegmentsSparseGradient sets the optional sparse_gradient attribute to value. If not specified, defaults to false

type SparseTensorDenseMatMulAttr 

type SparseTensorDenseMatMulAttr func(optionalAttr)

SparseTensorDenseMatMulAttr is an optional argument to SparseTensorDenseMatMul.

func SparseTensorDenseMatMulAdjointA 

func SparseTensorDenseMatMulAdjointA(value bool) SparseTensorDenseMatMulAttr

SparseTensorDenseMatMulAdjointA sets the optional adjoint_a attribute to value.

value: Use the adjoint of A in the matrix multiply. If A is complex, this is transpose(conj(A)). Otherwise it's transpose(A). If not specified, defaults to false

func SparseTensorDenseMatMulAdjointB 

func SparseTensorDenseMatMulAdjointB(value bool) SparseTensorDenseMatMulAttr

SparseTensorDenseMatMulAdjointB sets the optional adjoint_b attribute to value.

value: Use the adjoint of B in the matrix multiply. If B is complex, this is transpose(conj(B)). Otherwise it's transpose(B). If not specified, defaults to false

type SparseToDenseAttr 

type SparseToDenseAttr func(optionalAttr)

SparseToDenseAttr is an optional argument to SparseToDense.

func SparseToDenseValidateIndices 

func SparseToDenseValidateIndices(value bool) SparseToDenseAttr

SparseToDenseValidateIndices sets the optional validate_indices attribute to value.

value: If true, indices are checked to make sure they are sorted in lexicographic order and that there are no repeats. If not specified, defaults to true

type SparseToSparseSetOperationAttr 

type SparseToSparseSetOperationAttr func(optionalAttr)

SparseToSparseSetOperationAttr is an optional argument to SparseToSparseSetOperation.

func SparseToSparseSetOperationValidateIndices 

func SparseToSparseSetOperationValidateIndices(value bool) SparseToSparseSetOperationAttr

SparseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type SplitDedupDataAttr added in v0.5.0 

type SplitDedupDataAttr func(optionalAttr)

SplitDedupDataAttr is an optional argument to SplitDedupData.

func SplitDedupDataConfig added in v0.5.0 

func SplitDedupDataConfig(value string) SplitDedupDataAttr

SplitDedupDataConfig sets the optional config attribute to value. If not specified, defaults to ""

type SqueezeAttr 

type SqueezeAttr func(optionalAttr)

SqueezeAttr is an optional argument to Squeeze.

func SqueezeAxis 

func SqueezeAxis(value []int64) SqueezeAttr

SqueezeAxis sets the optional axis attribute to value.

value: If specified, only squeezes the dimensions listed. The dimension index starts at 0. It is an error to squeeze a dimension that is not 1. Must be in the range `[-rank(input), rank(input))`. If not specified, defaults to {}

REQUIRES: len(value) >= 0

type StackPushV2Attr 

type StackPushV2Attr func(optionalAttr)

StackPushV2Attr is an optional argument to StackPushV2.

func StackPushV2SwapMemory 

func StackPushV2SwapMemory(value bool) StackPushV2Attr

StackPushV2SwapMemory sets the optional swap_memory attribute to value.

value: Swap `elem` to CPU. Default to false. If not specified, defaults to false

type StackV2Attr 

type StackV2Attr func(optionalAttr)

StackV2Attr is an optional argument to StackV2.

func StackV2StackName 

func StackV2StackName(value string) StackV2Attr

StackV2StackName sets the optional stack_name attribute to value.

value: Overrides the name used for the temporary stack resource. Default value is the name of the 'Stack' op (which is guaranteed unique). If not specified, defaults to ""

type StageAttr 

type StageAttr func(optionalAttr)

StageAttr is an optional argument to Stage.

func StageCapacity 

func StageCapacity(value int64) StageAttr

StageCapacity sets the optional capacity attribute to value.

value: Maximum number of elements in the Staging Area. If > 0, inserts on the container will block when the capacity is reached. If not specified, defaults to 0

REQUIRES: value >= 0

func StageContainer 

func StageContainer(value string) StageAttr

StageContainer sets the optional container attribute to value.

value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func StageMemoryLimit 

func StageMemoryLimit(value int64) StageAttr

StageMemoryLimit sets the optional memory_limit attribute to value.

value: The maximum number of bytes allowed for Tensors in the Staging Area. If > 0, inserts will block until sufficient space is available. If not specified, defaults to 0

REQUIRES: value >= 0

func StageSharedName 

func StageSharedName(value string) StageAttr

StageSharedName sets the optional shared_name attribute to value.

value: It is necessary to match this name to the matching Unstage Op. If not specified, defaults to ""

type StageClearAttr 

type StageClearAttr func(optionalAttr)

StageClearAttr is an optional argument to StageClear.

func StageClearCapacity 

func StageClearCapacity(value int64) StageClearAttr

StageClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageClearContainer 

func StageClearContainer(value string) StageClearAttr

StageClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func StageClearMemoryLimit 

func StageClearMemoryLimit(value int64) StageClearAttr

StageClearMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageClearSharedName 

func StageClearSharedName(value string) StageClearAttr

StageClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StagePeekAttr 

type StagePeekAttr func(optionalAttr)

StagePeekAttr is an optional argument to StagePeek.

func StagePeekCapacity 

func StagePeekCapacity(value int64) StagePeekAttr

StagePeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StagePeekContainer 

func StagePeekContainer(value string) StagePeekAttr

StagePeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func StagePeekMemoryLimit 

func StagePeekMemoryLimit(value int64) StagePeekAttr

StagePeekMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StagePeekSharedName 

func StagePeekSharedName(value string) StagePeekAttr

StagePeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StageSizeAttr 

type StageSizeAttr func(optionalAttr)

StageSizeAttr is an optional argument to StageSize.

func StageSizeCapacity 

func StageSizeCapacity(value int64) StageSizeAttr

StageSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageSizeContainer 

func StageSizeContainer(value string) StageSizeAttr

StageSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func StageSizeMemoryLimit 

func StageSizeMemoryLimit(value int64) StageSizeAttr

StageSizeMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageSizeSharedName 

func StageSizeSharedName(value string) StageSizeAttr

StageSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StatefulStandardNormalAttr 

type StatefulStandardNormalAttr func(optionalAttr)

StatefulStandardNormalAttr is an optional argument to StatefulStandardNormal.

func StatefulStandardNormalDtype 

func StatefulStandardNormalDtype(value tf.DataType) StatefulStandardNormalAttr

StatefulStandardNormalDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatefulStandardNormalV2Attr 

type StatefulStandardNormalV2Attr func(optionalAttr)

StatefulStandardNormalV2Attr is an optional argument to StatefulStandardNormalV2.

func StatefulStandardNormalV2Dtype 

func StatefulStandardNormalV2Dtype(value tf.DataType) StatefulStandardNormalV2Attr

StatefulStandardNormalV2Dtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatefulTruncatedNormalAttr 

type StatefulTruncatedNormalAttr func(optionalAttr)

StatefulTruncatedNormalAttr is an optional argument to StatefulTruncatedNormal.

func StatefulTruncatedNormalDtype 

func StatefulTruncatedNormalDtype(value tf.DataType) StatefulTruncatedNormalAttr

StatefulTruncatedNormalDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatefulUniformAttr 

type StatefulUniformAttr func(optionalAttr)

StatefulUniformAttr is an optional argument to StatefulUniform.

func StatefulUniformDtype 

func StatefulUniformDtype(value tf.DataType) StatefulUniformAttr

StatefulUniformDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatefulUniformFullIntAttr 

type StatefulUniformFullIntAttr func(optionalAttr)

StatefulUniformFullIntAttr is an optional argument to StatefulUniformFullInt.

func StatefulUniformFullIntDtype 

func StatefulUniformFullIntDtype(value tf.DataType) StatefulUniformFullIntAttr

StatefulUniformFullIntDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_UINT64

type StatelessMultinomialAttr 

type StatelessMultinomialAttr func(optionalAttr)

StatelessMultinomialAttr is an optional argument to StatelessMultinomial.

func StatelessMultinomialOutputDtype 

func StatelessMultinomialOutputDtype(value tf.DataType) StatelessMultinomialAttr

StatelessMultinomialOutputDtype sets the optional output_dtype attribute to value. If not specified, defaults to DT_INT64

type StatelessRandomBinomialAttr 

type StatelessRandomBinomialAttr func(optionalAttr)

StatelessRandomBinomialAttr is an optional argument to StatelessRandomBinomial.

func StatelessRandomBinomialDtype 

func StatelessRandomBinomialDtype(value tf.DataType) StatelessRandomBinomialAttr

StatelessRandomBinomialDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_INT64

type StatelessRandomNormalAttr 

type StatelessRandomNormalAttr func(optionalAttr)

StatelessRandomNormalAttr is an optional argument to StatelessRandomNormal.

func StatelessRandomNormalDtype 

func StatelessRandomNormalDtype(value tf.DataType) StatelessRandomNormalAttr

StatelessRandomNormalDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatelessRandomNormalV2Attr 

type StatelessRandomNormalV2Attr func(optionalAttr)

StatelessRandomNormalV2Attr is an optional argument to StatelessRandomNormalV2.

func StatelessRandomNormalV2Dtype 

func StatelessRandomNormalV2Dtype(value tf.DataType) StatelessRandomNormalV2Attr

StatelessRandomNormalV2Dtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatelessRandomUniformAttr 

type StatelessRandomUniformAttr func(optionalAttr)

StatelessRandomUniformAttr is an optional argument to StatelessRandomUniform.

func StatelessRandomUniformDtype 

func StatelessRandomUniformDtype(value tf.DataType) StatelessRandomUniformAttr

StatelessRandomUniformDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatelessRandomUniformFullIntAttr 

type StatelessRandomUniformFullIntAttr func(optionalAttr)

StatelessRandomUniformFullIntAttr is an optional argument to StatelessRandomUniformFullInt.

func StatelessRandomUniformFullIntDtype 

func StatelessRandomUniformFullIntDtype(value tf.DataType) StatelessRandomUniformFullIntAttr

StatelessRandomUniformFullIntDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_UINT64

type StatelessRandomUniformFullIntV2Attr 

type StatelessRandomUniformFullIntV2Attr func(optionalAttr)

StatelessRandomUniformFullIntV2Attr is an optional argument to StatelessRandomUniformFullIntV2.

func StatelessRandomUniformFullIntV2Dtype 

func StatelessRandomUniformFullIntV2Dtype(value tf.DataType) StatelessRandomUniformFullIntV2Attr

StatelessRandomUniformFullIntV2Dtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_UINT64

type StatelessRandomUniformV2Attr 

type StatelessRandomUniformV2Attr func(optionalAttr)

StatelessRandomUniformV2Attr is an optional argument to StatelessRandomUniformV2.

func StatelessRandomUniformV2Dtype 

func StatelessRandomUniformV2Dtype(value tf.DataType) StatelessRandomUniformV2Attr

StatelessRandomUniformV2Dtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatelessSampleDistortedBoundingBoxAttr 

type StatelessSampleDistortedBoundingBoxAttr func(optionalAttr)

StatelessSampleDistortedBoundingBoxAttr is an optional argument to StatelessSampleDistortedBoundingBox.

func StatelessSampleDistortedBoundingBoxAreaRange 

func StatelessSampleDistortedBoundingBoxAreaRange(value []float32) StatelessSampleDistortedBoundingBoxAttr

StatelessSampleDistortedBoundingBoxAreaRange sets the optional area_range attribute to value.

value: The cropped area of the image must contain a fraction of the supplied image within this range. If not specified, defaults to {f:0.05 f:1}

func StatelessSampleDistortedBoundingBoxAspectRatioRange 

func StatelessSampleDistortedBoundingBoxAspectRatioRange(value []float32) StatelessSampleDistortedBoundingBoxAttr

StatelessSampleDistortedBoundingBoxAspectRatioRange sets the optional aspect_ratio_range attribute to value.

value: The cropped area of the image must have an aspect ratio = width / height within this range. If not specified, defaults to {f:0.75 f:1.33}

func StatelessSampleDistortedBoundingBoxMaxAttempts 

func StatelessSampleDistortedBoundingBoxMaxAttempts(value int64) StatelessSampleDistortedBoundingBoxAttr

StatelessSampleDistortedBoundingBoxMaxAttempts sets the optional max_attempts attribute to value.

value: Number of attempts at generating a cropped region of the image of the specified constraints. After `max_attempts` failures, return the entire image. If not specified, defaults to 100

func StatelessSampleDistortedBoundingBoxUseImageIfNoBoundingBoxes 

func StatelessSampleDistortedBoundingBoxUseImageIfNoBoundingBoxes(value bool) StatelessSampleDistortedBoundingBoxAttr

StatelessSampleDistortedBoundingBoxUseImageIfNoBoundingBoxes sets the optional use_image_if_no_bounding_boxes attribute to value.

value: Controls behavior if no bounding boxes supplied. If true, assume an implicit bounding box covering the whole input. If false, raise an error. If not specified, defaults to false

type StatelessTruncatedNormalAttr 

type StatelessTruncatedNormalAttr func(optionalAttr)

StatelessTruncatedNormalAttr is an optional argument to StatelessTruncatedNormal.

func StatelessTruncatedNormalDtype 

func StatelessTruncatedNormalDtype(value tf.DataType) StatelessTruncatedNormalAttr

StatelessTruncatedNormalDtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StatelessTruncatedNormalV2Attr 

type StatelessTruncatedNormalV2Attr func(optionalAttr)

StatelessTruncatedNormalV2Attr is an optional argument to StatelessTruncatedNormalV2.

func StatelessTruncatedNormalV2Dtype 

func StatelessTruncatedNormalV2Dtype(value tf.DataType) StatelessTruncatedNormalV2Attr

StatelessTruncatedNormalV2Dtype sets the optional dtype attribute to value.

value: The type of the output. If not specified, defaults to DT_FLOAT

type StaticRegexReplaceAttr 

type StaticRegexReplaceAttr func(optionalAttr)

StaticRegexReplaceAttr is an optional argument to StaticRegexReplace.

func StaticRegexReplaceReplaceGlobal 

func StaticRegexReplaceReplaceGlobal(value bool) StaticRegexReplaceAttr

StaticRegexReplaceReplaceGlobal sets the optional replace_global attribute to value.

value: If True, the replacement is global, otherwise the replacement is done only on the first match. If not specified, defaults to true

type StatsAggregatorHandleAttr 

type StatsAggregatorHandleAttr func(optionalAttr)

StatsAggregatorHandleAttr is an optional argument to StatsAggregatorHandle.

func StatsAggregatorHandleContainer 

func StatsAggregatorHandleContainer(value string) StatsAggregatorHandleAttr

StatsAggregatorHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func StatsAggregatorHandleSharedName 

func StatsAggregatorHandleSharedName(value string) StatsAggregatorHandleAttr

StatsAggregatorHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StridedSliceAttr 

type StridedSliceAttr func(optionalAttr)

StridedSliceAttr is an optional argument to StridedSlice.

func StridedSliceBeginMask 

func StridedSliceBeginMask(value int64) StridedSliceAttr

StridedSliceBeginMask sets the optional begin_mask attribute to value.

value: a bitmask where a bit i being 1 means to ignore the begin value and instead use the largest interval possible. At runtime begin[i] will be replaced with `[0, n-1)` if `stride[i] > 0` or `[-1, n-1]` if `stride[i] < 0` If not specified, defaults to 0

func StridedSliceEllipsisMask 

func StridedSliceEllipsisMask(value int64) StridedSliceAttr

StridedSliceEllipsisMask sets the optional ellipsis_mask attribute to value.

value: a bitmask where bit `i` being 1 means the `i`th position is actually an ellipsis. One bit at most can be 1. If `ellipsis_mask == 0`, then an implicit ellipsis mask of `1 << (m+1)` is provided. This means that `foo[3:5] == foo[3:5, ...]`. An ellipsis implicitly creates as many range specifications as necessary to fully specify the sliced range for every dimension. For example for a 4-dimensional tensor `foo` the slice `foo[2, ..., 5:8]` implies `foo[2, :, :, 5:8]`. If not specified, defaults to 0

func StridedSliceEndMask 

func StridedSliceEndMask(value int64) StridedSliceAttr

StridedSliceEndMask sets the optional end_mask attribute to value.

value: analogous to `begin_mask` If not specified, defaults to 0

func StridedSliceNewAxisMask 

func StridedSliceNewAxisMask(value int64) StridedSliceAttr

StridedSliceNewAxisMask sets the optional new_axis_mask attribute to value.

value: a bitmask where bit `i` being 1 means the `i`th specification creates a new shape 1 dimension. For example `foo[:4, tf.newaxis, :2]` would produce a shape `(4, 1, 2)` tensor. If not specified, defaults to 0

func StridedSliceShrinkAxisMask 

func StridedSliceShrinkAxisMask(value int64) StridedSliceAttr

StridedSliceShrinkAxisMask sets the optional shrink_axis_mask attribute to value.

value: a bitmask where bit `i` implies that the `i`th specification should shrink the dimensionality. begin and end must imply a slice of size 1 in the dimension. For example in python one might do `foo[:, 3, :]` which would result in `shrink_axis_mask` being 2. If not specified, defaults to 0

type StridedSliceGradAttr 

type StridedSliceGradAttr func(optionalAttr)

StridedSliceGradAttr is an optional argument to StridedSliceGrad.

func StridedSliceGradBeginMask 

func StridedSliceGradBeginMask(value int64) StridedSliceGradAttr

StridedSliceGradBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradEllipsisMask 

func StridedSliceGradEllipsisMask(value int64) StridedSliceGradAttr

StridedSliceGradEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradEndMask 

func StridedSliceGradEndMask(value int64) StridedSliceGradAttr

StridedSliceGradEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradNewAxisMask 

func StridedSliceGradNewAxisMask(value int64) StridedSliceGradAttr

StridedSliceGradNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradShrinkAxisMask 

func StridedSliceGradShrinkAxisMask(value int64) StridedSliceGradAttr

StridedSliceGradShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type StringFormatAttr 

type StringFormatAttr func(optionalAttr)

StringFormatAttr is an optional argument to StringFormat.

func StringFormatPlaceholder 

func StringFormatPlaceholder(value string) StringFormatAttr

StringFormatPlaceholder sets the optional placeholder attribute to value.

value: A string, at each placeholder in the template a subsequent tensor summary will be inserted. If not specified, defaults to "%s"

func StringFormatSummarize 

func StringFormatSummarize(value int64) StringFormatAttr

StringFormatSummarize sets the optional summarize attribute to value.

value: When formatting the tensor summaries print the first and last summarize entries of each tensor dimension. If not specified, defaults to 3

func StringFormatTemplate 

func StringFormatTemplate(value string) StringFormatAttr

StringFormatTemplate sets the optional template attribute to value.

value: A string, the template to format tensor summaries into. If not specified, defaults to "%s"

type StringJoinAttr 

type StringJoinAttr func(optionalAttr)

StringJoinAttr is an optional argument to StringJoin.

func StringJoinSeparator 

func StringJoinSeparator(value string) StringJoinAttr

StringJoinSeparator sets the optional separator attribute to value.

value: string, an optional join separator. If not specified, defaults to ""

type StringLengthAttr 

type StringLengthAttr func(optionalAttr)

StringLengthAttr is an optional argument to StringLength.

func StringLengthUnit 

func StringLengthUnit(value string) StringLengthAttr

StringLengthUnit sets the optional unit attribute to value.

value: The unit that is counted to compute string length. One of: `"BYTE"` (for the number of bytes in each string) or `"UTF8_CHAR"` (for the number of UTF-8 encoded Unicode code points in each string). Results are undefined if `unit=UTF8_CHAR` and the `input` strings do not contain structurally valid UTF-8. If not specified, defaults to "BYTE"

type StringLowerAttr 

type StringLowerAttr func(optionalAttr)

StringLowerAttr is an optional argument to StringLower.

func StringLowerEncoding 

func StringLowerEncoding(value string) StringLowerAttr

StringLowerEncoding sets the optional encoding attribute to value.

value: Character encoding of `input`. Allowed values are ” and 'utf-8'. Value ” is interpreted as ASCII. If not specified, defaults to ""

type StringSplitAttr 

type StringSplitAttr func(optionalAttr)

StringSplitAttr is an optional argument to StringSplit.

func StringSplitSkipEmpty 

func StringSplitSkipEmpty(value bool) StringSplitAttr

StringSplitSkipEmpty sets the optional skip_empty attribute to value.

value: A `bool`. If `True`, skip the empty strings from the result. If not specified, defaults to true

type StringSplitV2Attr 

type StringSplitV2Attr func(optionalAttr)

StringSplitV2Attr is an optional argument to StringSplitV2.

func StringSplitV2Maxsplit 

func StringSplitV2Maxsplit(value int64) StringSplitV2Attr

StringSplitV2Maxsplit sets the optional maxsplit attribute to value.

value: An `int`. If `maxsplit > 0`, limit of the split of the result. If not specified, defaults to -1

type StringToNumberAttr 

type StringToNumberAttr func(optionalAttr)

StringToNumberAttr is an optional argument to StringToNumber.

func StringToNumberOutType 

func StringToNumberOutType(value tf.DataType) StringToNumberAttr

StringToNumberOutType sets the optional out_type attribute to value.

value: The numeric type to interpret each string in `string_tensor` as. If not specified, defaults to DT_FLOAT

type StringUpperAttr 

type StringUpperAttr func(optionalAttr)

StringUpperAttr is an optional argument to StringUpper.

func StringUpperEncoding 

func StringUpperEncoding(value string) StringUpperAttr

StringUpperEncoding sets the optional encoding attribute to value.

value: Character encoding of `input`. Allowed values are ” and 'utf-8'. Value ” is interpreted as ASCII. If not specified, defaults to ""

type SubstrAttr 

type SubstrAttr func(optionalAttr)

SubstrAttr is an optional argument to Substr.

func SubstrUnit 

func SubstrUnit(value string) SubstrAttr

SubstrUnit sets the optional unit attribute to value.

value: The unit that is used to create the substring. One of: `"BYTE"` (for defining position and length by bytes) or `"UTF8_CHAR"` (for the UTF-8 encoded Unicode code points). The default is `"BYTE"`. Results are undefined if `unit=UTF8_CHAR` and the `input` strings do not contain structurally valid UTF-8. If not specified, defaults to "BYTE"

type SumAttr 

type SumAttr func(optionalAttr)

SumAttr is an optional argument to Sum.

func SumKeepDims 

func SumKeepDims(value bool) SumAttr

SumKeepDims sets the optional keep_dims attribute to value.

value: If true, retain reduced dimensions with length 1. If not specified, defaults to false

type SvdAttr 

type SvdAttr func(optionalAttr)

SvdAttr is an optional argument to Svd.

func SvdComputeUv 

func SvdComputeUv(value bool) SvdAttr

SvdComputeUv sets the optional compute_uv attribute to value.

value: If true, left and right singular vectors will be computed and returned in `u` and `v`, respectively. If false, `u` and `v` are not set and should never referenced. If not specified, defaults to true

func SvdFullMatrices 

func SvdFullMatrices(value bool) SvdAttr

SvdFullMatrices sets the optional full_matrices attribute to value.

value: If true, compute full-sized `u` and `v`. If false (the default), compute only the leading `P` singular vectors. Ignored if `compute_uv` is `False`. If not specified, defaults to false

type TFRecordDatasetAttr 

type TFRecordDatasetAttr func(optionalAttr)

TFRecordDatasetAttr is an optional argument to TFRecordDataset.

func TFRecordDatasetMetadata 

func TFRecordDatasetMetadata(value string) TFRecordDatasetAttr

TFRecordDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type TFRecordDatasetV2Attr added in v0.6.0 

type TFRecordDatasetV2Attr func(optionalAttr)

TFRecordDatasetV2Attr is an optional argument to TFRecordDatasetV2.

func TFRecordDatasetV2Metadata added in v0.6.0 

func TFRecordDatasetV2Metadata(value string) TFRecordDatasetV2Attr

TFRecordDatasetV2Metadata sets the optional metadata attribute to value. If not specified, defaults to ""

type TFRecordReaderV2Attr 

type TFRecordReaderV2Attr func(optionalAttr)

TFRecordReaderV2Attr is an optional argument to TFRecordReaderV2.

func TFRecordReaderV2CompressionType 

func TFRecordReaderV2CompressionType(value string) TFRecordReaderV2Attr

TFRecordReaderV2CompressionType sets the optional compression_type attribute to value. If not specified, defaults to ""

func TFRecordReaderV2Container 

func TFRecordReaderV2Container(value string) TFRecordReaderV2Attr

TFRecordReaderV2Container sets the optional container attribute to value.

value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func TFRecordReaderV2SharedName 

func TFRecordReaderV2SharedName(value string) TFRecordReaderV2Attr

TFRecordReaderV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

type TPUPartitionedInputAttr 

type TPUPartitionedInputAttr func(optionalAttr)

TPUPartitionedInputAttr is an optional argument to TPUPartitionedInput.

func TPUPartitionedInputPartitionDim 

func TPUPartitionedInputPartitionDim(value int64) TPUPartitionedInputAttr

TPUPartitionedInputPartitionDim sets the optional partition_dim attribute to value.

value: An integer describles which dimension is partitioned. -1 means those inputs are replicated. If not specified, defaults to 0

type TPUPartitionedInputV2Attr added in v0.4.0 

type TPUPartitionedInputV2Attr func(optionalAttr)

TPUPartitionedInputV2Attr is an optional argument to TPUPartitionedInputV2.

func TPUPartitionedInputV2IsPacked added in v0.4.0 

func TPUPartitionedInputV2IsPacked(value bool) TPUPartitionedInputV2Attr

TPUPartitionedInputV2IsPacked sets the optional is_packed attribute to value.

value: Indicates whether the input is a packed resource. If not specified, defaults to false

type TPUPartitionedOutputAttr 

type TPUPartitionedOutputAttr func(optionalAttr)

TPUPartitionedOutputAttr is an optional argument to TPUPartitionedOutput.

func TPUPartitionedOutputPartitionDim 

func TPUPartitionedOutputPartitionDim(value int64) TPUPartitionedOutputAttr

TPUPartitionedOutputPartitionDim sets the optional partition_dim attribute to value.

value: An integer describles which dimension is partitioned. If not specified, defaults to 0

type TPUReplicateMetadataAttr 

type TPUReplicateMetadataAttr func(optionalAttr)

TPUReplicateMetadataAttr is an optional argument to TPUReplicateMetadata.

func TPUReplicateMetadataAllowSoftPlacement 

func TPUReplicateMetadataAllowSoftPlacement(value bool) TPUReplicateMetadataAttr

TPUReplicateMetadataAllowSoftPlacement sets the optional allow_soft_placement attribute to value. If not specified, defaults to false

func TPUReplicateMetadataComputationShape 

func TPUReplicateMetadataComputationShape(value []int64) TPUReplicateMetadataAttr

TPUReplicateMetadataComputationShape sets the optional computation_shape attribute to value.

value: DEPRECATED. Use num_cores_per_replica instead. If not specified, defaults to {}

func TPUReplicateMetadataDeviceAssignment 

func TPUReplicateMetadataDeviceAssignment(value []int64) TPUReplicateMetadataAttr

TPUReplicateMetadataDeviceAssignment sets the optional device_assignment attribute to value.

value: The assignment of devices for the computation. If not specified, defaults to {}

func TPUReplicateMetadataHostComputeCore 

func TPUReplicateMetadataHostComputeCore(value []string) TPUReplicateMetadataAttr

TPUReplicateMetadataHostComputeCore sets the optional host_compute_core attribute to value. If not specified, defaults to {}

func TPUReplicateMetadataNumCoresPerReplica 

func TPUReplicateMetadataNumCoresPerReplica(value int64) TPUReplicateMetadataAttr

TPUReplicateMetadataNumCoresPerReplica sets the optional num_cores_per_replica attribute to value.

value: Number of cores per replica. Used for model parallelism. If not specified, defaults to 1

func TPUReplicateMetadataPaddingMap 

func TPUReplicateMetadataPaddingMap(value []string) TPUReplicateMetadataAttr

TPUReplicateMetadataPaddingMap sets the optional padding_map attribute to value. If not specified, defaults to {}

func TPUReplicateMetadataStepMarkerLocation 

func TPUReplicateMetadataStepMarkerLocation(value string) TPUReplicateMetadataAttr

TPUReplicateMetadataStepMarkerLocation sets the optional step_marker_location attribute to value. If not specified, defaults to "STEP_MARK_AT_ENTRY"

func TPUReplicateMetadataTopology 

func TPUReplicateMetadataTopology(value string) TPUReplicateMetadataAttr

TPUReplicateMetadataTopology sets the optional topology attribute to value.

value: TopologyProto indicating the topology of the TPU pod slice. If not specified, defaults to ""

func TPUReplicateMetadataTpuCompileOptionsProto added in v0.2.0 

func TPUReplicateMetadataTpuCompileOptionsProto(value string) TPUReplicateMetadataAttr

TPUReplicateMetadataTpuCompileOptionsProto sets the optional tpu_compile_options_proto attribute to value. If not specified, defaults to ""

func TPUReplicateMetadataUseSpmdForXlaPartitioning 

func TPUReplicateMetadataUseSpmdForXlaPartitioning(value bool) TPUReplicateMetadataAttr

TPUReplicateMetadataUseSpmdForXlaPartitioning sets the optional use_spmd_for_xla_partitioning attribute to value. If not specified, defaults to false

func TPUReplicateMetadataUseTpu 

func TPUReplicateMetadataUseTpu(value bool) TPUReplicateMetadataAttr

TPUReplicateMetadataUseTpu sets the optional use_tpu attribute to value.

value: Whether to place the computation on the TPU. If not specified, defaults to true

type TPUReplicatedInputAttr 

type TPUReplicatedInputAttr func(optionalAttr)

TPUReplicatedInputAttr is an optional argument to TPUReplicatedInput.

func TPUReplicatedInputIndex 

func TPUReplicatedInputIndex(value int64) TPUReplicatedInputAttr

TPUReplicatedInputIndex sets the optional index attribute to value. If not specified, defaults to -1

func TPUReplicatedInputIsMirroredVariable 

func TPUReplicatedInputIsMirroredVariable(value bool) TPUReplicatedInputAttr

TPUReplicatedInputIsMirroredVariable sets the optional is_mirrored_variable attribute to value. If not specified, defaults to false

func TPUReplicatedInputIsPacked 

func TPUReplicatedInputIsPacked(value bool) TPUReplicatedInputAttr

TPUReplicatedInputIsPacked sets the optional is_packed attribute to value. If not specified, defaults to false

type TakeDatasetAttr 

type TakeDatasetAttr func(optionalAttr)

TakeDatasetAttr is an optional argument to TakeDataset.

func TakeDatasetMetadata 

func TakeDatasetMetadata(value string) TakeDatasetAttr

TakeDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type TakeManySparseFromTensorsMapAttr 

type TakeManySparseFromTensorsMapAttr func(optionalAttr)

TakeManySparseFromTensorsMapAttr is an optional argument to TakeManySparseFromTensorsMap.

func TakeManySparseFromTensorsMapContainer 

func TakeManySparseFromTensorsMapContainer(value string) TakeManySparseFromTensorsMapAttr

TakeManySparseFromTensorsMapContainer sets the optional container attribute to value.

value: The container name for the `SparseTensorsMap` read by this op. If not specified, defaults to ""

func TakeManySparseFromTensorsMapSharedName 

func TakeManySparseFromTensorsMapSharedName(value string) TakeManySparseFromTensorsMapAttr

TakeManySparseFromTensorsMapSharedName sets the optional shared_name attribute to value.

value: The shared name for the `SparseTensorsMap` read by this op. It should not be blank; rather the `shared_name` or unique Operation name of the Op that created the original `SparseTensorsMap` should be used. If not specified, defaults to ""

type TensorArrayConcatV2Attr 

type TensorArrayConcatV2Attr func(optionalAttr)

TensorArrayConcatV2Attr is an optional argument to TensorArrayConcatV2.

func TensorArrayConcatV2ElementShapeExcept0 

func TensorArrayConcatV2ElementShapeExcept0(value tf.Shape) TensorArrayConcatV2Attr

TensorArrayConcatV2ElementShapeExcept0 sets the optional element_shape_except0 attribute to value. If not specified, defaults to {unknown_rank:true}

type TensorArrayConcatV3Attr 

type TensorArrayConcatV3Attr func(optionalAttr)

TensorArrayConcatV3Attr is an optional argument to TensorArrayConcatV3.

func TensorArrayConcatV3ElementShapeExcept0 

func TensorArrayConcatV3ElementShapeExcept0(value tf.Shape) TensorArrayConcatV3Attr

TensorArrayConcatV3ElementShapeExcept0 sets the optional element_shape_except0 attribute to value.

value: The expected shape of an element, if known, excluding the first dimension. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, concatenating zero-size TensorArrays is an error. If not specified, defaults to {unknown_rank:true}

type TensorArrayGatherV2Attr 

type TensorArrayGatherV2Attr func(optionalAttr)

TensorArrayGatherV2Attr is an optional argument to TensorArrayGatherV2.

func TensorArrayGatherV2ElementShape 

func TensorArrayGatherV2ElementShape(value tf.Shape) TensorArrayGatherV2Attr

TensorArrayGatherV2ElementShape sets the optional element_shape attribute to value. If not specified, defaults to {unknown_rank:true}

type TensorArrayGatherV3Attr 

type TensorArrayGatherV3Attr func(optionalAttr)

TensorArrayGatherV3Attr is an optional argument to TensorArrayGatherV3.

func TensorArrayGatherV3ElementShape 

func TensorArrayGatherV3ElementShape(value tf.Shape) TensorArrayGatherV3Attr

TensorArrayGatherV3ElementShape sets the optional element_shape attribute to value.

value: The expected shape of an element, if known. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, gathering zero-size TensorArrays is an error. If not specified, defaults to {unknown_rank:true}

type TensorArrayV2Attr 

type TensorArrayV2Attr func(optionalAttr)

TensorArrayV2Attr is an optional argument to TensorArrayV2.

func TensorArrayV2ClearAfterRead 

func TensorArrayV2ClearAfterRead(value bool) TensorArrayV2Attr

TensorArrayV2ClearAfterRead sets the optional clear_after_read attribute to value. If not specified, defaults to true

func TensorArrayV2DynamicSize 

func TensorArrayV2DynamicSize(value bool) TensorArrayV2Attr

TensorArrayV2DynamicSize sets the optional dynamic_size attribute to value. If not specified, defaults to false

func TensorArrayV2ElementShape 

func TensorArrayV2ElementShape(value tf.Shape) TensorArrayV2Attr

TensorArrayV2ElementShape sets the optional element_shape attribute to value. If not specified, defaults to {unknown_rank:true}

func TensorArrayV2TensorArrayName 

func TensorArrayV2TensorArrayName(value string) TensorArrayV2Attr

TensorArrayV2TensorArrayName sets the optional tensor_array_name attribute to value. If not specified, defaults to ""

type TensorArrayV3Attr 

type TensorArrayV3Attr func(optionalAttr)

TensorArrayV3Attr is an optional argument to TensorArrayV3.

func TensorArrayV3ClearAfterRead 

func TensorArrayV3ClearAfterRead(value bool) TensorArrayV3Attr

TensorArrayV3ClearAfterRead sets the optional clear_after_read attribute to value.

value: If true (default), Tensors in the TensorArray are cleared after being read. This disables multiple read semantics but allows early release of memory. If not specified, defaults to true

func TensorArrayV3DynamicSize 

func TensorArrayV3DynamicSize(value bool) TensorArrayV3Attr

TensorArrayV3DynamicSize sets the optional dynamic_size attribute to value.

value: A boolean that determines whether writes to the TensorArray are allowed to grow the size. By default, this is not allowed. If not specified, defaults to false

func TensorArrayV3ElementShape 

func TensorArrayV3ElementShape(value tf.Shape) TensorArrayV3Attr

TensorArrayV3ElementShape sets the optional element_shape attribute to value.

value: The expected shape of an element, if known. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, gathering zero-size TensorArrays is an error. If not specified, defaults to {unknown_rank:true}

func TensorArrayV3IdenticalElementShapes 

func TensorArrayV3IdenticalElementShapes(value bool) TensorArrayV3Attr

TensorArrayV3IdenticalElementShapes sets the optional identical_element_shapes attribute to value.

value: If true (default is false), then all elements in the TensorArray will be expected to have identical shapes. This allows certain behaviors, like dynamically checking for consistent shapes on write, and being able to fill in properly shaped zero tensors on stack -- even if the element_shape attribute is not fully defined. If not specified, defaults to false

func TensorArrayV3TensorArrayName 

func TensorArrayV3TensorArrayName(value string) TensorArrayV3Attr

TensorArrayV3TensorArrayName sets the optional tensor_array_name attribute to value.

value: Overrides the name used for the temporary tensor_array resource. Default value is the name of the 'TensorArray' op (which is guaranteed unique). If not specified, defaults to ""

type TensorDatasetAttr 

type TensorDatasetAttr func(optionalAttr)

TensorDatasetAttr is an optional argument to TensorDataset.

func TensorDatasetMetadata 

func TensorDatasetMetadata(value string) TensorDatasetAttr

TensorDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type TensorListConcatAttr 

type TensorListConcatAttr func(optionalAttr)

TensorListConcatAttr is an optional argument to TensorListConcat.

func TensorListConcatElementShape 

func TensorListConcatElementShape(value tf.Shape) TensorListConcatAttr

TensorListConcatElementShape sets the optional element_shape attribute to value. If not specified, defaults to {unknown_rank:true}

type TensorListSetItemAttr added in v0.5.0 

type TensorListSetItemAttr func(optionalAttr)

TensorListSetItemAttr is an optional argument to TensorListSetItem.

func TensorListSetItemResizeIfIndexOutOfBounds added in v0.5.0 

func TensorListSetItemResizeIfIndexOutOfBounds(value bool) TensorListSetItemAttr

TensorListSetItemResizeIfIndexOutOfBounds sets the optional resize_if_index_out_of_bounds attribute to value. If not specified, defaults to false

type TensorListStackAttr 

type TensorListStackAttr func(optionalAttr)

TensorListStackAttr is an optional argument to TensorListStack.

func TensorListStackNumElements 

func TensorListStackNumElements(value int64) TensorListStackAttr

TensorListStackNumElements sets the optional num_elements attribute to value. If not specified, defaults to -1

type TensorScatterAddAttr added in v0.8.2 

type TensorScatterAddAttr func(optionalAttr)

TensorScatterAddAttr is an optional argument to TensorScatterAdd.

func TensorScatterAddBadIndicesPolicy added in v0.8.2 

func TensorScatterAddBadIndicesPolicy(value string) TensorScatterAddAttr

TensorScatterAddBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type TensorScatterMaxAttr added in v0.8.2 

type TensorScatterMaxAttr func(optionalAttr)

TensorScatterMaxAttr is an optional argument to TensorScatterMax.

func TensorScatterMaxBadIndicesPolicy added in v0.8.2 

func TensorScatterMaxBadIndicesPolicy(value string) TensorScatterMaxAttr

TensorScatterMaxBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type TensorScatterSubAttr added in v0.8.2 

type TensorScatterSubAttr func(optionalAttr)

TensorScatterSubAttr is an optional argument to TensorScatterSub.

func TensorScatterSubBadIndicesPolicy added in v0.8.2 

func TensorScatterSubBadIndicesPolicy(value string) TensorScatterSubAttr

TensorScatterSubBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type TensorScatterUpdateAttr added in v0.8.2 

type TensorScatterUpdateAttr func(optionalAttr)

TensorScatterUpdateAttr is an optional argument to TensorScatterUpdate.

func TensorScatterUpdateBadIndicesPolicy added in v0.8.2 

func TensorScatterUpdateBadIndicesPolicy(value string) TensorScatterUpdateAttr

TensorScatterUpdateBadIndicesPolicy sets the optional bad_indices_policy attribute to value. If not specified, defaults to ""

type TensorSliceDatasetAttr 

type TensorSliceDatasetAttr func(optionalAttr)

TensorSliceDatasetAttr is an optional argument to TensorSliceDataset.

func TensorSliceDatasetIsFiles 

func TensorSliceDatasetIsFiles(value bool) TensorSliceDatasetAttr

TensorSliceDatasetIsFiles sets the optional is_files attribute to value. If not specified, defaults to false

func TensorSliceDatasetMetadata 

func TensorSliceDatasetMetadata(value string) TensorSliceDatasetAttr

TensorSliceDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

func TensorSliceDatasetReplicateOnSplit added in v0.2.0 

func TensorSliceDatasetReplicateOnSplit(value bool) TensorSliceDatasetAttr

TensorSliceDatasetReplicateOnSplit sets the optional replicate_on_split attribute to value. If not specified, defaults to false

type TensorStridedSliceUpdateAttr 

type TensorStridedSliceUpdateAttr func(optionalAttr)

TensorStridedSliceUpdateAttr is an optional argument to TensorStridedSliceUpdate.

func TensorStridedSliceUpdateBeginMask 

func TensorStridedSliceUpdateBeginMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateEllipsisMask 

func TensorStridedSliceUpdateEllipsisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateEndMask 

func TensorStridedSliceUpdateEndMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateNewAxisMask 

func TensorStridedSliceUpdateNewAxisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateShrinkAxisMask 

func TensorStridedSliceUpdateShrinkAxisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type TensorSummaryAttr 

type TensorSummaryAttr func(optionalAttr)

TensorSummaryAttr is an optional argument to TensorSummary.

func TensorSummaryDescription 

func TensorSummaryDescription(value string) TensorSummaryAttr

TensorSummaryDescription sets the optional description attribute to value.

value: A json-encoded SummaryDescription proto. If not specified, defaults to ""

func TensorSummaryDisplayName 

func TensorSummaryDisplayName(value string) TensorSummaryAttr

TensorSummaryDisplayName sets the optional display_name attribute to value.

value: An unused string. If not specified, defaults to ""

func TensorSummaryLabels 

func TensorSummaryLabels(value []string) TensorSummaryAttr

TensorSummaryLabels sets the optional labels attribute to value.

value: An unused list of strings. If not specified, defaults to {}

type TextLineDatasetAttr 

type TextLineDatasetAttr func(optionalAttr)

TextLineDatasetAttr is an optional argument to TextLineDataset.

func TextLineDatasetMetadata 

func TextLineDatasetMetadata(value string) TextLineDatasetAttr

TextLineDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type TextLineReaderV2Attr 

type TextLineReaderV2Attr func(optionalAttr)

TextLineReaderV2Attr is an optional argument to TextLineReaderV2.

func TextLineReaderV2Container 

func TextLineReaderV2Container(value string) TextLineReaderV2Attr

TextLineReaderV2Container sets the optional container attribute to value.

value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func TextLineReaderV2SharedName 

func TextLineReaderV2SharedName(value string) TextLineReaderV2Attr

TextLineReaderV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

func TextLineReaderV2SkipHeaderLines 

func TextLineReaderV2SkipHeaderLines(value int64) TextLineReaderV2Attr

TextLineReaderV2SkipHeaderLines sets the optional skip_header_lines attribute to value.

value: Number of lines to skip from the beginning of every file. If not specified, defaults to 0

type ThreadPoolHandleAttr 

type ThreadPoolHandleAttr func(optionalAttr)

ThreadPoolHandleAttr is an optional argument to ThreadPoolHandle.

func ThreadPoolHandleContainer 

func ThreadPoolHandleContainer(value string) ThreadPoolHandleAttr

ThreadPoolHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ThreadPoolHandleMaxIntraOpParallelism 

func ThreadPoolHandleMaxIntraOpParallelism(value int64) ThreadPoolHandleAttr

ThreadPoolHandleMaxIntraOpParallelism sets the optional max_intra_op_parallelism attribute to value.

value: The maximum degree of parallelism to use within operations that execute on this threadpool. If not specified, defaults to 1

func ThreadPoolHandleSharedName 

func ThreadPoolHandleSharedName(value string) ThreadPoolHandleAttr

ThreadPoolHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ThreadUnsafeUnigramCandidateSamplerAttr 

type ThreadUnsafeUnigramCandidateSamplerAttr func(optionalAttr)

ThreadUnsafeUnigramCandidateSamplerAttr is an optional argument to ThreadUnsafeUnigramCandidateSampler.

func ThreadUnsafeUnigramCandidateSamplerSeed 

func ThreadUnsafeUnigramCandidateSamplerSeed(value int64) ThreadUnsafeUnigramCandidateSamplerAttr

ThreadUnsafeUnigramCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func ThreadUnsafeUnigramCandidateSamplerSeed2 

func ThreadUnsafeUnigramCandidateSamplerSeed2(value int64) ThreadUnsafeUnigramCandidateSamplerAttr

ThreadUnsafeUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type TopKAttr 

type TopKAttr func(optionalAttr)

TopKAttr is an optional argument to TopK.

func TopKSorted 

func TopKSorted(value bool) TopKAttr

TopKSorted sets the optional sorted attribute to value.

value: If true the resulting `k` elements will be sorted by the values in descending order. If not specified, defaults to true

type TopKV2Attr 

type TopKV2Attr func(optionalAttr)

TopKV2Attr is an optional argument to TopKV2.

func TopKV2IndexType added in v0.5.0 

func TopKV2IndexType(value tf.DataType) TopKV2Attr

TopKV2IndexType sets the optional index_type attribute to value. If not specified, defaults to DT_INT32

func TopKV2Sorted 

func TopKV2Sorted(value bool) TopKV2Attr

TopKV2Sorted sets the optional sorted attribute to value.

value: If true the resulting `k` elements will be sorted by the values in descending order. If not specified, defaults to true

type TridiagonalSolveAttr 

type TridiagonalSolveAttr func(optionalAttr)

TridiagonalSolveAttr is an optional argument to TridiagonalSolve.

func TridiagonalSolvePartialPivoting 

func TridiagonalSolvePartialPivoting(value bool) TridiagonalSolveAttr

TridiagonalSolvePartialPivoting sets the optional partial_pivoting attribute to value.

value: Whether to apply partial pivoting. Partial pivoting makes the procedure more stable, but slower. If not specified, defaults to true

func TridiagonalSolvePerturbSingular 

func TridiagonalSolvePerturbSingular(value bool) TridiagonalSolveAttr

TridiagonalSolvePerturbSingular sets the optional perturb_singular attribute to value. If not specified, defaults to false

type TruncatedNormalAttr 

type TruncatedNormalAttr func(optionalAttr)

TruncatedNormalAttr is an optional argument to TruncatedNormal.

func TruncatedNormalSeed 

func TruncatedNormalSeed(value int64) TruncatedNormalAttr

TruncatedNormalSeed sets the optional seed attribute to value.

value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func TruncatedNormalSeed2 

func TruncatedNormalSeed2(value int64) TruncatedNormalAttr

TruncatedNormalSeed2 sets the optional seed2 attribute to value.

value: A second seed to avoid seed collision. If not specified, defaults to 0

type UnbatchAttr 

type UnbatchAttr func(optionalAttr)

UnbatchAttr is an optional argument to Unbatch.

func UnbatchContainer 

func UnbatchContainer(value string) UnbatchAttr

UnbatchContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnbatchSharedName 

func UnbatchSharedName(value string) UnbatchAttr

UnbatchSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UnbatchDatasetAttr 

type UnbatchDatasetAttr func(optionalAttr)

UnbatchDatasetAttr is an optional argument to UnbatchDataset.

func UnbatchDatasetMetadata 

func UnbatchDatasetMetadata(value string) UnbatchDatasetAttr

UnbatchDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type UnbatchGradAttr 

type UnbatchGradAttr func(optionalAttr)

UnbatchGradAttr is an optional argument to UnbatchGrad.

func UnbatchGradContainer 

func UnbatchGradContainer(value string) UnbatchGradAttr

UnbatchGradContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnbatchGradSharedName 

func UnbatchGradSharedName(value string) UnbatchGradAttr

UnbatchGradSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UnicodeDecodeAttr 

type UnicodeDecodeAttr func(optionalAttr)

UnicodeDecodeAttr is an optional argument to UnicodeDecode.

func UnicodeDecodeErrors 

func UnicodeDecodeErrors(value string) UnicodeDecodeAttr

UnicodeDecodeErrors sets the optional errors attribute to value.

value: Error handling policy when there is invalid formatting found in the input. The value of 'strict' will cause the operation to produce a InvalidArgument error on any invalid input formatting. A value of 'replace' (the default) will cause the operation to replace any invalid formatting in the input with the `replacement_char` codepoint. A value of 'ignore' will cause the operation to skip any invalid formatting in the input and produce no corresponding output character. If not specified, defaults to "replace"

func UnicodeDecodeReplaceControlCharacters 

func UnicodeDecodeReplaceControlCharacters(value bool) UnicodeDecodeAttr

UnicodeDecodeReplaceControlCharacters sets the optional replace_control_characters attribute to value.

value: Whether to replace the C0 control characters (00-1F) with the `replacement_char`. Default is false. If not specified, defaults to false

func UnicodeDecodeReplacementChar 

func UnicodeDecodeReplacementChar(value int64) UnicodeDecodeAttr

UnicodeDecodeReplacementChar sets the optional replacement_char attribute to value.

value: The replacement character codepoint to be used in place of any invalid formatting in the input when `errors='replace'`. Any valid unicode codepoint may be used. The default value is the default unicode replacement character is 0xFFFD or U+65533.) If not specified, defaults to 65533

func UnicodeDecodeTsplits 

func UnicodeDecodeTsplits(value tf.DataType) UnicodeDecodeAttr

UnicodeDecodeTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type UnicodeDecodeWithOffsetsAttr 

type UnicodeDecodeWithOffsetsAttr func(optionalAttr)

UnicodeDecodeWithOffsetsAttr is an optional argument to UnicodeDecodeWithOffsets.

func UnicodeDecodeWithOffsetsErrors 

func UnicodeDecodeWithOffsetsErrors(value string) UnicodeDecodeWithOffsetsAttr

UnicodeDecodeWithOffsetsErrors sets the optional errors attribute to value.

value: Error handling policy when there is invalid formatting found in the input. The value of 'strict' will cause the operation to produce a InvalidArgument error on any invalid input formatting. A value of 'replace' (the default) will cause the operation to replace any invalid formatting in the input with the `replacement_char` codepoint. A value of 'ignore' will cause the operation to skip any invalid formatting in the input and produce no corresponding output character. If not specified, defaults to "replace"

func UnicodeDecodeWithOffsetsReplaceControlCharacters 

func UnicodeDecodeWithOffsetsReplaceControlCharacters(value bool) UnicodeDecodeWithOffsetsAttr

UnicodeDecodeWithOffsetsReplaceControlCharacters sets the optional replace_control_characters attribute to value.

value: Whether to replace the C0 control characters (00-1F) with the `replacement_char`. Default is false. If not specified, defaults to false

func UnicodeDecodeWithOffsetsReplacementChar 

func UnicodeDecodeWithOffsetsReplacementChar(value int64) UnicodeDecodeWithOffsetsAttr

UnicodeDecodeWithOffsetsReplacementChar sets the optional replacement_char attribute to value.

value: The replacement character codepoint to be used in place of any invalid formatting in the input when `errors='replace'`. Any valid unicode codepoint may be used. The default value is the default unicode replacement character is 0xFFFD or U+65533.) If not specified, defaults to 65533

func UnicodeDecodeWithOffsetsTsplits 

func UnicodeDecodeWithOffsetsTsplits(value tf.DataType) UnicodeDecodeWithOffsetsAttr

UnicodeDecodeWithOffsetsTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type UnicodeEncodeAttr 

type UnicodeEncodeAttr func(optionalAttr)

UnicodeEncodeAttr is an optional argument to UnicodeEncode.

func UnicodeEncodeErrors 

func UnicodeEncodeErrors(value string) UnicodeEncodeAttr

UnicodeEncodeErrors sets the optional errors attribute to value.

value: Error handling policy when there is invalid formatting found in the input. The value of 'strict' will cause the operation to produce a InvalidArgument error on any invalid input formatting. A value of 'replace' (the default) will cause the operation to replace any invalid formatting in the input with the `replacement_char` codepoint. A value of 'ignore' will cause the operation to skip any invalid formatting in the input and produce no corresponding output character. If not specified, defaults to "replace"

func UnicodeEncodeReplacementChar 

func UnicodeEncodeReplacementChar(value int64) UnicodeEncodeAttr

UnicodeEncodeReplacementChar sets the optional replacement_char attribute to value.

value: The replacement character codepoint to be used in place of any invalid formatting in the input when `errors='replace'`. Any valid unicode codepoint may be used. The default value is the default unicode replacement character is 0xFFFD (U+65533). If not specified, defaults to 65533

type UnicodeTranscodeAttr 

type UnicodeTranscodeAttr func(optionalAttr)

UnicodeTranscodeAttr is an optional argument to UnicodeTranscode.

func UnicodeTranscodeErrors 

func UnicodeTranscodeErrors(value string) UnicodeTranscodeAttr

UnicodeTranscodeErrors sets the optional errors attribute to value.

value: Error handling policy when there is invalid formatting found in the input. The value of 'strict' will cause the operation to produce a InvalidArgument error on any invalid input formatting. A value of 'replace' (the default) will cause the operation to replace any invalid formatting in the input with the `replacement_char` codepoint. A value of 'ignore' will cause the operation to skip any invalid formatting in the input and produce no corresponding output character. If not specified, defaults to "replace"

func UnicodeTranscodeReplaceControlCharacters 

func UnicodeTranscodeReplaceControlCharacters(value bool) UnicodeTranscodeAttr

UnicodeTranscodeReplaceControlCharacters sets the optional replace_control_characters attribute to value.

value: Whether to replace the C0 control characters (00-1F) with the `replacement_char`. Default is false. If not specified, defaults to false

func UnicodeTranscodeReplacementChar 

func UnicodeTranscodeReplacementChar(value int64) UnicodeTranscodeAttr

UnicodeTranscodeReplacementChar sets the optional replacement_char attribute to value.

value: The replacement character codepoint to be used in place of any invalid formatting in the input when `errors='replace'`. Any valid unicode codepoint may be used. The default value is the default unicode replacement character is 0xFFFD or U+65533.)

Note that for UTF-8, passing a replacement character expressible in 1 byte, such as ' ', will preserve string alignment to the source since invalid bytes will be replaced with a 1-byte replacement. For UTF-16-BE and UTF-16-LE, any 1 or 2 byte replacement character will preserve byte alignment to the source. If not specified, defaults to 65533

type UniformCandidateSamplerAttr 

type UniformCandidateSamplerAttr func(optionalAttr)

UniformCandidateSamplerAttr is an optional argument to UniformCandidateSampler.

func UniformCandidateSamplerSeed 

func UniformCandidateSamplerSeed(value int64) UniformCandidateSamplerAttr

UniformCandidateSamplerSeed sets the optional seed attribute to value.

value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0

func UniformCandidateSamplerSeed2 

func UniformCandidateSamplerSeed2(value int64) UniformCandidateSamplerAttr

UniformCandidateSamplerSeed2 sets the optional seed2 attribute to value.

value: An second seed to avoid seed collision. If not specified, defaults to 0

type UniformDequantizeAttr added in v0.2.0 

type UniformDequantizeAttr func(optionalAttr)

UniformDequantizeAttr is an optional argument to UniformDequantize.

func UniformDequantizeQuantizationAxis added in v0.2.0 

func UniformDequantizeQuantizationAxis(value int64) UniformDequantizeAttr

UniformDequantizeQuantizationAxis sets the optional quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. Otherwise, it must be set within range [0, input.dims()). If not specified, defaults to -1

type UniformQuantizeAttr added in v0.3.0 

type UniformQuantizeAttr func(optionalAttr)

UniformQuantizeAttr is an optional argument to UniformQuantize.

func UniformQuantizeQuantizationAxis added in v0.3.0 

func UniformQuantizeQuantizationAxis(value int64) UniformQuantizeAttr

UniformQuantizeQuantizationAxis sets the optional quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. Otherwise, it must be set within range [0, input.dims()). If not specified, defaults to -1

type UniformQuantizedAddAttr added in v0.4.0 

type UniformQuantizedAddAttr func(optionalAttr)

UniformQuantizedAddAttr is an optional argument to UniformQuantizedAdd.

func UniformQuantizedAddLhsQuantizationAxis added in v0.4.0 

func UniformQuantizedAddLhsQuantizationAxis(value int64) UniformQuantizedAddAttr

UniformQuantizedAddLhsQuantizationAxis sets the optional lhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `lhs`, only per-tensor quantization is supported. Thus, this must be set to -1. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedAddOutputQuantizationAxis added in v0.4.0 

func UniformQuantizedAddOutputQuantizationAxis(value int64) UniformQuantizedAddAttr

UniformQuantizedAddOutputQuantizationAxis sets the optional output_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `output`, only per-tensor quantization or per-channel quantization along `output_feature_dimension` is supported. Thus, this must be set to -1 or `dimension_numbers.output_feature_dimension`. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedAddRhsQuantizationAxis added in v0.4.0 

func UniformQuantizedAddRhsQuantizationAxis(value int64) UniformQuantizedAddAttr

UniformQuantizedAddRhsQuantizationAxis sets the optional rhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `rhs`, only per-tensor quantization or per-channel quantization along `kernel_output_feature_dimension` is supported. Thus, this must be set to -1 or `dimension_numbers.kernel_output_feature_dimension`. Other values will raise error at OpKernel construction. If not specified, defaults to -1

type UniformQuantizedClipByValueAttr added in v0.3.0 

type UniformQuantizedClipByValueAttr func(optionalAttr)

UniformQuantizedClipByValueAttr is an optional argument to UniformQuantizedClipByValue.

func UniformQuantizedClipByValueQuantizationAxis added in v0.3.0 

func UniformQuantizedClipByValueQuantizationAxis(value int64) UniformQuantizedClipByValueAttr

UniformQuantizedClipByValueQuantizationAxis sets the optional quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. Otherwise, it must be set within range [0, operand.dims()). If not specified, defaults to -1

type UniformQuantizedConvolutionAttr added in v0.4.0 

type UniformQuantizedConvolutionAttr func(optionalAttr)

UniformQuantizedConvolutionAttr is an optional argument to UniformQuantizedConvolution.

func UniformQuantizedConvolutionBatchGroupCount added in v0.4.0 

func UniformQuantizedConvolutionBatchGroupCount(value int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionBatchGroupCount sets the optional batch_group_count attribute to value.

value: The number of batch groups. Used for grouped filters. Must be a divisor of `output_feature`. If not specified, defaults to 1

func UniformQuantizedConvolutionDimensionNumbers added in v0.4.0 

func UniformQuantizedConvolutionDimensionNumbers(value string) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionDimensionNumbers sets the optional dimension_numbers attribute to value.

value: Structure of dimension information for the convolution op. Must be an empty string (default) or a serialized string of `tensorflow.UniformQuantizedConvolutionDimensionNumbersAttr` proto. If empty string, the default is `("NCHW", "OIHW", "NCHW")` (for a 2D convolution). If not specified, defaults to ""

func UniformQuantizedConvolutionExplicitPadding added in v0.4.0 

func UniformQuantizedConvolutionExplicitPadding(value []int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionExplicitPadding sets the optional explicit_padding attribute to value.

value: If `padding` is `"EXPLICIT"`, must be set as a list indicating the explicit paddings at the start and end of each `lhs` spatial dimension. Otherwise, this must be empty.

(If used,) Must be a list of size `2 * (number of lhs spatial dimensions)`, where `(explicit_padding[2 * i], explicit_padding[2 * i + 1])` indicates `(start_padding, end_padding)` of `spatial_dimensions[i]`. If not specified, defaults to {}

func UniformQuantizedConvolutionFeatureGroupCount added in v0.4.0 

func UniformQuantizedConvolutionFeatureGroupCount(value int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionFeatureGroupCount sets the optional feature_group_count attribute to value.

value: The number of feature groups. Used for grouped convolutions. Must be a divisor of both `lhs_feature` and `output_feature`. If not specified, defaults to 1

func UniformQuantizedConvolutionLhsDilation added in v0.4.0 

func UniformQuantizedConvolutionLhsDilation(value []int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionLhsDilation sets the optional lhs_dilation attribute to value.

value: The dilation factor to apply in each spatial dimension of `lhs`. Must be an empty list (default) or a list of size (number of `lhs` spatial dimensions). If empty list, the dilation for each `lhs` spatial dimension is set to 1. If not specified, defaults to {}

func UniformQuantizedConvolutionLhsQuantizationAxis added in v0.4.0 

func UniformQuantizedConvolutionLhsQuantizationAxis(value int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionLhsQuantizationAxis sets the optional lhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `lhs`, only per-tensor quantization is supported. Thus, this must be set to -1. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedConvolutionOutputQuantizationAxis added in v0.4.0 

func UniformQuantizedConvolutionOutputQuantizationAxis(value int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionOutputQuantizationAxis sets the optional output_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `output`, only per-tensor quantization or per-channel quantization along `output_feature_dimension` is supported. Thus, this must be set to -1 or `dimension_numbers.output_feature_dimension`. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedConvolutionRhsDilation added in v0.4.0 

func UniformQuantizedConvolutionRhsDilation(value []int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionRhsDilation sets the optional rhs_dilation attribute to value.

value: The dilation factor to apply in each spatial dimension of `rhs`. Must be an empty list (default) or a list of size (number of `rhs` spatial dimensions). If empty list, the dilation for each `rhs` spatial dimension is set to 1. If not specified, defaults to {}

func UniformQuantizedConvolutionRhsQuantizationAxis added in v0.4.0 

func UniformQuantizedConvolutionRhsQuantizationAxis(value int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionRhsQuantizationAxis sets the optional rhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `rhs`, only per-tensor quantization or per-channel quantization along `kernel_output_feature_dimension` is supported. Thus, this must be set to -1 or `dimension_numbers.kernel_output_feature_dimension`. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedConvolutionWindowStrides added in v0.4.0 

func UniformQuantizedConvolutionWindowStrides(value []int64) UniformQuantizedConvolutionAttr

UniformQuantizedConvolutionWindowStrides sets the optional window_strides attribute to value.

value: The stride of the sliding window for each spatial dimension of `lhs`. Must be an empty list (default) or a list of size (number of spatial dimensions). If an empty list is provided, the stride for each spatial dimension is set to 1. If not specified, defaults to {}

type UniformQuantizedConvolutionHybridAttr added in v0.4.0 

type UniformQuantizedConvolutionHybridAttr func(optionalAttr)

UniformQuantizedConvolutionHybridAttr is an optional argument to UniformQuantizedConvolutionHybrid.

func UniformQuantizedConvolutionHybridBatchGroupCount added in v0.4.0 

func UniformQuantizedConvolutionHybridBatchGroupCount(value int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridBatchGroupCount sets the optional batch_group_count attribute to value.

value: The number of batch groups. Used for grouped filters. Must be a divisor of output_feature. If not specified, defaults to 1

func UniformQuantizedConvolutionHybridDimensionNumbers added in v0.4.0 

func UniformQuantizedConvolutionHybridDimensionNumbers(value string) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridDimensionNumbers sets the optional dimension_numbers attribute to value.

value: Structure of dimension information for the convolution op. Must be an empty string (default) or a serialized string of tensorflow.UniformQuantizedConvolutionDimensionNumbersAttr proto. If empty string, the default is `("NCHW", "OIHW", "NCHW")` (for a 2D convolution). If not specified, defaults to ""

func UniformQuantizedConvolutionHybridExplicitPadding added in v0.4.0 

func UniformQuantizedConvolutionHybridExplicitPadding(value []int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridExplicitPadding sets the optional explicit_padding attribute to value.

value: If `padding` Attr is `"EXPLICIT"`, must be set as a list indicating the explicit paddings at the start and end of each lhs spatial dimension. Otherwise, this Attr is must be empty.

(If used,) Must be a list of size 2 * (number of lhs spatial dimensions), where (explicit_padding[2 * i], explicit_padding[2 * i + 1]) indicates spatial_dimensions[i] (start_padding, end_padding). If not specified, defaults to {}

func UniformQuantizedConvolutionHybridFeatureGroupCount added in v0.4.0 

func UniformQuantizedConvolutionHybridFeatureGroupCount(value int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridFeatureGroupCount sets the optional feature_group_count attribute to value.

value: The number of feature groups. Used for grouped convolutions. Must be a divisor of both lhs_feature and output_feature. If not specified, defaults to 1

func UniformQuantizedConvolutionHybridLhsDilation added in v0.4.0 

func UniformQuantizedConvolutionHybridLhsDilation(value []int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridLhsDilation sets the optional lhs_dilation attribute to value.

value: The dilation factor to apply in each spatial dimension of `lhs`. Must be an empty list (default) or a list of size (number of lhs spatial dimensions). If empty list, the dilation for each lhs spatial dimension is set to 1. If not specified, defaults to {}

func UniformQuantizedConvolutionHybridRhsDilation added in v0.4.0 

func UniformQuantizedConvolutionHybridRhsDilation(value []int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridRhsDilation sets the optional rhs_dilation attribute to value.

value: The dilation factor to apply in each spatial dimension of `rhs`. Must be an empty list (default) or a list of size (number of rhs spatial dimensions). If empty list, the dilation for each rhs spatial dimension is set to 1. If not specified, defaults to {}

func UniformQuantizedConvolutionHybridRhsQuantizationAxis added in v0.4.0 

func UniformQuantizedConvolutionHybridRhsQuantizationAxis(value int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridRhsQuantizationAxis sets the optional rhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For the `rhs`, only per-tensor quantization or per-channel quantization along kernel_output_feature_dimension is supported. Thus, this attribute must be set to -1 or `dimension_numbers.kernel_output_feature_dimension`. Other values will raise error at OpKernel construction. If not specified, defaults to -1

func UniformQuantizedConvolutionHybridWindowStrides added in v0.4.0 

func UniformQuantizedConvolutionHybridWindowStrides(value []int64) UniformQuantizedConvolutionHybridAttr

UniformQuantizedConvolutionHybridWindowStrides sets the optional window_strides attribute to value.

value: The stride of the sliding window for each spatial dimension of `lhs`. Must be an empty list (default) or a list of size (number of spatial dimensions). If an empty list is provided, the stride for each spatial dimension is set to 1. If not specified, defaults to {}

type UniformQuantizedDotAttr added in v0.3.0 

type UniformQuantizedDotAttr func(optionalAttr)

UniformQuantizedDotAttr is an optional argument to UniformQuantizedDot.

func UniformQuantizedDotLhsQuantizationAxis added in v0.3.0 

func UniformQuantizedDotLhsQuantizationAxis(value int64) UniformQuantizedDotAttr

UniformQuantizedDotLhsQuantizationAxis sets the optional lhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For dot op lhs, only per-tensor quantization is supported. Thus, this attribute must be set to -1. Other values are rejected. If not specified, defaults to -1

func UniformQuantizedDotOutputQuantizationAxis added in v0.3.0 

func UniformQuantizedDotOutputQuantizationAxis(value int64) UniformQuantizedDotAttr

UniformQuantizedDotOutputQuantizationAxis sets the optional output_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For dot op output, only per-tensor quantization or per-channel quantization along dimension 1 is supported. Thus, this attribute must be set to -1 or 1. Other values are rejected. If not specified, defaults to -1

func UniformQuantizedDotRhsQuantizationAxis added in v0.3.0 

func UniformQuantizedDotRhsQuantizationAxis(value int64) UniformQuantizedDotAttr

UniformQuantizedDotRhsQuantizationAxis sets the optional rhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For dot op rhs, only per-tensor quantization or per-channel quantization along dimension 1 is supported. Thus, this attribute must be set to -1 or 1. Other values are rejected. If not specified, defaults to -1

type UniformQuantizedDotHybridAttr added in v0.2.0 

type UniformQuantizedDotHybridAttr func(optionalAttr)

UniformQuantizedDotHybridAttr is an optional argument to UniformQuantizedDotHybrid.

func UniformQuantizedDotHybridRhsQuantizationAxis added in v0.2.0 

func UniformQuantizedDotHybridRhsQuantizationAxis(value int64) UniformQuantizedDotHybridAttr

UniformQuantizedDotHybridRhsQuantizationAxis sets the optional rhs_quantization_axis attribute to value.

value: Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. For dot op rhs, only per-tensor quantization or per-channel quantization along dimension 1 is supported. Thus, this attribute must be set to -1 or 1. Other values are rejected. If not specified, defaults to -1

type UniformRequantizeAttr added in v0.3.0 

type UniformRequantizeAttr func(optionalAttr)

UniformRequantizeAttr is an optional argument to UniformRequantize.

func UniformRequantizeInputQuantizationAxis added in v0.3.0 

func UniformRequantizeInputQuantizationAxis(value int64) UniformRequantizeAttr

UniformRequantizeInputQuantizationAxis sets the optional input_quantization_axis attribute to value.

value: The quantization axis that was used when quantizing original data that `input` represents. Indicates the dimension index of the tensor where per-axis quantization is applied for the slices along that dimension. If set to -1 (default), this indicates per-tensor quantization. Otherwise, it must be set within range [0, input.dims()). If not specified, defaults to -1

func UniformRequantizeOutputQuantizationAxis added in v0.3.0 

func UniformRequantizeOutputQuantizationAxis(value int64) UniformRequantizeAttr

UniformRequantizeOutputQuantizationAxis sets the optional output_quantization_axis attribute to value.

value: The new quantization axis to use to quantize original data that `input` represents. If not specified, defaults to -1

type UniqueAttr 

type UniqueAttr func(optionalAttr)

UniqueAttr is an optional argument to Unique.

func UniqueOutIdx 

func UniqueOutIdx(value tf.DataType) UniqueAttr

UniqueOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueDatasetAttr 

type UniqueDatasetAttr func(optionalAttr)

UniqueDatasetAttr is an optional argument to UniqueDataset.

func UniqueDatasetMetadata 

func UniqueDatasetMetadata(value string) UniqueDatasetAttr

UniqueDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type UniqueV2Attr 

type UniqueV2Attr func(optionalAttr)

UniqueV2Attr is an optional argument to UniqueV2.

func UniqueV2OutIdx 

func UniqueV2OutIdx(value tf.DataType) UniqueV2Attr

UniqueV2OutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueWithCountsAttr 

type UniqueWithCountsAttr func(optionalAttr)

UniqueWithCountsAttr is an optional argument to UniqueWithCounts.

func UniqueWithCountsOutIdx 

func UniqueWithCountsOutIdx(value tf.DataType) UniqueWithCountsAttr

UniqueWithCountsOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueWithCountsV2Attr 

type UniqueWithCountsV2Attr func(optionalAttr)

UniqueWithCountsV2Attr is an optional argument to UniqueWithCountsV2.

func UniqueWithCountsV2OutIdx 

func UniqueWithCountsV2OutIdx(value tf.DataType) UniqueWithCountsV2Attr

UniqueWithCountsV2OutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UnpackAttr 

type UnpackAttr func(optionalAttr)

UnpackAttr is an optional argument to Unpack.

func UnpackAxis 

func UnpackAxis(value int64) UnpackAttr

UnpackAxis sets the optional axis attribute to value.

value: Dimension along which to unpack. Negative values wrap around, so the valid range is `[-R, R)`. If not specified, defaults to 0

type UnstageAttr 

type UnstageAttr func(optionalAttr)

UnstageAttr is an optional argument to Unstage.

func UnstageCapacity 

func UnstageCapacity(value int64) UnstageAttr

UnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func UnstageContainer 

func UnstageContainer(value string) UnstageAttr

UnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnstageMemoryLimit 

func UnstageMemoryLimit(value int64) UnstageAttr

UnstageMemoryLimit sets the optional memory_limit attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func UnstageSharedName 

func UnstageSharedName(value string) UnstageAttr

UnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UpperBoundAttr 

type UpperBoundAttr func(optionalAttr)

UpperBoundAttr is an optional argument to UpperBound.

func UpperBoundOutType 

func UpperBoundOutType(value tf.DataType) UpperBoundAttr

UpperBoundOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type VarHandleOpAttr 

type VarHandleOpAttr func(optionalAttr)

VarHandleOpAttr is an optional argument to VarHandleOp.

func VarHandleOpAllowedDevices 

func VarHandleOpAllowedDevices(value []string) VarHandleOpAttr

VarHandleOpAllowedDevices sets the optional allowed_devices attribute to value.

value: DEPRECATED. The allowed devices containing the resource variable. Set when the output ResourceHandle represents a per-replica/partitioned resource variable. If not specified, defaults to {}

func VarHandleOpContainer 

func VarHandleOpContainer(value string) VarHandleOpAttr

VarHandleOpContainer sets the optional container attribute to value.

value: the container this variable is placed in. If not specified, defaults to ""

func VarHandleOpDebugName added in v0.7.0 

func VarHandleOpDebugName(value string) VarHandleOpAttr

VarHandleOpDebugName sets the optional debug_name attribute to value.

value: the user-given name, which still applies in anonymous mode. If not specified, defaults to ""

func VarHandleOpSharedName 

func VarHandleOpSharedName(value string) VarHandleOpAttr

VarHandleOpSharedName sets the optional shared_name attribute to value.

value: the name by which this variable is referred to. If not specified, defaults to ""

type VariableShapeAttr 

type VariableShapeAttr func(optionalAttr)

VariableShapeAttr is an optional argument to VariableShape.

func VariableShapeOutType 

func VariableShapeOutType(value tf.DataType) VariableShapeAttr

VariableShapeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type WholeFileReaderV2Attr 

type WholeFileReaderV2Attr func(optionalAttr)

WholeFileReaderV2Attr is an optional argument to WholeFileReaderV2.

func WholeFileReaderV2Container 

func WholeFileReaderV2Container(value string) WholeFileReaderV2Attr

WholeFileReaderV2Container sets the optional container attribute to value.

value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""

func WholeFileReaderV2SharedName 

func WholeFileReaderV2SharedName(value string) WholeFileReaderV2Attr

WholeFileReaderV2SharedName sets the optional shared_name attribute to value.

value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""

type WindowDatasetAttr 

type WindowDatasetAttr func(optionalAttr)

WindowDatasetAttr is an optional argument to WindowDataset.

func WindowDatasetMetadata 

func WindowDatasetMetadata(value string) WindowDatasetAttr

WindowDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

type WriteAudioSummaryAttr 

type WriteAudioSummaryAttr func(optionalAttr)

WriteAudioSummaryAttr is an optional argument to WriteAudioSummary.

func WriteAudioSummaryMaxOutputs 

func WriteAudioSummaryMaxOutputs(value int64) WriteAudioSummaryAttr

WriteAudioSummaryMaxOutputs sets the optional max_outputs attribute to value. If not specified, defaults to 3

REQUIRES: value >= 1

type WriteImageSummaryAttr 

type WriteImageSummaryAttr func(optionalAttr)

WriteImageSummaryAttr is an optional argument to WriteImageSummary.

func WriteImageSummaryMaxImages 

func WriteImageSummaryMaxImages(value int64) WriteImageSummaryAttr

WriteImageSummaryMaxImages sets the optional max_images attribute to value. If not specified, defaults to 3

REQUIRES: value >= 1

type XlaConcatNDAttr 

type XlaConcatNDAttr func(optionalAttr)

XlaConcatNDAttr is an optional argument to XlaConcatND.

func XlaConcatNDPaddings 

func XlaConcatNDPaddings(value []int64) XlaConcatNDAttr

XlaConcatNDPaddings sets the optional paddings attribute to value.

value: Optional list of right paddings per dimension to strip from the final merged tensor. These paddings must not exceed the dimension size of the merged result prior to stripping paddings. If not specified, defaults to {}

type XlaConvV2Attr 

type XlaConvV2Attr func(optionalAttr)

XlaConvV2Attr is an optional argument to XlaConvV2.

func XlaConvV2BatchGroupCount 

func XlaConvV2BatchGroupCount(value int64) XlaConvV2Attr

XlaConvV2BatchGroupCount sets the optional batch_group_count attribute to value.

value: number of batch groups or grouped filters. If not specified, defaults to 1

type XlaRngBitGeneratorAttr 

type XlaRngBitGeneratorAttr func(optionalAttr)

XlaRngBitGeneratorAttr is an optional argument to XlaRngBitGenerator.

func XlaRngBitGeneratorDtype 

func XlaRngBitGeneratorDtype(value tf.DataType) XlaRngBitGeneratorAttr

XlaRngBitGeneratorDtype sets the optional dtype attribute to value.

value: The type of the tensor. If not specified, defaults to DT_UINT64

type XlaShardingAttr 

type XlaShardingAttr func(optionalAttr)

XlaShardingAttr is an optional argument to XlaSharding.

func XlaShardingSharding 

func XlaShardingSharding(value string) XlaShardingAttr

XlaShardingSharding sets the optional sharding attribute to value. If not specified, defaults to ""

func XlaShardingUnspecifiedDims 

func XlaShardingUnspecifiedDims(value []int64) XlaShardingAttr

XlaShardingUnspecifiedDims sets the optional unspecified_dims attribute to value. If not specified, defaults to {}

type XlaSplitNDAttr 

type XlaSplitNDAttr func(optionalAttr)

XlaSplitNDAttr is an optional argument to XlaSplitND.

func XlaSplitNDPaddings 

func XlaSplitNDPaddings(value []int64) XlaSplitNDAttr

XlaSplitNDPaddings sets the optional paddings attribute to value.

value: Optional list of right paddings per dimension of input tensor to apply before splitting. This can be used to make a dimension evenly divisible. If not specified, defaults to {}

type XlaSpmdFullToShardShapeAttr 

type XlaSpmdFullToShardShapeAttr func(optionalAttr)

XlaSpmdFullToShardShapeAttr is an optional argument to XlaSpmdFullToShardShape.

func XlaSpmdFullToShardShapeDim 

func XlaSpmdFullToShardShapeDim(value int64) XlaSpmdFullToShardShapeAttr

XlaSpmdFullToShardShapeDim sets the optional dim attribute to value. If not specified, defaults to -1

func XlaSpmdFullToShardShapeUnspecifiedDims 

func XlaSpmdFullToShardShapeUnspecifiedDims(value []int64) XlaSpmdFullToShardShapeAttr

XlaSpmdFullToShardShapeUnspecifiedDims sets the optional unspecified_dims attribute to value. If not specified, defaults to {}

type XlaSpmdShardToFullShapeAttr 

type XlaSpmdShardToFullShapeAttr func(optionalAttr)

XlaSpmdShardToFullShapeAttr is an optional argument to XlaSpmdShardToFullShape.

func XlaSpmdShardToFullShapeDim 

func XlaSpmdShardToFullShapeDim(value int64) XlaSpmdShardToFullShapeAttr

XlaSpmdShardToFullShapeDim sets the optional dim attribute to value. If not specified, defaults to -1

func XlaSpmdShardToFullShapeUnspecifiedDims 

func XlaSpmdShardToFullShapeUnspecifiedDims(value []int64) XlaSpmdShardToFullShapeAttr

XlaSpmdShardToFullShapeUnspecifiedDims sets the optional unspecified_dims attribute to value. If not specified, defaults to {}

type ZipDatasetAttr 

type ZipDatasetAttr func(optionalAttr)

ZipDatasetAttr is an optional argument to ZipDataset.

func ZipDatasetMetadata 

func ZipDatasetMetadata(value string) ZipDatasetAttr

ZipDatasetMetadata sets the optional metadata attribute to value. If not specified, defaults to ""

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.