op

package
v1.15.0 Latest Latest
Warning

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

Go to latest
Published: Oct 14, 2019 License: Apache-2.0 Imports: 3 Imported by: 0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.5.0

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 added in v1.1.0

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

Computes acos of x element-wise.

func Acosh added in v1.3.0

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 added in v1.1.0

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)

func AddManySparseToTensorsMap added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 interpretted 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 added in v1.1.0

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 interpretted 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 added in v1.1.0

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 added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 added in v1.12.1

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 added in v1.4.0

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.angle(input) ==> [2.0132, 1.056] ```

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

func AnonymousIterator added in v1.11.0

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 added in v1.12.1

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

func Any added in v1.1.0

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 ApproximateEqual added in v1.1.0

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 added in v1.1.0

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: 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 added in v1.1.0

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 added in v1.1.0

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.

func Asin added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 AssignAddVariableOp added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

func AssignVariableOp(scope *Scope, resource tf.Output, value tf.Output) (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 Atan added in v1.1.0

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 added in v1.2.0

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 \(r = \sqrt(x^2 + y^2) \).

func Atanh added in v1.3.0

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 added in v1.2.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

func AutoShardDataset(scope *Scope, input_dataset tf.Output, num_workers tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 added in v1.1.0

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.

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.9.0

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 added in v1.2.0

func BatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.11.0

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 added in v1.1.0

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 added in v1.12.1

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 BatchNormWithGlobalNormalization added in v1.1.0

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 added in v1.1.0

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 4D backprop tensor for input.1D backprop tensor for mean.1D backprop tensor for variance.1D backprop tensor for beta.1D backprop tensor for gamma.

func BatchToSpace added in v1.1.0

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 added in v1.1.0

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]]
  1. 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]]

  1. 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]]

  1. 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 BesselI0e added in v1.11.0

func BesselI0e(scope *Scope, x tf.Output) (y tf.Output)

Computes the Bessel i0e function of `x` element-wise.

Exponentially scaled modified Bessel function of order 0 defined as `bessel_i0e(x) = exp(-abs(x)) bessel_i0(x)`.

This function is faster and numerically stabler than `bessel_i0(x)`.

func BesselI1e added in v1.11.0

func BesselI1e(scope *Scope, x tf.Output) (y tf.Output)

Computes the Bessel i1e function of `x` element-wise.

Exponentially scaled modified Bessel function of order 0 defined as `bessel_i1e(x) = exp(-abs(x)) bessel_i1(x)`.

This function is faster and numerically stabler than `bessel_i1(x)`.

func Betainc added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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: ```python >>> a = [1., 2., 3.] >>> equality_bitcast = tf.bitcast(a,tf.complex128) tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot bitcast from float to complex128: shape [3] [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: ```python >>> 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: ```python >>> 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.

func BitwiseAnd added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.15.0

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 The input gate over the whole time sequence.The cell state before the tanh over the whole time sequence.The forget gate over the whole time sequence.The output gate over the whole time sequence.The cell input over the whole time sequence.The cell after the tanh over the whole time sequence.The output h vector over the whole time sequence.

func BlockLSTMGrad added in v1.15.0

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 The gradient of x to be back-propped.The gradient of cs_prev to be back-propped.The gradient of h_prev to be back-propped.The gradient for w to be back-propped.The gradient for wci to be back-propped.The gradient for wcf to be back-propped.The gradient for wco to be back-propped.The gradient for w to be back-propped.

func BoostedTreesAggregateStats added in v1.12.1

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 added in v1.12.0

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 added in v1.12.1

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: mininum 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 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.A Rank 1 tensors indicating the best gains for each feature to split for certain nodes. See above for details like shapes and sizes.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.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.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.A Rank 2 tensors, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.A Rank 1 tensors indicating the which direction to go if data is missing. See above for details like shapes and sizes.

func BoostedTreesCalculateBestGainsPerFeature added in v1.9.0

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: mininum 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 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.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.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.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.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 added in v1.11.0

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 added in v1.9.0

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 added in v1.12.0

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 added in v1.9.0

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 added in v1.9.0

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

Creates a handle to a BoostedTreesEnsembleResource

func BoostedTreesExampleDebugOutputs added in v1.11.0

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 BoostedTreesGetEnsembleStates added in v1.9.0

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

func BoostedTreesMakeQuantileSummaries added in v1.12.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.12.0

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 added in v1.12.1

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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

Creates a handle to a BoostedTreesQuantileStreamResource.

func BoostedTreesSerializeEnsemble added in v1.9.0

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 of the tree ensemble resource.Serialized proto of the ensemble.

func BoostedTreesSparseAggregateStats added in v1.15.0

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 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.output Rank 1 Tensor (shape=[number of non zero statistics])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 added in v1.15.0

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: mininum 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 A Rank 1 tensor indicating possible node ids that can be split.A Rank 1 tensor indicating the best gains to split each node.A Rank 1 tensor indicating the best feature dimension for each feature to split for each node.A Rank 1 tensor indicating the bucket id to compare with (as a threshold) for split in each node.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.A Rank 2 tensor, with the same shape/conditions as left_node_contribs_list, but just that the value is for the right node.A Rank 1 tensor indicating which direction to go if data is missing.

func BoostedTreesTrainingPredict added in v1.9.0

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 Rank 2 Tensor containing logits update (with respect to cached values stored) for each example.Rank 1 Tensor containing new tree ids for each example.Rank 1 Tensor containing new node ids in the new tree_ids.

func BoostedTreesUpdateEnsemble added in v1.9.0

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 BroadcastArgs added in v1.1.0

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 added in v1.1.0

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 added in v1.11.0

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. When trying to broadcast a Tensor to a shape, it starts with the trailing dimensions, and works its way forward.

For example,

```python >>> x = tf.constant([1, 2, 3]) >>> y = tf.broadcast_to(x, [3, 3]) >>> sess.run(y) array([[1, 2, 3],

[1, 2, 3],
[1, 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 `[3, 3]`.

Arguments:

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

Returns A Tensor.

func Bucketize added in v1.2.0

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 added in v1.5.0

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 CTCBeamSearchDecoder added in v1.1.0

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 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].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.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]]`.A matrix, shaped: `(batch_size x top_paths)`. The sequence log-probabilities.

func CTCGreedyDecoder added in v1.1.0

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 Indices matrix, size `(total_decoded_outputs x 2)`, of a `SparseTensor<int64, 2>`. The rows store: [batch, time].Values vector, size: `(total_decoded_outputs)`, of a `SparseTensor<int64, 2>`. The vector stores the decoded classes.Shape vector, size `(2)`, of the decoded SparseTensor. Values are: `[batch_size, max_decoded_length]`.Matrix, size `(batch_size x 1)`, containing sequence log-probabilities.

func CTCLoss added in v1.1.0

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

func CacheDataset added in v1.3.0

func CacheDataset(scope *Scope, input_dataset tf.Output, filename tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 added in v1.1.0

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

Returns element-wise smallest integer not less than x.

func CheckNumerics added in v1.1.0

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, passes `tensor` as-is.

Arguments:

message: Prefix of the error message.

func Cholesky added in v1.1.0

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 added in v1.1.0

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 added in v1.9.0

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 CollectiveBcastRecv added in v1.9.0

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

Receives a tensor value broadcast from another device.

func CollectiveBcastSend added in v1.9.0

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

Broadcasts a tensor value to one or more other devices.

func CollectiveGather added in v1.12.1

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

Mutually accumulates multiple tensors of identical type and shape.

func CollectivePermute added in v1.12.1

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 added in v1.9.0

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 CombinedNonMaxSuppression added in v1.12.1

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: A scalar representing maximum number of boxes retained over all classes.
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 [batch_size, max_detections, 4] float32 tensor containing the non-max suppressed boxes.A [batch_size, max_detections] float32 tensor containing the scores for the boxes.A [batch_size, max_detections] float32 tensor containing the classes for the boxes.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 CompareAndBitpack added in v1.4.0

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

Compare values of `input` to `threshold` and pack resulting bits into a `uint8`.

Each comparison returns a boolean `true` (if `input_value > threshold`) or and `false` otherwise.

This operation is useful for Locality-Sensitive-Hashing (LSH) and other algorithms that use hashing approximations of cosine and `L2` distances; codes can be generated from an input via:

```python codebook_size = 50 codebook_bits = codebook_size * 32 codebook = tf.get_variable('codebook', [x.shape[-1].value, codebook_bits],

dtype=x.dtype,
initializer=tf.orthogonal_initializer())

codes = compare_and_threshold(tf.matmul(x, codebook), threshold=0.) codes = tf.bitcast(codes, tf.int32) # go from uint8 to int32 # now codes has shape x.shape[:-1] + [codebook_size] ```

**NOTE**: Currently, the innermost dimension of the tensor must be divisible by 8.

Given an `input` shaped `[s0, s1, ..., s_n]`, the output is a `uint8` tensor shaped `[s0, s1, ..., s_n / 8]`.

Arguments:

input: Values to compare against `threshold` and bitpack.
threshold: Threshold to compare against.

Returns The bitpacked comparisons.

func Complex added in v1.1.0

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 added in v1.1.0

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}\\).

func ComputeAccidentalHits added in v1.1.0

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 A vector of indices corresponding to rows of true_candidates.A vector of IDs of positions in sampled_candidates that match a true_label for the row with the corresponding index in indices.A vector of the same length as indices and ids, in which each element is -FLOAT_MAX.

func Concat added in v1.1.0

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 added in v1.1.0

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' is [2, 2, 7] # 'y' is [2, 3, 7] # 'z' is [2, 5, 7] concat_offset(2, [x, y, z]) => [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 vectors representing shape of tensors being concatenated.

Returns The `N` int32 vectors representing the starting offset of input tensors within the concatenated output.

func ConcatV2 added in v1.1.0

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 added in v1.3.0

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

Creates a dataset that concatenates `input_dataset` with `another_dataset`.

func ConfigureDistributedTPU added in v1.12.1

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 Conj added in v1.1.0

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 added in v1.5.0

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 Const

func Const(scope *Scope, value interface{}) (output tf.Output)

Const adds an operation to graph that produces value as output.

func ConsumeMutexLock added in v1.9.0

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 added in v1.1.0

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 Conv2D added in v1.1.0

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 added in v1.1.0

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 Conv2DBackpropInput added in v1.1.0

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 Conv3D added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 Cos added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

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 added in v1.9.0

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 inferenece 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 added in v1.9.0

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 added in v1.11.0

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 added in v1.12.1

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 added in v1.9.0

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 added in v1.15.0

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_weigths: 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 added in v1.9.0

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 added in v1.9.0

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 added in v1.15.0

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_weigths: 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 added in v1.11.0

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 inferenece 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 added in v1.12.1

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 inferenece 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 added in v1.1.0

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 added in v1.1.0

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 DataFormatDimMap added in v1.6.0

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 added in v1.6.0

func DataFormatVecPermute(scope *Scope, x tf.Output, optional ...DataFormatVecPermuteAttr) (y tf.Output)

Returns the permuted vector/tensor in the destination data format given the

one in the source data format.

Arguments:

x: Vector of size 4 or Tensor of shape (4, 2) in source data format.

Returns Vector of size 4 or Tensor of shape (4, 2) in destination data format.

func DatasetCardinality added in v1.15.0

func DatasetCardinality(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 DatasetToGraph added in v1.11.0

func DatasetToGraph(scope *Scope, input_dataset tf.Output) (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 added in v1.5.0

func DatasetToSingleElement(scope *Scope, dataset tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.11.0

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 added in v1.4.0

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 DecodeAndCropJpeg added in v1.4.0

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 added in v1.1.0

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 for padding. Web-safe means that input must use - and _ instead of + and /.

Arguments:

input: Base64 strings to decode.

Returns Decoded strings.

func DecodeBmp added in v1.3.0

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 added in v1.1.0

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 added in v1.6.0

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 added in v1.1.0

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.image.decode_image`.

Arguments:

contents: 0-D.  The GIF-encoded image.

Returns 4-D with shape `[num_frames, height, width, 3]`. RGB channel order.

func DecodeJSONExample added in v1.1.0

func DecodeJSONExample(scope *Scope, json_examples tf.Output) (binary_examples tf.Output)

Convert JSON-encoded Example records to binary protocol buffer strings.

This op translates a tensor containing Example records, encoded using the [standard JSON mapping](https://developers.google.com/protocol-buffers/docs/proto3#json), into a tensor containing the same records encoded as binary protocol buffers. The resulting tensor can then be fed to any of the other Example-parsing ops.

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 added in v1.1.0

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.image.decode_image`.

Arguments:

contents: 0-D.  The JPEG-encoded image.

Returns 3-D with shape `[height, width, channels]`..

func DecodePaddedRaw added in v1.14.0

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 added in v1.1.0

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.image.decode_image`.

Arguments:

contents: 0-D.  The PNG-encoded image.

Returns 3-D with shape `[height, width, channels]`.

func DecodeProtoV2 added in v1.9.0

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.

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.

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 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.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 added in v1.1.0

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 added in v1.2.0

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 2-D with shape `[length, channels]`.Scalar holding the sample rate found in the WAV header.

func DeepCopy added in v1.9.0

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 added in v1.12.1

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 DeleteSessionTensor added in v1.1.0

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 DenseToDenseSetOperation added in v1.1.0

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 2D indices of a `SparseTensor`.1D values of a `SparseTensor`.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 added in v1.2.0

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 added in v1.1.0

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 2D indices of a `SparseTensor`.1D values of a `SparseTensor`.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 added in v1.1.0

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 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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 Tensor.

[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.

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) ```

*SCALED mode Example*

`SCALED` mode matches the quantization approach used in `QuantizeAndDequantize{V2|V3}`.

If the mode is `SCALED`, we do not use the full range of the output type, choosing to elide the lowest possible value for symmetry (e.g., output range is -127 to 127, not -128 to 127 for signed 8 bit quantization), so that 0.0 maps to 0.

We first find the range of values in our tensor. The range we use is always centered on 0, so we find m such that ```c++

m = max(abs(input_min), abs(input_max))

```

Our input tensor range is then `[-m, m]`.

Next, we choose our fixed-point quantization buckets, `[min_fixed, max_fixed]`. If T is signed, this is ```

num_bits = sizeof(T) * 8
[min_fixed, max_fixed] =
    [-(1 << (num_bits - 1) - 1), (1 << (num_bits - 1)) - 1]

```

Otherwise, if T is unsigned, the fixed-point range is ```

[min_fixed, max_fixed] = [0, (1 << num_bits) - 1]

```

From this we compute our scaling factor, s: ```c++

s = (2 * m) / (max_fixed - min_fixed)

```

Now we can dequantize the elements of our tensor: ```c++ result = input * s ```

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.5.0

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 added in v1.1.0

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 Diag added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

func DirectedInterleaveDataset(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 Div added in v1.1.0

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.12.1

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 DynamicPartition added in v1.1.0

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>

Arguments:

partitions: Any shape.  Indices in the range `[0, num_partitions)`.
num_partitions: The number of partitions to output.

func DynamicStitch added in v1.1.0

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)] + 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 added in v1.6.0

func EagerPyFunc(scope *Scope, input []tf.Output, token string, Tout []tf.DataType) (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 added in v1.1.0

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 Einsum added in v1.15.0

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 subcripts 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 subcripts 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 subcripts 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 added in v1.1.0

func Elu(scope *Scope, features tf.Output) (activations tf.Output)

Computes exponential linear: `exp(features) - 1` if < 0, `features` otherwise.

See [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) ](http://arxiv.org/abs/1511.07289)

func EluGrad added in v1.1.0

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 added in v1.9.0

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 added in v1.9.0

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 EncodeBase64 added in v1.1.0

func EncodeBase64(scope *Scope, input tf.Output, optional ...EncodeBase64Attr) (output tf.Output)

Encode strings into web-safe base64 format.

Refer to the following article for more information on base64 format: en.wikipedia.org/wiki/Base64. 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 added in v1.1.0

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.9.0

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 added in v1.2.0

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 EnqueueTPUEmbeddingIntegerBatch added in v1.12.1

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 EnqueueTPUEmbeddingSparseBatch added in v1.12.1

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 added in v1.12.1

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.1.0

func Equal(scope *Scope, x tf.Output, y tf.Output) (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 added in v1.1.0

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

Computes the Gauss error function of `x` element-wise.

func Erfc added in v1.1.0

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

Computes the complementary error function of `x` element-wise.

func EuclideanNorm added in v1.12.1

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 Exit added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

func ExperimentalAutoShardDataset(scope *Scope, input_dataset tf.Output, num_workers tf.Output, index tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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

Creates a dataset that contains the elements of `input_dataset` ignoring errors.

func ExperimentalIteratorGetDevice added in v1.12.1

func ExperimentalIteratorGetDevice(scope *Scope, resource tf.Output) (device tf.Output)

Returns the name of the device on which `resource` has been placed.

func ExperimentalLatencyStatsDataset added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

func ExperimentalRebatchDataset(scope *Scope, input_dataset tf.Output, num_workers 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 changes the batch size of the dataset to current batch size // num_workers.

Arguments:

input_dataset: A variant tensor representing the input dataset.
num_workers: A scalar representing the number of workers 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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

func ExperimentalStatsAggregatorHandle(scope *Scope, optional ...ExperimentalStatsAggregatorHandleAttr) (handle tf.Output)

Creates a statistics manager resource.

func ExperimentalStatsAggregatorSummary added in v1.12.1

func ExperimentalStatsAggregatorSummary(scope *Scope, iterator tf.Output) (summary tf.Output)

Produces a summary of any statistics recorded by the given statistics manager.

func ExperimentalThreadPoolDataset added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 ExtractImagePatches added in v1.1.0

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 added in v1.4.0

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 added in v1.12.0

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.

We specify the size-related attributes as:

```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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 complex64 tensor.

Returns A complex64 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 FIFOQueueV2 added in v1.1.0

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 added in v1.1.0

func Fact(scope *Scope) (fact tf.Output)

Output a fact about factorials.

func FakeParam added in v1.11.0

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 added in v1.1.0

func FakeQuantWithMinMaxArgs(scope *Scope, inputs tf.Output, optional ...FakeQuantWithMinMaxArgsAttr) (outputs tf.Output)

Fake-quantize the 'inputs' tensor, type float to 'outputs' tensor of same type.

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`.

Quantization is called fake since the output is still in floating point.

func FakeQuantWithMinMaxArgsGradient added in v1.1.0

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)`.

func FakeQuantWithMinMaxVars added in v1.1.0

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 `min`

and `max` to 'outputs' tensor of same shape as `inputs`.

`[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 FakeQuantWithMinMaxVarsGradient added in v1.1.0

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 Backpropagated gradients w.r.t. inputs: `gradients * (inputs >= min && inputs <= max)`.Backpropagated gradients w.r.t. min parameter: `sum(gradients * (inputs < min))`.Backpropagated gradients w.r.t. max parameter: `sum(gradients * (inputs > max))`.

func FakeQuantWithMinMaxVarsPerChannel added in v1.1.0

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 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`.

`[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 added in v1.1.0

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 Backpropagated gradients w.r.t. inputs, shape same as `inputs`:

`gradients * (inputs >= min && inputs <= max)`.Backpropagated gradients w.r.t. min parameter, shape `[d]`:

`sum_per_d(gradients * (inputs < min))`.Backpropagated gradients w.r.t. max parameter, shape `[d]`: `sum_per_d(gradients * (inputs > max))`.

func Fill added in v1.1.0

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 added in v1.12.0

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 Fingerprint added in v1.14.0

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 added in v1.2.0

func FixedLengthRecordDataset(scope *Scope, filenames tf.Output, header_bytes tf.Output, record_bytes tf.Output, footer_bytes tf.Output, buffer_size tf.Output) (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 added in v1.1.0

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 added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 added in v1.1.0

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

Returns element-wise largest integer not greater than x.

func FloorDiv added in v1.1.0

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 added in v1.1.0

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

Returns element-wise remainder of division. When `x < 0` xor `y < 0` is

true, this follows Python semantics in that the result here is consistent with a flooring divide. E.g. `floor(x / y) * y + mod(x, y) = x`.

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

func FractionalAvgPool added in v1.1.0

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 tensor after fractional avg pooling.row pooling sequence, needed to calculate gradient.column pooling sequence, needed to calculate gradient.

func FractionalAvgPoolGrad added in v1.1.0

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 added in v1.1.0

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 tensor after fractional max pooling.row pooling sequence, needed to calculate gradient.column pooling sequence, needed to calculate gradient.

func FractionalMaxPoolGrad added in v1.1.0

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 added in v1.1.0

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 A 4D Tensor for output data.A 1D Tensor for the computed batch mean, to be used by TensorFlow to compute the running mean.A 1D Tensor for the computed batch variance, to be used by TensorFlow to compute the running variance.A 1D Tensor for the computed batch mean, to be reused in the gradient computation.A 1D Tensor for the computed batch variance (inverted variance in the cuDNN case), to be reused in the gradient computation.

func FusedBatchNormGrad added in v1.1.0

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 A 4D Tensor for the gradient with respect to x.A 1D Tensor for the gradient with respect to scale.A 1D Tensor for the gradient with respect to offset.Unused placeholder to match the mean input in FusedBatchNorm.Unused placeholder to match the variance input in FusedBatchNorm.

func FusedBatchNormGradV2 added in v1.4.0

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 A 4D Tensor for the gradient with respect to x.A 1D Tensor for the gradient with respect to scale.A 1D Tensor for the gradient with respect to offset.Unused placeholder to match the mean input in FusedBatchNorm.Unused placeholder to match the variance input in FusedBatchNorm.

func FusedBatchNormGradV3 added in v1.14.0

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 A 4D Tensor for the gradient with respect to x.A 1D Tensor for the gradient with respect to scale.A 1D Tensor for the gradient with respect to offset.Unused placeholder to match the mean input in FusedBatchNorm.Unused placeholder to match the variance input in FusedBatchNorm.

func FusedBatchNormV2 added in v1.4.0

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 A 4D Tensor for output data.A 1D Tensor for the computed batch mean, to be used by TensorFlow to compute the running mean.A 1D Tensor for the computed batch variance, to be used by TensorFlow to compute the running variance.A 1D Tensor for the computed batch mean, to be reused in the gradient computation.A 1D Tensor for the computed batch variance (inverted variance in the cuDNN case), to be reused in the gradient computation.

func FusedBatchNormV3 added in v1.14.0

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 A 4D Tensor for output data.A 1D Tensor for the computed batch mean, to be used by TensorFlow to compute the running mean.A 1D Tensor for the computed batch variance, to be used by TensorFlow to compute the running variance.A 1D Tensor for the computed batch mean, to be reused in the gradient computation.A 1D Tensor for the computed batch variance (inverted variance in the cuDNN case), to be reused in the gradient computation.A 1D Tensor for some intermediate results, to be reused in the gradient computation for better efficiency.

func FusedPadConv2D added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.1.0

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 added in v1.1.0

func GatherNd(scope *Scope, params tf.Output, indices tf.Output) (output tf.Output)

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

`indices` is an 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 first 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]:]

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.

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 added in v1.3.0

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 + 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.

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 GenerateVocabRemapping added in v1.4.0

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 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.Number of new vocab entries found in old vocab.

func GetSessionHandle added in v1.2.0

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 added in v1.1.0

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 added in v1.1.0

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 Gradients added in v1.12.1

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 added in v1.1.0

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)

func GreaterEqual added in v1.1.0

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)

func GuaranteeConst added in v1.6.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 complex64 tensor.

Returns A complex64 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 IRFFT added in v1.1.0

func IRFFT(scope *Scope, input tf.Output, fft_length tf.Output) (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 complex64 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 added in v1.1.0

func IRFFT2D(scope *Scope, input tf.Output, fft_length tf.Output) (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 complex64 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 added in v1.1.0

func IRFFT3D(scope *Scope, input tf.Output, fft_length tf.Output) (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 complex64 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 Identity added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.11.0

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

Computes the gradient of `igamma(a, x)` wrt `a`.

func Igammac added in v1.1.0

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 Gama function.

Note, above `P(a, x)` (`Igamma`) is the lower regularized complete Gamma function.

func IgnoreErrorsDataset added in v1.3.0

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

Creates a dataset that contains the elements of `input_dataset` ignoring errors.

func Imag added in v1.1.0

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 ImageSummary added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

func InplaceUpdate(scope *Scope, x tf.Output, i tf.Output, v tf.Output) (y tf.Output)
Updates specified rows with values in `v`.

Computes `x[i, :] = v; return 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 added in v1.1.0

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

Computes the reciprocal of x element-wise.

I.e., \\(y = 1 / x\\).

func InvGrad added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.9.0

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 resouce.

Returns output boolean on whether it is initialized or not.

func IsBoostedTreesQuantileStreamResourceInitialized added in v1.12.0

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 added in v1.1.0

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

func IsInf added in v1.1.0

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

func IsNan added in v1.1.0

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

func Iterator added in v1.2.0

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 added in v1.3.0

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 added in v1.15.0

func IteratorGetDevice(scope *Scope, resource tf.Output) (device tf.Output)

Returns the name of the device on which `resource` has been placed.

func IteratorGetNext added in v1.2.0

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 added in v1.11.0

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 added in v1.9.0

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 added in v1.3.0

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 added in v1.12.1

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 added in v1.12.1

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 L2Loss added in v1.1.0

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 LRN added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

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 The input gate.The cell state before the tanh.The forget gate.The output gate.The cell input.The cell after the tanh.The output h vector.

func LSTMBlockCellGrad added in v1.15.0

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 The gradient of cs to be back-propped.The derivative wrt to [i, cs, f, o].The gradient for wci to be back-propped.The gradient for wcf to be back-propped.The gradient for wco to be back-propped.

func LatencyStatsDataset added in v1.5.0

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 added in v1.12.1

func LeakyRelu(scope *Scope, features tf.Output, optional ...LeakyReluAttr) (activations tf.Output)

Computes rectified linear: `max(features, features * alpha)`.

func LeakyReluGrad added in v1.12.1

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 * (featurs <= 0)`.

func LearnedUnigramCandidateSampler added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 added in v1.5.0

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: id=139, shape=(4,), dtype=int8, numpy=array([ -2, 64, 101, 32], dtype=int8)> ```

func Less added in v1.1.0

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)

func LessEqual added in v1.1.0

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)

func Lgamma added in v1.1.0

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

Computes the log of the absolute value of `Gamma(x)` element-wise.

func LinSpace added in v1.1.0

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 ListDiff added in v1.1.0

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 1-D. Values present in `x` but not in `y`.1-D. Positions of `x` values preserved in `out`.

func LoadAndRemapMatrix added in v1.4.0

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 added in v1.12.1

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 LoadTPUEmbeddingADAMParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingADAMParametersGradAccumDebug(scope *Scope, parameters tf.Output, momenta tf.Output, velocities tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingADAMParametersGradAccumDebugAttr) (o *tf.Operation)

Load ADAM embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the ADAM optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingAdadeltaParameters added in v1.12.1

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 LoadTPUEmbeddingAdadeltaParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingAdadeltaParametersGradAccumDebug(scope *Scope, parameters tf.Output, accumulators tf.Output, updates tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr) (o *tf.Operation)

Load Adadelta parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the Adadelta optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingAdagradParameters added in v1.12.1

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 LoadTPUEmbeddingAdagradParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingAdagradParametersGradAccumDebug(scope *Scope, parameters tf.Output, accumulators tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr) (o *tf.Operation)

Load Adagrad embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the Adagrad optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingCenteredRMSPropParameters added in v1.12.1

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 added in v1.12.1

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 LoadTPUEmbeddingFTRLParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingFTRLParametersGradAccumDebug(scope *Scope, parameters tf.Output, accumulators tf.Output, linears tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr) (o *tf.Operation)

Load FTRL embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the FTRL optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingMDLAdagradLightParameters added in v1.12.1

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 added in v1.12.1

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 LoadTPUEmbeddingMomentumParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingMomentumParametersGradAccumDebug(scope *Scope, parameters tf.Output, momenta tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr) (o *tf.Operation)

Load Momentum embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the Momentum optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingProximalAdagradParameters added in v1.12.1

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 LoadTPUEmbeddingProximalAdagradParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingProximalAdagradParametersGradAccumDebug(scope *Scope, parameters tf.Output, accumulators tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr) (o *tf.Operation)

Load proximal Adagrad embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the proximal Adagrad optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingRMSPropParameters added in v1.12.1

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 LoadTPUEmbeddingRMSPropParametersGradAccumDebug added in v1.12.1

func LoadTPUEmbeddingRMSPropParametersGradAccumDebug(scope *Scope, parameters tf.Output, ms tf.Output, mom tf.Output, gradient_accumulators tf.Output, num_shards int64, shard_id int64, optional ...LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr) (o *tf.Operation)

Load RMSProp embedding parameters with debug support.

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.
gradient_accumulators: Value of gradient_accumulators used in the RMSProp optimization algorithm.

Returns the created operation.

func LoadTPUEmbeddingStochasticGradientDescentParameters added in v1.12.1

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 added in v1.1.0

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

Computes natural logarithm of x element-wise.

I.e., \\(y = \log_e x\\).

func Log1p added in v1.1.0

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)\\).

func LogMatrixDeterminant added in v1.4.0

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 the 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 The signs of the log determinants of the inputs. Shape is `[N]`.The logs of the absolute values of the determinants of the N input matrices. Shape is `[N]`.

func LogSoftmax added in v1.1.0

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 added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 added in v1.1.0

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 added in v1.1.0

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

Returns the truth value of NOT x element-wise.

func LogicalOr added in v1.1.0

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 added in v1.2.0

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 Vector of all keys present in the table.Tensor of all values in the table. Indexed in parallel with `keys`.

func LookupTableFindV2 added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.12.1

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 added in v1.2.0

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 added in v1.1.0

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 added in v1.12.0

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 added in v1.12.1

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 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`.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 added in v1.2.0

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 MapClear added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

func MapSize(scope *Scope, dtypes []tf.DataType, optional ...MapSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func MapStage added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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.matrix_band_part(input, 1, -1) ==> [[ 0, 1, 2, 3]

[-1,  0,  1, 2]
[ 0, -1,  0, 1]
[ 0,  0, -1, 0]],

tf.matrix_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.matrix_band_part(input, 0, -1) ==> Upper triangular part.
tf.matrix_band_part(input, -1, 0) ==> Lower triangular part.
tf.matrix_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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

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] ; when 0 <= n-y < M and 0 <= n-x < N,
  0                             ; 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] ; when 0 <= n-y < M and 0 <= n-x < N,
  0                             ; 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 = 9 tf.matrix_diag_part(input, k = (1, 3), padding = 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 MatrixDiagV2 added in v1.15.0

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
  output[i, j, ..., l, m, n]                ; 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, k[1]-d, n-max(d, 0)] ; if d_lower <= d <= d_upper
  input[i, j, ..., l, m, n]                   ; otherwise

``` where `d = n - m`

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 = 9. tf.matrix_diag(diagonal, k = -1, num_rows = 3, padding = 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 added in v1.5.0

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 added in v1.1.0

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 added in v1.9.0

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 added in v1.1.0

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 added in v1.15.0

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]
  output[i, j, ..., l, m, n]             ; otherwise

```

Otherwise,

``` output[i, j, ..., l, m, n]

= diagonal[i, j, ..., l, k[1]-d, n-max(d, 0)] ; if d_lower <= d <= d_upper
  input[i, j, ..., l, m, n]                   ; otherwise

``` where `d = n - m`

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_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_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_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 MatrixSolve added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

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 added in v1.1.0

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, K]`.

The output is a tensor of shape `[..., M, K]`. 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]`.

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: id=257, 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: id=263, 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 added in v1.1.0

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 added in v1.15.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 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 added in v1.2.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.4.0

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 added in v1.2.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 The max pooled output tensor.4-D. The flattened indices of the max values chosen for each output.

func Maximum added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 Will be set to the available input tensor.The index of the chosen input tensor in `inputs`.

func MergeSummary added in v1.1.0

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 added in v1.1.0

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.

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 added in v1.2.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 Mod added in v1.1.0

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.12.1

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 Matrix of shape (n, min(m, k)). Each row contains the indices of the centers closest to the corresponding point, ordered by increasing distance.Matrix of shape (n, min(m, k)). Each row contains the squared L2 distance to the corresponding center in nearest_center_indices.

func Neg added in v1.1.0

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

Computes numerical negative value element-wise.

I.e., \\(y = -x\\).

func NextAfter added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

func NoOp(scope *Scope) (o *tf.Operation)

Does nothing. Only useful as a placeholder for control edges.

Returns the created operation.

func NonDeterministicInts added in v1.12.1

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.11.0

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 added in v1.11.0

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 A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.A 0-D integer tensor representing the number of valid elements in `selected_indices`, with the valid elements appearing first.

func NonMaxSuppressionV5 added in v1.15.0

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 A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.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`)A 0-D integer tensor representing the number of valid elements in `selected_indices`, with the valid elements appearing first.

func NonMaxSuppressionWithOverlaps added in v1.11.0

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 added in v1.1.0

func NotEqual(scope *Scope, x tf.Output, y tf.Output) (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 added in v1.5.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.11.0

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 OptionalFromValue added in v1.11.0

func OptionalFromValue(scope *Scope, components []tf.Output) (optional tf.Output)

Constructs an Optional variant from a tuple of tensors.

func OptionalGetValue added in v1.11.0

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 added in v1.11.0

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 added in v1.11.0

func OptionalNone(scope *Scope) (optional tf.Output)

Creates an Optional variant with no value.

func OrderedMapClear added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

func OrderedMapSize(scope *Scope, dtypes []tf.DataType, optional ...OrderedMapSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func OrderedMapStage added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.12.1

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 added in v1.12.1

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 OutfeedEnqueue added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.2.0

func PaddedBatchDataset(scope *Scope, input_dataset tf.Output, batch_size tf.Output, padded_shapes []tf.Output, padding_values []tf.Output, output_shapes []tf.Shape) (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 added in v1.11.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.0

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 ParseSequenceExample added in v1.12.0

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 ParseSingleExample added in v1.6.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.0

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 added in v1.1.0

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 added in v1.15.0

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 added in v1.1.0

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 added in v1.1.0

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[..., :,:])`

```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 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]`.Triangular factor. If `full_matrices` is `False` then shape is `[..., P, N]`. If `full_matrices` is `True` then shape is `[..., M, N]`.

func QuantizeAndDequantize added in v1.1.0

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 added in v1.1.0

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 respresentable 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 added in v1.3.0

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 QuantizeDownAndShrinkRange added in v1.1.0

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 The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizeV2 added in v1.1.0

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`, we do not use the full range of the output type, choosing to elide the lowest possible value for symmetry (e.g., output range is -127 to 127, not -128 to 127 for signed 8 bit quantization), so that 0.0 maps to 0.

We first find the range of values in our tensor. The range we use is always centered on 0, so we find m such that

```c++

m = max(abs(input_min), abs(input_max))

```

Our input tensor range is then `[-m, m]`.

Next, we choose our fixed-point quantization buckets, `[min_fixed, max_fixed]`. If T is signed, this is

```

num_bits = sizeof(T) * 8
[min_fixed, max_fixed] =
    [-(1 << (num_bits - 1) - 1), (1 << (num_bits - 1)) - 1]

```

Otherwise, if T is unsigned, the fixed-point range is

```

[min_fixed, max_fixed] = [0, (1 << num_bits) - 1]

```

From this we compute our scaling factor, s:

```c++

s = (max_fixed - min_fixed) / (2 * m)

```

Now we can quantize the elements of our tensor:

```c++ result = round(input * s) ```

One thing to watch out for is that the operator may choose to adjust the requested minimum and maximum values slightly during the quantization process, so you should always use the output ports as the range for further calculations. For example, if the requested minimum and maximum values are close to equal, they will be separated by a small epsilon value to prevent ill-formed quantized buffers from being created. Otherwise, you can end up with buffers where all the quantized values map to the same float value, which causes problems for operations that have to perform further calculations on them.

Arguments:

min_range: The minimum scalar value possibly produced for the input.
max_range: The maximum scalar value possibly produced for the input.

Returns The quantized data produced from the float input.The actual minimum scalar value used for the output.The actual maximum scalar value used for the output.

func QuantizedAdd added in v1.3.0

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 The float value that the lowest quantized output value represents.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 added in v1.1.0

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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedBatchNormWithGlobalNormalization added in v1.1.0

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 added in v1.1.0

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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedConcat added in v1.1.0

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 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.The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizedConv2D added in v1.1.0

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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedConv2DPerChannel added in v1.12.1

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 The output tensor.The minimum value of the final output tensor.The maximum value of the final output tensor.

func QuantizedDepthwiseConv2D added in v1.12.1

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 The output tensor.The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBias added in v1.12.1

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 The output tensor.The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBiasAndRelu added in v1.12.1

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 The output tensor.The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize added in v1.12.1

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 The output tensor.The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.

func QuantizedInstanceNorm added in v1.1.0

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 A 4D Tensor.The value represented by the lowest quantized output.The value represented by the highest quantized output.

func QuantizedMatMul added in v1.1.0

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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedMatMulWithBias added in v1.15.0

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 mulplication 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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedMatMulWithBiasAndRelu added in v1.15.0

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 mulplication 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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedMatMulWithBiasAndReluAndRequantize added in v1.15.0

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 mulplication 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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedMaxPool added in v1.1.0

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 The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.

func QuantizedMul added in v1.1.0

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 The float value that the lowest quantized output value represents.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 added in v1.1.0

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 Has the same output shape as "features".The float value that the lowest quantized value represents.The float value that the highest quantized value represents.

func QuantizedRelu6 added in v1.1.0

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 Has the same output shape as "features".The float value that the lowest quantized value represents.The float value that the highest quantized value represents.

func QuantizedReluX added in v1.1.0

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 Has the same output shape as "features".The float value that the lowest quantized value represents.The float value that the highest quantized value represents.

func QuantizedReshape added in v1.1.0

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 This value is copied from input_min.This value is copied from input_max.

func QuantizedResizeBilinear added in v1.3.0

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 4-D with shape `[batch, new_height, new_width, channels]`.

func QueueCloseV2 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.1.0

func RFFT(scope *Scope, input tf.Output, fft_length tf.Output) (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 added in v1.1.0

func RFFT2D(scope *Scope, input tf.Output, fft_length tf.Output) (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 added in v1.1.0

func RFFT3D(scope *Scope, input tf.Output, fft_length tf.Output) (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 RGBToHSV added in v1.1.0

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.

Arguments:

images: 1-D or higher rank. RGB data to convert. Last dimension must be size 3.

Returns `images` converted to HSV.

func RaggedGather added in v1.12.1

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 The `nested_row_splits` tensors that define the row-partitioning for the returned RaggedTensor.The `flat_values` for the returned RaggedTensor.

func RaggedRange added in v1.12.1

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) = gen_ragged_ops.ragged_range( ... starts=[2, 5, 8], limits=[3, 5, 12], deltas=1) >>> result = ragged.from_nested_row_splits(rt_dense_values, rt_nested_splits) >>> print result.eval().tolist() [[2], # result[0] = range(2, 3)

[],                # result[1] = range(5, 5)
[8, 9, 10, 11]]    # result[2] = range(8, 12)

```

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 The `row_splits` for the returned `RaggedTensor`.The `flat_values` for the returned `RaggedTensor`.

func RaggedTensorFromVariant added in v1.14.0

func RaggedTensorFromVariant(scope *Scope, encoded_ragged tf.Output, input_ragged_rank int64, output_ragged_rank int64, Tvalues tf.DataType, Tsplits tf.DataType) (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 A list of one or more Tensors representing the splits of the output `RaggedTensor`.A Tensor representing the values of the output `RaggedTensor`.

func RaggedTensorToSparse added in v1.12.1

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 The indices for the `SparseTensor`.The values of the `SparseTensor`.`sparse_dense_shape` is a tight bounding box of the input `RaggedTensor`.

func RaggedTensorToVariant added in v1.14.0

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 RandomCrop added in v1.1.0

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 added in v1.6.0

func RandomDataset(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 RandomGamma added in v1.1.0

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 added in v1.11.0

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 RandomPoisson added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

func RangeDataset(scope *Scope, start tf.Output, stop tf.Output, step tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 added in v1.1.0

func ReadFile(scope *Scope, filename tf.Output) (contents tf.Output)

Reads and outputs the entire contents of the input filename.

func ReadVariableOp added in v1.1.0

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 ReaderNumRecordsProducedV2 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 A 1-D tensor.A 1-D tensor.

func ReaderReadV2 added in v1.1.0

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 A scalar.A scalar.

func ReaderResetV2 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

func RebatchDataset(scope *Scope, input_dataset tf.Output, num_workers 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 changes the batch size of the dataset to current batch size // num_workers.

Arguments:

input_dataset: A variant tensor representing the input dataset.
num_workers: A scalar representing the number of workers to distribute this batch across. As

a result of this transformation the current batch size would end up being divided by this parameter.

func Reciprocal added in v1.1.0

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

Computes the reciprocal of x element-wise.

I.e., \\(y = 1 / x\\).

func ReciprocalGrad added in v1.1.0

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 added in v1.1.0

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 RecvTPUEmbeddingActivations added in v1.12.1

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 added in v1.1.0

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 added in v1.11.0

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)

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 added in v1.9.0

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 Relu added in v1.1.0

func Relu(scope *Scope, features tf.Output) (activations tf.Output)

Computes rectified linear: `max(features, 0)`.

func Relu6 added in v1.1.0

func Relu6(scope *Scope, features tf.Output) (activations tf.Output)

Computes rectified linear 6: `min(max(features, 0), 6)`.

func Relu6Grad added in v1.1.0

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 added in v1.1.0

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 RemoteFusedGraphExecute added in v1.1.0

func RemoteFusedGraphExecute(scope *Scope, inputs []tf.Output, Toutputs []tf.DataType, serialized_remote_fused_graph_execute_info string) (outputs []tf.Output)

Execute a sub graph on a remote processor.

The graph specifications(such as graph itself, input tensors and output names) are stored as a serialized protocol buffer of RemoteFusedGraphExecuteInfo as serialized_remote_fused_graph_execute_info. The specifications will be passed to a dedicated registered remote fused graph executor. The executor will send the graph specifications to a remote processor and execute that graph. The execution results will be passed to consumer nodes as outputs of this node.

Arguments:

inputs: Arbitrary number of tensors with arbitrary data types

serialized_remote_fused_graph_execute_info: Serialized protocol buffer

of RemoteFusedGraphExecuteInfo which contains graph specifications.

Returns Arbitrary number of tensors with arbitrary data types

func RepeatDataset added in v1.2.0

func RepeatDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 The computed min output.the computed max output.

func RequantizationRangePerChannel added in v1.12.1

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 The minimum value of the final output tensorThe maximum value of the final output tensor.

func Requantize added in v1.1.0

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 The requested_output_min value is copied into this output.The requested_output_max value is copied into this output.

func RequantizePerChannel added in v1.12.1

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 tensor.The minimum value of the final output tensorThe maximum value of the final output tensor.

func Reshape added in v1.1.0

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 `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` can be -1.

If `shape` is 1-D or higher, then 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`.

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.11.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 ResourceApplyAdam added in v1.1.0

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.

$$lr_t := \text{learning\_rate} * \sqrt{1 - beta_2^t} / (1 - beta_1^t)$$ $$m_t := beta_1 * m_{t-1} + (1 - beta_1) * g$$ $$v_t := beta_2 * v_{t-1} + (1 - beta_2) * g * g$$ $$variable := variable - lr_t * m_t / (\sqrt{v_t} + \epsilon)$$

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.
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 added in v1.12.1

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.

$$lr_t := \text{learning\_rate} * \sqrt{1 - beta_2^t} / (1 - beta_1^t)$$ $$m_t := beta_1 * m_{t-1} + (1 - beta_1) * g$$ $$v_t := beta_2 * v_{t-1} + (1 - beta_2) * g * g$$ $$vhat_t := max{vhat_{t-1}, v_t}$$ $$variable := variable - lr_t * m_t / (\sqrt{vhat_t} + \epsilon)$$

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 added in v1.5.0

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 added in v1.1.0

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.

epsilon: Ridge term. Must be a scalar.
grad: The gradient.

Returns the created operation.

func ResourceApplyFtrl added in v1.1.0

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 regulariation. Must be a scalar.
l2: L2 regulariation. Must be a scalar.
lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceApplyFtrlV2 added in v1.3.0

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.

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.
lr: Scaling factor. Must be a scalar.
l1: L1 regulariation. Must be a scalar.
l2: L2 shrinkage regulariation. Must be a scalar.

lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceApplyGradientDescent added in v1.1.0

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.11.0

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 added in v1.12.1

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 added in v1.6.0

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 added in v1.9.0

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 added in v1.5.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 ResourceSparseApplyCenteredRMSProp added in v1.1.0

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 added in v1.1.0

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 added in v1.3.0

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 regulariation. Must be a scalar.

lr_power: Scaling factor. Must be a scalar.

Returns the created operation.

func ResourceSparseApplyKerasMomentum added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 RetrieveTPUEmbeddingADAMParameters added in v1.12.1

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 Parameter parameters updated by the ADAM optimization algorithm.Parameter momenta updated by the ADAM optimization algorithm.Parameter velocities updated by the ADAM optimization algorithm.

func RetrieveTPUEmbeddingADAMParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingADAMParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr) (parameters tf.Output, momenta tf.Output, velocities tf.Output, gradient_accumulators tf.Output)

Retrieve ADAM embedding parameters with debug support.

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 ADAM optimization algorithm.Parameter momenta updated by the ADAM optimization algorithm.Parameter velocities updated by the ADAM optimization algorithm.Parameter gradient_accumulators updated by the ADAM optimization algorithm.

func RetrieveTPUEmbeddingAdadeltaParameters added in v1.12.1

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 Parameter parameters updated by the Adadelta optimization algorithm.Parameter accumulators updated by the Adadelta optimization algorithm.Parameter updates updated by the Adadelta optimization algorithm.

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr) (parameters tf.Output, accumulators tf.Output, updates tf.Output, gradient_accumulators tf.Output)

Retrieve Adadelta embedding parameters with debug support.

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 Adadelta optimization algorithm.Parameter accumulators updated by the Adadelta optimization algorithm.Parameter updates updated by the Adadelta optimization algorithm.Parameter gradient_accumulators updated by the Adadelta optimization algorithm.

func RetrieveTPUEmbeddingAdagradParameters added in v1.12.1

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 Parameter parameters updated by the Adagrad optimization algorithm.Parameter accumulators updated by the Adagrad optimization algorithm.

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr) (parameters tf.Output, accumulators tf.Output, gradient_accumulators tf.Output)

Retrieve Adagrad embedding parameters with debug support.

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 Adagrad optimization algorithm.Parameter accumulators updated by the Adagrad optimization algorithm.Parameter gradient_accumulators updated by the Adagrad optimization algorithm.

func RetrieveTPUEmbeddingCenteredRMSPropParameters added in v1.12.1

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 Parameter parameters updated by the centered RMSProp optimization algorithm.Parameter ms updated by the centered RMSProp optimization algorithm.Parameter mom updated by the centered RMSProp optimization algorithm.Parameter mg updated by the centered RMSProp optimization algorithm.

func RetrieveTPUEmbeddingFTRLParameters added in v1.12.1

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 Parameter parameters updated by the FTRL optimization algorithm.Parameter accumulators updated by the FTRL optimization algorithm.Parameter linears updated by the FTRL optimization algorithm.

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr) (parameters tf.Output, accumulators tf.Output, linears tf.Output, gradient_accumulators tf.Output)

Retrieve FTRL embedding parameters with debug support.

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 FTRL optimization algorithm.Parameter accumulators updated by the FTRL optimization algorithm.Parameter linears updated by the FTRL optimization algorithm.Parameter gradient_accumulators updated by the FTRL optimization algorithm.

func RetrieveTPUEmbeddingMDLAdagradLightParameters added in v1.12.1

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 Parameter parameters updated by the MDL Adagrad Light optimization algorithm.Parameter accumulators updated by the MDL Adagrad Light optimization algorithm.Parameter weights updated by the MDL Adagrad Light optimization algorithm.Parameter benefits updated by the MDL Adagrad Light optimization algorithm.

func RetrieveTPUEmbeddingMomentumParameters added in v1.12.1

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 Parameter parameters updated by the Momentum optimization algorithm.Parameter momenta updated by the Momentum optimization algorithm.

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr) (parameters tf.Output, momenta tf.Output, gradient_accumulators tf.Output)

Retrieve Momentum embedding parameters with debug support.

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 Momentum optimization algorithm.Parameter momenta updated by the Momentum optimization algorithm.Parameter gradient_accumulators updated by the Momentum optimization algorithm.

func RetrieveTPUEmbeddingProximalAdagradParameters added in v1.12.1

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 Parameter parameters updated by the proximal Adagrad optimization algorithm.Parameter accumulators updated by the proximal Adagrad optimization algorithm.

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr) (parameters tf.Output, accumulators tf.Output, gradient_accumulators tf.Output)

Retrieve proximal Adagrad embedding parameters with debug support.

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 proximal Adagrad optimization algorithm.Parameter accumulators updated by the proximal Adagrad optimization algorithm.Parameter gradient_accumulators updated by the proximal Adagrad optimization algorithm.

func RetrieveTPUEmbeddingRMSPropParameters added in v1.12.1

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 Parameter parameters updated by the RMSProp optimization algorithm.Parameter ms updated by the RMSProp optimization algorithm.Parameter mom updated by the RMSProp optimization algorithm.

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebug added in v1.12.1

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebug(scope *Scope, num_shards int64, shard_id int64, optional ...RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr) (parameters tf.Output, ms tf.Output, mom tf.Output, gradient_accumulators tf.Output)

Retrieve RMSProp embedding parameters with debug support.

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 RMSProp optimization algorithm.Parameter ms updated by the RMSProp optimization algorithm.Parameter mom updated by the RMSProp optimization algorithm.Parameter gradient_accumulators updated by the RMSProp optimization algorithm.

func RetrieveTPUEmbeddingStochasticGradientDescentParameters added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

func ReverseV2(scope *Scope, tensor tf.Output, axis tf.Output) (output tf.Output)

Reverses specific dimensions of a tensor.

NOTE `tf.reverse` has now changed behavior in preparation for 1.0. `tf.reverse_v2` is currently an alias that will be deprecated before TF 1.0.

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 added in v1.5.0

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: id=151, shape=(4,), dtype=int8, numpy=array([ -2, 64, 101, 32], dtype=int8)> ```

func Rint added in v1.1.0

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 RngSkip added in v1.14.0

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 added in v1.9.0

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 added in v1.1.0

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 Rpc added in v1.9.0

func Rpc(scope *Scope, address tf.Output, method tf.Output, request tf.Output, optional ...RpcAttr) (response tf.Output)

Perform batches of RPC requests.

This op asynchronously performs either a single RPC request, or a batch of requests. RPC requests are defined by three main parameters:

  • `address` (the host+port or BNS address of the request)
  • `method` (the RPC method name for the request)
  • `request` (the serialized proto string, or vector of strings, of the RPC request argument).

For example, if you have an RPC service running on port localhost:2345, and its interface is configured with the following proto declaration:

```

service MyService {
  rpc MyMethod(MyRequestProto) returns (MyResponseProto) {
  }
};

```

then call this op with arguments:

``` address = "localhost:2345" method = "MyService/MyMethod" ```

The `request` tensor is a string tensor representing serialized `MyRequestProto` strings; and the output string tensor `response` will have the same shape and contain (upon successful completion) corresponding serialized `MyResponseProto` strings.

For example, to send a single, empty, `MyRequestProto`, call this op with `request = ""`. To send 5 **parallel** empty requests, call this op with `request = ["", "", "", "", ""]`.

More generally, one can create a batch of `MyRequestProto` serialized protos from regular batched tensors using the `encode_proto` op, and convert the response `MyResponseProto` serialized protos to batched tensors using the `decode_proto` op.

**NOTE** Working with serialized proto strings is faster than instantiating actual proto objects in memory, so no performance degradation is expected compared to writing custom kernels for this workflow.

If the connection fails or the remote worker returns an error status, the op reraises this exception locally.

See the `TryRpc` op if you prefer to handle RPC failures manually in the graph.

Arguments:

address: `0-D` or `1-D`.  The address (i.e. host_name:port) of the RPC server.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `method` and `request`.

method: `0-D` or `1-D`.  The method address on the RPC server.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `address` and `request`.

request: `0-D` or `1-D`.  Serialized proto strings: the rpc request argument.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `address` and `method`.

Returns Same shape as `request`. Serialized proto strings: the rpc responses.

func Rsqrt added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 1-D, containing `[offset_height, offset_width, 0]`. Provide as input to `tf.slice`.1-D, containing `[target_height, target_width, -1]`. Provide as input to `tf.slice`.3-D with shape `[1, 1, 4]` containing the distorted bounding box. Provide as input to `tf.image.draw_bounding_boxes`.

func SampleDistortedBoundingBoxV2 added in v1.3.0

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 1-D, containing `[offset_height, offset_width, 0]`. Provide as input to `tf.slice`.1-D, containing `[target_height, target_width, -1]`. Provide as input to `tf.slice`.3-D with shape `[1, 1, 4]` containing the distorted bounding box. Provide as input to `tf.image.draw_bounding_boxes`.

func SamplingDataset added in v1.12.1

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 contains `rate` elements from the `input_dataset`.

Arguments:

rate: A scalar representing the sample rate of elements from the `input_dataset`

that should be taken.

seed: A scalar representing seed of random number generator.
seed2: A scalar representing seed2 of random number generator.

func Save added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

func ScatterNd(scope *Scope, indices tf.Output, updates tf.Output, shape tf.Output) (output tf.Output)

Scatter `updates` into a new tensor according to `indices`.

Creates a new tensor by applying sparse `updates` to individual values or slices within a tensor (initially zero for numeric, empty for string) of the given `shape` according to indices. This operator is the inverse of the `tf.gather_nd` operator which extracts values or slices from a given tensor.

This operation is similar to tensor_scatter_add, except that the tensor is zero-initialized. Calling `tf.scatter_nd(indices, values, shape)` is identical to `tensor_scatter_add(tf.zeros(shape, values.dtype), indices, values)`

If `indices` contains duplicates, then their updates are accumulated (summed).

**WARNING**: The order in which updates are applied is nondeterministic, so the output will be nondeterministic if `indices` contains duplicates -- because of some numerical approximation issues, numbers summed in different order may yield different results.

`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 scatter is to insert individual elements in a tensor by index. For example, say we 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)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[0, 11, 0, 10, 9, 0, 0, 12]

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.

<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([[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]]])
shape = tf.constant([4, 4, 4])
scatter = tf.scatter_nd(indices, updates, shape)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[[[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]],
 [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]]

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:

indices: Index tensor.
updates: Updates to scatter into output.
shape: 1-D. The shape of the resulting tensor.

Returns A new tensor with the given shape and updates applied according to the indices.

func ScatterNdNonAliasingAdd added in v1.3.0

func ScatterNdNonAliasingAdd(scope *Scope, input tf.Output, indices tf.Output, updates tf.Output) (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 added in v1.1.0

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 added in v1.1.0

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 a list of vectors containing the updated example state data.a list of vectors where each value is the delta weights associated with a sparse feature group.a list of vectors where the values are the delta weights associated with a dense feature group.

func SdcaOptimizerV2 added in v1.12.1

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 a list of vectors containing the updated example state data.a list of vectors where each value is the delta weights associated with a sparse feature group.a list of vectors where the values are the delta weights associated with a dense feature group.

func SegmentMax added in v1.1.0

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`.

<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.segment_max(c, tf.constant([0, 0, 1])) # ==> [[4, 3, 3, 4], # [5, 6, 7, 8]] ```

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.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentMean added in v1.1.0

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`.

<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.segment_mean(c, tf.constant([0, 0, 1])) # ==> [[2.5, 2.5, 2.5, 2.5], # [5, 6, 7, 8]] ```

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.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentMin added in v1.1.0

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`.

<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.segment_min(c, tf.constant([0, 0, 1])) # ==> [[1, 2, 2, 1], # [5, 6, 7, 8]] ```

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.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentProd added in v1.1.0

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`.

<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.segment_prod(c, tf.constant([0, 0, 1])) # ==> [[4, 6, 6, 4], # [5, 6, 7, 8]] ```

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.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func SegmentSum added in v1.1.0

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`.

<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.segment_sum(c, tf.constant([0, 0, 1])) # ==> [[5, 5, 5, 5], # [5, 6, 7, 8]] ```

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.

Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.

func Select added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 Eigenvalues. Shape is `[N]`.Eigenvectors. Shape is `[N, N]`.

func Selu added in v1.4.0

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 added in v1.4.0

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 SendTPUEmbeddingGradients added in v1.12.1

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 added in v1.5.0

func SerializeIterator(scope *Scope, resource_handle tf.Output) (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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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.

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.6.0

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) (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 added in v1.2.0

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 added in v1.12.1

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 Sigmoid added in v1.1.0

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

Computes sigmoid of `x` element-wise.

Specifically, `y = 1 / (1 + exp(-x))`.

func SigmoidGrad added in v1.1.0

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 added in v1.1.0

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`.

func Sin added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.2.0

func SkipDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 A vector of words in the corpus.Frequencies of words. Sorted in the non-ascending order.Number of words per epoch in the data file.The current epoch number.The total number of words processed so far.A vector of word ids.A vector of word ids.

func Slice added in v1.1.0

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 added in v1.15.0

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) (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 added in v1.6.0

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

Returns a copy of the input tensor.

func SnapshotDataset added in v1.14.0

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 Softmax added in v1.1.0

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 added in v1.1.0

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 Per example loss (batch_size vector).backpropagated gradients (batch_size x num_classes matrix).

func Softplus added in v1.1.0

func Softplus(scope *Scope, features tf.Output) (activations tf.Output)

Computes softplus: `log(exp(features) + 1)`.

func SoftplusGrad added in v1.1.0

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 added in v1.1.0

func Softsign(scope *Scope, features tf.Output) (activations tf.Output)

Computes softsign: `features / (abs(features) + 1)`.

func SoftsignGrad added in v1.1.0

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 added in v1.1.0

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.

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

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.

func SpaceToBatchND added in v1.1.0

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.

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`.

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
  1. 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

  1. 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.

func SpaceToDepth added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 1-D with shape `[nnz(A)]`. The gradient with respect to the non-empty values of A.1-D with shape `[nnz(B)]`. The gradient with respect to the non-empty values of B.

func SparseConcat added in v1.1.0

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 2-D. Indices of the concatenated `SparseTensor`.1-D. Non-empty values of the concatenated `SparseTensor`.1-D. Shape of the concatenated `SparseTensor`.

func SparseCross added in v1.2.0

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 2-D. Indices of the concatenated `SparseTensor`.1-D. Non-empty values of the concatenated or hashed `SparseTensor`.1-D. Shape of the concatenated `SparseTensor`.

func SparseDenseCwiseAdd added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.3.0

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 1-D. the values of the filled sparse tensor.1-D. whether the dense row was missing in the input sparse tensor.1-D. a map from the input indices to the output indices.

func SparseFillEmptyRowsGrad added in v1.3.0

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 1-D. The backprop into values.0-D. The backprop into default_value.

func SparseMatMul added in v1.1.0

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 SparseReduceMax added in v1.3.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 2-D. `N x R` matrix with the same indices as input_indices, but in canonical row-major ordering.1-D. `N` non-empty values corresponding to `output_indices`.

func SparseReshape added in v1.1.0

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 2-D. `N x R_out` matrix with the updated indices of non-empty values in the output SparseTensor.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 added in v1.1.0

func SparseSegmentMean(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output) (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 added in v1.1.0

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 SparseSegmentMeanWithNumSegments added in v1.6.0

func SparseSegmentMeanWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output) (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 misisng, 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 added in v1.1.0

func SparseSegmentSqrtN(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output) (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 added in v1.1.0

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 SparseSegmentSqrtNWithNumSegments added in v1.6.0

func SparseSegmentSqrtNWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output) (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 misisng, 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 added in v1.1.0

func SparseSegmentSum(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output) (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 SparseSegmentSumWithNumSegments added in v1.6.0

func SparseSegmentSumWithNumSegments(scope *Scope, data tf.Output, indices tf.Output, segment_ids tf.Output, num_segments tf.Output) (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 misisng, 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 added in v1.3.0

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 A list of 1-D tensors represents the values of the output sparse tensors.A list of 1-D tensors represents the shape of the output sparse tensors.

func SparseSliceGrad added in v1.11.0

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 added in v1.1.0

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 added in v1.1.0

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 Per example loss (batch_size vector).backpropagated gradients (batch_size x num_classes matrix).

func SparseSparseMaximum added in v1.1.0

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 2-D. The indices of the output SparseTensor.1-D. The values of the output SparseTensor.

func SparseSparseMinimum added in v1.1.0

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 2-D. The indices of the output SparseTensor.1-D. The values of the output SparseTensor.

func SparseSplit added in v1.1.0

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 A list of 1-D tensors represents the values of the output sparse tensors.A list of 1-D tensors represents the shape of the output sparse tensors.

func SparseTensorDenseAdd added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

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 SparseToDense added in v1.1.0

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 added in v1.1.0

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 2D indices of a `SparseTensor`.1D values of a `SparseTensor`.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 added in v1.1.0

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 SplitV added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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

Returns (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 added in v1.1.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

func StageSize(scope *Scope, dtypes []tf.DataType, optional ...StageSizeAttr) (size tf.Output)

Op returns the number of elements in the underlying container.

func StatefulStandardNormal added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.9.0

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 StatelessRandomNormal added in v1.2.0

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 StatelessRandomUniform added in v1.2.0

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 StatelessRandomUniformInt added in v1.12.1

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 StatelessTruncatedNormal added in v1.2.0

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 StaticRegexFullMatch added in v1.12.0

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 added in v1.12.0

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 added in v1.5.0

func StatsAggregatorHandle(scope *Scope, optional ...StatsAggregatorHandleAttr) (handle tf.Output)

Creates a statistics manager resource.

func StatsAggregatorSetSummaryWriter added in v1.12.1

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 added in v1.5.0

func StatsAggregatorSummary(scope *Scope, iterator tf.Output) (summary tf.Output)

Produces a summary of any statistics recorded by the given statistics manager.

func StopGradient added in v1.1.0

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. Some 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 added in v1.1.0

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 ```

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 added in v1.1.0

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 added in v1.12.0

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 added in v1.1.0

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).

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 added in v1.12.0

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.

Arguments:

input: The string for which to compute the length.

Returns Integer tensor that has the same shape as `input`. The output contains the element-wise string lengths of `input`.

func StringSplit added in v1.1.0

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 A dense matrix of int64 representing the indices of the sparse tensor.A vector of strings corresponding to the splited values.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 added in v1.11.0

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 added in v1.11.0

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

Strip leading and trailing whitespaces from the Tensor.

Arguments:

input: A string `Tensor` of any shape.

Returns A string `Tensor` of the same shape as the input.

func StringToHashBucket added in v1.1.0

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 added in v1.1.0

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`.

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 added in v1.1.0

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`.

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 added in v1.1.0

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.)

Returns A Tensor of the same shape as the input `string_tensor`.

func Sub added in v1.1.0

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 added in v1.1.0

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, 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'] ```

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 added in v1.1.0

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 added in v1.1.0

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 of left singular vectors for each matrix. # v is the tensor containing of 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 Singular values. Shape is `[..., P]`.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`.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 added in v1.1.0

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 If `pred` is false, data will be forwarded to this output.If `pred` is true, data will be forwarded to this output.

func TFRecordDataset added in v1.2.0

func TFRecordDataset(scope *Scope, filenames tf.Output, compression_type tf.Output, buffer_size tf.Output) (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 TFRecordReaderV2 added in v1.1.0

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 added in v1.12.1

func TPUCompilationResult(scope *Scope) (output tf.Output)

CompilationResultProto indicating the status of the TPU compilation.

func TPUEmbeddingActivations added in v1.12.1

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 TPUOrdinalSelector added in v1.12.1

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 TPUReplicateMetadata added in v1.12.1

func TPUReplicateMetadata(scope *Scope, num_replicas int64, optional ...TPUReplicateMetadataAttr) (o *tf.Operation)

Metadata indicaitng how the TPU computation should be replicated.

Arguments:

num_replicas: Number of replicas of the computation

Returns the created operation.

func TPUReplicatedInput added in v1.12.1

func TPUReplicatedInput(scope *Scope, inputs []tf.Output) (output tf.Output)

Connects N inputs to an N-way replicated TPU computation.

func TPUReplicatedOutput added in v1.12.1

func TPUReplicatedOutput(scope *Scope, input tf.Output, num_replicas int64) (outputs []tf.Output)

Connects outputs of an N-way replicated computation to N outputs.

func TakeDataset added in v1.2.0

func TakeDataset(scope *Scope, input_dataset tf.Output, count tf.Output, output_types []tf.DataType, output_shapes []tf.Shape) (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 added in v1.1.0

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 2-D. The `indices` of the minibatch `SparseTensor`.1-D. The `values` of the minibatch `SparseTensor`.1-D. The `shape` of the minibatch `SparseTensor`.

func Tan added in v1.1.0

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 added in v1.1.0

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]`.

```python
x = tf.constant([-float("inf"), -5, -0.5, 1, 1.2, 2, 3, float("inf")])
tf.math.tanh(x) ==> [-1. -0.99990916 -0.46211717 0.7615942 0.8336547 0.9640276 0.9950547 1.]
```

func TanhGrad added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 All of the elements in the TensorArray, concatenated along the first axis.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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.11.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 The handle to the TensorArray.A scalar used to control gradient flow.

func TensorArrayWriteV2 added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

func TensorDataset(scope *Scope, components []tf.Output, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that emits `components` as a tuple of tensors once.

func TensorForestCreateTreeVariable added in v1.12.1

func TensorForestCreateTreeVariable(scope *Scope, tree_handle tf.Output, tree_config tf.Output) (o *tf.Operation)

Creates a tree resource and returns a handle to it.

Arguments:

tree_handle: Handle to the tree resource to be created.
tree_config: Serialized proto string of the boosted_trees.Tree.

Returns the created operation.

func TensorForestTreeDeserialize added in v1.12.1

func TensorForestTreeDeserialize(scope *Scope, tree_handle tf.Output, tree_config tf.Output) (o *tf.Operation)

Deserializes a proto into the tree handle

Arguments:

tree_handle: Handle to the tree resource to be restored.
tree_config: Serialied proto string of the boosted_trees.Tree proto.

Returns the created operation.

func TensorForestTreeIsInitializedOp added in v1.12.1

func TensorForestTreeIsInitializedOp(scope *Scope, tree_handle tf.Output) (is_initialized tf.Output)

Checks whether a tree has been initialized.

Arguments:

tree_handle: Handle to the tree.

Returns Whether the tree is initialized.

func TensorForestTreePredict added in v1.12.1

func TensorForestTreePredict(scope *Scope, tree_handle tf.Output, dense_features tf.Output, logits_dimension int64) (logits tf.Output)

Output the logits for the given input data

Arguments:

tree_handle: Handle to the tree resource.
dense_features: Rank 2 dense features tensor.
logits_dimension: Scalar, dimension of the logits.

Returns The logits predictions from the tree for each instance in the batch.

func TensorForestTreeResourceHandleOp added in v1.12.1

func TensorForestTreeResourceHandleOp(scope *Scope, optional ...TensorForestTreeResourceHandleOpAttr) (resource tf.Output)

Creates a handle to a TensorForestTreeResource

func TensorForestTreeSerialize added in v1.12.1

func TensorForestTreeSerialize(scope *Scope, tree_handle tf.Output) (tree_config tf.Output)

Serializes the tree handle to a proto

Arguments:

tree_handle: Handle to the tree resource to be serialized.

Returns Serialied proto string of the tree resource.

func TensorForestTreeSize added in v1.12.1

func TensorForestTreeSize(scope *Scope, tree_handle tf.Output) (tree_size tf.Output)

Get the number of nodes in a tree

Arguments:

tree_handle: Handle to the tree resource.

Returns The size of the tree.

func TensorListConcat added in v1.12.1

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 added in v1.12.1

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.12.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

func TensorListPushBack(scope *Scope, input_handle tf.Output, tensor tf.Output) (output_handle tf.Output)

Returns a list 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 added in v1.9.0

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 added in v1.12.1

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 added in v1.12.0

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.9.0

func TensorListSetItem(scope *Scope, input_handle tf.Output, index tf.Output, item tf.Output) (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 added in v1.12.1

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 added in v1.9.0

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 TensorScatterAdd added in v1.12.1

func TensorScatterAdd(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output) (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.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 `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_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:

```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_add(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[1, 12, 1, 11, 10, 1, 1, 13]

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])
updated = tf.tensor_scatter_add(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[[[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]]]

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 added according to the indices.

func TensorScatterSub added in v1.12.1

func TensorScatterSub(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output) (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_sub(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

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])
updated = tf.tensor_scatter_sub(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

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 added in v1.12.1

func TensorScatterUpdate(scope *Scope, tensor tf.Output, indices tf.Output, updates tf.Output) (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 their updates are accumulated (summed).

**WARNING**: The order in which updates are applied is nondeterministic, so the output will be nondeterministic if `indices` contains duplicates -- because of some numerical approximation issues, numbers summed in different order may yield different results.

`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 scatter is to insert individual elements in a tensor by index. For example, say we 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])
tensor = tf.ones([8], dtype=tf.int32)
updated = tf.tensor_scatter_update(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[1, 11, 1, 10, 9, 1, 1, 12]

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 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])
updated = tf.tensor_scatter_update(tensor, indices, updates)
with tf.Session() as sess:
  print(sess.run(scatter))

```

The resulting tensor would look like this:

[[[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]],
 [[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]],
 [[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]],
 [[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 with the given shape and updates applied according to the indices.

func TensorSliceDataset added in v1.2.0

func TensorSliceDataset(scope *Scope, components []tf.Output, output_shapes []tf.Shape) (handle tf.Output)

Creates a dataset that emits each dim-0 slice of `components` once.

func TensorStridedSliceUpdate added in v1.12.1

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.2.0

func TextLineDataset(scope *Scope, filenames tf.Output, compression_type tf.Output, buffer_size tf.Output) (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 added in v1.1.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 added in v1.1.0

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]`.

Arguments:

input: 1-D or higher.
multiples: 1-D. Length must be the same as the number of dimensions in `input`

func TileGrad added in v1.1.0

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 added in v1.9.0

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.

Note: the timestamp is computed when the op is executed, not when it is added to the graph.

func TopK added in v1.1.0

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 The `k` largest elements along each last dimensional slice.The indices of `values` within the last dimension of `input`.

func TopKV2 added in v1.1.0

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 The `k` largest elements along each last dimensional slice.The indices of `values` within the last dimension of `input`.

func Transpose added in v1.1.0

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 added in v1.14.0

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 ingored.

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 ingored.

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 added in v1.12.1

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

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 added in v1.1.0

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

Returns x / y element-wise for integer types.

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 added in v1.1.0

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 added in v1.1.0

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 TryRpc added in v1.9.0

func TryRpc(scope *Scope, address tf.Output, method tf.Output, request tf.Output, optional ...TryRpcAttr) (response tf.Output, status_code tf.Output, status_message tf.Output)

Perform batches of RPC requests.

This op asynchronously performs either a single RPC request, or a batch of requests. RPC requests are defined by three main parameters:

  • `address` (the host+port or BNS address of the request)
  • `method` (the method name for the request)
  • `request` (the serialized proto string, or vector of strings, of the RPC request argument).

For example, if you have an RPC service running on port localhost:2345, and its interface is configured with the following proto declaration:

```

service MyService {
  rpc MyMethod(MyRequestProto) returns (MyResponseProto) {
  }
};

```

then call this op with arguments:

``` address = "localhost:2345" method = "MyService/MyMethod" ```

The `request` tensor is a string tensor representing serialized `MyRequestProto` strings; and the output string tensor `response` will have the same shape and contain (upon successful completion) corresponding serialized `MyResponseProto` strings.

For example, to send a single, empty, `MyRequestProto`, call this op with `request = ""`. To send 5 **parallel** empty requests, call this op with `request = ["", "", "", "", ""]`.

More generally, one can create a batch of `MyRequestProto` serialized protos from regular batched tensors using the `encode_proto` op, and convert the response `MyResponseProto` serialized protos to batched tensors using the `decode_proto` op.

**NOTE** Working with serialized proto strings is faster than instantiating actual proto objects in memory, so no performance degradation is expected compared to writing custom kernels for this workflow.

Unlike the standard `Rpc` op, if the connection fails or the remote worker returns an error status, this op does **not** reraise the exception. Instead, the `status_code` and `status_message` entry for the corresponding RPC call is set with the error returned from the RPC call. The `response` tensor will contain valid response values for those minibatch entries whose RPCs did not fail; the rest of the entries will have empty strings.

Arguments:

address: `0-D` or `1-D`.  The address (i.e. host_name:port) of the RPC server.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `method` and `request`.

method: `0-D` or `1-D`.  The method address on the RPC server.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `address` and `request`.

request: `0-D` or `1-D`.  Serialized proto strings: the rpc request argument.

If this tensor has more than 1 element, then multiple parallel rpc requests are sent. This argument broadcasts with `address` and `method`.

Returns Same shape as `request`. Serialized proto strings: the rpc responses.Same shape as `request`. Values correspond to tensorflow Status enum codes.Same shape as `request`. Values correspond to Status messages returned from the RPC calls.

func Unbatch added in v1.9.0

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 added in v1.11.0

func UnbatchDataset(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 UnbatchGrad added in v1.9.0

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 UnicodeDecode added in v1.12.1

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 A 1D int32 tensor containing the row splits.A 1D int32 Tensor containing the decoded codepoints.

func UnicodeDecodeWithOffsets added in v1.12.1

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 A 1D int32 tensor containing the row splits.A 1D int32 Tensor containing the decoded codepoints.A 1D int32 Tensor containing the byte index in the input string where each character in `char_values` starts.

func UnicodeEncode added in v1.12.1

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 added in v1.12.1

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 http://icu-project.org/apiref/icu4c/uscript_8h.html. Returns -1 (USCRIPT_INVALID_CODE) for invalid codepoints. Output shape will match input shape.

Arguments:

input: A Tensor of int32 Unicode code points.

Returns A Tensor of int32 script codes corresponding to each input code point.

func UnicodeTranscode added in v1.12.1

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.

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 added in v1.1.0

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 A vector of length num_sampled, in which each element is the ID of a sampled candidate.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.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 Unique added in v1.1.0

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`. 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]`

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] ```

Arguments:

x: 1-D.

Returns 1-D.1-D.

func UniqueDataset added in v1.6.0

func UniqueDataset(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 UniqueV2 added in v1.6.0

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 A `Tensor`. Unique elements along the `axis` of `Tensor` x.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 added in v1.1.0

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 1-D.1-D.1-D.

func UniqueWithCountsV2 added in v1.9.0

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:

``` # 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] ```

For an `2-D` tensor `x` with `axis = 0`:

``` # tensor 'x' is [[1, 0, 0], # [1, 0, 0], # [2, 0, 0]] y, idx, count = unique_with_counts(x, axis=0) y ==> [[1, 0, 0],

[2, 0, 0]]

idx ==> [0, 0, 1] count ==> [2, 1] ```

For an `2-D` tensor `x` with `axis = 1`:

``` # tensor 'x' is [[1, 0, 0], # [1, 0, 0], # [2, 0, 0]] y, idx, count = unique_with_counts(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 A `Tensor`. Unique elements along the `axis` of `Tensor` x.A 1-D Tensor. Has the same type as x that contains the index of each value of x in the output y.A 1-D Tensor. The count of each value of x in the output y.

func Unpack added in v1.1.0

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 added in v1.9.0

func UnravelIndex(scope *Scope, indices tf.Output, dims tf.Output) (output tf.Output)

Converts a flat index or array of flat indices into a tuple of

coordinate arrays.

@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 UnsortedSegmentJoin added in v1.15.0

func UnsortedSegmentJoin(scope *Scope, inputs tf.Output, segment_ids tf.Output, num_segments tf.Output, optional ...UnsortedSegmentJoinAttr) (output tf.Output)

Joins the elements of `inputs` based on `segment_ids`.

Computes the string join along segments of a tensor. Given `segment_ids` with rank `N` and `data` with rank `N+M`:

`output[i, k1...kM] = strings.join([data[j1...jN, k1...kM])`

where the join is over all [j1...jN] such that segment_ids[j1...jN] = i. Strings are joined in row-major order.

For example:

```python inputs = [['Y', 'q', 'c'], ['Y', '6', '6'], ['p', 'G', 'a']] output_array = string_ops.unsorted_segment_join(inputs=inputs,

segment_ids=[1, 0, 1],
num_segments=2,
separator=':'))

output_array ==> [['Y', '6', '6'], ['Y:p', 'q:G', 'c:a']]

inputs = ['this', 'is', 'a', 'test'] output_array = string_ops.unsorted_segment_join(inputs=inputs,

segment_ids=[0, 0, 0, 0],
num_segments=1,
separator=':'))

# output_array ==> ['this:is:a:test'] ```

Arguments:

inputs: The input to be joined.
segment_ids: A tensor whose shape is a prefix of data.shape.  Negative segment ids are not

supported.

num_segments: A scalar.

func UnsortedSegmentMax added in v1.1.0

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 the unsorted segment sum operator found [(here)](../../../api_docs/python/math_ops.md#UnsortedSegmentSum). 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.

<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:

``` python c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) tf.unsorted_segment_max(c, tf.constant([0, 1, 0]), num_segments=2) # ==> [[ 4, 3, 3, 4], # [5, 6, 7, 8]] ```

Arguments:

segment_ids: A tensor whose shape is a prefix of `data.shape`.

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 added in v1.9.0

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 the unsorted segment sum operator found [(here)](../../../api_docs/python/math_ops.md#UnsortedSegmentSum). 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:

``` python c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) tf.unsorted_segment_min(c, tf.constant([0, 1, 0]), num_segments=2) # ==> [[ 1, 2, 2, 1], # [5, 6, 7, 8]] ```

If the given segment ID `i` is negative, then the corresponding value is dropped, and will not be included in the result.

Arguments:

segment_ids: A tensor whose shape is a prefix of `data.shape`.

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 added in v1.9.0

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 the unsorted segment sum operator found [(here)](../../../api_docs/python/math_ops.md#UnsortedSegmentSum). 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:

``` python c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) tf.unsorted_segment_prod(c, tf.constant([0, 1, 0]), num_segments=2) # ==> [[ 4, 6, 6, 4], # [5, 6, 7, 8]] ```

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.

Arguments:

segment_ids: A tensor whose shape is a prefix of `data.shape`.

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 added in v1.1.0

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.

<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>

``` python c = tf.constant([[1,2,3,4], [5,6,7,8], [4,3,2,1]]) tf.unsorted_segment_sum(c, tf.constant([0, 1, 0]), num_segments=2) # ==> [[ 5, 5, 5, 5], # [5, 6, 7, 8]] ```

Arguments:

segment_ids: A tensor whose shape is a prefix of `data.shape`.

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 added in v1.1.0

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 UpperBound added in v1.12.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.11.0

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) (handle tf.Output)

A dataset that creates window datasets from the input dataset.

Arguments:

size: A scalar representing the number of elements to accumulate in a window.
shift: A scalar representing the steps moving the sliding window forward in one

iteration. It must be positive.

stride: A scalar representing the stride of the input elements of the sliding window.

It must be positive.

drop_remainder: A scalar representing whether a window should be dropped in case its size is

smaller than desired.

func WorkerHeartbeat added in v1.12.1

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 WriteFile added in v1.1.0

func WriteFile(scope *Scope, filename tf.Output, contents tf.Output) (o *tf.Operation)

Writes contents to the file at input filename. Creates file and recursively

creates directory if not existing.

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 Xdivy added in v1.12.1

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

Returns 0 if x == 0, and x / y otherwise, elementwise.

func Xlogy added in v1.12.1

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

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

Creates a dataset that zips together `input_datasets`.

Types

type AbortAttr added in v1.1.0

type AbortAttr func(optionalAttr)

AbortAttr is an optional argument to Abort.

func AbortErrorMsg added in v1.1.0

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 added in v1.1.0

func AbortExitWithoutError(value bool) AbortAttr

AbortExitWithoutError sets the optional exit_without_error attribute to value. If not specified, defaults to false

type AddManySparseToTensorsMapAttr added in v1.1.0

type AddManySparseToTensorsMapAttr func(optionalAttr)

AddManySparseToTensorsMapAttr is an optional argument to AddManySparseToTensorsMap.

func AddManySparseToTensorsMapContainer added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type AddSparseToTensorsMapAttr func(optionalAttr)

AddSparseToTensorsMapAttr is an optional argument to AddSparseToTensorsMap.

func AddSparseToTensorsMapContainer added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type AllAttr func(optionalAttr)

AllAttr is an optional argument to All.

func AllKeepDims added in v1.1.0

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 added in v1.1.0

type AllCandidateSamplerAttr func(optionalAttr)

AllCandidateSamplerAttr is an optional argument to AllCandidateSampler.

func AllCandidateSamplerSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

type AngleAttr func(optionalAttr)

AngleAttr is an optional argument to Angle.

func AngleTout added in v1.4.0

func AngleTout(value tf.DataType) AngleAttr

AngleTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type AnyAttr added in v1.1.0

type AnyAttr func(optionalAttr)

AnyAttr is an optional argument to Any.

func AnyKeepDims added in v1.1.0

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 ApproximateEqualAttr added in v1.1.0

type ApproximateEqualAttr func(optionalAttr)

ApproximateEqualAttr is an optional argument to ApproximateEqual.

func ApproximateEqualTolerance added in v1.1.0

func ApproximateEqualTolerance(value float32) ApproximateEqualAttr

ApproximateEqualTolerance sets the optional tolerance attribute to value. If not specified, defaults to 1e-05

type ArgMaxAttr added in v1.3.0

type ArgMaxAttr func(optionalAttr)

ArgMaxAttr is an optional argument to ArgMax.

func ArgMaxOutputType added in v1.3.0

func ArgMaxOutputType(value tf.DataType) ArgMaxAttr

ArgMaxOutputType sets the optional output_type attribute to value. If not specified, defaults to DT_INT64

type ArgMinAttr added in v1.3.0

type ArgMinAttr func(optionalAttr)

ArgMinAttr is an optional argument to ArgMin.

func ArgMinOutputType added in v1.3.0

func ArgMinOutputType(value tf.DataType) ArgMinAttr

ArgMinOutputType sets the optional output_type attribute to value. If not specified, defaults to DT_INT64

type AsStringAttr added in v1.1.0

type AsStringAttr func(optionalAttr)

AsStringAttr is an optional argument to AsString.

func AsStringFill added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type AssertAttr func(optionalAttr)

AssertAttr is an optional argument to Assert.

func AssertSummarize added in v1.1.0

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 AudioSpectrogramAttr added in v1.2.0

type AudioSpectrogramAttr func(optionalAttr)

AudioSpectrogramAttr is an optional argument to AudioSpectrogram.

func AudioSpectrogramMagnitudeSquared added in v1.2.0

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 added in v1.1.0

type AudioSummaryAttr func(optionalAttr)

AudioSummaryAttr is an optional argument to AudioSummary.

func AudioSummaryMaxOutputs added in v1.1.0

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 added in v1.1.0

type AudioSummaryV2Attr func(optionalAttr)

AudioSummaryV2Attr is an optional argument to AudioSummaryV2.

func AudioSummaryV2MaxOutputs added in v1.1.0

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 AvgPool3DAttr added in v1.2.0

type AvgPool3DAttr func(optionalAttr)

AvgPool3DAttr is an optional argument to AvgPool3D.

func AvgPool3DDataFormat added in v1.2.0

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 added in v1.2.0

type AvgPool3DGradAttr func(optionalAttr)

AvgPool3DGradAttr is an optional argument to AvgPool3DGrad.

func AvgPool3DGradDataFormat added in v1.2.0

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 added in v1.1.0

type AvgPoolAttr func(optionalAttr)

AvgPoolAttr is an optional argument to AvgPool.

func AvgPoolDataFormat added in v1.1.0

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 added in v1.1.0

type AvgPoolGradAttr func(optionalAttr)

AvgPoolGradAttr is an optional argument to AvgPoolGrad.

func AvgPoolGradDataFormat added in v1.1.0

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 added in v1.9.0

type BatchAttr func(optionalAttr)

BatchAttr is an optional argument to Batch.

func BatchAllowedBatchSizes added in v1.9.0

func BatchAllowedBatchSizes(value []int64) BatchAttr

BatchAllowedBatchSizes sets the optional allowed_batch_sizes attribute to value. If not specified, defaults to <>

func BatchBatchingQueue added in v1.9.0

func BatchBatchingQueue(value string) BatchAttr

BatchBatchingQueue sets the optional batching_queue attribute to value. If not specified, defaults to ""

func BatchContainer added in v1.9.0

func BatchContainer(value string) BatchAttr

BatchContainer sets the optional container attribute to value. If not specified, defaults to ""

func BatchMaxEnqueuedBatches added in v1.9.0

func BatchMaxEnqueuedBatches(value int64) BatchAttr

BatchMaxEnqueuedBatches sets the optional max_enqueued_batches attribute to value. If not specified, defaults to 10

func BatchSharedName added in v1.9.0

func BatchSharedName(value string) BatchAttr

BatchSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BatchDatasetV2Attr added in v1.14.0

type BatchDatasetV2Attr func(optionalAttr)

BatchDatasetV2Attr is an optional argument to BatchDatasetV2.

func BatchDatasetV2ParallelCopy added in v1.14.0

func BatchDatasetV2ParallelCopy(value bool) BatchDatasetV2Attr

BatchDatasetV2ParallelCopy sets the optional parallel_copy attribute to value. If not specified, defaults to false

type BatchMatMulAttr added in v1.1.0

type BatchMatMulAttr func(optionalAttr)

BatchMatMulAttr is an optional argument to BatchMatMul.

func BatchMatMulAdjX added in v1.1.0

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 added in v1.1.0

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

type BatchMatMulV2Attr added in v1.12.1

type BatchMatMulV2Attr func(optionalAttr)

BatchMatMulV2Attr is an optional argument to BatchMatMulV2.

func BatchMatMulV2AdjX added in v1.12.1

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 added in v1.12.1

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

type BiasAddAttr added in v1.1.0

type BiasAddAttr func(optionalAttr)

BiasAddAttr is an optional argument to BiasAdd.

func BiasAddDataFormat added in v1.1.0

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 added in v1.1.0

type BiasAddGradAttr func(optionalAttr)

BiasAddGradAttr is an optional argument to BiasAddGrad.

func BiasAddGradDataFormat added in v1.1.0

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 added in v1.15.0

type BlockLSTMAttr func(optionalAttr)

BlockLSTMAttr is an optional argument to BlockLSTM.

func BlockLSTMCellClip added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

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 BoostedTreesCalculateBestFeatureSplitAttr added in v1.12.1

type BoostedTreesCalculateBestFeatureSplitAttr func(optionalAttr)

BoostedTreesCalculateBestFeatureSplitAttr is an optional argument to BoostedTreesCalculateBestFeatureSplit.

func BoostedTreesCalculateBestFeatureSplitSplitType added in v1.12.1

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 added in v1.12.0

type BoostedTreesCreateQuantileStreamResourceAttr func(optionalAttr)

BoostedTreesCreateQuantileStreamResourceAttr is an optional argument to BoostedTreesCreateQuantileStreamResource.

func BoostedTreesCreateQuantileStreamResourceMaxElements added in v1.12.0

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 added in v1.9.0

type BoostedTreesEnsembleResourceHandleOpAttr func(optionalAttr)

BoostedTreesEnsembleResourceHandleOpAttr is an optional argument to BoostedTreesEnsembleResourceHandleOp.

func BoostedTreesEnsembleResourceHandleOpContainer added in v1.9.0

func BoostedTreesEnsembleResourceHandleOpContainer(value string) BoostedTreesEnsembleResourceHandleOpAttr

BoostedTreesEnsembleResourceHandleOpContainer sets the optional container attribute to value. If not specified, defaults to ""

func BoostedTreesEnsembleResourceHandleOpSharedName added in v1.9.0

func BoostedTreesEnsembleResourceHandleOpSharedName(value string) BoostedTreesEnsembleResourceHandleOpAttr

BoostedTreesEnsembleResourceHandleOpSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BoostedTreesQuantileStreamResourceFlushAttr added in v1.12.0

type BoostedTreesQuantileStreamResourceFlushAttr func(optionalAttr)

BoostedTreesQuantileStreamResourceFlushAttr is an optional argument to BoostedTreesQuantileStreamResourceFlush.

func BoostedTreesQuantileStreamResourceFlushGenerateQuantiles added in v1.12.0

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 added in v1.12.0

type BoostedTreesQuantileStreamResourceHandleOpAttr func(optionalAttr)

BoostedTreesQuantileStreamResourceHandleOpAttr is an optional argument to BoostedTreesQuantileStreamResourceHandleOp.

func BoostedTreesQuantileStreamResourceHandleOpContainer added in v1.12.0

func BoostedTreesQuantileStreamResourceHandleOpContainer(value string) BoostedTreesQuantileStreamResourceHandleOpAttr

BoostedTreesQuantileStreamResourceHandleOpContainer sets the optional container attribute to value. If not specified, defaults to ""

func BoostedTreesQuantileStreamResourceHandleOpSharedName added in v1.12.0

func BoostedTreesQuantileStreamResourceHandleOpSharedName(value string) BoostedTreesQuantileStreamResourceHandleOpAttr

BoostedTreesQuantileStreamResourceHandleOpSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type BoostedTreesSparseCalculateBestFeatureSplitAttr added in v1.15.0

type BoostedTreesSparseCalculateBestFeatureSplitAttr func(optionalAttr)

BoostedTreesSparseCalculateBestFeatureSplitAttr is an optional argument to BoostedTreesSparseCalculateBestFeatureSplit.

func BoostedTreesSparseCalculateBestFeatureSplitSplitType added in v1.15.0

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 CTCBeamSearchDecoderAttr added in v1.1.0

type CTCBeamSearchDecoderAttr func(optionalAttr)

CTCBeamSearchDecoderAttr is an optional argument to CTCBeamSearchDecoder.

func CTCBeamSearchDecoderMergeRepeated added in v1.1.0

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 added in v1.1.0

type CTCGreedyDecoderAttr func(optionalAttr)

CTCGreedyDecoderAttr is an optional argument to CTCGreedyDecoder.

func CTCGreedyDecoderMergeRepeated added in v1.1.0

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 added in v1.1.0

type CTCLossAttr func(optionalAttr)

CTCLossAttr is an optional argument to CTCLoss.

func CTCLossCtcMergeRepeated added in v1.1.0

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 added in v1.2.0

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 added in v1.1.0

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 CastAttr added in v1.11.0

type CastAttr func(optionalAttr)

CastAttr is an optional argument to Cast.

func CastTruncate added in v1.11.0

func CastTruncate(value bool) CastAttr

CastTruncate sets the optional Truncate attribute to value. If not specified, defaults to false

type CollectiveReduceAttr added in v1.12.1

type CollectiveReduceAttr func(optionalAttr)

CollectiveReduceAttr is an optional argument to CollectiveReduce.

func CollectiveReduceWaitFor added in v1.12.1

func CollectiveReduceWaitFor(value []int64) CollectiveReduceAttr

CollectiveReduceWaitFor sets the optional wait_for attribute to value. If not specified, defaults to <>

type CombinedNonMaxSuppressionAttr added in v1.12.1

type CombinedNonMaxSuppressionAttr func(optionalAttr)

CombinedNonMaxSuppressionAttr is an optional argument to CombinedNonMaxSuppression.

func CombinedNonMaxSuppressionClipBoxes added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

type ComplexAbsAttr func(optionalAttr)

ComplexAbsAttr is an optional argument to ComplexAbs.

func ComplexAbsTout added in v1.1.0

func ComplexAbsTout(value tf.DataType) ComplexAbsAttr

ComplexAbsTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type ComplexAttr added in v1.1.0

type ComplexAttr func(optionalAttr)

ComplexAttr is an optional argument to Complex.

func ComplexTout added in v1.1.0

func ComplexTout(value tf.DataType) ComplexAttr

ComplexTout sets the optional Tout attribute to value. If not specified, defaults to DT_COMPLEX64

type ComputeAccidentalHitsAttr added in v1.1.0

type ComputeAccidentalHitsAttr func(optionalAttr)

ComputeAccidentalHitsAttr is an optional argument to ComputeAccidentalHits.

func ComputeAccidentalHitsSeed added in v1.1.0

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 added in v1.1.0

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 ConfigureDistributedTPUAttr added in v1.12.1

type ConfigureDistributedTPUAttr func(optionalAttr)

ConfigureDistributedTPUAttr is an optional argument to ConfigureDistributedTPU.

func ConfigureDistributedTPUEmbeddingConfig added in v1.12.1

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 ConfigureDistributedTPUIsGlobalInit added in v1.12.1

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 ConfigureDistributedTPUTpuEmbeddingConfig added in v1.12.1

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 added in v1.1.0

type Conv2DAttr func(optionalAttr)

Conv2DAttr is an optional argument to Conv2D.

func Conv2DDataFormat added in v1.1.0

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 added in v1.6.0

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 added in v1.12.1

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 added in v1.1.0

func Conv2DUseCudnnOnGpu(value bool) Conv2DAttr

Conv2DUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropFilterAttr added in v1.1.0

type Conv2DBackpropFilterAttr func(optionalAttr)

Conv2DBackpropFilterAttr is an optional argument to Conv2DBackpropFilter.

func Conv2DBackpropFilterDataFormat added in v1.1.0

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 added in v1.6.0

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 added in v1.12.1

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 added in v1.1.0

func Conv2DBackpropFilterUseCudnnOnGpu(value bool) Conv2DBackpropFilterAttr

Conv2DBackpropFilterUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv2DBackpropInputAttr added in v1.1.0

type Conv2DBackpropInputAttr func(optionalAttr)

Conv2DBackpropInputAttr is an optional argument to Conv2DBackpropInput.

func Conv2DBackpropInputDataFormat added in v1.1.0

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 added in v1.6.0

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 added in v1.12.1

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 added in v1.1.0

func Conv2DBackpropInputUseCudnnOnGpu(value bool) Conv2DBackpropInputAttr

Conv2DBackpropInputUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true

type Conv3DAttr added in v1.2.0

type Conv3DAttr func(optionalAttr)

Conv3DAttr is an optional argument to Conv3D.

func Conv3DDataFormat added in v1.2.0

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 added in v1.6.0

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 added in v1.9.0

type Conv3DBackpropFilterAttr func(optionalAttr)

Conv3DBackpropFilterAttr is an optional argument to Conv3DBackpropFilter.

func Conv3DBackpropFilterDilations added in v1.9.0

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 added in v1.2.0

type Conv3DBackpropFilterV2Attr func(optionalAttr)

Conv3DBackpropFilterV2Attr is an optional argument to Conv3DBackpropFilterV2.

func Conv3DBackpropFilterV2DataFormat added in v1.2.0

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 added in v1.6.0

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 added in v1.9.0

type Conv3DBackpropInputAttr func(optionalAttr)

Conv3DBackpropInputAttr is an optional argument to Conv3DBackpropInput.

func Conv3DBackpropInputDilations added in v1.9.0

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 added in v1.2.0

type Conv3DBackpropInputV2Attr func(optionalAttr)

Conv3DBackpropInputV2Attr is an optional argument to Conv3DBackpropInputV2.

func Conv3DBackpropInputV2DataFormat added in v1.2.0

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 added in v1.6.0

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 CropAndResizeAttr added in v1.1.0

type CropAndResizeAttr func(optionalAttr)

CropAndResizeAttr is an optional argument to CropAndResize.

func CropAndResizeExtrapolationValue added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type CropAndResizeGradBoxesAttr func(optionalAttr)

CropAndResizeGradBoxesAttr is an optional argument to CropAndResizeGradBoxes.

func CropAndResizeGradBoxesMethod added in v1.1.0

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 added in v1.1.0

type CropAndResizeGradImageAttr func(optionalAttr)

CropAndResizeGradImageAttr is an optional argument to CropAndResizeGradImage.

func CropAndResizeGradImageMethod added in v1.1.0

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 added in v1.9.0

type CudnnRNNAttr func(optionalAttr)

CudnnRNNAttr is an optional argument to CudnnRNN.

func CudnnRNNDirection added in v1.9.0

func CudnnRNNDirection(value string) CudnnRNNAttr

CudnnRNNDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNDropout added in v1.9.0

func CudnnRNNDropout(value float32) CudnnRNNAttr

CudnnRNNDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNInputMode added in v1.9.0

func CudnnRNNInputMode(value string) CudnnRNNAttr

CudnnRNNInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNIsTraining added in v1.9.0

func CudnnRNNIsTraining(value bool) CudnnRNNAttr

CudnnRNNIsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNRnnMode added in v1.9.0

func CudnnRNNRnnMode(value string) CudnnRNNAttr

CudnnRNNRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNSeed added in v1.9.0

func CudnnRNNSeed(value int64) CudnnRNNAttr

CudnnRNNSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNSeed2 added in v1.9.0

func CudnnRNNSeed2(value int64) CudnnRNNAttr

CudnnRNNSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropAttr added in v1.9.0

type CudnnRNNBackpropAttr func(optionalAttr)

CudnnRNNBackpropAttr is an optional argument to CudnnRNNBackprop.

func CudnnRNNBackpropDirection added in v1.9.0

func CudnnRNNBackpropDirection(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropDropout added in v1.9.0

func CudnnRNNBackpropDropout(value float32) CudnnRNNBackpropAttr

CudnnRNNBackpropDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropInputMode added in v1.9.0

func CudnnRNNBackpropInputMode(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropRnnMode added in v1.9.0

func CudnnRNNBackpropRnnMode(value string) CudnnRNNBackpropAttr

CudnnRNNBackpropRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropSeed added in v1.9.0

func CudnnRNNBackpropSeed(value int64) CudnnRNNBackpropAttr

CudnnRNNBackpropSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropSeed2 added in v1.9.0

func CudnnRNNBackpropSeed2(value int64) CudnnRNNBackpropAttr

CudnnRNNBackpropSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropV2Attr added in v1.11.0

type CudnnRNNBackpropV2Attr func(optionalAttr)

CudnnRNNBackpropV2Attr is an optional argument to CudnnRNNBackpropV2.

func CudnnRNNBackpropV2Direction added in v1.11.0

func CudnnRNNBackpropV2Direction(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropV2Dropout added in v1.11.0

func CudnnRNNBackpropV2Dropout(value float32) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV2InputMode added in v1.11.0

func CudnnRNNBackpropV2InputMode(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropV2RnnMode added in v1.11.0

func CudnnRNNBackpropV2RnnMode(value string) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropV2Seed added in v1.11.0

func CudnnRNNBackpropV2Seed(value int64) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV2Seed2 added in v1.11.0

func CudnnRNNBackpropV2Seed2(value int64) CudnnRNNBackpropV2Attr

CudnnRNNBackpropV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNBackpropV3Attr added in v1.12.1

type CudnnRNNBackpropV3Attr func(optionalAttr)

CudnnRNNBackpropV3Attr is an optional argument to CudnnRNNBackpropV3.

func CudnnRNNBackpropV3Direction added in v1.12.1

func CudnnRNNBackpropV3Direction(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNBackpropV3Dropout added in v1.12.1

func CudnnRNNBackpropV3Dropout(value float32) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3InputMode added in v1.12.1

func CudnnRNNBackpropV3InputMode(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNBackpropV3NumProj added in v1.15.0

func CudnnRNNBackpropV3NumProj(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3RnnMode added in v1.12.1

func CudnnRNNBackpropV3RnnMode(value string) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNBackpropV3Seed added in v1.12.1

func CudnnRNNBackpropV3Seed(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3Seed2 added in v1.12.1

func CudnnRNNBackpropV3Seed2(value int64) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

func CudnnRNNBackpropV3TimeMajor added in v1.12.1

func CudnnRNNBackpropV3TimeMajor(value bool) CudnnRNNBackpropV3Attr

CudnnRNNBackpropV3TimeMajor sets the optional time_major attribute to value. If not specified, defaults to true

type CudnnRNNCanonicalToParamsAttr added in v1.9.0

type CudnnRNNCanonicalToParamsAttr func(optionalAttr)

CudnnRNNCanonicalToParamsAttr is an optional argument to CudnnRNNCanonicalToParams.

func CudnnRNNCanonicalToParamsDirection added in v1.9.0

func CudnnRNNCanonicalToParamsDirection(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNCanonicalToParamsDropout added in v1.9.0

func CudnnRNNCanonicalToParamsDropout(value float32) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsInputMode added in v1.9.0

func CudnnRNNCanonicalToParamsInputMode(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNCanonicalToParamsRnnMode added in v1.9.0

func CudnnRNNCanonicalToParamsRnnMode(value string) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNCanonicalToParamsSeed added in v1.9.0

func CudnnRNNCanonicalToParamsSeed(value int64) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsSeed2 added in v1.9.0

func CudnnRNNCanonicalToParamsSeed2(value int64) CudnnRNNCanonicalToParamsAttr

CudnnRNNCanonicalToParamsSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNCanonicalToParamsV2Attr added in v1.15.0

type CudnnRNNCanonicalToParamsV2Attr func(optionalAttr)

CudnnRNNCanonicalToParamsV2Attr is an optional argument to CudnnRNNCanonicalToParamsV2.

func CudnnRNNCanonicalToParamsV2Direction added in v1.15.0

func CudnnRNNCanonicalToParamsV2Direction(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNCanonicalToParamsV2Dropout added in v1.15.0

func CudnnRNNCanonicalToParamsV2Dropout(value float32) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2InputMode added in v1.15.0

func CudnnRNNCanonicalToParamsV2InputMode(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNCanonicalToParamsV2NumProj added in v1.15.0

func CudnnRNNCanonicalToParamsV2NumProj(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2RnnMode added in v1.15.0

func CudnnRNNCanonicalToParamsV2RnnMode(value string) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNCanonicalToParamsV2Seed added in v1.15.0

func CudnnRNNCanonicalToParamsV2Seed(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNCanonicalToParamsV2Seed2 added in v1.15.0

func CudnnRNNCanonicalToParamsV2Seed2(value int64) CudnnRNNCanonicalToParamsV2Attr

CudnnRNNCanonicalToParamsV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsSizeAttr added in v1.9.0

type CudnnRNNParamsSizeAttr func(optionalAttr)

CudnnRNNParamsSizeAttr is an optional argument to CudnnRNNParamsSize.

func CudnnRNNParamsSizeDirection added in v1.9.0

func CudnnRNNParamsSizeDirection(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsSizeDropout added in v1.9.0

func CudnnRNNParamsSizeDropout(value float32) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeInputMode added in v1.9.0

func CudnnRNNParamsSizeInputMode(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsSizeNumProj added in v1.15.0

func CudnnRNNParamsSizeNumProj(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeNumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeRnnMode added in v1.9.0

func CudnnRNNParamsSizeRnnMode(value string) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsSizeSeed added in v1.9.0

func CudnnRNNParamsSizeSeed(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsSizeSeed2 added in v1.9.0

func CudnnRNNParamsSizeSeed2(value int64) CudnnRNNParamsSizeAttr

CudnnRNNParamsSizeSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsToCanonicalAttr added in v1.9.0

type CudnnRNNParamsToCanonicalAttr func(optionalAttr)

CudnnRNNParamsToCanonicalAttr is an optional argument to CudnnRNNParamsToCanonical.

func CudnnRNNParamsToCanonicalDirection added in v1.9.0

func CudnnRNNParamsToCanonicalDirection(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalDirection sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsToCanonicalDropout added in v1.9.0

func CudnnRNNParamsToCanonicalDropout(value float32) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalDropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalInputMode added in v1.9.0

func CudnnRNNParamsToCanonicalInputMode(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalInputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsToCanonicalRnnMode added in v1.9.0

func CudnnRNNParamsToCanonicalRnnMode(value string) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalRnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsToCanonicalSeed added in v1.9.0

func CudnnRNNParamsToCanonicalSeed(value int64) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalSeed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalSeed2 added in v1.9.0

func CudnnRNNParamsToCanonicalSeed2(value int64) CudnnRNNParamsToCanonicalAttr

CudnnRNNParamsToCanonicalSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNParamsToCanonicalV2Attr added in v1.15.0

type CudnnRNNParamsToCanonicalV2Attr func(optionalAttr)

CudnnRNNParamsToCanonicalV2Attr is an optional argument to CudnnRNNParamsToCanonicalV2.

func CudnnRNNParamsToCanonicalV2Direction added in v1.15.0

func CudnnRNNParamsToCanonicalV2Direction(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNParamsToCanonicalV2Dropout added in v1.15.0

func CudnnRNNParamsToCanonicalV2Dropout(value float32) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2InputMode added in v1.15.0

func CudnnRNNParamsToCanonicalV2InputMode(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNParamsToCanonicalV2NumProj added in v1.15.0

func CudnnRNNParamsToCanonicalV2NumProj(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2RnnMode added in v1.15.0

func CudnnRNNParamsToCanonicalV2RnnMode(value string) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNParamsToCanonicalV2Seed added in v1.15.0

func CudnnRNNParamsToCanonicalV2Seed(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNParamsToCanonicalV2Seed2 added in v1.15.0

func CudnnRNNParamsToCanonicalV2Seed2(value int64) CudnnRNNParamsToCanonicalV2Attr

CudnnRNNParamsToCanonicalV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNV2Attr added in v1.11.0

type CudnnRNNV2Attr func(optionalAttr)

CudnnRNNV2Attr is an optional argument to CudnnRNNV2.

func CudnnRNNV2Direction added in v1.11.0

func CudnnRNNV2Direction(value string) CudnnRNNV2Attr

CudnnRNNV2Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNV2Dropout added in v1.11.0

func CudnnRNNV2Dropout(value float32) CudnnRNNV2Attr

CudnnRNNV2Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNV2InputMode added in v1.11.0

func CudnnRNNV2InputMode(value string) CudnnRNNV2Attr

CudnnRNNV2InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNV2IsTraining added in v1.11.0

func CudnnRNNV2IsTraining(value bool) CudnnRNNV2Attr

CudnnRNNV2IsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNV2RnnMode added in v1.11.0

func CudnnRNNV2RnnMode(value string) CudnnRNNV2Attr

CudnnRNNV2RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNV2Seed added in v1.11.0

func CudnnRNNV2Seed(value int64) CudnnRNNV2Attr

CudnnRNNV2Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNV2Seed2 added in v1.11.0

func CudnnRNNV2Seed2(value int64) CudnnRNNV2Attr

CudnnRNNV2Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type CudnnRNNV3Attr added in v1.12.1

type CudnnRNNV3Attr func(optionalAttr)

CudnnRNNV3Attr is an optional argument to CudnnRNNV3.

func CudnnRNNV3Direction added in v1.12.1

func CudnnRNNV3Direction(value string) CudnnRNNV3Attr

CudnnRNNV3Direction sets the optional direction attribute to value. If not specified, defaults to "unidirectional"

func CudnnRNNV3Dropout added in v1.12.1

func CudnnRNNV3Dropout(value float32) CudnnRNNV3Attr

CudnnRNNV3Dropout sets the optional dropout attribute to value. If not specified, defaults to 0

func CudnnRNNV3InputMode added in v1.12.1

func CudnnRNNV3InputMode(value string) CudnnRNNV3Attr

CudnnRNNV3InputMode sets the optional input_mode attribute to value. If not specified, defaults to "linear_input"

func CudnnRNNV3IsTraining added in v1.12.1

func CudnnRNNV3IsTraining(value bool) CudnnRNNV3Attr

CudnnRNNV3IsTraining sets the optional is_training attribute to value. If not specified, defaults to true

func CudnnRNNV3NumProj added in v1.15.0

func CudnnRNNV3NumProj(value int64) CudnnRNNV3Attr

CudnnRNNV3NumProj sets the optional num_proj attribute to value. If not specified, defaults to 0

func CudnnRNNV3RnnMode added in v1.12.1

func CudnnRNNV3RnnMode(value string) CudnnRNNV3Attr

CudnnRNNV3RnnMode sets the optional rnn_mode attribute to value. If not specified, defaults to "lstm"

func CudnnRNNV3Seed added in v1.12.1

func CudnnRNNV3Seed(value int64) CudnnRNNV3Attr

CudnnRNNV3Seed sets the optional seed attribute to value. If not specified, defaults to 0

func CudnnRNNV3Seed2 added in v1.12.1

func CudnnRNNV3Seed2(value int64) CudnnRNNV3Attr

CudnnRNNV3Seed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

func CudnnRNNV3TimeMajor added in v1.12.1

func CudnnRNNV3TimeMajor(value bool) CudnnRNNV3Attr

CudnnRNNV3TimeMajor sets the optional time_major attribute to value. If not specified, defaults to true

type CumprodAttr added in v1.1.0

type CumprodAttr func(optionalAttr)

CumprodAttr is an optional argument to Cumprod.

func CumprodExclusive added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type CumsumAttr func(optionalAttr)

CumsumAttr is an optional argument to Cumsum.

func CumsumExclusive added in v1.1.0

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 added in v1.1.0

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 DataFormatDimMapAttr added in v1.6.0

type DataFormatDimMapAttr func(optionalAttr)

DataFormatDimMapAttr is an optional argument to DataFormatDimMap.

func DataFormatDimMapDstFormat added in v1.6.0

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 added in v1.6.0

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 added in v1.6.0

type DataFormatVecPermuteAttr func(optionalAttr)

DataFormatVecPermuteAttr is an optional argument to DataFormatVecPermute.

func DataFormatVecPermuteDstFormat added in v1.6.0

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 added in v1.6.0

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 DecodeAndCropJpegAttr added in v1.4.0

type DecodeAndCropJpegAttr func(optionalAttr)

DecodeAndCropJpegAttr is an optional argument to DecodeAndCropJpeg.

func DecodeAndCropJpegAcceptableFraction added in v1.4.0

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 added in v1.4.0

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 added in v1.4.0

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 added in v1.4.0

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 added in v1.4.0

func DecodeAndCropJpegRatio(value int64) DecodeAndCropJpegAttr

DecodeAndCropJpegRatio sets the optional ratio attribute to value.

value: Downscaling ratio. If not specified, defaults to 1

func DecodeAndCropJpegTryRecoverTruncated added in v1.4.0

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 added in v1.3.0

type DecodeBmpAttr func(optionalAttr)

DecodeBmpAttr is an optional argument to DecodeBmp.

func DecodeBmpChannels added in v1.3.0

func DecodeBmpChannels(value int64) DecodeBmpAttr

DecodeBmpChannels sets the optional channels attribute to value. If not specified, defaults to 0

type DecodeCSVAttr added in v1.1.0

type DecodeCSVAttr func(optionalAttr)

DecodeCSVAttr is an optional argument to DecodeCSV.

func DecodeCSVFieldDelim added in v1.1.0

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 added in v1.5.0

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 added in v1.9.0

func DecodeCSVSelectCols(value []int64) DecodeCSVAttr

DecodeCSVSelectCols sets the optional select_cols attribute to value. If not specified, defaults to <>

func DecodeCSVUseQuoteDelim added in v1.3.0

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 added in v1.6.0

type DecodeCompressedAttr func(optionalAttr)

DecodeCompressedAttr is an optional argument to DecodeCompressed.

func DecodeCompressedCompressionType added in v1.6.0

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 DecodeJpegAttr added in v1.1.0

type DecodeJpegAttr func(optionalAttr)

DecodeJpegAttr is an optional argument to DecodeJpeg.

func DecodeJpegAcceptableFraction added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

func DecodeJpegRatio(value int64) DecodeJpegAttr

DecodeJpegRatio sets the optional ratio attribute to value.

value: Downscaling ratio. If not specified, defaults to 1

func DecodeJpegTryRecoverTruncated added in v1.1.0

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 added in v1.14.0

type DecodePaddedRawAttr func(optionalAttr)

DecodePaddedRawAttr is an optional argument to DecodePaddedRaw.

func DecodePaddedRawLittleEndian added in v1.14.0

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 added in v1.1.0

type DecodePngAttr func(optionalAttr)

DecodePngAttr is an optional argument to DecodePng.

func DecodePngChannels added in v1.1.0

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 added in v1.1.0

func DecodePngDtype(value tf.DataType) DecodePngAttr

DecodePngDtype sets the optional dtype attribute to value. If not specified, defaults to DT_UINT8

type DecodeProtoV2Attr added in v1.9.0

type DecodeProtoV2Attr func(optionalAttr)

DecodeProtoV2Attr is an optional argument to DecodeProtoV2.

func DecodeProtoV2DescriptorSource added in v1.9.0

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 added in v1.9.0

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 added in v1.9.0

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 added in v1.1.0

type DecodeRawAttr func(optionalAttr)

DecodeRawAttr is an optional argument to DecodeRaw.

func DecodeRawLittleEndian added in v1.1.0

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 added in v1.2.0

type DecodeWavAttr func(optionalAttr)

DecodeWavAttr is an optional argument to DecodeWav.

func DecodeWavDesiredChannels added in v1.2.0

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 added in v1.2.0

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 DenseToDenseSetOperationAttr added in v1.1.0

type DenseToDenseSetOperationAttr func(optionalAttr)

DenseToDenseSetOperationAttr is an optional argument to DenseToDenseSetOperation.

func DenseToDenseSetOperationValidateIndices added in v1.1.0

func DenseToDenseSetOperationValidateIndices(value bool) DenseToDenseSetOperationAttr

DenseToDenseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type DenseToSparseSetOperationAttr added in v1.1.0

type DenseToSparseSetOperationAttr func(optionalAttr)

DenseToSparseSetOperationAttr is an optional argument to DenseToSparseSetOperation.

func DenseToSparseSetOperationValidateIndices added in v1.1.0

func DenseToSparseSetOperationValidateIndices(value bool) DenseToSparseSetOperationAttr

DenseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type DepthToSpaceAttr added in v1.4.0

type DepthToSpaceAttr func(optionalAttr)

DepthToSpaceAttr is an optional argument to DepthToSpace.

func DepthToSpaceDataFormat added in v1.4.0

func DepthToSpaceDataFormat(value string) DepthToSpaceAttr

DepthToSpaceDataFormat sets the optional data_format attribute to value. If not specified, defaults to "NHWC"

type DepthwiseConv2dNativeAttr added in v1.1.0

type DepthwiseConv2dNativeAttr func(optionalAttr)

DepthwiseConv2dNativeAttr is an optional argument to DepthwiseConv2dNative.

func DepthwiseConv2dNativeDataFormat added in v1.1.0

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 added in v1.6.0

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 >

type DepthwiseConv2dNativeBackpropFilterAttr added in v1.1.0

type DepthwiseConv2dNativeBackpropFilterAttr func(optionalAttr)

DepthwiseConv2dNativeBackpropFilterAttr is an optional argument to DepthwiseConv2dNativeBackpropFilter.

func DepthwiseConv2dNativeBackpropFilterDataFormat added in v1.1.0

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 added in v1.6.0

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 >

type DepthwiseConv2dNativeBackpropInputAttr added in v1.1.0

type DepthwiseConv2dNativeBackpropInputAttr func(optionalAttr)

DepthwiseConv2dNativeBackpropInputAttr is an optional argument to DepthwiseConv2dNativeBackpropInput.

func DepthwiseConv2dNativeBackpropInputDataFormat added in v1.1.0

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 added in v1.6.0

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 >

type DequantizeAttr added in v1.1.0

type DequantizeAttr func(optionalAttr)

DequantizeAttr is an optional argument to Dequantize.

func DequantizeMode added in v1.1.0

func DequantizeMode(value string) DequantizeAttr

DequantizeMode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"

type DestroyResourceOpAttr added in v1.1.0

type DestroyResourceOpAttr func(optionalAttr)

DestroyResourceOpAttr is an optional argument to DestroyResourceOp.

func DestroyResourceOpIgnoreLookupError added in v1.1.0

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 EditDistanceAttr added in v1.1.0

type EditDistanceAttr func(optionalAttr)

EditDistanceAttr is an optional argument to EditDistance.

func EditDistanceNormalize added in v1.1.0

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 EmptyAttr added in v1.9.0

type EmptyAttr func(optionalAttr)

EmptyAttr is an optional argument to Empty.

func EmptyInit added in v1.9.0

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 added in v1.1.0

type EncodeBase64Attr func(optionalAttr)

EncodeBase64Attr is an optional argument to EncodeBase64.

func EncodeBase64Pad added in v1.1.0

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 added in v1.1.0

type EncodeJpegAttr func(optionalAttr)

EncodeJpegAttr is an optional argument to EncodeJpeg.

func EncodeJpegChromaDownsampling added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type EncodePngAttr func(optionalAttr)

EncodePngAttr is an optional argument to EncodePng.

func EncodePngCompression added in v1.1.0

func EncodePngCompression(value int64) EncodePngAttr

EncodePngCompression sets the optional compression attribute to value.

value: Compression level. If not specified, defaults to -1

type EncodeProtoAttr added in v1.9.0

type EncodeProtoAttr func(optionalAttr)

EncodeProtoAttr is an optional argument to EncodeProto.

func EncodeProtoDescriptorSource added in v1.9.0

func EncodeProtoDescriptorSource(value string) EncodeProtoAttr

EncodeProtoDescriptorSource sets the optional descriptor_source attribute to value. If not specified, defaults to "local://"

type EnqueueTPUEmbeddingIntegerBatchAttr added in v1.12.1

type EnqueueTPUEmbeddingIntegerBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingIntegerBatchAttr is an optional argument to EnqueueTPUEmbeddingIntegerBatch.

func EnqueueTPUEmbeddingIntegerBatchDeviceOrdinal added in v1.12.1

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 EnqueueTPUEmbeddingSparseBatchAttr added in v1.12.1

type EnqueueTPUEmbeddingSparseBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingSparseBatchAttr is an optional argument to EnqueueTPUEmbeddingSparseBatch.

func EnqueueTPUEmbeddingSparseBatchCombiners added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type EnqueueTPUEmbeddingSparseTensorBatchAttr func(optionalAttr)

EnqueueTPUEmbeddingSparseTensorBatchAttr is an optional argument to EnqueueTPUEmbeddingSparseTensorBatch.

func EnqueueTPUEmbeddingSparseTensorBatchCombiners added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

func EnqueueTPUEmbeddingSparseTensorBatchMaxSequenceLengths(value []int64) EnqueueTPUEmbeddingSparseTensorBatchAttr

EnqueueTPUEmbeddingSparseTensorBatchMaxSequenceLengths sets the optional max_sequence_lengths attribute to value. If not specified, defaults to <>

type EnterAttr added in v1.1.0

type EnterAttr func(optionalAttr)

EnterAttr is an optional argument to Enter.

func EnterIsConstant added in v1.1.0

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 added in v1.1.0

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 EuclideanNormAttr added in v1.12.1

type EuclideanNormAttr func(optionalAttr)

EuclideanNormAttr is an optional argument to EuclideanNorm.

func EuclideanNormKeepDims added in v1.12.1

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 ExperimentalParseExampleDatasetAttr added in v1.12.1

type ExperimentalParseExampleDatasetAttr func(optionalAttr)

ExperimentalParseExampleDatasetAttr is an optional argument to ExperimentalParseExampleDataset.

func ExperimentalParseExampleDatasetSloppy added in v1.12.1

func ExperimentalParseExampleDatasetSloppy(value bool) ExperimentalParseExampleDatasetAttr

ExperimentalParseExampleDatasetSloppy sets the optional sloppy attribute to value. If not specified, defaults to false

type ExperimentalStatsAggregatorHandleAttr added in v1.12.1

type ExperimentalStatsAggregatorHandleAttr func(optionalAttr)

ExperimentalStatsAggregatorHandleAttr is an optional argument to ExperimentalStatsAggregatorHandle.

func ExperimentalStatsAggregatorHandleContainer added in v1.12.1

func ExperimentalStatsAggregatorHandleContainer(value string) ExperimentalStatsAggregatorHandleAttr

ExperimentalStatsAggregatorHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ExperimentalStatsAggregatorHandleSharedName added in v1.12.1

func ExperimentalStatsAggregatorHandleSharedName(value string) ExperimentalStatsAggregatorHandleAttr

ExperimentalStatsAggregatorHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ExperimentalThreadPoolHandleAttr added in v1.12.1

type ExperimentalThreadPoolHandleAttr func(optionalAttr)

ExperimentalThreadPoolHandleAttr is an optional argument to ExperimentalThreadPoolHandle.

func ExperimentalThreadPoolHandleContainer added in v1.12.1

func ExperimentalThreadPoolHandleContainer(value string) ExperimentalThreadPoolHandleAttr

ExperimentalThreadPoolHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ExperimentalThreadPoolHandleMaxIntraOpParallelism added in v1.12.1

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 added in v1.12.1

func ExperimentalThreadPoolHandleSharedName(value string) ExperimentalThreadPoolHandleAttr

ExperimentalThreadPoolHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ExtractGlimpseAttr added in v1.1.0

type ExtractGlimpseAttr func(optionalAttr)

ExtractGlimpseAttr is an optional argument to ExtractGlimpse.

func ExtractGlimpseCentered added in v1.1.0

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 added in v1.12.1

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 the noise type will be decided by `uniform_noise`. If not specified, defaults to "uniform"

func ExtractGlimpseNormalized added in v1.1.0

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 added in v1.1.0

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 ExtractJpegShapeAttr added in v1.4.0

type ExtractJpegShapeAttr func(optionalAttr)

ExtractJpegShapeAttr is an optional argument to ExtractJpegShape.

func ExtractJpegShapeOutputType added in v1.4.0

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 added in v1.1.0

type FIFOQueueV2Attr func(optionalAttr)

FIFOQueueV2Attr is an optional argument to FIFOQueueV2.

func FIFOQueueV2Capacity added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type FakeQuantWithMinMaxArgsAttr func(optionalAttr)

FakeQuantWithMinMaxArgsAttr is an optional argument to FakeQuantWithMinMaxArgs.

func FakeQuantWithMinMaxArgsMax added in v1.1.0

func FakeQuantWithMinMaxArgsMax(value float32) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsMax sets the optional max attribute to value. If not specified, defaults to 6

func FakeQuantWithMinMaxArgsMin added in v1.1.0

func FakeQuantWithMinMaxArgsMin(value float32) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsMin sets the optional min attribute to value. If not specified, defaults to -6

func FakeQuantWithMinMaxArgsNarrowRange added in v1.3.0

func FakeQuantWithMinMaxArgsNarrowRange(value bool) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxArgsNumBits added in v1.2.0

func FakeQuantWithMinMaxArgsNumBits(value int64) FakeQuantWithMinMaxArgsAttr

FakeQuantWithMinMaxArgsNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxArgsGradientAttr added in v1.1.0

type FakeQuantWithMinMaxArgsGradientAttr func(optionalAttr)

FakeQuantWithMinMaxArgsGradientAttr is an optional argument to FakeQuantWithMinMaxArgsGradient.

func FakeQuantWithMinMaxArgsGradientMax added in v1.1.0

func FakeQuantWithMinMaxArgsGradientMax(value float32) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientMax sets the optional max attribute to value. If not specified, defaults to 6

func FakeQuantWithMinMaxArgsGradientMin added in v1.1.0

func FakeQuantWithMinMaxArgsGradientMin(value float32) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientMin sets the optional min attribute to value. If not specified, defaults to -6

func FakeQuantWithMinMaxArgsGradientNarrowRange added in v1.3.0

func FakeQuantWithMinMaxArgsGradientNarrowRange(value bool) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxArgsGradientNumBits added in v1.2.0

func FakeQuantWithMinMaxArgsGradientNumBits(value int64) FakeQuantWithMinMaxArgsGradientAttr

FakeQuantWithMinMaxArgsGradientNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsAttr added in v1.2.0

type FakeQuantWithMinMaxVarsAttr func(optionalAttr)

FakeQuantWithMinMaxVarsAttr is an optional argument to FakeQuantWithMinMaxVars.

func FakeQuantWithMinMaxVarsNarrowRange added in v1.3.0

func FakeQuantWithMinMaxVarsNarrowRange(value bool) FakeQuantWithMinMaxVarsAttr

FakeQuantWithMinMaxVarsNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxVarsNumBits added in v1.2.0

func FakeQuantWithMinMaxVarsNumBits(value int64) FakeQuantWithMinMaxVarsAttr

FakeQuantWithMinMaxVarsNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsGradientAttr added in v1.2.0

type FakeQuantWithMinMaxVarsGradientAttr func(optionalAttr)

FakeQuantWithMinMaxVarsGradientAttr is an optional argument to FakeQuantWithMinMaxVarsGradient.

func FakeQuantWithMinMaxVarsGradientNarrowRange added in v1.3.0

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 added in v1.2.0

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 added in v1.2.0

type FakeQuantWithMinMaxVarsPerChannelAttr func(optionalAttr)

FakeQuantWithMinMaxVarsPerChannelAttr is an optional argument to FakeQuantWithMinMaxVarsPerChannel.

func FakeQuantWithMinMaxVarsPerChannelNarrowRange added in v1.3.0

func FakeQuantWithMinMaxVarsPerChannelNarrowRange(value bool) FakeQuantWithMinMaxVarsPerChannelAttr

FakeQuantWithMinMaxVarsPerChannelNarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func FakeQuantWithMinMaxVarsPerChannelNumBits added in v1.2.0

func FakeQuantWithMinMaxVarsPerChannelNumBits(value int64) FakeQuantWithMinMaxVarsPerChannelAttr

FakeQuantWithMinMaxVarsPerChannelNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

type FakeQuantWithMinMaxVarsPerChannelGradientAttr added in v1.2.0

type FakeQuantWithMinMaxVarsPerChannelGradientAttr func(optionalAttr)

FakeQuantWithMinMaxVarsPerChannelGradientAttr is an optional argument to FakeQuantWithMinMaxVarsPerChannelGradient.

func FakeQuantWithMinMaxVarsPerChannelGradientNarrowRange added in v1.3.0

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 added in v1.2.0

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 FixedLengthRecordReaderV2Attr added in v1.1.0

type FixedLengthRecordReaderV2Attr func(optionalAttr)

FixedLengthRecordReaderV2Attr is an optional argument to FixedLengthRecordReaderV2.

func FixedLengthRecordReaderV2Container added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

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 added in v1.1.0

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 added in v1.1.0

type FixedUnigramCandidateSamplerAttr func(optionalAttr)

FixedUnigramCandidateSamplerAttr is an optional argument to FixedUnigramCandidateSampler.

func FixedUnigramCandidateSamplerDistortion added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type FractionalAvgPoolAttr func(optionalAttr)

FractionalAvgPoolAttr is an optional argument to FractionalAvgPool.

func FractionalAvgPoolDeterministic added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type FractionalAvgPoolGradAttr func(optionalAttr)

FractionalAvgPoolGradAttr is an optional argument to FractionalAvgPoolGrad.

func FractionalAvgPoolGradOverlapping added in v1.1.0

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 added in v1.1.0

type FractionalMaxPoolAttr func(optionalAttr)

FractionalMaxPoolAttr is an optional argument to FractionalMaxPool.

func FractionalMaxPoolDeterministic added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type FractionalMaxPoolGradAttr func(optionalAttr)

FractionalMaxPoolGradAttr is an optional argument to FractionalMaxPoolGrad.

func FractionalMaxPoolGradOverlapping added in v1.1.0

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 added in v1.1.0

type FusedBatchNormAttr func(optionalAttr)

FusedBatchNormAttr is an optional argument to FusedBatchNorm.

func FusedBatchNormDataFormat added in v1.1.0

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 added in v1.1.0

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 FusedBatchNormIsTraining added in v1.1.0

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 added in v1.1.0

type FusedBatchNormGradAttr func(optionalAttr)

FusedBatchNormGradAttr is an optional argument to FusedBatchNormGrad.

func FusedBatchNormGradDataFormat added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.4.0

type FusedBatchNormGradV2Attr func(optionalAttr)

FusedBatchNormGradV2Attr is an optional argument to FusedBatchNormGradV2.

func FusedBatchNormGradV2DataFormat added in v1.4.0

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 added in v1.4.0

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 added in v1.4.0

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 added in v1.14.0

type FusedBatchNormGradV3Attr func(optionalAttr)

FusedBatchNormGradV3Attr is an optional argument to FusedBatchNormGradV3.

func FusedBatchNormGradV3DataFormat added in v1.14.0

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 added in v1.14.0

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 added in v1.14.0

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 added in v1.4.0

type FusedBatchNormV2Attr func(optionalAttr)

FusedBatchNormV2Attr is an optional argument to FusedBatchNormV2.

func FusedBatchNormV2DataFormat added in v1.4.0

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 added in v1.4.0

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 FusedBatchNormV2IsTraining added in v1.4.0

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 added in v1.14.0

type FusedBatchNormV3Attr func(optionalAttr)

FusedBatchNormV3Attr is an optional argument to FusedBatchNormV3.

func FusedBatchNormV3DataFormat added in v1.14.0

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 added in v1.14.0

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 FusedBatchNormV3IsTraining added in v1.14.0

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 added in v1.1.0

type FusedResizeAndPadConv2DAttr func(optionalAttr)

FusedResizeAndPadConv2DAttr is an optional argument to FusedResizeAndPadConv2D.

func FusedResizeAndPadConv2DResizeAlignCorners added in v1.1.0

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 added in v1.1.0

type GatherAttr func(optionalAttr)

GatherAttr is an optional argument to Gather.

func GatherValidateIndices added in v1.1.0

func GatherValidateIndices(value bool) GatherAttr

GatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type GatherV2Attr added in v1.14.0

type GatherV2Attr func(optionalAttr)

GatherV2Attr is an optional argument to GatherV2.

func GatherV2BatchDims added in v1.14.0

func GatherV2BatchDims(value int64) GatherV2Attr

GatherV2BatchDims sets the optional batch_dims attribute to value. If not specified, defaults to 0

type GenerateVocabRemappingAttr added in v1.5.0

type GenerateVocabRemappingAttr func(optionalAttr)

GenerateVocabRemappingAttr is an optional argument to GenerateVocabRemapping.

func GenerateVocabRemappingOldVocabSize added in v1.5.0

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 added in v1.2.0

type HashTableV2Attr func(optionalAttr)

HashTableV2Attr is an optional argument to HashTableV2.

func HashTableV2Container added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.5.0

type HistogramFixedWidthAttr func(optionalAttr)

HistogramFixedWidthAttr is an optional argument to HistogramFixedWidth.

func HistogramFixedWidthDtype added in v1.5.0

func HistogramFixedWidthDtype(value tf.DataType) HistogramFixedWidthAttr

HistogramFixedWidthDtype sets the optional dtype attribute to value. If not specified, defaults to DT_INT32

type IdentityReaderV2Attr added in v1.1.0

type IdentityReaderV2Attr func(optionalAttr)

IdentityReaderV2Attr is an optional argument to IdentityReaderV2.

func IdentityReaderV2Container added in v1.1.0

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 added in v1.1.0

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 ImagAttr added in v1.1.0

type ImagAttr func(optionalAttr)

ImagAttr is an optional argument to Imag.

func ImagTout added in v1.1.0

func ImagTout(value tf.DataType) ImagAttr

ImagTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type ImageSummaryAttr added in v1.1.0

type ImageSummaryAttr func(optionalAttr)

ImageSummaryAttr is an optional argument to ImageSummary.

func ImageSummaryBadColor added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

type InfeedEnqueueAttr func(optionalAttr)

InfeedEnqueueAttr is an optional argument to InfeedEnqueue.

func InfeedEnqueueDeviceOrdinal added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type InfeedEnqueuePrelinearizedBufferAttr func(optionalAttr)

InfeedEnqueuePrelinearizedBufferAttr is an optional argument to InfeedEnqueuePrelinearizedBuffer.

func InfeedEnqueuePrelinearizedBufferDeviceOrdinal added in v1.12.1

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 added in v1.12.1

type InfeedEnqueueTupleAttr func(optionalAttr)

InfeedEnqueueTupleAttr is an optional argument to InfeedEnqueueTuple.

func InfeedEnqueueTupleDeviceOrdinal added in v1.12.1

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 added in v1.12.1

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 added in v1.2.0

type InitializeTableFromTextFileV2Attr func(optionalAttr)

InitializeTableFromTextFileV2Attr is an optional argument to InitializeTableFromTextFileV2.

func InitializeTableFromTextFileV2Delimiter added in v1.2.0

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 InitializeTableFromTextFileV2VocabSize added in v1.2.0

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 IteratorFromStringHandleAttr added in v1.4.0

type IteratorFromStringHandleAttr func(optionalAttr)

IteratorFromStringHandleAttr is an optional argument to IteratorFromStringHandle.

func IteratorFromStringHandleOutputShapes added in v1.4.0

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 added in v1.4.0

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 added in v1.1.0

type LRNAttr func(optionalAttr)

LRNAttr is an optional argument to LRN.

func LRNAlpha added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type LRNGradAttr func(optionalAttr)

LRNGradAttr is an optional argument to LRNGrad.

func LRNGradAlpha added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

type LSTMBlockCellAttr func(optionalAttr)

LSTMBlockCellAttr is an optional argument to LSTMBlockCell.

func LSTMBlockCellCellClip added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

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 added in v1.12.1

type LeakyReluAttr func(optionalAttr)

LeakyReluAttr is an optional argument to LeakyRelu.

func LeakyReluAlpha added in v1.12.1

func LeakyReluAlpha(value float32) LeakyReluAttr

LeakyReluAlpha sets the optional alpha attribute to value. If not specified, defaults to 0.2

type LeakyReluGradAttr added in v1.12.1

type LeakyReluGradAttr func(optionalAttr)

LeakyReluGradAttr is an optional argument to LeakyReluGrad.

func LeakyReluGradAlpha added in v1.12.1

func LeakyReluGradAlpha(value float32) LeakyReluGradAttr

LeakyReluGradAlpha sets the optional alpha attribute to value. If not specified, defaults to 0.2

type LearnedUnigramCandidateSamplerAttr added in v1.1.0

type LearnedUnigramCandidateSamplerAttr func(optionalAttr)

LearnedUnigramCandidateSamplerAttr is an optional argument to LearnedUnigramCandidateSampler.

func LearnedUnigramCandidateSamplerSeed added in v1.1.0

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 added in v1.1.0

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 ListDiffAttr added in v1.1.0

type ListDiffAttr func(optionalAttr)

ListDiffAttr is an optional argument to ListDiff.

func ListDiffOutIdx added in v1.1.0

func ListDiffOutIdx(value tf.DataType) ListDiffAttr

ListDiffOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type LoadAndRemapMatrixAttr added in v1.4.0

type LoadAndRemapMatrixAttr func(optionalAttr)

LoadAndRemapMatrixAttr is an optional argument to LoadAndRemapMatrix.

func LoadAndRemapMatrixMaxRowsInMemory added in v1.4.0

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 added in v1.12.1

type LoadTPUEmbeddingADAMParametersAttr func(optionalAttr)

LoadTPUEmbeddingADAMParametersAttr is an optional argument to LoadTPUEmbeddingADAMParameters.

func LoadTPUEmbeddingADAMParametersTableId added in v1.12.1

func LoadTPUEmbeddingADAMParametersTableId(value int64) LoadTPUEmbeddingADAMParametersAttr

LoadTPUEmbeddingADAMParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingADAMParametersTableName added in v1.12.1

func LoadTPUEmbeddingADAMParametersTableName(value string) LoadTPUEmbeddingADAMParametersAttr

LoadTPUEmbeddingADAMParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingADAMParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingADAMParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingADAMParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingADAMParametersGradAccumDebug.

func LoadTPUEmbeddingADAMParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingADAMParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingADAMParametersGradAccumDebugAttr

LoadTPUEmbeddingADAMParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingADAMParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingADAMParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingADAMParametersGradAccumDebugAttr

LoadTPUEmbeddingADAMParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdadeltaParametersAttr added in v1.12.1

type LoadTPUEmbeddingAdadeltaParametersAttr func(optionalAttr)

LoadTPUEmbeddingAdadeltaParametersAttr is an optional argument to LoadTPUEmbeddingAdadeltaParameters.

func LoadTPUEmbeddingAdadeltaParametersTableId added in v1.12.1

func LoadTPUEmbeddingAdadeltaParametersTableId(value int64) LoadTPUEmbeddingAdadeltaParametersAttr

LoadTPUEmbeddingAdadeltaParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingAdadeltaParametersTableName added in v1.12.1

func LoadTPUEmbeddingAdadeltaParametersTableName(value string) LoadTPUEmbeddingAdadeltaParametersAttr

LoadTPUEmbeddingAdadeltaParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingAdadeltaParametersGradAccumDebug.

func LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr

LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingAdadeltaParametersGradAccumDebugAttr

LoadTPUEmbeddingAdadeltaParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdagradParametersAttr added in v1.12.1

type LoadTPUEmbeddingAdagradParametersAttr func(optionalAttr)

LoadTPUEmbeddingAdagradParametersAttr is an optional argument to LoadTPUEmbeddingAdagradParameters.

func LoadTPUEmbeddingAdagradParametersTableId added in v1.12.1

func LoadTPUEmbeddingAdagradParametersTableId(value int64) LoadTPUEmbeddingAdagradParametersAttr

LoadTPUEmbeddingAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingAdagradParametersTableName added in v1.12.1

func LoadTPUEmbeddingAdagradParametersTableName(value string) LoadTPUEmbeddingAdagradParametersAttr

LoadTPUEmbeddingAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingAdagradParametersGradAccumDebug.

func LoadTPUEmbeddingAdagradParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingAdagradParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr

LoadTPUEmbeddingAdagradParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingAdagradParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingAdagradParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingAdagradParametersGradAccumDebugAttr

LoadTPUEmbeddingAdagradParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingCenteredRMSPropParametersAttr added in v1.12.1

type LoadTPUEmbeddingCenteredRMSPropParametersAttr func(optionalAttr)

LoadTPUEmbeddingCenteredRMSPropParametersAttr is an optional argument to LoadTPUEmbeddingCenteredRMSPropParameters.

func LoadTPUEmbeddingCenteredRMSPropParametersTableId added in v1.12.1

func LoadTPUEmbeddingCenteredRMSPropParametersTableId(value int64) LoadTPUEmbeddingCenteredRMSPropParametersAttr

LoadTPUEmbeddingCenteredRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingCenteredRMSPropParametersTableName added in v1.12.1

func LoadTPUEmbeddingCenteredRMSPropParametersTableName(value string) LoadTPUEmbeddingCenteredRMSPropParametersAttr

LoadTPUEmbeddingCenteredRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingFTRLParametersAttr added in v1.12.1

type LoadTPUEmbeddingFTRLParametersAttr func(optionalAttr)

LoadTPUEmbeddingFTRLParametersAttr is an optional argument to LoadTPUEmbeddingFTRLParameters.

func LoadTPUEmbeddingFTRLParametersTableId added in v1.12.1

func LoadTPUEmbeddingFTRLParametersTableId(value int64) LoadTPUEmbeddingFTRLParametersAttr

LoadTPUEmbeddingFTRLParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingFTRLParametersTableName added in v1.12.1

func LoadTPUEmbeddingFTRLParametersTableName(value string) LoadTPUEmbeddingFTRLParametersAttr

LoadTPUEmbeddingFTRLParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingFTRLParametersGradAccumDebug.

func LoadTPUEmbeddingFTRLParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingFTRLParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr

LoadTPUEmbeddingFTRLParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingFTRLParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingFTRLParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingFTRLParametersGradAccumDebugAttr

LoadTPUEmbeddingFTRLParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingMDLAdagradLightParametersAttr added in v1.12.1

type LoadTPUEmbeddingMDLAdagradLightParametersAttr func(optionalAttr)

LoadTPUEmbeddingMDLAdagradLightParametersAttr is an optional argument to LoadTPUEmbeddingMDLAdagradLightParameters.

func LoadTPUEmbeddingMDLAdagradLightParametersTableId added in v1.12.1

func LoadTPUEmbeddingMDLAdagradLightParametersTableId(value int64) LoadTPUEmbeddingMDLAdagradLightParametersAttr

LoadTPUEmbeddingMDLAdagradLightParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingMDLAdagradLightParametersTableName added in v1.12.1

func LoadTPUEmbeddingMDLAdagradLightParametersTableName(value string) LoadTPUEmbeddingMDLAdagradLightParametersAttr

LoadTPUEmbeddingMDLAdagradLightParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingMomentumParametersAttr added in v1.12.1

type LoadTPUEmbeddingMomentumParametersAttr func(optionalAttr)

LoadTPUEmbeddingMomentumParametersAttr is an optional argument to LoadTPUEmbeddingMomentumParameters.

func LoadTPUEmbeddingMomentumParametersTableId added in v1.12.1

func LoadTPUEmbeddingMomentumParametersTableId(value int64) LoadTPUEmbeddingMomentumParametersAttr

LoadTPUEmbeddingMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingMomentumParametersTableName added in v1.12.1

func LoadTPUEmbeddingMomentumParametersTableName(value string) LoadTPUEmbeddingMomentumParametersAttr

LoadTPUEmbeddingMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingMomentumParametersGradAccumDebug.

func LoadTPUEmbeddingMomentumParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingMomentumParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr

LoadTPUEmbeddingMomentumParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingMomentumParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingMomentumParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingMomentumParametersGradAccumDebugAttr

LoadTPUEmbeddingMomentumParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingProximalAdagradParametersAttr added in v1.12.1

type LoadTPUEmbeddingProximalAdagradParametersAttr func(optionalAttr)

LoadTPUEmbeddingProximalAdagradParametersAttr is an optional argument to LoadTPUEmbeddingProximalAdagradParameters.

func LoadTPUEmbeddingProximalAdagradParametersTableId added in v1.12.1

func LoadTPUEmbeddingProximalAdagradParametersTableId(value int64) LoadTPUEmbeddingProximalAdagradParametersAttr

LoadTPUEmbeddingProximalAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingProximalAdagradParametersTableName added in v1.12.1

func LoadTPUEmbeddingProximalAdagradParametersTableName(value string) LoadTPUEmbeddingProximalAdagradParametersAttr

LoadTPUEmbeddingProximalAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingProximalAdagradParametersGradAccumDebug.

func LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr

LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr

LoadTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingRMSPropParametersAttr added in v1.12.1

type LoadTPUEmbeddingRMSPropParametersAttr func(optionalAttr)

LoadTPUEmbeddingRMSPropParametersAttr is an optional argument to LoadTPUEmbeddingRMSPropParameters.

func LoadTPUEmbeddingRMSPropParametersTableId added in v1.12.1

func LoadTPUEmbeddingRMSPropParametersTableId(value int64) LoadTPUEmbeddingRMSPropParametersAttr

LoadTPUEmbeddingRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingRMSPropParametersTableName added in v1.12.1

func LoadTPUEmbeddingRMSPropParametersTableName(value string) LoadTPUEmbeddingRMSPropParametersAttr

LoadTPUEmbeddingRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr added in v1.12.1

type LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr func(optionalAttr)

LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr is an optional argument to LoadTPUEmbeddingRMSPropParametersGradAccumDebug.

func LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableId added in v1.12.1

func LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableId(value int64) LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr

LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableName added in v1.12.1

func LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableName(value string) LoadTPUEmbeddingRMSPropParametersGradAccumDebugAttr

LoadTPUEmbeddingRMSPropParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LoadTPUEmbeddingStochasticGradientDescentParametersAttr added in v1.12.1

type LoadTPUEmbeddingStochasticGradientDescentParametersAttr func(optionalAttr)

LoadTPUEmbeddingStochasticGradientDescentParametersAttr is an optional argument to LoadTPUEmbeddingStochasticGradientDescentParameters.

func LoadTPUEmbeddingStochasticGradientDescentParametersTableId added in v1.12.1

func LoadTPUEmbeddingStochasticGradientDescentParametersTableId(value int64) LoadTPUEmbeddingStochasticGradientDescentParametersAttr

LoadTPUEmbeddingStochasticGradientDescentParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func LoadTPUEmbeddingStochasticGradientDescentParametersTableName added in v1.12.1

func LoadTPUEmbeddingStochasticGradientDescentParametersTableName(value string) LoadTPUEmbeddingStochasticGradientDescentParametersAttr

LoadTPUEmbeddingStochasticGradientDescentParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type LogUniformCandidateSamplerAttr added in v1.1.0

type LogUniformCandidateSamplerAttr func(optionalAttr)

LogUniformCandidateSamplerAttr is an optional argument to LogUniformCandidateSampler.

func LogUniformCandidateSamplerSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.12.0

type LowerBoundAttr func(optionalAttr)

LowerBoundAttr is an optional argument to LowerBound.

func LowerBoundOutType added in v1.12.0

func LowerBoundOutType(value tf.DataType) LowerBoundAttr

LowerBoundOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type LuAttr added in v1.12.1

type LuAttr func(optionalAttr)

LuAttr is an optional argument to Lu.

func LuOutputIdxType added in v1.12.1

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 added in v1.3.0

type MapClearAttr func(optionalAttr)

MapClearAttr is an optional argument to MapClear.

func MapClearCapacity added in v1.3.0

func MapClearCapacity(value int64) MapClearAttr

MapClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapClearContainer added in v1.3.0

func MapClearContainer(value string) MapClearAttr

MapClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapClearMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapClearSharedName(value string) MapClearAttr

MapClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapIncompleteSizeAttr added in v1.3.0

type MapIncompleteSizeAttr func(optionalAttr)

MapIncompleteSizeAttr is an optional argument to MapIncompleteSize.

func MapIncompleteSizeCapacity added in v1.3.0

func MapIncompleteSizeCapacity(value int64) MapIncompleteSizeAttr

MapIncompleteSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapIncompleteSizeContainer added in v1.3.0

func MapIncompleteSizeContainer(value string) MapIncompleteSizeAttr

MapIncompleteSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapIncompleteSizeMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapIncompleteSizeSharedName(value string) MapIncompleteSizeAttr

MapIncompleteSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapPeekAttr added in v1.3.0

type MapPeekAttr func(optionalAttr)

MapPeekAttr is an optional argument to MapPeek.

func MapPeekCapacity added in v1.3.0

func MapPeekCapacity(value int64) MapPeekAttr

MapPeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapPeekContainer added in v1.3.0

func MapPeekContainer(value string) MapPeekAttr

MapPeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapPeekMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapPeekSharedName(value string) MapPeekAttr

MapPeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapSizeAttr added in v1.3.0

type MapSizeAttr func(optionalAttr)

MapSizeAttr is an optional argument to MapSize.

func MapSizeCapacity added in v1.3.0

func MapSizeCapacity(value int64) MapSizeAttr

MapSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapSizeContainer added in v1.3.0

func MapSizeContainer(value string) MapSizeAttr

MapSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapSizeMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapSizeSharedName(value string) MapSizeAttr

MapSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapStageAttr added in v1.3.0

type MapStageAttr func(optionalAttr)

MapStageAttr is an optional argument to MapStage.

func MapStageCapacity added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

type MapUnstageAttr func(optionalAttr)

MapUnstageAttr is an optional argument to MapUnstage.

func MapUnstageCapacity added in v1.3.0

func MapUnstageCapacity(value int64) MapUnstageAttr

MapUnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageContainer added in v1.3.0

func MapUnstageContainer(value string) MapUnstageAttr

MapUnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapUnstageMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapUnstageSharedName(value string) MapUnstageAttr

MapUnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MapUnstageNoKeyAttr added in v1.3.0

type MapUnstageNoKeyAttr func(optionalAttr)

MapUnstageNoKeyAttr is an optional argument to MapUnstageNoKey.

func MapUnstageNoKeyCapacity added in v1.3.0

func MapUnstageNoKeyCapacity(value int64) MapUnstageNoKeyAttr

MapUnstageNoKeyCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func MapUnstageNoKeyContainer added in v1.3.0

func MapUnstageNoKeyContainer(value string) MapUnstageNoKeyAttr

MapUnstageNoKeyContainer sets the optional container attribute to value. If not specified, defaults to ""

func MapUnstageNoKeyMemoryLimit added in v1.3.0

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 added in v1.3.0

func MapUnstageNoKeySharedName(value string) MapUnstageNoKeyAttr

MapUnstageNoKeySharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type MatMulAttr added in v1.1.0

type MatMulAttr func(optionalAttr)

MatMulAttr is an optional argument to MatMul.

func MatMulTransposeA added in v1.1.0

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 added in v1.1.0

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 MatrixInverseAttr added in v1.1.0

type MatrixInverseAttr func(optionalAttr)

MatrixInverseAttr is an optional argument to MatrixInverse.

func MatrixInverseAdjoint added in v1.1.0

func MatrixInverseAdjoint(value bool) MatrixInverseAttr

MatrixInverseAdjoint sets the optional adjoint attribute to value. If not specified, defaults to false

type MatrixSolveAttr added in v1.1.0

type MatrixSolveAttr func(optionalAttr)

MatrixSolveAttr is an optional argument to MatrixSolve.

func MatrixSolveAdjoint added in v1.1.0

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 added in v1.1.0

type MatrixSolveLsAttr func(optionalAttr)

MatrixSolveLsAttr is an optional argument to MatrixSolveLs.

func MatrixSolveLsFast added in v1.1.0

func MatrixSolveLsFast(value bool) MatrixSolveLsAttr

MatrixSolveLsFast sets the optional fast attribute to value. If not specified, defaults to true

type MatrixTriangularSolveAttr added in v1.1.0

type MatrixTriangularSolveAttr func(optionalAttr)

MatrixTriangularSolveAttr is an optional argument to MatrixTriangularSolve.

func MatrixTriangularSolveAdjoint added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type MaxAttr func(optionalAttr)

MaxAttr is an optional argument to Max.

func MaxKeepDims added in v1.1.0

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 added in v1.2.0

type MaxPool3DAttr func(optionalAttr)

MaxPool3DAttr is an optional argument to MaxPool3D.

func MaxPool3DDataFormat added in v1.2.0

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 added in v1.2.0

type MaxPool3DGradAttr func(optionalAttr)

MaxPool3DGradAttr is an optional argument to MaxPool3DGrad.

func MaxPool3DGradDataFormat added in v1.2.0

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 added in v1.2.0

type MaxPool3DGradGradAttr func(optionalAttr)

MaxPool3DGradGradAttr is an optional argument to MaxPool3DGradGrad.

func MaxPool3DGradGradDataFormat added in v1.2.0

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 added in v1.1.0

type MaxPoolAttr func(optionalAttr)

MaxPoolAttr is an optional argument to MaxPool.

func MaxPoolDataFormat added in v1.1.0

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"

type MaxPoolGradAttr added in v1.1.0

type MaxPoolGradAttr func(optionalAttr)

MaxPoolGradAttr is an optional argument to MaxPoolGrad.

func MaxPoolGradDataFormat added in v1.1.0

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"

type MaxPoolGradGradAttr added in v1.2.0

type MaxPoolGradGradAttr func(optionalAttr)

MaxPoolGradGradAttr is an optional argument to MaxPoolGradGrad.

func MaxPoolGradGradDataFormat added in v1.2.0

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 added in v1.4.0

type MaxPoolGradGradV2Attr func(optionalAttr)

MaxPoolGradGradV2Attr is an optional argument to MaxPoolGradGradV2.

func MaxPoolGradGradV2DataFormat added in v1.4.0

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 added in v1.12.1

type MaxPoolGradGradWithArgmaxAttr func(optionalAttr)

MaxPoolGradGradWithArgmaxAttr is an optional argument to MaxPoolGradGradWithArgmax.

func MaxPoolGradGradWithArgmaxIncludeBatchInIndex added in v1.12.1

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 added in v1.4.0

type MaxPoolGradV2Attr func(optionalAttr)

MaxPoolGradV2Attr is an optional argument to MaxPoolGradV2.

func MaxPoolGradV2DataFormat added in v1.4.0

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 added in v1.12.1

type MaxPoolGradWithArgmaxAttr func(optionalAttr)

MaxPoolGradWithArgmaxAttr is an optional argument to MaxPoolGradWithArgmax.

func MaxPoolGradWithArgmaxIncludeBatchInIndex added in v1.12.1

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 added in v1.4.0

type MaxPoolV2Attr func(optionalAttr)

MaxPoolV2Attr is an optional argument to MaxPoolV2.

func MaxPoolV2DataFormat added in v1.4.0

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 added in v1.1.0

type MaxPoolWithArgmaxAttr func(optionalAttr)

MaxPoolWithArgmaxAttr is an optional argument to MaxPoolWithArgmax.

func MaxPoolWithArgmaxIncludeBatchInIndex added in v1.12.1

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 added in v1.1.0

func MaxPoolWithArgmaxTargmax(value tf.DataType) MaxPoolWithArgmaxAttr

MaxPoolWithArgmaxTargmax sets the optional Targmax attribute to value. If not specified, defaults to DT_INT64

type MeanAttr added in v1.1.0

type MeanAttr func(optionalAttr)

MeanAttr is an optional argument to Mean.

func MeanKeepDims added in v1.1.0

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 MergeV2CheckpointsAttr added in v1.1.0

type MergeV2CheckpointsAttr func(optionalAttr)

MergeV2CheckpointsAttr is an optional argument to MergeV2Checkpoints.

func MergeV2CheckpointsDeleteOldDirs added in v1.1.0

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 added in v1.2.0

type MfccAttr func(optionalAttr)

MfccAttr is an optional argument to Mfcc.

func MfccDctCoefficientCount added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.1.0

type MinAttr func(optionalAttr)

MinAttr is an optional argument to Min.

func MinKeepDims added in v1.1.0

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 added in v1.12.1

type ModelDatasetAttr func(optionalAttr)

ModelDatasetAttr is an optional argument to ModelDataset.

func ModelDatasetAlgorithm added in v1.15.0

func ModelDatasetAlgorithm(value int64) ModelDatasetAttr

ModelDatasetAlgorithm sets the optional algorithm attribute to value. If not specified, defaults to 0

func ModelDatasetCpuBudget added in v1.12.1

func ModelDatasetCpuBudget(value int64) ModelDatasetAttr

ModelDatasetCpuBudget sets the optional cpu_budget attribute to value. If not specified, defaults to 0

type MultiDeviceIteratorFromStringHandleAttr added in v1.12.0

type MultiDeviceIteratorFromStringHandleAttr func(optionalAttr)

MultiDeviceIteratorFromStringHandleAttr is an optional argument to MultiDeviceIteratorFromStringHandle.

func MultiDeviceIteratorFromStringHandleOutputShapes added in v1.12.0

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 added in v1.12.0

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 added in v1.1.0

type MultinomialAttr func(optionalAttr)

MultinomialAttr is an optional argument to Multinomial.

func MultinomialOutputDtype added in v1.6.0

func MultinomialOutputDtype(value tf.DataType) MultinomialAttr

MultinomialOutputDtype sets the optional output_dtype attribute to value. If not specified, defaults to DT_INT64

func MultinomialSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.2.0

type MutableDenseHashTableV2Attr func(optionalAttr)

MutableDenseHashTableV2Attr is an optional argument to MutableDenseHashTableV2.

func MutableDenseHashTableV2Container added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

func MutableDenseHashTableV2UseNodeNameSharing(value bool) MutableDenseHashTableV2Attr

MutableDenseHashTableV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value. If not specified, defaults to false

func MutableDenseHashTableV2ValueShape added in v1.2.0

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 added in v1.2.0

type MutableHashTableOfTensorsV2Attr func(optionalAttr)

MutableHashTableOfTensorsV2Attr is an optional argument to MutableHashTableOfTensorsV2.

func MutableHashTableOfTensorsV2Container added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

func MutableHashTableOfTensorsV2UseNodeNameSharing(value bool) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2UseNodeNameSharing sets the optional use_node_name_sharing attribute to value. If not specified, defaults to false

func MutableHashTableOfTensorsV2ValueShape added in v1.2.0

func MutableHashTableOfTensorsV2ValueShape(value tf.Shape) MutableHashTableOfTensorsV2Attr

MutableHashTableOfTensorsV2ValueShape sets the optional value_shape attribute to value. If not specified, defaults to <>

type MutableHashTableV2Attr added in v1.2.0

type MutableHashTableV2Attr func(optionalAttr)

MutableHashTableV2Attr is an optional argument to MutableHashTableV2.

func MutableHashTableV2Container added in v1.2.0

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 added in v1.2.0

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 added in v1.2.0

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 added in v1.9.0

type MutexV2Attr func(optionalAttr)

MutexV2Attr is an optional argument to MutexV2.

func MutexV2Container added in v1.9.0

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 added in v1.9.0

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 added in v1.12.1

type NonDeterministicIntsAttr func(optionalAttr)

NonDeterministicIntsAttr is an optional argument to NonDeterministicInts.

func NonDeterministicIntsDtype added in v1.12.1

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 added in v1.1.0

type NonMaxSuppressionAttr func(optionalAttr)

NonMaxSuppressionAttr is an optional argument to NonMaxSuppression.

func NonMaxSuppressionIouThreshold added in v1.1.0

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 added in v1.11.0

type NonMaxSuppressionV4Attr func(optionalAttr)

NonMaxSuppressionV4Attr is an optional argument to NonMaxSuppressionV4.

func NonMaxSuppressionV4PadToMaxOutputSize added in v1.11.0

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 added in v1.15.0

type NonMaxSuppressionV5Attr func(optionalAttr)

NonMaxSuppressionV5Attr is an optional argument to NonMaxSuppressionV5.

func NonMaxSuppressionV5PadToMaxOutputSize added in v1.15.0

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 NthElementAttr added in v1.5.0

type NthElementAttr func(optionalAttr)

NthElementAttr is an optional argument to NthElement.

func NthElementReverse added in v1.5.0

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 added in v1.1.0

type OneHotAttr func(optionalAttr)

OneHotAttr is an optional argument to OneHot.

func OneHotAxis added in v1.1.0

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 added in v1.12.1

type OptimizeDatasetAttr func(optionalAttr)

OptimizeDatasetAttr is an optional argument to OptimizeDataset.

func OptimizeDatasetOptimizationConfigs added in v1.12.1

func OptimizeDatasetOptimizationConfigs(value []string) OptimizeDatasetAttr

OptimizeDatasetOptimizationConfigs sets the optional optimization_configs attribute to value. If not specified, defaults to <>

type OrderedMapClearAttr added in v1.3.0

type OrderedMapClearAttr func(optionalAttr)

OrderedMapClearAttr is an optional argument to OrderedMapClear.

func OrderedMapClearCapacity added in v1.3.0

func OrderedMapClearCapacity(value int64) OrderedMapClearAttr

OrderedMapClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapClearContainer added in v1.3.0

func OrderedMapClearContainer(value string) OrderedMapClearAttr

OrderedMapClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapClearMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapClearSharedName(value string) OrderedMapClearAttr

OrderedMapClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapIncompleteSizeAttr added in v1.3.0

type OrderedMapIncompleteSizeAttr func(optionalAttr)

OrderedMapIncompleteSizeAttr is an optional argument to OrderedMapIncompleteSize.

func OrderedMapIncompleteSizeCapacity added in v1.3.0

func OrderedMapIncompleteSizeCapacity(value int64) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapIncompleteSizeContainer added in v1.3.0

func OrderedMapIncompleteSizeContainer(value string) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapIncompleteSizeMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapIncompleteSizeSharedName(value string) OrderedMapIncompleteSizeAttr

OrderedMapIncompleteSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapPeekAttr added in v1.3.0

type OrderedMapPeekAttr func(optionalAttr)

OrderedMapPeekAttr is an optional argument to OrderedMapPeek.

func OrderedMapPeekCapacity added in v1.3.0

func OrderedMapPeekCapacity(value int64) OrderedMapPeekAttr

OrderedMapPeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapPeekContainer added in v1.3.0

func OrderedMapPeekContainer(value string) OrderedMapPeekAttr

OrderedMapPeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapPeekMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapPeekSharedName(value string) OrderedMapPeekAttr

OrderedMapPeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapSizeAttr added in v1.3.0

type OrderedMapSizeAttr func(optionalAttr)

OrderedMapSizeAttr is an optional argument to OrderedMapSize.

func OrderedMapSizeCapacity added in v1.3.0

func OrderedMapSizeCapacity(value int64) OrderedMapSizeAttr

OrderedMapSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapSizeContainer added in v1.3.0

func OrderedMapSizeContainer(value string) OrderedMapSizeAttr

OrderedMapSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapSizeMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapSizeSharedName(value string) OrderedMapSizeAttr

OrderedMapSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapStageAttr added in v1.3.0

type OrderedMapStageAttr func(optionalAttr)

OrderedMapStageAttr is an optional argument to OrderedMapStage.

func OrderedMapStageCapacity added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

type OrderedMapUnstageAttr func(optionalAttr)

OrderedMapUnstageAttr is an optional argument to OrderedMapUnstage.

func OrderedMapUnstageCapacity added in v1.3.0

func OrderedMapUnstageCapacity(value int64) OrderedMapUnstageAttr

OrderedMapUnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageContainer added in v1.3.0

func OrderedMapUnstageContainer(value string) OrderedMapUnstageAttr

OrderedMapUnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapUnstageMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapUnstageSharedName(value string) OrderedMapUnstageAttr

OrderedMapUnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OrderedMapUnstageNoKeyAttr added in v1.3.0

type OrderedMapUnstageNoKeyAttr func(optionalAttr)

OrderedMapUnstageNoKeyAttr is an optional argument to OrderedMapUnstageNoKey.

func OrderedMapUnstageNoKeyCapacity added in v1.3.0

func OrderedMapUnstageNoKeyCapacity(value int64) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeyCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func OrderedMapUnstageNoKeyContainer added in v1.3.0

func OrderedMapUnstageNoKeyContainer(value string) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeyContainer sets the optional container attribute to value. If not specified, defaults to ""

func OrderedMapUnstageNoKeyMemoryLimit added in v1.3.0

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 added in v1.3.0

func OrderedMapUnstageNoKeySharedName(value string) OrderedMapUnstageNoKeyAttr

OrderedMapUnstageNoKeySharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type OutfeedDequeueAttr added in v1.12.1

type OutfeedDequeueAttr func(optionalAttr)

OutfeedDequeueAttr is an optional argument to OutfeedDequeue.

func OutfeedDequeueDeviceOrdinal added in v1.12.1

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 added in v1.12.1

type OutfeedDequeueTupleAttr func(optionalAttr)

OutfeedDequeueTupleAttr is an optional argument to OutfeedDequeueTuple.

func OutfeedDequeueTupleDeviceOrdinal added in v1.12.1

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 added in v1.1.0

type PackAttr func(optionalAttr)

PackAttr is an optional argument to Pack.

func PackAxis added in v1.1.0

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 PaddedBatchDatasetV2Attr added in v1.14.0

type PaddedBatchDatasetV2Attr func(optionalAttr)

PaddedBatchDatasetV2Attr is an optional argument to PaddedBatchDatasetV2.

func PaddedBatchDatasetV2ParallelCopy added in v1.14.0

func PaddedBatchDatasetV2ParallelCopy(value bool) PaddedBatchDatasetV2Attr

PaddedBatchDatasetV2ParallelCopy sets the optional parallel_copy attribute to value. If not specified, defaults to false

type PaddingFIFOQueueV2Attr added in v1.1.0

type PaddingFIFOQueueV2Attr func(optionalAttr)

PaddingFIFOQueueV2Attr is an optional argument to PaddingFIFOQueueV2.

func PaddingFIFOQueueV2Capacity added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type ParameterizedTruncatedNormalAttr func(optionalAttr)

ParameterizedTruncatedNormalAttr is an optional argument to ParameterizedTruncatedNormal.

func ParameterizedTruncatedNormalSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

type ParseExampleDatasetAttr func(optionalAttr)

ParseExampleDatasetAttr is an optional argument to ParseExampleDataset.

func ParseExampleDatasetSloppy added in v1.15.0

func ParseExampleDatasetSloppy(value bool) ParseExampleDatasetAttr

ParseExampleDatasetSloppy sets the optional sloppy attribute to value. If not specified, defaults to false

type ParseSequenceExampleAttr added in v1.12.0

type ParseSequenceExampleAttr func(optionalAttr)

ParseSequenceExampleAttr is an optional argument to ParseSequenceExample.

func ParseSequenceExampleContextDenseShapes added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 ParseSingleSequenceExampleAttr added in v1.1.0

type ParseSingleSequenceExampleAttr func(optionalAttr)

ParseSingleSequenceExampleAttr is an optional argument to ParseSingleSequenceExample.

func ParseSingleSequenceExampleContextDenseShapes added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type PlaceholderAttr func(optionalAttr)

PlaceholderAttr is an optional argument to Placeholder.

func PlaceholderShape added in v1.1.0

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 added in v1.14.0

type PrefetchDatasetAttr func(optionalAttr)

PrefetchDatasetAttr is an optional argument to PrefetchDataset.

func PrefetchDatasetSlackPeriod added in v1.14.0

func PrefetchDatasetSlackPeriod(value int64) PrefetchDatasetAttr

PrefetchDatasetSlackPeriod sets the optional slack_period attribute to value. If not specified, defaults to 0

type PrelinearizeAttr added in v1.12.1

type PrelinearizeAttr func(optionalAttr)

PrelinearizeAttr is an optional argument to Prelinearize.

func PrelinearizeLayout added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type PrelinearizeTupleAttr func(optionalAttr)

PrelinearizeTupleAttr is an optional argument to PrelinearizeTuple.

func PrelinearizeTupleLayouts added in v1.12.1

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 added in v1.1.0

type PreventGradientAttr func(optionalAttr)

PreventGradientAttr is an optional argument to PreventGradient.

func PreventGradientMessage added in v1.1.0

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 added in v1.1.0

type PrintAttr func(optionalAttr)

PrintAttr is an optional argument to Print.

func PrintFirstN added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.0

type PrintV2Attr func(optionalAttr)

PrintV2Attr is an optional argument to PrintV2.

func PrintV2End added in v1.14.0

func PrintV2End(value string) PrintV2Attr

PrintV2End sets the optional end attribute to value. If not specified, defaults to "\n"

func PrintV2OutputStream added in v1.12.0

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 added in v1.1.0

type PriorityQueueV2Attr func(optionalAttr)

PriorityQueueV2Attr is an optional argument to PriorityQueueV2.

func PriorityQueueV2Capacity added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type ProdAttr func(optionalAttr)

ProdAttr is an optional argument to Prod.

func ProdKeepDims added in v1.1.0

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 added in v1.1.0

type QrAttr func(optionalAttr)

QrAttr is an optional argument to Qr.

func QrFullMatrices added in v1.1.0

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 added in v1.1.0

type QuantizeAndDequantizeAttr func(optionalAttr)

QuantizeAndDequantizeAttr is an optional argument to QuantizeAndDequantize.

func QuantizeAndDequantizeInputMax added in v1.1.0

func QuantizeAndDequantizeInputMax(value float32) QuantizeAndDequantizeAttr

QuantizeAndDequantizeInputMax sets the optional input_max attribute to value. If not specified, defaults to 0

func QuantizeAndDequantizeInputMin added in v1.1.0

func QuantizeAndDequantizeInputMin(value float32) QuantizeAndDequantizeAttr

QuantizeAndDequantizeInputMin sets the optional input_min attribute to value. If not specified, defaults to 0

func QuantizeAndDequantizeNumBits added in v1.1.0

func QuantizeAndDequantizeNumBits(value int64) QuantizeAndDequantizeAttr

QuantizeAndDequantizeNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8

func QuantizeAndDequantizeRangeGiven added in v1.1.0

func QuantizeAndDequantizeRangeGiven(value bool) QuantizeAndDequantizeAttr

QuantizeAndDequantizeRangeGiven sets the optional range_given attribute to value. If not specified, defaults to false

func QuantizeAndDequantizeSignedInput added in v1.1.0

func QuantizeAndDequantizeSignedInput(value bool) QuantizeAndDequantizeAttr

QuantizeAndDequantizeSignedInput sets the optional signed_input attribute to value. If not specified, defaults to true

type QuantizeAndDequantizeV2Attr added in v1.1.0

type QuantizeAndDequantizeV2Attr func(optionalAttr)

QuantizeAndDequantizeV2Attr is an optional argument to QuantizeAndDequantizeV2.

func QuantizeAndDequantizeV2NarrowRange added in v1.15.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.12.1

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 added in v1.1.0

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 added in v1.3.0

type QuantizeAndDequantizeV3Attr func(optionalAttr)

QuantizeAndDequantizeV3Attr is an optional argument to QuantizeAndDequantizeV3.

func QuantizeAndDequantizeV3NarrowRange added in v1.15.0

func QuantizeAndDequantizeV3NarrowRange(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3NarrowRange sets the optional narrow_range attribute to value. If not specified, defaults to false

func QuantizeAndDequantizeV3RangeGiven added in v1.3.0

func QuantizeAndDequantizeV3RangeGiven(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3RangeGiven sets the optional range_given attribute to value. If not specified, defaults to true

func QuantizeAndDequantizeV3SignedInput added in v1.3.0

func QuantizeAndDequantizeV3SignedInput(value bool) QuantizeAndDequantizeV3Attr

QuantizeAndDequantizeV3SignedInput sets the optional signed_input attribute to value. If not specified, defaults to true

type QuantizeV2Attr added in v1.1.0

type QuantizeV2Attr func(optionalAttr)

QuantizeV2Attr is an optional argument to QuantizeV2.

func QuantizeV2Mode added in v1.1.0

func QuantizeV2Mode(value string) QuantizeV2Attr

QuantizeV2Mode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"

func QuantizeV2RoundMode added in v1.5.0

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 added in v1.3.0

type QuantizedAddAttr func(optionalAttr)

QuantizedAddAttr is an optional argument to QuantizedAdd.

func QuantizedAddToutput added in v1.3.0

func QuantizedAddToutput(value tf.DataType) QuantizedAddAttr

QuantizedAddToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

type QuantizedConv2DAttr added in v1.1.0

type QuantizedConv2DAttr func(optionalAttr)

QuantizedConv2DAttr is an optional argument to QuantizedConv2D.

func QuantizedConv2DDilations added in v1.6.0

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 added in v1.1.0

func QuantizedConv2DOutType(value tf.DataType) QuantizedConv2DAttr

QuantizedConv2DOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QINT32

type QuantizedConv2DPerChannelAttr added in v1.12.1

type QuantizedConv2DPerChannelAttr func(optionalAttr)

QuantizedConv2DPerChannelAttr is an optional argument to QuantizedConv2DPerChannel.

func QuantizedConv2DPerChannelDilations added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type QuantizedDepthwiseConv2DAttr func(optionalAttr)

QuantizedDepthwiseConv2DAttr is an optional argument to QuantizedDepthwiseConv2D.

func QuantizedDepthwiseConv2DDilations added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeAttr is an optional argument to QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize.

func QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeDilations added in v1.12.1

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 added in v1.12.1

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

type QuantizedDepthwiseConv2DWithBiasAndReluAttr added in v1.12.1

type QuantizedDepthwiseConv2DWithBiasAndReluAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAndReluAttr is an optional argument to QuantizedDepthwiseConv2DWithBiasAndRelu.

func QuantizedDepthwiseConv2DWithBiasAndReluDilations added in v1.12.1

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 added in v1.12.1

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

type QuantizedDepthwiseConv2DWithBiasAttr added in v1.12.1

type QuantizedDepthwiseConv2DWithBiasAttr func(optionalAttr)

QuantizedDepthwiseConv2DWithBiasAttr is an optional argument to QuantizedDepthwiseConv2DWithBias.

func QuantizedDepthwiseConv2DWithBiasDilations added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

type QuantizedInstanceNormAttr func(optionalAttr)

QuantizedInstanceNormAttr is an optional argument to QuantizedInstanceNorm.

func QuantizedInstanceNormGivenYMax added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type QuantizedMatMulAttr func(optionalAttr)

QuantizedMatMulAttr is an optional argument to QuantizedMatMul.

func QuantizedMatMulTactivation added in v1.1.0

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 added in v1.1.0

func QuantizedMatMulToutput(value tf.DataType) QuantizedMatMulAttr

QuantizedMatMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulTransposeA added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

type QuantizedMatMulWithBiasAndReluAndRequantizeAttr func(optionalAttr)

QuantizedMatMulWithBiasAndReluAndRequantizeAttr is an optional argument to QuantizedMatMulWithBiasAndReluAndRequantize.

func QuantizedMatMulWithBiasAndReluAndRequantizeInputQuantMode added in v1.15.0

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 added in v1.15.0

func QuantizedMatMulWithBiasAndReluAndRequantizeToutput(value tf.DataType) QuantizedMatMulWithBiasAndReluAndRequantizeAttr

QuantizedMatMulWithBiasAndReluAndRequantizeToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QUINT8

func QuantizedMatMulWithBiasAndReluAndRequantizeTransposeA added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

type QuantizedMatMulWithBiasAndReluAttr func(optionalAttr)

QuantizedMatMulWithBiasAndReluAttr is an optional argument to QuantizedMatMulWithBiasAndRelu.

func QuantizedMatMulWithBiasAndReluInputQuantMode added in v1.15.0

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 added in v1.15.0

func QuantizedMatMulWithBiasAndReluToutput(value tf.DataType) QuantizedMatMulWithBiasAndReluAttr

QuantizedMatMulWithBiasAndReluToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulWithBiasAndReluTransposeA added in v1.15.0

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 added in v1.15.0

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 added in v1.15.0

type QuantizedMatMulWithBiasAttr func(optionalAttr)

QuantizedMatMulWithBiasAttr is an optional argument to QuantizedMatMulWithBias.

func QuantizedMatMulWithBiasInputQuantMode added in v1.15.0

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 added in v1.15.0

func QuantizedMatMulWithBiasToutput(value tf.DataType) QuantizedMatMulWithBiasAttr

QuantizedMatMulWithBiasToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

func QuantizedMatMulWithBiasTransposeA added in v1.15.0

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 added in v1.15.0

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 added in v1.1.0

type QuantizedMulAttr func(optionalAttr)

QuantizedMulAttr is an optional argument to QuantizedMul.

func QuantizedMulToutput added in v1.1.0

func QuantizedMulToutput(value tf.DataType) QuantizedMulAttr

QuantizedMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32

type QuantizedRelu6Attr added in v1.1.0

type QuantizedRelu6Attr func(optionalAttr)

QuantizedRelu6Attr is an optional argument to QuantizedRelu6.

func QuantizedRelu6OutType added in v1.1.0

func QuantizedRelu6OutType(value tf.DataType) QuantizedRelu6Attr

QuantizedRelu6OutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedReluAttr added in v1.1.0

type QuantizedReluAttr func(optionalAttr)

QuantizedReluAttr is an optional argument to QuantizedRelu.

func QuantizedReluOutType added in v1.1.0

func QuantizedReluOutType(value tf.DataType) QuantizedReluAttr

QuantizedReluOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedReluXAttr added in v1.1.0

type QuantizedReluXAttr func(optionalAttr)

QuantizedReluXAttr is an optional argument to QuantizedReluX.

func QuantizedReluXOutType added in v1.1.0

func QuantizedReluXOutType(value tf.DataType) QuantizedReluXAttr

QuantizedReluXOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8

type QuantizedResizeBilinearAttr added in v1.3.0

type QuantizedResizeBilinearAttr func(optionalAttr)

QuantizedResizeBilinearAttr is an optional argument to QuantizedResizeBilinear.

func QuantizedResizeBilinearAlignCorners added in v1.3.0

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 added in v1.12.1

func QuantizedResizeBilinearHalfPixelCenters(value bool) QuantizedResizeBilinearAttr

QuantizedResizeBilinearHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type QueueCloseV2Attr added in v1.1.0

type QueueCloseV2Attr func(optionalAttr)

QueueCloseV2Attr is an optional argument to QueueCloseV2.

func QueueCloseV2CancelPendingEnqueues added in v1.1.0

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 added in v1.1.0

type QueueDequeueManyV2Attr func(optionalAttr)

QueueDequeueManyV2Attr is an optional argument to QueueDequeueManyV2.

func QueueDequeueManyV2TimeoutMs added in v1.1.0

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 added in v1.1.0

type QueueDequeueUpToV2Attr func(optionalAttr)

QueueDequeueUpToV2Attr is an optional argument to QueueDequeueUpToV2.

func QueueDequeueUpToV2TimeoutMs added in v1.1.0

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 added in v1.1.0

type QueueDequeueV2Attr func(optionalAttr)

QueueDequeueV2Attr is an optional argument to QueueDequeueV2.

func QueueDequeueV2TimeoutMs added in v1.1.0

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 added in v1.1.0

type QueueEnqueueManyV2Attr func(optionalAttr)

QueueEnqueueManyV2Attr is an optional argument to QueueEnqueueManyV2.

func QueueEnqueueManyV2TimeoutMs added in v1.1.0

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 added in v1.1.0

type QueueEnqueueV2Attr func(optionalAttr)

QueueEnqueueV2Attr is an optional argument to QueueEnqueueV2.

func QueueEnqueueV2TimeoutMs added in v1.1.0

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 RaggedRangeAttr added in v1.14.0

type RaggedRangeAttr func(optionalAttr)

RaggedRangeAttr is an optional argument to RaggedRange.

func RaggedRangeTsplits added in v1.14.0

func RaggedRangeTsplits(value tf.DataType) RaggedRangeAttr

RaggedRangeTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type RandomCropAttr added in v1.1.0

type RandomCropAttr func(optionalAttr)

RandomCropAttr is an optional argument to RandomCrop.

func RandomCropSeed added in v1.1.0

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 added in v1.1.0

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 RandomGammaAttr added in v1.1.0

type RandomGammaAttr func(optionalAttr)

RandomGammaAttr is an optional argument to RandomGamma.

func RandomGammaSeed added in v1.1.0

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 added in v1.1.0

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 RandomPoissonAttr added in v1.1.0

type RandomPoissonAttr func(optionalAttr)

RandomPoissonAttr is an optional argument to RandomPoisson.

func RandomPoissonSeed added in v1.1.0

func RandomPoissonSeed(value int64) RandomPoissonAttr

RandomPoissonSeed sets the optional seed attribute to value. If not specified, defaults to 0

func RandomPoissonSeed2 added in v1.1.0

func RandomPoissonSeed2(value int64) RandomPoissonAttr

RandomPoissonSeed2 sets the optional seed2 attribute to value. If not specified, defaults to 0

type RandomPoissonV2Attr added in v1.4.0

type RandomPoissonV2Attr func(optionalAttr)

RandomPoissonV2Attr is an optional argument to RandomPoissonV2.

func RandomPoissonV2Dtype added in v1.4.0

func RandomPoissonV2Dtype(value tf.DataType) RandomPoissonV2Attr

RandomPoissonV2Dtype sets the optional dtype attribute to value. If not specified, defaults to DT_INT64

func RandomPoissonV2Seed added in v1.4.0

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 added in v1.4.0

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 added in v1.1.0

type RandomShuffleAttr func(optionalAttr)

RandomShuffleAttr is an optional argument to RandomShuffle.

func RandomShuffleSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type RandomShuffleQueueV2Attr func(optionalAttr)

RandomShuffleQueueV2Attr is an optional argument to RandomShuffleQueueV2.

func RandomShuffleQueueV2Capacity added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type RandomStandardNormalAttr func(optionalAttr)

RandomStandardNormalAttr is an optional argument to RandomStandardNormal.

func RandomStandardNormalSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type RandomUniformAttr func(optionalAttr)

RandomUniformAttr is an optional argument to RandomUniform.

func RandomUniformSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type RandomUniformIntAttr func(optionalAttr)

RandomUniformIntAttr is an optional argument to RandomUniformInt.

func RandomUniformIntSeed added in v1.1.0

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 added in v1.1.0

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 RealAttr added in v1.1.0

type RealAttr func(optionalAttr)

RealAttr is an optional argument to Real.

func RealTout added in v1.1.0

func RealTout(value tf.DataType) RealAttr

RealTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT

type RecordInputAttr added in v1.1.0

type RecordInputAttr func(optionalAttr)

RecordInputAttr is an optional argument to RecordInput.

func RecordInputBatchSize added in v1.1.0

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 added in v1.6.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 ReduceJoinAttr added in v1.1.0

type ReduceJoinAttr func(optionalAttr)

ReduceJoinAttr is an optional argument to ReduceJoin.

func ReduceJoinKeepDims added in v1.1.0

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 added in v1.1.0

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 added in v1.9.0

type RegexReplaceAttr func(optionalAttr)

RegexReplaceAttr is an optional argument to RegexReplace.

func RegexReplaceReplaceGlobal added in v1.9.0

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 RequantizePerChannelAttr added in v1.12.1

type RequantizePerChannelAttr func(optionalAttr)

RequantizePerChannelAttr is an optional argument to RequantizePerChannel.

func RequantizePerChannelOutType added in v1.12.1

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 added in v1.1.0

type ResizeAreaAttr func(optionalAttr)

ResizeAreaAttr is an optional argument to ResizeArea.

func ResizeAreaAlignCorners added in v1.1.0

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 added in v1.1.0

type ResizeBicubicAttr func(optionalAttr)

ResizeBicubicAttr is an optional argument to ResizeBicubic.

func ResizeBicubicAlignCorners added in v1.1.0

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 added in v1.12.1

func ResizeBicubicHalfPixelCenters(value bool) ResizeBicubicAttr

ResizeBicubicHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBicubicGradAttr added in v1.4.0

type ResizeBicubicGradAttr func(optionalAttr)

ResizeBicubicGradAttr is an optional argument to ResizeBicubicGrad.

func ResizeBicubicGradAlignCorners added in v1.4.0

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 added in v1.12.1

func ResizeBicubicGradHalfPixelCenters(value bool) ResizeBicubicGradAttr

ResizeBicubicGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBilinearAttr added in v1.1.0

type ResizeBilinearAttr func(optionalAttr)

ResizeBilinearAttr is an optional argument to ResizeBilinear.

func ResizeBilinearAlignCorners added in v1.1.0

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 added in v1.12.1

func ResizeBilinearHalfPixelCenters(value bool) ResizeBilinearAttr

ResizeBilinearHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeBilinearGradAttr added in v1.1.0

type ResizeBilinearGradAttr func(optionalAttr)

ResizeBilinearGradAttr is an optional argument to ResizeBilinearGrad.

func ResizeBilinearGradAlignCorners added in v1.1.0

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 added in v1.12.1

func ResizeBilinearGradHalfPixelCenters(value bool) ResizeBilinearGradAttr

ResizeBilinearGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeNearestNeighborAttr added in v1.1.0

type ResizeNearestNeighborAttr func(optionalAttr)

ResizeNearestNeighborAttr is an optional argument to ResizeNearestNeighbor.

func ResizeNearestNeighborAlignCorners added in v1.1.0

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 added in v1.12.1

func ResizeNearestNeighborHalfPixelCenters(value bool) ResizeNearestNeighborAttr

ResizeNearestNeighborHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResizeNearestNeighborGradAttr added in v1.1.0

type ResizeNearestNeighborGradAttr func(optionalAttr)

ResizeNearestNeighborGradAttr is an optional argument to ResizeNearestNeighborGrad.

func ResizeNearestNeighborGradAlignCorners added in v1.1.0

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 added in v1.12.1

func ResizeNearestNeighborGradHalfPixelCenters(value bool) ResizeNearestNeighborGradAttr

ResizeNearestNeighborGradHalfPixelCenters sets the optional half_pixel_centers attribute to value. If not specified, defaults to false

type ResourceApplyAdaMaxAttr added in v1.11.0

type ResourceApplyAdaMaxAttr func(optionalAttr)

ResourceApplyAdaMaxAttr is an optional argument to ResourceApplyAdaMax.

func ResourceApplyAdaMaxUseLocking added in v1.11.0

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 added in v1.1.0

type ResourceApplyAdadeltaAttr func(optionalAttr)

ResourceApplyAdadeltaAttr is an optional argument to ResourceApplyAdadelta.

func ResourceApplyAdadeltaUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceApplyAdagradAttr func(optionalAttr)

ResourceApplyAdagradAttr is an optional argument to ResourceApplyAdagrad.

func ResourceApplyAdagradUpdateSlots added in v1.9.0

func ResourceApplyAdagradUpdateSlots(value bool) ResourceApplyAdagradAttr

ResourceApplyAdagradUpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceApplyAdagradUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceApplyAdagradDAAttr func(optionalAttr)

ResourceApplyAdagradDAAttr is an optional argument to ResourceApplyAdagradDA.

func ResourceApplyAdagradDAUseLocking added in v1.1.0

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 ResourceApplyAdamAttr added in v1.1.0

type ResourceApplyAdamAttr func(optionalAttr)

ResourceApplyAdamAttr is an optional argument to ResourceApplyAdam.

func ResourceApplyAdamUseLocking added in v1.1.0

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 added in v1.3.0

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 added in v1.12.1

type ResourceApplyAdamWithAmsgradAttr func(optionalAttr)

ResourceApplyAdamWithAmsgradAttr is an optional argument to ResourceApplyAdamWithAmsgrad.

func ResourceApplyAdamWithAmsgradUseLocking added in v1.12.1

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 added in v1.5.0

type ResourceApplyAddSignAttr func(optionalAttr)

ResourceApplyAddSignAttr is an optional argument to ResourceApplyAddSign.

func ResourceApplyAddSignUseLocking added in v1.5.0

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 added in v1.1.0

type ResourceApplyCenteredRMSPropAttr func(optionalAttr)

ResourceApplyCenteredRMSPropAttr is an optional argument to ResourceApplyCenteredRMSProp.

func ResourceApplyCenteredRMSPropUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceApplyFtrlAttr func(optionalAttr)

ResourceApplyFtrlAttr is an optional argument to ResourceApplyFtrl.

func ResourceApplyFtrlUseLocking added in v1.1.0

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 added in v1.3.0

type ResourceApplyFtrlV2Attr func(optionalAttr)

ResourceApplyFtrlV2Attr is an optional argument to ResourceApplyFtrlV2.

func ResourceApplyFtrlV2UseLocking added in v1.3.0

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 added in v1.1.0

type ResourceApplyGradientDescentAttr func(optionalAttr)

ResourceApplyGradientDescentAttr is an optional argument to ResourceApplyGradientDescent.

func ResourceApplyGradientDescentUseLocking added in v1.1.0

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 added in v1.12.1

type ResourceApplyKerasMomentumAttr func(optionalAttr)

ResourceApplyKerasMomentumAttr is an optional argument to ResourceApplyKerasMomentum.

func ResourceApplyKerasMomentumUseLocking added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

type ResourceApplyMomentumAttr func(optionalAttr)

ResourceApplyMomentumAttr is an optional argument to ResourceApplyMomentum.

func ResourceApplyMomentumUseLocking added in v1.1.0

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 added in v1.1.0

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 added in v1.5.0

type ResourceApplyPowerSignAttr func(optionalAttr)

ResourceApplyPowerSignAttr is an optional argument to ResourceApplyPowerSign.

func ResourceApplyPowerSignUseLocking added in v1.5.0

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 added in v1.1.0

type ResourceApplyProximalAdagradAttr func(optionalAttr)

ResourceApplyProximalAdagradAttr is an optional argument to ResourceApplyProximalAdagrad.

func ResourceApplyProximalAdagradUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceApplyProximalGradientDescentAttr func(optionalAttr)

ResourceApplyProximalGradientDescentAttr is an optional argument to ResourceApplyProximalGradientDescent.

func ResourceApplyProximalGradientDescentUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceApplyRMSPropAttr func(optionalAttr)

ResourceApplyRMSPropAttr is an optional argument to ResourceApplyRMSProp.

func ResourceApplyRMSPropUseLocking added in v1.1.0

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 added in v1.15.0

type ResourceConditionalAccumulatorAttr func(optionalAttr)

ResourceConditionalAccumulatorAttr is an optional argument to ResourceConditionalAccumulator.

func ResourceConditionalAccumulatorContainer added in v1.15.0

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 added in v1.15.0

func ResourceConditionalAccumulatorReductionType(value string) ResourceConditionalAccumulatorAttr

ResourceConditionalAccumulatorReductionType sets the optional reduction_type attribute to value. If not specified, defaults to "MEAN"

func ResourceConditionalAccumulatorSharedName added in v1.15.0

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 added in v1.1.0

type ResourceGatherAttr func(optionalAttr)

ResourceGatherAttr is an optional argument to ResourceGather.

func ResourceGatherBatchDims added in v1.12.1

func ResourceGatherBatchDims(value int64) ResourceGatherAttr

ResourceGatherBatchDims sets the optional batch_dims attribute to value. If not specified, defaults to 0

func ResourceGatherValidateIndices added in v1.1.0

func ResourceGatherValidateIndices(value bool) ResourceGatherAttr

ResourceGatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type ResourceScatterNdAddAttr added in v1.11.0

type ResourceScatterNdAddAttr func(optionalAttr)

ResourceScatterNdAddAttr is an optional argument to ResourceScatterNdAdd.

func ResourceScatterNdAddUseLocking added in v1.11.0

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 added in v1.12.1

type ResourceScatterNdSubAttr func(optionalAttr)

ResourceScatterNdSubAttr is an optional argument to ResourceScatterNdSub.

func ResourceScatterNdSubUseLocking added in v1.12.1

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 added in v1.6.0

type ResourceScatterNdUpdateAttr func(optionalAttr)

ResourceScatterNdUpdateAttr is an optional argument to ResourceScatterNdUpdate.

func ResourceScatterNdUpdateUseLocking added in v1.6.0

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 added in v1.1.0

type ResourceSparseApplyAdadeltaAttr func(optionalAttr)

ResourceSparseApplyAdadeltaAttr is an optional argument to ResourceSparseApplyAdadelta.

func ResourceSparseApplyAdadeltaUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyAdagradAttr func(optionalAttr)

ResourceSparseApplyAdagradAttr is an optional argument to ResourceSparseApplyAdagrad.

func ResourceSparseApplyAdagradUpdateSlots added in v1.9.0

func ResourceSparseApplyAdagradUpdateSlots(value bool) ResourceSparseApplyAdagradAttr

ResourceSparseApplyAdagradUpdateSlots sets the optional update_slots attribute to value. If not specified, defaults to true

func ResourceSparseApplyAdagradUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyAdagradDAAttr func(optionalAttr)

ResourceSparseApplyAdagradDAAttr is an optional argument to ResourceSparseApplyAdagradDA.

func ResourceSparseApplyAdagradDAUseLocking added in v1.1.0

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 ResourceSparseApplyCenteredRMSPropAttr added in v1.1.0

type ResourceSparseApplyCenteredRMSPropAttr func(optionalAttr)

ResourceSparseApplyCenteredRMSPropAttr is an optional argument to ResourceSparseApplyCenteredRMSProp.

func ResourceSparseApplyCenteredRMSPropUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyFtrlAttr func(optionalAttr)

ResourceSparseApplyFtrlAttr is an optional argument to ResourceSparseApplyFtrl.

func ResourceSparseApplyFtrlUseLocking added in v1.1.0

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 added in v1.3.0

type ResourceSparseApplyFtrlV2Attr func(optionalAttr)

ResourceSparseApplyFtrlV2Attr is an optional argument to ResourceSparseApplyFtrlV2.

func ResourceSparseApplyFtrlV2UseLocking added in v1.3.0

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 added in v1.12.1

type ResourceSparseApplyKerasMomentumAttr func(optionalAttr)

ResourceSparseApplyKerasMomentumAttr is an optional argument to ResourceSparseApplyKerasMomentum.

func ResourceSparseApplyKerasMomentumUseLocking added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

type ResourceSparseApplyMomentumAttr func(optionalAttr)

ResourceSparseApplyMomentumAttr is an optional argument to ResourceSparseApplyMomentum.

func ResourceSparseApplyMomentumUseLocking added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyProximalAdagradAttr func(optionalAttr)

ResourceSparseApplyProximalAdagradAttr is an optional argument to ResourceSparseApplyProximalAdagrad.

func ResourceSparseApplyProximalAdagradUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyProximalGradientDescentAttr func(optionalAttr)

ResourceSparseApplyProximalGradientDescentAttr is an optional argument to ResourceSparseApplyProximalGradientDescent.

func ResourceSparseApplyProximalGradientDescentUseLocking added in v1.1.0

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 added in v1.1.0

type ResourceSparseApplyRMSPropAttr func(optionalAttr)

ResourceSparseApplyRMSPropAttr is an optional argument to ResourceSparseApplyRMSProp.

func ResourceSparseApplyRMSPropUseLocking added in v1.1.0

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 added in v1.2.0

type ResourceStridedSliceAssignAttr func(optionalAttr)

ResourceStridedSliceAssignAttr is an optional argument to ResourceStridedSliceAssign.

func ResourceStridedSliceAssignBeginMask added in v1.2.0

func ResourceStridedSliceAssignBeginMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignEllipsisMask added in v1.2.0

func ResourceStridedSliceAssignEllipsisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignEndMask added in v1.2.0

func ResourceStridedSliceAssignEndMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignNewAxisMask added in v1.2.0

func ResourceStridedSliceAssignNewAxisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func ResourceStridedSliceAssignShrinkAxisMask added in v1.2.0

func ResourceStridedSliceAssignShrinkAxisMask(value int64) ResourceStridedSliceAssignAttr

ResourceStridedSliceAssignShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type RestoreAttr added in v1.1.0

type RestoreAttr func(optionalAttr)

RestoreAttr is an optional argument to Restore.

func RestorePreferredShard added in v1.1.0

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 added in v1.1.0

type RestoreSliceAttr func(optionalAttr)

RestoreSliceAttr is an optional argument to RestoreSlice.

func RestoreSlicePreferredShard added in v1.1.0

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 added in v1.12.1

type RetrieveTPUEmbeddingADAMParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingADAMParametersAttr is an optional argument to RetrieveTPUEmbeddingADAMParameters.

func RetrieveTPUEmbeddingADAMParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingADAMParametersTableId(value int64) RetrieveTPUEmbeddingADAMParametersAttr

RetrieveTPUEmbeddingADAMParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingADAMParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingADAMParametersTableName(value string) RetrieveTPUEmbeddingADAMParametersAttr

RetrieveTPUEmbeddingADAMParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingADAMParametersGradAccumDebug.

func RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr

RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingADAMParametersGradAccumDebugAttr

RetrieveTPUEmbeddingADAMParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdadeltaParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingAdadeltaParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingAdadeltaParametersAttr is an optional argument to RetrieveTPUEmbeddingAdadeltaParameters.

func RetrieveTPUEmbeddingAdadeltaParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingAdadeltaParametersTableId(value int64) RetrieveTPUEmbeddingAdadeltaParametersAttr

RetrieveTPUEmbeddingAdadeltaParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingAdadeltaParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingAdadeltaParametersTableName(value string) RetrieveTPUEmbeddingAdadeltaParametersAttr

RetrieveTPUEmbeddingAdadeltaParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebug.

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr

RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugAttr

RetrieveTPUEmbeddingAdadeltaParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdagradParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingAdagradParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingAdagradParametersAttr is an optional argument to RetrieveTPUEmbeddingAdagradParameters.

func RetrieveTPUEmbeddingAdagradParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingAdagradParametersTableId(value int64) RetrieveTPUEmbeddingAdagradParametersAttr

RetrieveTPUEmbeddingAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingAdagradParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingAdagradParametersTableName(value string) RetrieveTPUEmbeddingAdagradParametersAttr

RetrieveTPUEmbeddingAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingAdagradParametersGradAccumDebug.

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr

RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingAdagradParametersGradAccumDebugAttr

RetrieveTPUEmbeddingAdagradParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingCenteredRMSPropParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingCenteredRMSPropParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingCenteredRMSPropParametersAttr is an optional argument to RetrieveTPUEmbeddingCenteredRMSPropParameters.

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableId(value int64) RetrieveTPUEmbeddingCenteredRMSPropParametersAttr

RetrieveTPUEmbeddingCenteredRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingCenteredRMSPropParametersTableName(value string) RetrieveTPUEmbeddingCenteredRMSPropParametersAttr

RetrieveTPUEmbeddingCenteredRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingFTRLParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingFTRLParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingFTRLParametersAttr is an optional argument to RetrieveTPUEmbeddingFTRLParameters.

func RetrieveTPUEmbeddingFTRLParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingFTRLParametersTableId(value int64) RetrieveTPUEmbeddingFTRLParametersAttr

RetrieveTPUEmbeddingFTRLParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingFTRLParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingFTRLParametersTableName(value string) RetrieveTPUEmbeddingFTRLParametersAttr

RetrieveTPUEmbeddingFTRLParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingFTRLParametersGradAccumDebug.

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr

RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingFTRLParametersGradAccumDebugAttr

RetrieveTPUEmbeddingFTRLParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingMDLAdagradLightParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingMDLAdagradLightParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingMDLAdagradLightParametersAttr is an optional argument to RetrieveTPUEmbeddingMDLAdagradLightParameters.

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableId(value int64) RetrieveTPUEmbeddingMDLAdagradLightParametersAttr

RetrieveTPUEmbeddingMDLAdagradLightParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingMDLAdagradLightParametersTableName(value string) RetrieveTPUEmbeddingMDLAdagradLightParametersAttr

RetrieveTPUEmbeddingMDLAdagradLightParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingMomentumParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingMomentumParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingMomentumParametersAttr is an optional argument to RetrieveTPUEmbeddingMomentumParameters.

func RetrieveTPUEmbeddingMomentumParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingMomentumParametersTableId(value int64) RetrieveTPUEmbeddingMomentumParametersAttr

RetrieveTPUEmbeddingMomentumParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingMomentumParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingMomentumParametersTableName(value string) RetrieveTPUEmbeddingMomentumParametersAttr

RetrieveTPUEmbeddingMomentumParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingMomentumParametersGradAccumDebug.

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr

RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingMomentumParametersGradAccumDebugAttr

RetrieveTPUEmbeddingMomentumParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingProximalAdagradParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingProximalAdagradParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingProximalAdagradParametersAttr is an optional argument to RetrieveTPUEmbeddingProximalAdagradParameters.

func RetrieveTPUEmbeddingProximalAdagradParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingProximalAdagradParametersTableId(value int64) RetrieveTPUEmbeddingProximalAdagradParametersAttr

RetrieveTPUEmbeddingProximalAdagradParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingProximalAdagradParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingProximalAdagradParametersTableName(value string) RetrieveTPUEmbeddingProximalAdagradParametersAttr

RetrieveTPUEmbeddingProximalAdagradParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebug.

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr

RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugAttr

RetrieveTPUEmbeddingProximalAdagradParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingRMSPropParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingRMSPropParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingRMSPropParametersAttr is an optional argument to RetrieveTPUEmbeddingRMSPropParameters.

func RetrieveTPUEmbeddingRMSPropParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingRMSPropParametersTableId(value int64) RetrieveTPUEmbeddingRMSPropParametersAttr

RetrieveTPUEmbeddingRMSPropParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingRMSPropParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingRMSPropParametersTableName(value string) RetrieveTPUEmbeddingRMSPropParametersAttr

RetrieveTPUEmbeddingRMSPropParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr added in v1.12.1

type RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr func(optionalAttr)

RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr is an optional argument to RetrieveTPUEmbeddingRMSPropParametersGradAccumDebug.

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableId added in v1.12.1

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableId(value int64) RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr

RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableName added in v1.12.1

func RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableName(value string) RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugAttr

RetrieveTPUEmbeddingRMSPropParametersGradAccumDebugTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr added in v1.12.1

type RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr func(optionalAttr)

RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr is an optional argument to RetrieveTPUEmbeddingStochasticGradientDescentParameters.

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId added in v1.12.1

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId(value int64) RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr

RetrieveTPUEmbeddingStochasticGradientDescentParametersTableId sets the optional table_id attribute to value. If not specified, defaults to -1

REQUIRES: value >= -1

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName added in v1.12.1

func RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName(value string) RetrieveTPUEmbeddingStochasticGradientDescentParametersAttr

RetrieveTPUEmbeddingStochasticGradientDescentParametersTableName sets the optional table_name attribute to value. If not specified, defaults to ""

type ReverseSequenceAttr added in v1.1.0

type ReverseSequenceAttr func(optionalAttr)

ReverseSequenceAttr is an optional argument to ReverseSequence.

func ReverseSequenceBatchDim added in v1.1.0

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 RpcAttr added in v1.9.0

type RpcAttr func(optionalAttr)

RpcAttr is an optional argument to Rpc.

func RpcFailFast added in v1.9.0

func RpcFailFast(value bool) RpcAttr

RpcFailFast sets the optional fail_fast attribute to value.

value: `boolean`. If `true` (default), then failures to connect (i.e., the server does not immediately respond) cause an RPC failure. If not specified, defaults to true

func RpcProtocol added in v1.9.0

func RpcProtocol(value string) RpcAttr

RpcProtocol sets the optional protocol attribute to value.

value: RPC protocol to use. Empty string means use the default protocol. Options include 'grpc'. If not specified, defaults to ""

func RpcTimeoutInMs added in v1.9.0

func RpcTimeoutInMs(value int64) RpcAttr

RpcTimeoutInMs sets the optional timeout_in_ms attribute to value.

value: `int`. If `0` (default), then the kernel will run the RPC request and only time out if the RPC deadline passes or the session times out. If this value is greater than `0`, then the op will raise an exception if the RPC takes longer than `timeout_in_ms`. If not specified, defaults to 0

type SampleDistortedBoundingBoxAttr added in v1.1.0

type SampleDistortedBoundingBoxAttr func(optionalAttr)

SampleDistortedBoundingBoxAttr is an optional argument to SampleDistortedBoundingBox.

func SampleDistortedBoundingBoxAreaRange added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.3.0

type SampleDistortedBoundingBoxV2Attr func(optionalAttr)

SampleDistortedBoundingBoxV2Attr is an optional argument to SampleDistortedBoundingBoxV2.

func SampleDistortedBoundingBoxV2AreaRange added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 added in v1.3.0

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 Scope

type Scope struct {
	// contains filtered or unexported fields
}

Scope encapsulates common operation properties when building a Graph.

A Scope object (and its derivates, 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 derivates (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 added in v1.3.0

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 added in v1.6.0

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 added in v1.10.0

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 added in v1.1.0

type SdcaOptimizerAttr func(optionalAttr)

SdcaOptimizerAttr is an optional argument to SdcaOptimizer.

func SdcaOptimizerAdaptative added in v1.1.0

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 added in v1.12.1

type SdcaOptimizerV2Attr func(optionalAttr)

SdcaOptimizerV2Attr is an optional argument to SdcaOptimizerV2.

func SdcaOptimizerV2Adaptive added in v1.12.1

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 added in v1.1.0

type SelfAdjointEigV2Attr func(optionalAttr)

SelfAdjointEigV2Attr is an optional argument to SelfAdjointEigV2.

func SelfAdjointEigV2ComputeV added in v1.1.0

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 SerializeManySparseAttr added in v1.6.0

type SerializeManySparseAttr func(optionalAttr)

SerializeManySparseAttr is an optional argument to SerializeManySparse.

func SerializeManySparseOutType added in v1.6.0

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 added in v1.6.0

type SerializeSparseAttr func(optionalAttr)

SerializeSparseAttr is an optional argument to SerializeSparse.

func SerializeSparseOutType added in v1.6.0

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 added in v1.1.0

type SetSizeAttr func(optionalAttr)

SetSizeAttr is an optional argument to SetSize.

func SetSizeValidateIndices added in v1.1.0

func SetSizeValidateIndices(value bool) SetSizeAttr

SetSizeValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type ShapeAttr added in v1.1.0

type ShapeAttr func(optionalAttr)

ShapeAttr is an optional argument to Shape.

func ShapeOutType added in v1.1.0

func ShapeOutType(value tf.DataType) ShapeAttr

ShapeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type ShapeNAttr added in v1.1.0

type ShapeNAttr func(optionalAttr)

ShapeNAttr is an optional argument to ShapeN.

func ShapeNOutType added in v1.1.0

func ShapeNOutType(value tf.DataType) ShapeNAttr

ShapeNOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type ShardDatasetAttr added in v1.14.0

type ShardDatasetAttr func(optionalAttr)

ShardDatasetAttr is an optional argument to ShardDataset.

func ShardDatasetRequireNonEmpty added in v1.14.0

func ShardDatasetRequireNonEmpty(value bool) ShardDatasetAttr

ShardDatasetRequireNonEmpty sets the optional require_non_empty attribute to value. If not specified, defaults to false

type ShuffleDatasetAttr added in v1.4.0

type ShuffleDatasetAttr func(optionalAttr)

ShuffleDatasetAttr is an optional argument to ShuffleDataset.

func ShuffleDatasetReshuffleEachIteration added in v1.4.0

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 added in v1.1.0

type SizeAttr func(optionalAttr)

SizeAttr is an optional argument to Size.

func SizeOutType added in v1.1.0

func SizeOutType(value tf.DataType) SizeAttr

SizeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type SkipgramAttr added in v1.1.0

type SkipgramAttr func(optionalAttr)

SkipgramAttr is an optional argument to Skipgram.

func SkipgramMinCount added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 SnapshotDatasetAttr added in v1.15.0

type SnapshotDatasetAttr func(optionalAttr)

SnapshotDatasetAttr is an optional argument to SnapshotDataset.

func SnapshotDatasetCompression added in v1.15.0

func SnapshotDatasetCompression(value string) SnapshotDatasetAttr

SnapshotDatasetCompression sets the optional compression attribute to value. If not specified, defaults to ""

func SnapshotDatasetPendingSnapshotExpirySeconds added in v1.15.0

func SnapshotDatasetPendingSnapshotExpirySeconds(value int64) SnapshotDatasetAttr

SnapshotDatasetPendingSnapshotExpirySeconds sets the optional pending_snapshot_expiry_seconds attribute to value. If not specified, defaults to 86400

func SnapshotDatasetReaderPathPrefix added in v1.15.0

func SnapshotDatasetReaderPathPrefix(value string) SnapshotDatasetAttr

SnapshotDatasetReaderPathPrefix sets the optional reader_path_prefix attribute to value. If not specified, defaults to ""

func SnapshotDatasetShardSizeBytes added in v1.15.0

func SnapshotDatasetShardSizeBytes(value int64) SnapshotDatasetAttr

SnapshotDatasetShardSizeBytes sets the optional shard_size_bytes attribute to value. If not specified, defaults to 10737418240

func SnapshotDatasetWriterPathPrefix added in v1.15.0

func SnapshotDatasetWriterPathPrefix(value string) SnapshotDatasetAttr

SnapshotDatasetWriterPathPrefix sets the optional writer_path_prefix attribute to value. If not specified, defaults to ""

type SpaceToDepthAttr added in v1.4.0

type SpaceToDepthAttr func(optionalAttr)

SpaceToDepthAttr is an optional argument to SpaceToDepth.

func SpaceToDepthDataFormat added in v1.4.0

func SpaceToDepthDataFormat(value string) SpaceToDepthAttr

SpaceToDepthDataFormat sets the optional data_format attribute to value. If not specified, defaults to "NHWC"

type SparseMatMulAttr added in v1.1.0

type SparseMatMulAttr func(optionalAttr)

SparseMatMulAttr is an optional argument to SparseMatMul.

func SparseMatMulAIsSparse added in v1.1.0

func SparseMatMulAIsSparse(value bool) SparseMatMulAttr

SparseMatMulAIsSparse sets the optional a_is_sparse attribute to value. If not specified, defaults to false

func SparseMatMulBIsSparse added in v1.1.0

func SparseMatMulBIsSparse(value bool) SparseMatMulAttr

SparseMatMulBIsSparse sets the optional b_is_sparse attribute to value. If not specified, defaults to false

func SparseMatMulTransposeA added in v1.1.0

func SparseMatMulTransposeA(value bool) SparseMatMulAttr

SparseMatMulTransposeA sets the optional transpose_a attribute to value. If not specified, defaults to false

func SparseMatMulTransposeB added in v1.1.0

func SparseMatMulTransposeB(value bool) SparseMatMulAttr

SparseMatMulTransposeB sets the optional transpose_b attribute to value. If not specified, defaults to false

type SparseReduceMaxAttr added in v1.3.0

type SparseReduceMaxAttr func(optionalAttr)

SparseReduceMaxAttr is an optional argument to SparseReduceMax.

func SparseReduceMaxKeepDims added in v1.3.0

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 added in v1.3.0

type SparseReduceMaxSparseAttr func(optionalAttr)

SparseReduceMaxSparseAttr is an optional argument to SparseReduceMaxSparse.

func SparseReduceMaxSparseKeepDims added in v1.3.0

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 added in v1.1.0

type SparseReduceSumAttr func(optionalAttr)

SparseReduceSumAttr is an optional argument to SparseReduceSum.

func SparseReduceSumKeepDims added in v1.1.0

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 added in v1.1.0

type SparseReduceSumSparseAttr func(optionalAttr)

SparseReduceSumSparseAttr is an optional argument to SparseReduceSumSparse.

func SparseReduceSumSparseKeepDims added in v1.1.0

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 SparseTensorDenseMatMulAttr added in v1.1.0

type SparseTensorDenseMatMulAttr func(optionalAttr)

SparseTensorDenseMatMulAttr is an optional argument to SparseTensorDenseMatMul.

func SparseTensorDenseMatMulAdjointA added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type SparseToDenseAttr func(optionalAttr)

SparseToDenseAttr is an optional argument to SparseToDense.

func SparseToDenseValidateIndices added in v1.1.0

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 added in v1.1.0

type SparseToSparseSetOperationAttr func(optionalAttr)

SparseToSparseSetOperationAttr is an optional argument to SparseToSparseSetOperation.

func SparseToSparseSetOperationValidateIndices added in v1.1.0

func SparseToSparseSetOperationValidateIndices(value bool) SparseToSparseSetOperationAttr

SparseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true

type SqueezeAttr added in v1.1.0

type SqueezeAttr func(optionalAttr)

SqueezeAttr is an optional argument to Squeeze.

func SqueezeAxis added in v1.6.0

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 added in v1.3.0

type StackPushV2Attr func(optionalAttr)

StackPushV2Attr is an optional argument to StackPushV2.

func StackPushV2SwapMemory added in v1.3.0

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 added in v1.3.0

type StackV2Attr func(optionalAttr)

StackV2Attr is an optional argument to StackV2.

func StackV2StackName added in v1.3.0

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 added in v1.1.0

type StageAttr func(optionalAttr)

StageAttr is an optional argument to Stage.

func StageCapacity added in v1.3.0

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 added in v1.1.0

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 added in v1.3.0

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 added in v1.1.0

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 added in v1.3.0

type StageClearAttr func(optionalAttr)

StageClearAttr is an optional argument to StageClear.

func StageClearCapacity added in v1.3.0

func StageClearCapacity(value int64) StageClearAttr

StageClearCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageClearContainer added in v1.3.0

func StageClearContainer(value string) StageClearAttr

StageClearContainer sets the optional container attribute to value. If not specified, defaults to ""

func StageClearMemoryLimit added in v1.3.0

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 added in v1.3.0

func StageClearSharedName(value string) StageClearAttr

StageClearSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StagePeekAttr added in v1.3.0

type StagePeekAttr func(optionalAttr)

StagePeekAttr is an optional argument to StagePeek.

func StagePeekCapacity added in v1.3.0

func StagePeekCapacity(value int64) StagePeekAttr

StagePeekCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StagePeekContainer added in v1.3.0

func StagePeekContainer(value string) StagePeekAttr

StagePeekContainer sets the optional container attribute to value. If not specified, defaults to ""

func StagePeekMemoryLimit added in v1.3.0

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 added in v1.3.0

func StagePeekSharedName(value string) StagePeekAttr

StagePeekSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StageSizeAttr added in v1.3.0

type StageSizeAttr func(optionalAttr)

StageSizeAttr is an optional argument to StageSize.

func StageSizeCapacity added in v1.3.0

func StageSizeCapacity(value int64) StageSizeAttr

StageSizeCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func StageSizeContainer added in v1.3.0

func StageSizeContainer(value string) StageSizeAttr

StageSizeContainer sets the optional container attribute to value. If not specified, defaults to ""

func StageSizeMemoryLimit added in v1.3.0

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 added in v1.3.0

func StageSizeSharedName(value string) StageSizeAttr

StageSizeSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StatefulStandardNormalAttr added in v1.12.1

type StatefulStandardNormalAttr func(optionalAttr)

StatefulStandardNormalAttr is an optional argument to StatefulStandardNormal.

func StatefulStandardNormalDtype added in v1.12.1

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 added in v1.12.1

type StatefulStandardNormalV2Attr func(optionalAttr)

StatefulStandardNormalV2Attr is an optional argument to StatefulStandardNormalV2.

func StatefulStandardNormalV2Dtype added in v1.12.1

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 added in v1.12.1

type StatefulTruncatedNormalAttr func(optionalAttr)

StatefulTruncatedNormalAttr is an optional argument to StatefulTruncatedNormal.

func StatefulTruncatedNormalDtype added in v1.12.1

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 added in v1.12.1

type StatefulUniformAttr func(optionalAttr)

StatefulUniformAttr is an optional argument to StatefulUniform.

func StatefulUniformDtype added in v1.12.1

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 added in v1.12.1

type StatefulUniformFullIntAttr func(optionalAttr)

StatefulUniformFullIntAttr is an optional argument to StatefulUniformFullInt.

func StatefulUniformFullIntDtype added in v1.12.1

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 added in v1.9.0

type StatelessMultinomialAttr func(optionalAttr)

StatelessMultinomialAttr is an optional argument to StatelessMultinomial.

func StatelessMultinomialOutputDtype added in v1.9.0

func StatelessMultinomialOutputDtype(value tf.DataType) StatelessMultinomialAttr

StatelessMultinomialOutputDtype sets the optional output_dtype attribute to value. If not specified, defaults to DT_INT64

type StatelessRandomNormalAttr added in v1.2.0

type StatelessRandomNormalAttr func(optionalAttr)

StatelessRandomNormalAttr is an optional argument to StatelessRandomNormal.

func StatelessRandomNormalDtype added in v1.2.0

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 StatelessRandomUniformAttr added in v1.2.0

type StatelessRandomUniformAttr func(optionalAttr)

StatelessRandomUniformAttr is an optional argument to StatelessRandomUniform.

func StatelessRandomUniformDtype added in v1.2.0

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 StatelessTruncatedNormalAttr added in v1.2.0

type StatelessTruncatedNormalAttr func(optionalAttr)

StatelessTruncatedNormalAttr is an optional argument to StatelessTruncatedNormal.

func StatelessTruncatedNormalDtype added in v1.2.0

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 StaticRegexReplaceAttr added in v1.12.0

type StaticRegexReplaceAttr func(optionalAttr)

StaticRegexReplaceAttr is an optional argument to StaticRegexReplace.

func StaticRegexReplaceReplaceGlobal added in v1.12.0

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 added in v1.5.0

type StatsAggregatorHandleAttr func(optionalAttr)

StatsAggregatorHandleAttr is an optional argument to StatsAggregatorHandle.

func StatsAggregatorHandleContainer added in v1.5.0

func StatsAggregatorHandleContainer(value string) StatsAggregatorHandleAttr

StatsAggregatorHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func StatsAggregatorHandleSharedName added in v1.5.0

func StatsAggregatorHandleSharedName(value string) StatsAggregatorHandleAttr

StatsAggregatorHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type StridedSliceAttr added in v1.1.0

type StridedSliceAttr func(optionalAttr)

StridedSliceAttr is an optional argument to StridedSlice.

func StridedSliceBeginMask added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type StridedSliceGradAttr func(optionalAttr)

StridedSliceGradAttr is an optional argument to StridedSliceGrad.

func StridedSliceGradBeginMask added in v1.1.0

func StridedSliceGradBeginMask(value int64) StridedSliceGradAttr

StridedSliceGradBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradEllipsisMask added in v1.1.0

func StridedSliceGradEllipsisMask(value int64) StridedSliceGradAttr

StridedSliceGradEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradEndMask added in v1.1.0

func StridedSliceGradEndMask(value int64) StridedSliceGradAttr

StridedSliceGradEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradNewAxisMask added in v1.1.0

func StridedSliceGradNewAxisMask(value int64) StridedSliceGradAttr

StridedSliceGradNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func StridedSliceGradShrinkAxisMask added in v1.1.0

func StridedSliceGradShrinkAxisMask(value int64) StridedSliceGradAttr

StridedSliceGradShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type StringFormatAttr added in v1.12.0

type StringFormatAttr func(optionalAttr)

StringFormatAttr is an optional argument to StringFormat.

func StringFormatPlaceholder added in v1.12.0

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 added in v1.12.0

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 added in v1.12.0

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 added in v1.1.0

type StringJoinAttr func(optionalAttr)

StringJoinAttr is an optional argument to StringJoin.

func StringJoinSeparator added in v1.1.0

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 added in v1.12.0

type StringLengthAttr func(optionalAttr)

StringLengthAttr is an optional argument to StringLength.

func StringLengthUnit added in v1.12.0

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 StringSplitAttr added in v1.4.0

type StringSplitAttr func(optionalAttr)

StringSplitAttr is an optional argument to StringSplit.

func StringSplitSkipEmpty added in v1.4.0

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 added in v1.11.0

type StringSplitV2Attr func(optionalAttr)

StringSplitV2Attr is an optional argument to StringSplitV2.

func StringSplitV2Maxsplit added in v1.11.0

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 added in v1.1.0

type StringToNumberAttr func(optionalAttr)

StringToNumberAttr is an optional argument to StringToNumber.

func StringToNumberOutType added in v1.1.0

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 SubstrAttr added in v1.12.1

type SubstrAttr func(optionalAttr)

SubstrAttr is an optional argument to Substr.

func SubstrUnit added in v1.12.1

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 added in v1.1.0

type SumAttr func(optionalAttr)

SumAttr is an optional argument to Sum.

func SumKeepDims added in v1.1.0

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 added in v1.1.0

type SvdAttr func(optionalAttr)

SvdAttr is an optional argument to Svd.

func SvdComputeUv added in v1.1.0

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 added in v1.1.0

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 TFRecordReaderV2Attr added in v1.1.0

type TFRecordReaderV2Attr func(optionalAttr)

TFRecordReaderV2Attr is an optional argument to TFRecordReaderV2.

func TFRecordReaderV2CompressionType added in v1.1.0

func TFRecordReaderV2CompressionType(value string) TFRecordReaderV2Attr

TFRecordReaderV2CompressionType sets the optional compression_type attribute to value. If not specified, defaults to ""

func TFRecordReaderV2Container added in v1.1.0

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 added in v1.1.0

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 TPUReplicateMetadataAttr added in v1.12.1

type TPUReplicateMetadataAttr func(optionalAttr)

TPUReplicateMetadataAttr is an optional argument to TPUReplicateMetadata.

func TPUReplicateMetadataComputationShape added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

func TPUReplicateMetadataHostComputeCore(value []string) TPUReplicateMetadataAttr

TPUReplicateMetadataHostComputeCore sets the optional host_compute_core attribute to value. If not specified, defaults to <>

func TPUReplicateMetadataNumCoresPerReplica added in v1.12.1

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 added in v1.12.1

func TPUReplicateMetadataPaddingMap(value []string) TPUReplicateMetadataAttr

TPUReplicateMetadataPaddingMap sets the optional padding_map attribute to value. If not specified, defaults to <>

func TPUReplicateMetadataStepMarkerLocation added in v1.12.1

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 added in v1.12.1

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 TPUReplicateMetadataUseTpu added in v1.12.1

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 TakeManySparseFromTensorsMapAttr added in v1.1.0

type TakeManySparseFromTensorsMapAttr func(optionalAttr)

TakeManySparseFromTensorsMapAttr is an optional argument to TakeManySparseFromTensorsMap.

func TakeManySparseFromTensorsMapContainer added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type TensorArrayConcatV2Attr func(optionalAttr)

TensorArrayConcatV2Attr is an optional argument to TensorArrayConcatV2.

func TensorArrayConcatV2ElementShapeExcept0 added in v1.1.0

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 added in v1.1.0

type TensorArrayConcatV3Attr func(optionalAttr)

TensorArrayConcatV3Attr is an optional argument to TensorArrayConcatV3.

func TensorArrayConcatV3ElementShapeExcept0 added in v1.1.0

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 added in v1.1.0

type TensorArrayGatherV2Attr func(optionalAttr)

TensorArrayGatherV2Attr is an optional argument to TensorArrayGatherV2.

func TensorArrayGatherV2ElementShape added in v1.1.0

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 added in v1.1.0

type TensorArrayGatherV3Attr func(optionalAttr)

TensorArrayGatherV3Attr is an optional argument to TensorArrayGatherV3.

func TensorArrayGatherV3ElementShape added in v1.1.0

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 added in v1.1.0

type TensorArrayV2Attr func(optionalAttr)

TensorArrayV2Attr is an optional argument to TensorArrayV2.

func TensorArrayV2ClearAfterRead added in v1.1.0

func TensorArrayV2ClearAfterRead(value bool) TensorArrayV2Attr

TensorArrayV2ClearAfterRead sets the optional clear_after_read attribute to value. If not specified, defaults to true

func TensorArrayV2DynamicSize added in v1.1.0

func TensorArrayV2DynamicSize(value bool) TensorArrayV2Attr

TensorArrayV2DynamicSize sets the optional dynamic_size attribute to value. If not specified, defaults to false

func TensorArrayV2ElementShape added in v1.1.0

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 added in v1.1.0

func TensorArrayV2TensorArrayName(value string) TensorArrayV2Attr

TensorArrayV2TensorArrayName sets the optional tensor_array_name attribute to value. If not specified, defaults to ""

type TensorArrayV3Attr added in v1.1.0

type TensorArrayV3Attr func(optionalAttr)

TensorArrayV3Attr is an optional argument to TensorArrayV3.

func TensorArrayV3ClearAfterRead added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.5.0

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 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 added in v1.1.0

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 TensorForestTreeResourceHandleOpAttr added in v1.12.1

type TensorForestTreeResourceHandleOpAttr func(optionalAttr)

TensorForestTreeResourceHandleOpAttr is an optional argument to TensorForestTreeResourceHandleOp.

func TensorForestTreeResourceHandleOpContainer added in v1.12.1

func TensorForestTreeResourceHandleOpContainer(value string) TensorForestTreeResourceHandleOpAttr

TensorForestTreeResourceHandleOpContainer sets the optional container attribute to value. If not specified, defaults to ""

func TensorForestTreeResourceHandleOpSharedName added in v1.12.1

func TensorForestTreeResourceHandleOpSharedName(value string) TensorForestTreeResourceHandleOpAttr

TensorForestTreeResourceHandleOpSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type TensorListConcatAttr added in v1.12.1

type TensorListConcatAttr func(optionalAttr)

TensorListConcatAttr is an optional argument to TensorListConcat.

func TensorListConcatElementShape added in v1.12.1

func TensorListConcatElementShape(value tf.Shape) TensorListConcatAttr

TensorListConcatElementShape sets the optional element_shape attribute to value. If not specified, defaults to <unknown_rank:true >

type TensorListStackAttr added in v1.9.0

type TensorListStackAttr func(optionalAttr)

TensorListStackAttr is an optional argument to TensorListStack.

func TensorListStackNumElements added in v1.9.0

func TensorListStackNumElements(value int64) TensorListStackAttr

TensorListStackNumElements sets the optional num_elements attribute to value. If not specified, defaults to -1

type TensorStridedSliceUpdateAttr added in v1.12.1

type TensorStridedSliceUpdateAttr func(optionalAttr)

TensorStridedSliceUpdateAttr is an optional argument to TensorStridedSliceUpdate.

func TensorStridedSliceUpdateBeginMask added in v1.12.1

func TensorStridedSliceUpdateBeginMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateEllipsisMask added in v1.12.1

func TensorStridedSliceUpdateEllipsisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateEndMask added in v1.12.1

func TensorStridedSliceUpdateEndMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateNewAxisMask added in v1.12.1

func TensorStridedSliceUpdateNewAxisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0

func TensorStridedSliceUpdateShrinkAxisMask added in v1.12.1

func TensorStridedSliceUpdateShrinkAxisMask(value int64) TensorStridedSliceUpdateAttr

TensorStridedSliceUpdateShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0

type TensorSummaryAttr added in v1.1.0

type TensorSummaryAttr func(optionalAttr)

TensorSummaryAttr is an optional argument to TensorSummary.

func TensorSummaryDescription added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 TextLineReaderV2Attr added in v1.1.0

type TextLineReaderV2Attr func(optionalAttr)

TextLineReaderV2Attr is an optional argument to TextLineReaderV2.

func TextLineReaderV2Container added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

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 added in v1.15.0

type ThreadPoolHandleAttr func(optionalAttr)

ThreadPoolHandleAttr is an optional argument to ThreadPoolHandle.

func ThreadPoolHandleContainer added in v1.15.0

func ThreadPoolHandleContainer(value string) ThreadPoolHandleAttr

ThreadPoolHandleContainer sets the optional container attribute to value. If not specified, defaults to ""

func ThreadPoolHandleMaxIntraOpParallelism added in v1.15.0

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 added in v1.15.0

func ThreadPoolHandleSharedName(value string) ThreadPoolHandleAttr

ThreadPoolHandleSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type ThreadUnsafeUnigramCandidateSamplerAttr added in v1.1.0

type ThreadUnsafeUnigramCandidateSamplerAttr func(optionalAttr)

ThreadUnsafeUnigramCandidateSamplerAttr is an optional argument to ThreadUnsafeUnigramCandidateSampler.

func ThreadUnsafeUnigramCandidateSamplerSeed added in v1.1.0

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 added in v1.1.0

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 added in v1.1.0

type TopKAttr func(optionalAttr)

TopKAttr is an optional argument to TopK.

func TopKSorted added in v1.1.0

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 added in v1.1.0

type TopKV2Attr func(optionalAttr)

TopKV2Attr is an optional argument to TopKV2.

func TopKV2Sorted added in v1.1.0

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 added in v1.14.0

type TridiagonalSolveAttr func(optionalAttr)

TridiagonalSolveAttr is an optional argument to TridiagonalSolve.

func TridiagonalSolvePartialPivoting added in v1.14.0

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

type TruncatedNormalAttr added in v1.1.0

type TruncatedNormalAttr func(optionalAttr)

TruncatedNormalAttr is an optional argument to TruncatedNormal.

func TruncatedNormalSeed added in v1.1.0

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 added in v1.1.0

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 TryRpcAttr added in v1.9.0

type TryRpcAttr func(optionalAttr)

TryRpcAttr is an optional argument to TryRpc.

func TryRpcFailFast added in v1.9.0

func TryRpcFailFast(value bool) TryRpcAttr

TryRpcFailFast sets the optional fail_fast attribute to value.

value: `boolean`. If `true` (default), then failures to connect (i.e., the server does not immediately respond) cause an RPC failure. If not specified, defaults to true

func TryRpcProtocol added in v1.9.0

func TryRpcProtocol(value string) TryRpcAttr

TryRpcProtocol sets the optional protocol attribute to value.

value: RPC protocol to use. Empty string means use the default protocol. Options include 'grpc'. If not specified, defaults to ""

func TryRpcTimeoutInMs added in v1.9.0

func TryRpcTimeoutInMs(value int64) TryRpcAttr

TryRpcTimeoutInMs sets the optional timeout_in_ms attribute to value.

value: `int`. If `0` (default), then the kernel will run the RPC request and only time out if the RPC deadline passes or the session times out. If this value is greater than `0`, then the op will raise an exception if the RPC takes longer than `timeout_in_ms`. If not specified, defaults to 0

type UnbatchAttr added in v1.9.0

type UnbatchAttr func(optionalAttr)

UnbatchAttr is an optional argument to Unbatch.

func UnbatchContainer added in v1.9.0

func UnbatchContainer(value string) UnbatchAttr

UnbatchContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnbatchSharedName added in v1.9.0

func UnbatchSharedName(value string) UnbatchAttr

UnbatchSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UnbatchGradAttr added in v1.9.0

type UnbatchGradAttr func(optionalAttr)

UnbatchGradAttr is an optional argument to UnbatchGrad.

func UnbatchGradContainer added in v1.9.0

func UnbatchGradContainer(value string) UnbatchGradAttr

UnbatchGradContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnbatchGradSharedName added in v1.9.0

func UnbatchGradSharedName(value string) UnbatchGradAttr

UnbatchGradSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UnicodeDecodeAttr added in v1.12.1

type UnicodeDecodeAttr func(optionalAttr)

UnicodeDecodeAttr is an optional argument to UnicodeDecode.

func UnicodeDecodeErrors added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.14.0

func UnicodeDecodeTsplits(value tf.DataType) UnicodeDecodeAttr

UnicodeDecodeTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type UnicodeDecodeWithOffsetsAttr added in v1.12.1

type UnicodeDecodeWithOffsetsAttr func(optionalAttr)

UnicodeDecodeWithOffsetsAttr is an optional argument to UnicodeDecodeWithOffsets.

func UnicodeDecodeWithOffsetsErrors added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.14.0

func UnicodeDecodeWithOffsetsTsplits(value tf.DataType) UnicodeDecodeWithOffsetsAttr

UnicodeDecodeWithOffsetsTsplits sets the optional Tsplits attribute to value. If not specified, defaults to DT_INT64

type UnicodeEncodeAttr added in v1.12.1

type UnicodeEncodeAttr func(optionalAttr)

UnicodeEncodeAttr is an optional argument to UnicodeEncode.

func UnicodeEncodeErrors added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

type UnicodeTranscodeAttr func(optionalAttr)

UnicodeTranscodeAttr is an optional argument to UnicodeTranscode.

func UnicodeTranscodeErrors added in v1.12.1

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 added in v1.12.1

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 added in v1.12.1

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 added in v1.1.0

type UniformCandidateSamplerAttr func(optionalAttr)

UniformCandidateSamplerAttr is an optional argument to UniformCandidateSampler.

func UniformCandidateSamplerSeed added in v1.1.0

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 added in v1.1.0

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 UniqueAttr added in v1.1.0

type UniqueAttr func(optionalAttr)

UniqueAttr is an optional argument to Unique.

func UniqueOutIdx added in v1.1.0

func UniqueOutIdx(value tf.DataType) UniqueAttr

UniqueOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueV2Attr added in v1.6.0

type UniqueV2Attr func(optionalAttr)

UniqueV2Attr is an optional argument to UniqueV2.

func UniqueV2OutIdx added in v1.6.0

func UniqueV2OutIdx(value tf.DataType) UniqueV2Attr

UniqueV2OutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueWithCountsAttr added in v1.1.0

type UniqueWithCountsAttr func(optionalAttr)

UniqueWithCountsAttr is an optional argument to UniqueWithCounts.

func UniqueWithCountsOutIdx added in v1.1.0

func UniqueWithCountsOutIdx(value tf.DataType) UniqueWithCountsAttr

UniqueWithCountsOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UniqueWithCountsV2Attr added in v1.9.0

type UniqueWithCountsV2Attr func(optionalAttr)

UniqueWithCountsV2Attr is an optional argument to UniqueWithCountsV2.

func UniqueWithCountsV2OutIdx added in v1.9.0

func UniqueWithCountsV2OutIdx(value tf.DataType) UniqueWithCountsV2Attr

UniqueWithCountsV2OutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32

type UnpackAttr added in v1.1.0

type UnpackAttr func(optionalAttr)

UnpackAttr is an optional argument to Unpack.

func UnpackAxis added in v1.1.0

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 UnsortedSegmentJoinAttr added in v1.15.0

type UnsortedSegmentJoinAttr func(optionalAttr)

UnsortedSegmentJoinAttr is an optional argument to UnsortedSegmentJoin.

func UnsortedSegmentJoinSeparator added in v1.15.0

func UnsortedSegmentJoinSeparator(value string) UnsortedSegmentJoinAttr

UnsortedSegmentJoinSeparator sets the optional separator attribute to value.

value: The separator to use when joining. If not specified, defaults to ""

type UnstageAttr added in v1.1.0

type UnstageAttr func(optionalAttr)

UnstageAttr is an optional argument to Unstage.

func UnstageCapacity added in v1.3.0

func UnstageCapacity(value int64) UnstageAttr

UnstageCapacity sets the optional capacity attribute to value. If not specified, defaults to 0

REQUIRES: value >= 0

func UnstageContainer added in v1.1.0

func UnstageContainer(value string) UnstageAttr

UnstageContainer sets the optional container attribute to value. If not specified, defaults to ""

func UnstageMemoryLimit added in v1.3.0

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 added in v1.1.0

func UnstageSharedName(value string) UnstageAttr

UnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""

type UpperBoundAttr added in v1.12.0

type UpperBoundAttr func(optionalAttr)

UpperBoundAttr is an optional argument to UpperBound.

func UpperBoundOutType added in v1.12.0

func UpperBoundOutType(value tf.DataType) UpperBoundAttr

UpperBoundOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type VarHandleOpAttr added in v1.1.0

type VarHandleOpAttr func(optionalAttr)

VarHandleOpAttr is an optional argument to VarHandleOp.

func VarHandleOpContainer added in v1.1.0

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 VarHandleOpSharedName added in v1.1.0

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 added in v1.4.0

type VariableShapeAttr func(optionalAttr)

VariableShapeAttr is an optional argument to VariableShape.

func VariableShapeOutType added in v1.4.0

func VariableShapeOutType(value tf.DataType) VariableShapeAttr

VariableShapeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32

type WholeFileReaderV2Attr added in v1.1.0

type WholeFileReaderV2Attr func(optionalAttr)

WholeFileReaderV2Attr is an optional argument to WholeFileReaderV2.

func WholeFileReaderV2Container added in v1.1.0

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 added in v1.1.0

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 ""

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL