ActiveGo 1.8Package op
Overview ▹
▹ Example
Index ▹
func Abort ¶
func Abort(scope *Scope, optional ...AbortAttr) (o *tf.Operation)
Raise a exception to abort the process when called. If exit_without_error is true, the process will exit normally, otherwise it will exit with a SIGABORT signal.
Returns nothing but an exception.
Returns the created operation.
func Abs ¶
func Abs(scope *Scope, x tf.Output) (y tf.Output)
Computes the absolute value of a tensor.
Given a tensor `x`, this operation returns a tensor containing the absolute value of each element in `x`. For example, if x is an input element and y is an output element, this operation computes \\(y = |x|\\).
func Acos ¶
func Acos(scope *Scope, x tf.Output) (y tf.Output)
Computes acos of x element-wise.
func Add ¶
func Add(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x + y element-wise.
*NOTE*: `Add` supports broadcasting. `AddN` does not. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func AddManySparseToTensorsMap ¶
func AddManySparseToTensorsMap(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...AddManySparseToTensorsMapAttr) (sparse_handles tf.Output)
Add an `N`-minibatch `SparseTensor` to a `SparseTensorsMap`, return `N` handles.
A `SparseTensor` of rank `R` is represented by three tensors: `sparse_indices`, `sparse_values`, and `sparse_shape`, where
“`sparse_indices.shape[1] == sparse_shape.shape[0] == R“`
An `N`-minibatch of `SparseTensor` objects is represented as a `SparseTensor` having a first `sparse_indices` column taking values between `[0, N)`, where the minibatch size `N == sparse_shape[0]`.
The input `SparseTensor` must have rank `R` greater than 1, and the first dimension is treated as the minibatch dimension. Elements of the `SparseTensor` must be sorted in increasing order of this first dimension. The stored `SparseTensor` objects pointed to by each row of the output `sparse_handles` will have rank `R-1`.
The `SparseTensor` values can then be read out as part of a minibatch by passing the given keys as vector elements to `TakeManySparseFromTensorsMap`. To ensure the correct `SparseTensorsMap` is accessed, ensure that the same `container` and `shared_name` are passed to that Op. If no `shared_name` is provided here, instead use the *name* of the Operation created by calling `AddManySparseToTensorsMap` as the `shared_name` passed to `TakeManySparseFromTensorsMap`. Ensure the Operations are colocated.
Arguments:
sparse_indices: 2-D. The `indices` of the minibatch `SparseTensor`.
`sparse_indices[:, 0]` must be ordered values in `[0, N)`.
sparse_values: 1-D. The `values` of the minibatch `SparseTensor`. sparse_shape: 1-D. The `shape` of the minibatch `SparseTensor`.
The minibatch size `N == sparse_shape[0]`.
Returns 1-D. The handles of the `SparseTensor` now stored in the `SparseTensorsMap`. Shape: `[N]`.
func AddN ¶
func AddN(scope *Scope, inputs []tf.Output) (sum tf.Output)
Add all input tensors element wise.
Arguments:
inputs: Must all be the same size and shape.
func AddSparseToTensorsMap ¶
func AddSparseToTensorsMap(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output, optional ...AddSparseToTensorsMapAttr) (sparse_handle tf.Output)
Add a `SparseTensor` to a `SparseTensorsMap` return its handle.
A `SparseTensor` is represented by three tensors: `sparse_indices`, `sparse_values`, and `sparse_shape`.
This operator takes the given `SparseTensor` and adds it to a container object (a `SparseTensorsMap`). A unique key within this container is generated in the form of an `int64`, and this is the value that is returned.
The `SparseTensor` can then be read out as part of a minibatch by passing the key as a vector element to `TakeManySparseFromTensorsMap`. To ensure the correct `SparseTensorsMap` is accessed, ensure that the same `container` and `shared_name` are passed to that Op. If no `shared_name` is provided here, instead use the *name* of the Operation created by calling `AddSparseToTensorsMap` as the `shared_name` passed to `TakeManySparseFromTensorsMap`. Ensure the Operations are colocated.
Arguments:
sparse_indices: 2-D. The `indices` of the `SparseTensor`. sparse_values: 1-D. The `values` of the `SparseTensor`. sparse_shape: 1-D. The `shape` of the `SparseTensor`.
Returns 0-D. The handle of the `SparseTensor` now stored in the `SparseTensorsMap`.
func AdjustContrast ¶
func AdjustContrast(scope *Scope, images tf.Output, contrast_factor tf.Output, min_value tf.Output, max_value tf.Output) (output tf.Output)
Deprecated. Disallowed in GraphDef version >= 2.
DEPRECATED at GraphDef version 2: Use AdjustContrastv2 instead
func AdjustContrastv2 ¶
func AdjustContrastv2(scope *Scope, images tf.Output, contrast_factor tf.Output) (output tf.Output)
Adjust the contrast of one or more images.
`images` is a tensor of at least 3 dimensions. The last 3 dimensions are interpreted as `[height, width, channels]`. The other dimensions only represent a collection of images, such as `[batch, height, width, channels].`
Contrast is adjusted independently for each channel of each image.
For each channel, the Op first computes the mean of the image pixels in the channel and then adjusts each component of each pixel to `(x - mean) * contrast_factor + mean`.
Arguments:
images: Images to adjust. At least 3-D. contrast_factor: A float multiplier for adjusting contrast.
Returns The contrast-adjusted image or images.
func AdjustHue ¶
func AdjustHue(scope *Scope, images tf.Output, delta tf.Output) (output tf.Output)
Adjust the hue of one or more images.
`images` is a tensor of at least 3 dimensions. The last dimension is 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 ¶
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 ¶
func All(scope *Scope, input tf.Output, reduction_indices 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 `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func AllCandidateSampler ¶
func AllCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, optional ...AllCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a learned unigram distribution.
See explanations of candidate sampling and the data formats at go/candidate-sampling.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to produce per batch. 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 Any ¶
func Any(scope *Scope, input tf.Output, reduction_indices 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 `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func ApproximateEqual ¶
func ApproximateEqual(scope *Scope, x tf.Output, y tf.Output, optional ...ApproximateEqualAttr) (z tf.Output)
Returns the truth value of abs(x-y) < tolerance element-wise.
func ArgMax ¶
func ArgMax(scope *Scope, input tf.Output, dimension tf.Output) (output tf.Output)
Returns the index with the largest value across dimensions of a tensor.
Arguments:
dimension: int32, 0 <= dimension < rank(input). Describes which dimension
of the input Tensor to reduce across. For vectors, use dimension = 0.
func ArgMin ¶
func ArgMin(scope *Scope, input tf.Output, dimension tf.Output) (output tf.Output)
Returns the index with the smallest value across dimensions of a tensor.
Arguments:
dimension: int32, 0 <= dimension < rank(input). Describes which dimension
of the input Tensor to reduce across. For vectors, use dimension = 0.
func AsString ¶
func AsString(scope *Scope, input tf.Output, optional ...AsStringAttr) (output tf.Output)
Converts each entry in the given tensor to strings. Supports many numeric
types and boolean.
func Asin ¶
func Asin(scope *Scope, x tf.Output) (y tf.Output)
Computes asin of x element-wise.
func Assert ¶
func Assert(scope *Scope, condition tf.Output, data []tf.Output, optional ...AssertAttr) (o *tf.Operation)
Asserts that the given condition is true.
If `condition` evaluates to false, print the list of tensors in `data`. `summarize` determines how many entries of the tensors to print.
Arguments:
condition: The condition to evaluate. data: The tensors to print out when condition is false.
Returns the created operation.
func AssignAddVariableOp ¶
func AssignAddVariableOp(scope *Scope, resource tf.Output, value tf.Output) (o *tf.Operation)
Adds a value to the current value of a variable.
Any ReadVariableOp which depends directly or indirectly on this assign is guaranteed to see the incremented value or a subsequent newer one.
Outputs the incremented value, which can be used to totally order the increments to this variable.
Arguments:
resource: handle to the resource in which to store the variable. value: the value by which the variable will be incremented.
Returns the created operation.
func AssignSubVariableOp ¶
func AssignSubVariableOp(scope *Scope, resource tf.Output, value tf.Output) (o *tf.Operation)
Subtracts a value from the current value of a variable.
Any ReadVariableOp which depends directly or indirectly on this assign is guaranteed to see the incremented value or a subsequent newer one.
Outputs the incremented value, which can be used to totally order the increments to this variable.
Arguments:
resource: handle to the resource in which to store the variable. value: the value by which the variable will be incremented.
Returns the created operation.
func AssignVariableOp ¶
func AssignVariableOp(scope *Scope, resource tf.Output, value tf.Output) (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 ¶
func Atan(scope *Scope, x tf.Output) (y tf.Output)
Computes atan of x element-wise.
func AudioSummary ¶
func AudioSummary(scope *Scope, tag tf.Output, tensor tf.Output, sample_rate float32, optional ...AudioSummaryAttr) (summary tf.Output)
Outputs a `Summary` protocol buffer with audio.
DEPRECATED at GraphDef version 15: Use AudioSummaryV2.
The summary has up to `max_outputs` summary values containing audio. The audio is built from `tensor` which must be 3-D with shape `[batch_size, frames, channels]` or 2-D with shape `[batch_size, frames]`. The values are assumed to be in the range of `[-1.0, 1.0]` with a sample rate of `sample_rate`.
The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:
* If `max_outputs` is 1, the summary value tag is '*tag*/audio'. * If `max_outputs` is greater than 1, the summary value tags are
generated sequentially as '*tag*/audio/0', '*tag*/audio/1', etc.
Arguments:
tag: Scalar. Used to build the `tag` attribute of the summary values. tensor: 2-D of shape `[batch_size, frames]`. sample_rate: The sample rate of the signal in hertz.
Returns Scalar. Serialized `Summary` protocol buffer.
func AudioSummaryV2 ¶
func AudioSummaryV2(scope *Scope, tag tf.Output, tensor tf.Output, sample_rate tf.Output, optional ...AudioSummaryV2Attr) (summary tf.Output)
Outputs a `Summary` protocol buffer with audio.
The summary has up to `max_outputs` summary values containing audio. The audio is built from `tensor` which must be 3-D with shape `[batch_size, frames, channels]` or 2-D with shape `[batch_size, frames]`. The values are assumed to be in the range of `[-1.0, 1.0]` with a sample rate of `sample_rate`.
The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:
* If `max_outputs` is 1, the summary value tag is '*tag*/audio'. * If `max_outputs` is greater than 1, the summary value tags are
generated sequentially as '*tag*/audio/0', '*tag*/audio/1', etc.
Arguments:
tag: Scalar. Used to build the `tag` attribute of the summary values. tensor: 2-D of shape `[batch_size, frames]`. sample_rate: The sample rate of the signal in hertz.
Returns Scalar. Serialized `Summary` protocol buffer.
func AvgPool ¶
func AvgPool(scope *Scope, value tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPoolAttr) (output tf.Output)
Performs average pooling on the input.
Each entry in `output` is the mean of the corresponding size `ksize` window in `value`.
Arguments:
value: 4-D with shape `[batch, height, width, channels]`. ksize: The size of the sliding window for each dimension of `value`. strides: The stride of the sliding window for each dimension of `value`. padding: The type of padding algorithm to use.
Returns The average pooled output tensor.
func AvgPool3D ¶
func AvgPool3D(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string) (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 ¶
func AvgPool3DGrad(scope *Scope, orig_input_shape tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output)
Computes gradients of average pooling function.
Arguments:
orig_input_shape: The original input dimensions. grad: Output backprop of shape `[batch, depth, rows, cols, channels]`. ksize: 1-D tensor of length 5. The size of the window for each dimension of
the input tensor. Must have `ksize[0] = ksize[4] = 1`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
Returns The backprop for input.
func AvgPoolGrad ¶
func AvgPoolGrad(scope *Scope, orig_input_shape tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...AvgPoolGradAttr) (output tf.Output)
Computes gradients of the average pooling function.
Arguments:
orig_input_shape: 1-D. Shape of the original input to `avg_pool`. grad: 4-D with shape `[batch, height, width, channels]`. Gradients w.r.t.
the output of `avg_pool`.
ksize: The size of the sliding window for each dimension of the input. strides: The stride of the sliding window for each dimension of the input. padding: The type of padding algorithm to use.
Returns 4-D. Gradients w.r.t. the input of `avg_pool`.
func BatchMatMul ¶
func BatchMatMul(scope *Scope, x tf.Output, y tf.Output, optional ...BatchMatMulAttr) (output tf.Output)
Multiplies slices of two tensors in batches.
Multiplies all slices of `Tensor` `x` and `y` (each slice can be viewed as an element of a batch), and arranges the individual results in a single output tensor of the same batch size. Each of the individual slices can optionally be adjointed (to adjoint a matrix means to transpose and conjugate it) before multiplication by setting the `adj_x` or `adj_y` flag to `True`, which are by default `False`.
The input tensors `x` and `y` are 3-D or higher with shape `[..., r_x, c_x]` and `[..., r_y, c_y]`.
The output tensor is 3-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: 3-D or higher with shape `[..., r_x, c_x]`. y: 3-D or higher with shape `[..., r_y, c_y]`.
Returns 3-D or higher with shape `[..., r_o, c_o]`
func BatchNormWithGlobalNormalization ¶
func BatchNormWithGlobalNormalization(scope *Scope, t tf.Output, m tf.Output, v tf.Output, beta tf.Output, gamma tf.Output, variance_epsilon float32, scale_after_normalization bool) (result tf.Output)
Batch normalization.
DEPRECATED at GraphDef version 9: Use tf.nn.batch_normalization()
This op is deprecated. Prefer `tf.nn.batch_normalization`.
Arguments:
t: A 4D input Tensor. m: A 1D mean Tensor with size matching the last dimension of t.
This is the first output from tf.nn.moments, or a saved moving average thereof.
v: A 1D variance Tensor with size matching the last dimension of t.
This is the second output from tf.nn.moments, or a saved moving average thereof.
beta: A 1D beta Tensor with size matching the last dimension of t.
An offset to be added to the normalized tensor.
gamma: A 1D gamma Tensor with size matching the last dimension of t.
If "scale_after_normalization" is true, this tensor will be multiplied with the normalized tensor.
variance_epsilon: A small float number to avoid dividing by 0. scale_after_normalization: A bool indicating whether the resulted tensor
needs to be multiplied with gamma.
func BatchNormWithGlobalNormalizationGrad ¶
func BatchNormWithGlobalNormalizationGrad(scope *Scope, t tf.Output, m tf.Output, v tf.Output, gamma tf.Output, backprop tf.Output, variance_epsilon float32, scale_after_normalization bool) (dx tf.Output, dm tf.Output, dv tf.Output, db tf.Output, dg tf.Output)
Gradients for batch normalization.
DEPRECATED at GraphDef version 9: Use tf.nn.batch_normalization()
This op is deprecated. See `tf.nn.batch_normalization`.
Arguments:
t: A 4D input Tensor. m: A 1D mean Tensor with size matching the last dimension of t.
This is the first output from tf.nn.moments, or a saved moving average thereof.
v: A 1D variance Tensor with size matching the last dimension of t.
This is the second output from tf.nn.moments, or a saved moving average thereof.
gamma: A 1D gamma Tensor with size matching the last dimension of t.
If "scale_after_normalization" is true, this Tensor will be multiplied with the normalized Tensor.
backprop: 4D backprop Tensor. variance_epsilon: A small float number to avoid dividing by 0. scale_after_normalization: A bool indicating whether the resulted tensor
needs to be multiplied with gamma.
Returns 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 ¶
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:
“`prettyprint [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] “`
The output tensor has shape `[1, 2, 2, 1]` and value:
“`prettyprint x = [[[[1], [2]], [[3], [4]]]] “`
(2) For the following input of shape `[4, 1, 1, 3]` and block_size of 2:
“`prettyprint [[[1, 2, 3]], [[4, 5, 6]], [[7, 8, 9]], [[10, 11, 12]]] “`
The output tensor has shape `[1, 2, 2, 3]` and value:
“`prettyprint 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:
“`prettyprint 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:
“`prettyprint 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:
“`prettyprint 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:
“`prettyprint x = [[[[1], [3]], [[5], [7]]],
[[[2], [4]], [[10], [12]]], [[[5], [7]], [[13], [15]]], [[[6], [8]], [[14], [16]]]]
“`
func BatchToSpaceND ¶
func BatchToSpaceND(scope *Scope, input tf.Output, block_shape tf.Output, crops tf.Output) (output tf.Output)
BatchToSpace for N-D tensors of type T.
This operation reshapes the "batch" dimension 0 into `M + 1` dimensions of shape `block_shape + [batch]`, interleaves these blocks back into the grid defined by the spatial dimensions `[1, ..., M]`, to obtain a result with the same rank as the input. The spatial dimensions of this intermediate result are then optionally cropped according to `crops` to produce the output. This is the reverse of SpaceToBatch. See below for a precise description.
Arguments:
input: N-D with shape `input_shape = [batch] + spatial_shape + remaining_shape`,
where spatial_shape has M dimensions.
block_shape: 1-D with shape `[M]`, all values must be >= 1. crops: 2-D with shape `[M, 2]`, all values must be >= 0. `crops[i] = [crop_start, crop_end]` specifies the amount to crop from input dimension `i + 1`, which corresponds to spatial dimension `i`. It is required that `crop_start[i] + crop_end[i] <= block_shape[i] * input_shape[i + 1]`.
This operation is equivalent to the following steps:
1. Reshape `input` to `reshaped` of shape:
[block_shape[0], ..., block_shape[M-1], batch / prod(block_shape), input_shape[1], ..., input_shape[N-1]]
2. Permute dimensions of `reshaped` to produce `permuted` of shape
[batch / prod(block_shape), input_shape[1], block_shape[0], ..., input_shape[M], block_shape[M-1], input_shape[M+1], ..., input_shape[N-1]]
3. Reshape `permuted` to produce `reshaped_permuted` of shape
[batch / prod(block_shape), input_shape[1] * block_shape[0], ..., input_shape[M] * block_shape[M-1], input_shape[M+1], ..., input_shape[N-1]]
4. Crop the start and end of dimensions `[1, ..., M]` of
`reshaped_permuted` according to `crops` to produce the output of shape: [batch / prod(block_shape), input_shape[1] * block_shape[0] - crops[0,0] - crops[0,1], ..., input_shape[M] * block_shape[M-1] - crops[M-1,0] - crops[M-1,1], input_shape[M+1], ..., input_shape[N-1]]
Some examples:
(1) For the following input of shape `[4, 1, 1, 1]`, `block_shape = [2, 2]`, and
`crops = [[0, 0], [0, 0]]`:
“`prettyprint [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] “`
The output tensor has shape `[1, 2, 2, 1]` and value:
“`prettyprint 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]]`:
“`prettyprint [[[1, 2, 3]], [[4, 5, 6]], [[7, 8, 9]], [[10, 11, 12]]] “`
The output tensor has shape `[1, 2, 2, 3]` and value:
“`prettyprint 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]]`:
“`prettyprint 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:
“`prettyprint 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]]`:
“`prettyprint 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:
“`prettyprint x = [[[[1], [2], [3], [4]],
[[5], [6], [7], [8]]], [[[9], [10], [11], [12]], [[13], [14], [15], [16]]]]
“`
func Betainc ¶
func Betainc(scope *Scope, a tf.Output, b tf.Output, x tf.Output) (z tf.Output)
Compute the regularized incomplete beta integral \\(I_x(a, b)\\).
The regularized incomplete beta integral is defined as:
“` I_x(a, b) = \frac{B(x; a, b)}{B(a, b)} “` where
“` B(x; a, b) = \int_0^x t^{a-1} (1 - t)^{b-1} dt “`
is the incomplete beta function and \\(B(a, b)\\) is the *complete* beta function.
func BiasAdd ¶
func BiasAdd(scope *Scope, value tf.Output, bias tf.Output, optional ...BiasAddAttr) (output tf.Output)
Adds `bias` to `value`.
This is a special case of `tf.add` where `bias` is restricted to be 1-D. Broadcasting is supported, so `value` may have any number of dimensions.
Arguments:
value: Any number of dimensions. bias: 1-D with size the last dimension of `value`.
Returns Broadcasted sum of `value` and `bias`.
func BiasAddGrad ¶
func BiasAddGrad(scope *Scope, out_backprop tf.Output, optional ...BiasAddGradAttr) (output tf.Output)
The backward operation for "BiasAdd" on the "bias" tensor.
It accumulates all the values from out_backprop into the feature dimension. For NHWC data format, the feature dimension is the last. For NCHW data format, the feature dimension is the third-to-last.
Arguments:
out_backprop: Any number of dimensions.
Returns 1-D with size the feature dimension of `out_backprop`.
func BiasAddV1 ¶
func BiasAddV1(scope *Scope, value tf.Output, bias tf.Output) (output tf.Output)
Adds `bias` to `value`.
This is a deprecated version of BiasAdd and will be soon removed.
This is a special case of `tf.add` where `bias` is restricted to be 1-D. Broadcasting is supported, so `value` may have any number of dimensions.
Arguments:
value: Any number of dimensions. bias: 1-D with size the last dimension of `value`.
Returns Broadcasted sum of `value` and `bias`.
func Bincount ¶
func Bincount(scope *Scope, arr tf.Output, size tf.Output, weights tf.Output) (bins tf.Output)
Counts the number of occurrences of each value in an integer array.
Outputs a vector with length `size` and the same dtype as `weights`. If `weights` are empty, then index `i` stores the number of times the value `i` is counted in `arr`. If `weights` are non-empty, then index `i` stores the sum of the value in `weights` at each index where the corresponding value in `arr` is `i`.
Values in `arr` outside of the range [0, size) are ignored.
Arguments:
arr: int32 `Tensor`. size: non-negative int32 scalar `Tensor`. weights: is an int32, int64, float32, or float64 `Tensor` with the same
shape as `arr`, or a length-0 `Tensor`, in which case it acts as all weights equal to 1.
Returns 1D `Tensor` with length equal to `size`. The counts or summed weights for each value in the range [0, size).
func Bitcast ¶
func Bitcast(scope *Scope, input tf.Output, type_ tf.DataType) (output tf.Output)
Bitcasts a tensor from one type to another without copying data.
Given a tensor `input`, this operation returns a tensor that has the same buffer data as `input` with datatype `type`.
If the input datatype `T` is larger than the output datatype `type` then the shape changes from [...] to [..., sizeof(`T`)/sizeof(`type`)].
If `T` is smaller than `type`, the operator requires that the rightmost dimension be equal to sizeof(`type`)/sizeof(`T`). The shape then goes from [..., sizeof(`type`)/sizeof(`T`)] to [...].
*NOTE*: Bitcast is implemented as a low-level cast, so machines with different endian orderings will give different results.
func BroadcastArgs ¶
func BroadcastArgs(scope *Scope, s0 tf.Output, s1 tf.Output) (r0 tf.Output)
Return the shape of s0 op s1 with broadcast.
Given `s0` and `s1`, tensors that represent shapes, compute `r0`, the broadcasted shape. `s0`, `s1` and `r0` are all integer vectors.
func BroadcastGradientArgs ¶
func BroadcastGradientArgs(scope *Scope, s0 tf.Output, s1 tf.Output) (r0 tf.Output, r1 tf.Output)
Return the reduction indices for computing gradients of s0 op s1 with broadcast.
This is typically used by gradient computations for a broadcasting operation.
func CTCBeamSearchDecoder ¶
func CTCBeamSearchDecoder(scope *Scope, inputs tf.Output, sequence_length tf.Output, beam_width int64, top_paths int64, optional ...CTCBeamSearchDecoderAttr) (decoded_indices []tf.Output, decoded_values []tf.Output, decoded_shape []tf.Output, log_probability tf.Output)
Performs beam search decoding on the logits given in input.
A note about the attribute merge_repeated: For the beam search decoder, this means that if consecutive entries in a beam are the same, only the first of these is emitted. That is, when the top path is "A B B B B", "A B" is returned if merge_repeated = True but "A B B B B" is returned if merge_repeated = False.
Arguments:
inputs: 3-D, shape: `(max_time x batch_size x num_classes)`, the logits. sequence_length: A vector containing sequence lengths, size `(batch)`. beam_width: A scalar >= 0 (beam search beam width). top_paths: A scalar >= 0, <= beam_width (controls output size).
Returns 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 ¶
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 ¶
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 Cast ¶
func Cast(scope *Scope, x tf.Output, DstT tf.DataType) (y tf.Output)
Cast x of type SrcT to y of DstT.
func Ceil ¶
func Ceil(scope *Scope, x tf.Output) (y tf.Output)
Returns element-wise smallest integer in not less than x.
func CheckNumerics ¶
func CheckNumerics(scope *Scope, tensor tf.Output, message string) (output tf.Output)
Checks a tensor for NaN and Inf values.
When run, reports an `InvalidArgument` error if `tensor` has any values that are not a number (NaN) or infinity (Inf). Otherwise, passes `tensor` as-is.
Arguments:
message: Prefix of the error message.
func Cholesky ¶
func Cholesky(scope *Scope, input tf.Output) (output tf.Output)
Computes the Cholesky decomposition of one or more square matrices.
The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices, with the same constraints as the single matrix Cholesky decomposition above. The output is a tensor of the same shape as the input containing the Cholesky decompositions for all input submatrices `[..., :, :]`.
Arguments:
input: Shape is `[..., M, M]`.
Returns Shape is `[..., M, M]`.
func CholeskyGrad ¶
func CholeskyGrad(scope *Scope, l tf.Output, grad tf.Output) (output tf.Output)
Computes the reverse mode backpropagated gradient of the Cholesky algorithm.
For an explanation see "Differentiation of the Cholesky algorithm" by Iain Murray http://arxiv.org/abs/1602.07527.
Arguments:
l: Output of batch Cholesky algorithm l = cholesky(A). Shape is `[..., M, M]`.
Algorithm depends only on lower triangular part of the innermost matrices of this tensor.
grad: df/dl where f is some scalar function. Shape is `[..., M, M]`.
Algorithm depends only on lower triangular part of the innermost matrices of this tensor.
Returns Symmetrized version of df/dA . Shape is `[..., M, M]`
func Complex ¶
func Complex(scope *Scope, real tf.Output, imag tf.Output, optional ...ComplexAttr) (out tf.Output)
Converts two real numbers to a complex number.
Given a tensor `real` representing the real part of a complex number, and a tensor `imag` representing the imaginary part of a complex number, this operation returns complex numbers elementwise of the form \\(a + bj\\), where *a* represents the `real` part and *b* represents the `imag` part.
The input tensors `real` and `imag` must have the same shape.
For example:
“` # tensor 'real' is [2.25, 3.25] # tensor `imag` is [4.75, 5.75] tf.complex(real, imag) ==> [[2.25 + 4.75j], [3.25 + 5.75j]] “`
func ComplexAbs ¶
func ComplexAbs(scope *Scope, x tf.Output, optional ...ComplexAbsAttr) (y tf.Output)
Computes the complex absolute value of a tensor.
Given a tensor `x` of complex numbers, this operation returns a tensor of type `float` or `double` that is the absolute value of each element in `x`. All elements in `x` must be complex numbers of the form \\(a + bj\\). The absolute value is computed as \\( \sqrt{a^2 + b^2}\\).
func ComputeAccidentalHits ¶
func ComputeAccidentalHits(scope *Scope, true_classes tf.Output, sampled_candidates tf.Output, num_true int64, optional ...ComputeAccidentalHitsAttr) (indices tf.Output, ids tf.Output, weights tf.Output)
Computes the ids of the positions in sampled_candidates that match true_labels.
When doing log-odds NCE, the result of this op should be passed through a SparseToDense op, then added to the logits of the sampled candidates. This has the effect of 'removing' the sampled labels that match the true labels by making the classifier sure that they are sampled labels.
Arguments:
true_classes: The true_classes output of UnpackSparseLabels. sampled_candidates: The sampled_candidates output of CandidateSampler. num_true: Number of true labels per context.
Returns 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 ¶
func Concat(scope *Scope, concat_dim tf.Output, values []tf.Output) (output tf.Output)
Concatenates tensors along one dimension.
Arguments:
concat_dim: 0-D. The dimension along which to concatenate. Must be in the
range [0, rank(values)).
values: The `N` Tensors to concatenate. Their ranks and types must match,
and their sizes must match in all dimensions except `concat_dim`.
Returns A `Tensor` with the concatenation of values stacked along the `concat_dim` dimension. This tensor's shape matches that of `values` except in `concat_dim` where it has the sum of the sizes.
func ConcatOffset ¶
func ConcatOffset(scope *Scope, concat_dim tf.Output, shape []tf.Output) (offset []tf.Output)
Computes offsets of concat inputs within its output.
For example:
“`prettyprint # '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] “`
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.
This is typically used by gradient computations for a concat operation.
func ConcatV2 ¶
func ConcatV2(scope *Scope, values []tf.Output, axis tf.Output) (output tf.Output)
Concatenates tensors along one dimension.
Arguments:
values: List of `N` Tensors to concatenate. Their ranks and types must match,
and their sizes must match in all dimensions except `concat_dim`.
axis: 0-D. The dimension along which to concatenate. Must be in the
range [-rank(values), rank(values)).
Returns A `Tensor` with the concatenation of values stacked along the `concat_dim` dimension. This tensor's shape matches that of `values` except in `concat_dim` where it has the sum of the sizes.
func Conj ¶
func Conj(scope *Scope, input tf.Output) (output tf.Output)
Returns the complex conjugate of a complex number.
Given a tensor `input` of complex numbers, this operation returns a tensor of complex numbers that are the complex conjugate of each element in `input`. The complex numbers in `input` must be of the form \\(a + bj\\), where *a* is the real part and *b* is the imaginary part.
The complex conjugate returned by this operation is of the form \\(a - bj\\).
For example:
“` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.conj(input) ==> [-2.25 - 4.75j, 3.25 - 5.75j] “`
func Const ¶
func Const(scope *Scope, value interface{}) (output tf.Output)
Const adds an operation to graph that produces value as output.
func ControlTrigger ¶
func ControlTrigger(scope *Scope) (o *tf.Operation)
Does nothing. Serves as a control trigger for scheduling.
Only useful as a placeholder for control edges.
Returns the created operation.
func Conv2D ¶
func Conv2D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...Conv2DAttr) (output tf.Output)
Computes a 2-D convolution given 4-D `input` and `filter` tensors.
Given an input tensor of shape `[batch, in_height, in_width, in_channels]` and a filter / kernel tensor of shape `[filter_height, filter_width, in_channels, out_channels]`, this op performs the following:
1. Flattens the filter to a 2-D matrix with shape
`[filter_height * filter_width * in_channels, output_channels]`.
2. Extracts image patches from the input tensor to form a *virtual*
tensor of shape `[batch, out_height, out_width, filter_height * filter_width * in_channels]`.
3. For each patch, right-multiplies the filter matrix and the image patch
vector.
In detail, with the default NHWC format,
output[b, i, j, k] =
sum_{di, dj, q} input[b, strides[1] * i + di, strides[2] * j + dj, q] *
filter[di, dj, q, k]
Must have `strides[0] = strides[3] = 1`. For the most common case of the same horizontal and vertices strides, `strides = [1, stride, stride, 1]`.
Arguments:
input: A 4-D tensor. The dimension order is interpreted according to the value
of `data_format`, see below for details.
filter: A 4-D tensor of shape
`[filter_height, filter_width, in_channels, out_channels]`
strides: 1-D tensor of length 4. The stride of the sliding window for each
dimension of `input`. The dimension order is determined by the value of
`data_format`, see below for details. padding: The type of padding algorithm to use.
Returns A 4-D tensor. The dimension order is determined by the value of `data_format`, see below for details.
func Conv2DBackpropFilter ¶
func Conv2DBackpropFilter(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropFilterAttr) (output tf.Output)
Computes the gradients of convolution with respect to the filter.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, in_channels]`. filter_sizes: An integer vector representing the tensor shape of `filter`,
where `filter` is a 4-D `[filter_height, filter_width, in_channels, out_channels]` tensor.
out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.
Gradients w.r.t. the output of the convolution.
strides: The stride of the sliding window for each dimension of the input
of the convolution. Must be in the same order as the dimension specified with format.
padding: The type of padding algorithm to use.
Returns 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`. Gradient w.r.t. the `filter` input of the convolution.
func Conv2DBackpropInput ¶
func Conv2DBackpropInput(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...Conv2DBackpropInputAttr) (output tf.Output)
Computes the gradients of convolution with respect to the input.
Arguments:
input_sizes: An integer vector representing the shape of `input`,
where `input` is a 4-D `[batch, height, width, channels]` tensor.
filter: 4-D with shape
`[filter_height, filter_width, in_channels, out_channels]`.
out_backprop: 4-D with shape `[batch, out_height, out_width, out_channels]`.
Gradients w.r.t. the output of the convolution.
strides: The stride of the sliding window for each dimension of the input
of the convolution. Must be in the same order as the dimension specified with format.
padding: The type of padding algorithm to use.
Returns 4-D with shape `[batch, in_height, in_width, in_channels]`. Gradient w.r.t. the input of the convolution.
func Conv3D ¶
func Conv3D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string) (output tf.Output)
Computes a 3-D convolution given 5-D `input` and `filter` tensors.
In signal processing, cross-correlation is a measure of similarity of two waveforms as a function of a time-lag applied to one of them. This is also known as a sliding dot product or sliding inner-product.
Our Conv3D implements a form of cross-correlation.
Arguments:
input: Shape `[batch, in_depth, in_height, in_width, in_channels]`. filter: Shape `[filter_depth, filter_height, filter_width, in_channels,
out_channels]`. `in_channels` must match between `input` and `filter`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
func Conv3DBackpropFilter ¶
func Conv3DBackpropFilter(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string) (output tf.Output)
Computes the gradients of 3-D convolution with respect to the filter.
DEPRECATED at GraphDef version 10: Use Conv3DBackpropFilterV2
Arguments:
input: Shape `[batch, depth, rows, cols, in_channels]`. filter: Shape `[depth, rows, cols, in_channels, out_channels]`.
`in_channels` must match between `input` and `filter`.
out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,
out_channels]`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
func Conv3DBackpropFilterV2 ¶
func Conv3DBackpropFilterV2(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string) (output tf.Output)
Computes the gradients of 3-D convolution with respect to the filter.
Arguments:
input: Shape `[batch, depth, rows, cols, in_channels]`. filter_sizes: An integer vector representing the tensor shape of `filter`,
where `filter` is a 5-D `[filter_depth, filter_height, filter_width, in_channels, out_channels]` tensor.
out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,
out_channels]`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
func Conv3DBackpropInput ¶
func Conv3DBackpropInput(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string) (output tf.Output)
Computes the gradients of 3-D convolution with respect to the input.
DEPRECATED at GraphDef version 10: Use Conv3DBackpropInputV2
Arguments:
input: Shape `[batch, depth, rows, cols, in_channels]`. filter: Shape `[depth, rows, cols, in_channels, out_channels]`.
`in_channels` must match between `input` and `filter`.
out_backprop: Backprop signal of shape `[batch, out_depth, out_rows, out_cols,
out_channels]`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
func Conv3DBackpropInputV2 ¶
func Conv3DBackpropInputV2(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string) (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 ¶
func Cos(scope *Scope, x tf.Output) (y tf.Output)
Computes cos of x element-wise.
func CropAndResize ¶
func CropAndResize(scope *Scope, image tf.Output, boxes tf.Output, box_ind tf.Output, crop_size tf.Output, optional ...CropAndResizeAttr) (crops tf.Output)
Extracts crops from the input image tensor and bilinearly resizes them (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 interpolation) to a fixed `size = [crop_height, crop_width]`. The result is a 4-D tensor `[num_boxes, crop_height, crop_width, depth]`.
Arguments:
image: A 4-D tensor of shape `[batch, image_height, image_width, depth]`.
Both `image_height` and `image_width` need to be positive.
boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values.
box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
crop_size: A 1-D tensor of 2 elements, `size = [crop_height, crop_width]`. All
cropped image patches are resized to this size. The aspect ratio of the image content is not preserved. Both `crop_height` and `crop_width` need to be positive.
Returns A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.
func CropAndResizeGradBoxes ¶
func CropAndResizeGradBoxes(scope *Scope, grads tf.Output, image tf.Output, boxes tf.Output, box_ind tf.Output, optional ...CropAndResizeGradBoxesAttr) (output tf.Output)
Computes the gradient of the crop_and_resize op wrt the input boxes tensor.
Arguments:
grads: A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`. image: A 4-D tensor of shape `[batch, image_height, image_width, depth]`.
Both `image_height` and `image_width` need to be positive.
boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values.
box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
Returns A 2-D tensor of shape `[num_boxes, 4]`.
func CropAndResizeGradImage ¶
func CropAndResizeGradImage(scope *Scope, grads tf.Output, boxes tf.Output, box_ind tf.Output, image_size tf.Output, T tf.DataType, optional ...CropAndResizeGradImageAttr) (output tf.Output)
Computes the gradient of the crop_and_resize op wrt the input image tensor.
Arguments:
grads: A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`. boxes: A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of `y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the `[0, 1]` interval of normalized image height is mapped to `[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in which case the sampled crop is an up-down flipped version of the original image. The width dimension is treated similarly. Normalized coordinates outside the `[0, 1]` range are allowed, in which case we use `extrapolation_value` to extrapolate the input image values.
box_ind: A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
image_size: A 1-D tensor with value `[batch, image_height, image_width, depth]`
containing the original image size. Both `image_height` and `image_width` need to be positive.
Returns A 4-D tensor of shape `[batch, image_height, image_width, depth]`.
func Cross ¶
func Cross(scope *Scope, a tf.Output, b tf.Output) (product tf.Output)
Compute the pairwise cross product.
`a` and `b` must be the same shape; they can either be simple 3-element vectors, or any shape where the innermost dimension is 3. In the latter case, each pair of corresponding 3-element vectors is cross-multiplied independently.
Arguments:
a: A tensor containing 3-element vectors. b: Another tensor, of same type and shape as `a`.
Returns Pairwise cross product of the vectors in `a` and `b`.
func Cumprod ¶
func Cumprod(scope *Scope, x tf.Output, axis tf.Output, optional ...CumprodAttr) (out tf.Output)
Compute the cumulative product of the tensor `x` along `axis`.
By default, this op performs an inclusive cumprod, which means that the first element of the input is identical to the first element of the output: “`prettyprint tf.cumprod([a, b, c]) ==> [a, a * b, a * b * c] “`
By setting the `exclusive` kwarg to `True`, an exclusive cumprod is performed instead: “`prettyprint 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: “`prettyprint 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: “`prettyprint tf.cumprod([a, b, c], exclusive=True, reverse=True) ==> [b * c, c, 1] “`
func Cumsum ¶
func Cumsum(scope *Scope, x tf.Output, axis tf.Output, optional ...CumsumAttr) (out tf.Output)
Compute the cumulative sum of the tensor `x` along `axis`.
By default, this op performs an inclusive cumsum, which means that the first element of the input is identical to the first element of the output: “`prettyprint tf.cumsum([a, b, c]) ==> [a, a + b, a + b + c] “`
By setting the `exclusive` kwarg to `True`, an exclusive cumsum is performed instead: “`prettyprint 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: “`prettyprint 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: “`prettyprint tf.cumsum([a, b, c], exclusive=True, reverse=True) ==> [b + c, c, 0] “`
func DecodeBase64 ¶
func DecodeBase64(scope *Scope, input tf.Output) (output tf.Output)
Decode web-safe base64-encoded strings.
Input may or may not have padding at the end. See EncodeBase64 for padding. Web-safe means that input must use - and _ instead of + and /.
Arguments:
input: Base64 strings to decode.
Returns Decoded strings.
func DecodeCSV ¶
func DecodeCSV(scope *Scope, records tf.Output, record_defaults []tf.Output, optional ...DecodeCSVAttr) (output []tf.Output)
Convert CSV records to tensors. Each column maps to one tensor.
RFC 4180 format is expected for the CSV records. (https://tools.ietf.org/html/rfc4180) Note that we allow leading and trailing spaces with int or float field.
Arguments:
records: Each string is a record/row in the csv and all records should have
the same format.
record_defaults: One tensor per column of the input record, with either a
scalar default value for that column or empty if the column is required.
Returns Each tensor will have the same shape as records.
func DecodeGif ¶
func DecodeGif(scope *Scope, contents tf.Output) (image tf.Output)
Decode the first frame of a GIF-encoded image to a uint8 tensor.
GIF with frame or transparency compression are not supported convert animated GIF from compressed to uncompressed by:
convert $src.gif -coalesce $dst.gif
Arguments:
contents: 0-D. The GIF-encoded image.
Returns 4-D with shape `[num_frames, height, width, 3]`. RGB order
func DecodeJSONExample ¶
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 ¶
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.
Arguments:
contents: 0-D. The JPEG-encoded image.
Returns 3-D with shape `[height, width, channels]`..
func DecodePng ¶
func DecodePng(scope *Scope, contents tf.Output, optional ...DecodePngAttr) (image tf.Output)
Decode a PNG-encoded image to a uint8 or uint16 tensor.
The attr `channels` indicates the desired number of color channels for the decoded image.
Accepted values are:
* 0: Use the number of channels in the PNG-encoded image. * 1: output a grayscale image. * 3: output an RGB image. * 4: output an RGBA image.
If needed, the PNG-encoded image is transformed to match the requested number of color channels.
Arguments:
contents: 0-D. The PNG-encoded image.
Returns 3-D with shape `[height, width, channels]`.
func DecodeRaw ¶
func DecodeRaw(scope *Scope, bytes tf.Output, out_type tf.DataType, optional ...DecodeRawAttr) (output tf.Output)
Reinterpret the bytes of a string as a vector of numbers.
Arguments:
bytes: All the elements must have the same length.
Returns A Tensor with one more dimension than the input `bytes`. The added dimension will have size equal to the length of the elements of `bytes` divided by the number of bytes to represent `out_type`.
func DeleteSessionTensor ¶
func DeleteSessionTensor(scope *Scope, handle tf.Output) (o *tf.Operation)
Delete the tensor specified by its handle in the session.
Arguments:
handle: The handle for a tensor stored in the session state.
Returns the created operation.
func DenseToDenseSetOperation ¶
func DenseToDenseSetOperation(scope *Scope, set1 tf.Output, set2 tf.Output, set_operation string, optional ...DenseToDenseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)
Applies set operation along last dimension of 2 `Tensor` inputs.
See SetOperationOp::SetOperationFromContext for values of `set_operation`.
Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.
Arguments:
set1: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set2`.
Dimension `n` contains values in a set, duplicates are allowed but ignored.
set2: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set1`.
Dimension `n` contains values in a set, duplicates are allowed but ignored.
Returns 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 DenseToSparseSetOperation ¶
func DenseToSparseSetOperation(scope *Scope, set1 tf.Output, set2_indices tf.Output, set2_values tf.Output, set2_shape tf.Output, set_operation string, optional ...DenseToSparseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)
Applies set operation along last dimension of `Tensor` and `SparseTensor`.
See SetOperationOp::SetOperationFromContext for values of `set_operation`.
Input `set2` is a `SparseTensor` represented by `set2_indices`, `set2_values`, and `set2_shape`. For `set2` ranked `n`, 1st `n-1` dimensions must be the same as `set1`. Dimension `n` contains values in a set, duplicates are allowed but ignored.
If `validate_indices` is `True`, this op validates the order and range of `set2` indices.
Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.
Arguments:
set1: `Tensor` with rank `n`. 1st `n-1` dimensions must be the same as `set2`.
Dimension `n` contains values in a set, duplicates are allowed but ignored.
set2_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major
order.
set2_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major
order.
set2_shape: 1D `Tensor`, shape of a `SparseTensor`. `set2_shape[0...n-1]` must
be the same as the 1st `n-1` dimensions of `set1`, `result_shape[n]` is the max set size across `n-1` dimensions.
Returns 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 ¶
func DepthToSpace(scope *Scope, input tf.Output, block_size int64) (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 depth of the input tensor must be divisible by `block_size * block_size`.
That is, assuming the input is in the shape: `[batch, height, width, depth]`, the shape of the output will be: `[batch, height*block_size, width*block_size, depth/(block_size*block_size)]`
This operation requires that the input tensor be of rank 4, and that `block_size` be >=1 and that `block_size * block_size` be a divisor of the input depth.
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 this input of shape `[1, 1, 1, 4]`, and a block size of 2:
“`prettyprint x = [[[[1, 2, 3, 4]]]]
“`
This operation will output a tensor of shape `[1, 2, 2, 1]`:
“`prettyprint
[[[[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.
“`prettyprint 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]`
“`prettyprint
[[[[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:
“`prettyprint 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]`:
“`prettyprint x = [[ [1], [2], [5], [6]],
[ [3], [4], [7], [8]], [ [9], [10], [13], [14]], [ [11], [12], [15], [16]]]
“`
Arguments:
block_size: The size of the spatial block, same as in Space2Depth.
func DepthwiseConv2dNative ¶
func DepthwiseConv2dNative(scope *Scope, input tf.Output, filter tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeAttr) (output tf.Output)
Computes a 2-D depthwise convolution given 4-D `input` and `filter` tensors.
Given an input tensor of shape `[batch, in_height, in_width, in_channels]` and a filter / kernel tensor of shape `[filter_height, filter_width, in_channels, channel_multiplier]`, containing `in_channels` convolutional filters of depth 1, `depthwise_conv2d` applies a different filter to each input channel (expanding from 1 channel to `channel_multiplier` channels for each), then concatenates the results together. Thus, the output has `in_channels * channel_multiplier` channels.
for k in 0..in_channels-1
for q in 0..channel_multiplier-1
output[b, i, j, k * channel_multiplier + q] =
sum_{di, dj} input[b, strides[1] * i + di, strides[2] * j + dj, k] *
filter[di, dj, k, q]
Must have `strides[0] = strides[3] = 1`. For the most common case of the same horizontal and vertices strides, `strides = [1, stride, stride, 1]`.
Arguments:
strides: 1-D of length 4. The stride of the sliding window for each dimension
of `input`.
padding: The type of padding algorithm to use.
func DepthwiseConv2dNativeBackpropFilter ¶
func DepthwiseConv2dNativeBackpropFilter(scope *Scope, input tf.Output, filter_sizes tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeBackpropFilterAttr) (output tf.Output)
Computes the gradients of depthwise convolution with respect to the filter.
Arguments:
input: 4-D with shape based on `data_format`. For example, if
`data_format` is 'NHWC' then `input` is a 4-D `[batch, in_height, in_width, in_channels]` tensor.
filter_sizes: An integer vector representing the tensor shape of `filter`,
where `filter` is a 4-D `[filter_height, filter_width, in_channels, depthwise_multiplier]` tensor.
out_backprop: 4-D with shape based on `data_format`.
For example, if `data_format` is 'NHWC' then out_backprop shape is `[batch, out_height, out_width, out_channels]`. Gradients w.r.t. the output of the convolution.
strides: The stride of the sliding window for each dimension of the input
of the convolution.
padding: The type of padding algorithm to use.
Returns 4-D with shape `[filter_height, filter_width, in_channels, out_channels]`. Gradient w.r.t. the `filter` input of the convolution.
func DepthwiseConv2dNativeBackpropInput ¶
func DepthwiseConv2dNativeBackpropInput(scope *Scope, input_sizes tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, padding string, optional ...DepthwiseConv2dNativeBackpropInputAttr) (output tf.Output)
Computes the gradients of depthwise convolution with respect to the input.
Arguments:
input_sizes: An integer vector representing the shape of `input`, based
on `data_format`. For example, if `data_format` is 'NHWC' then
`input` is a 4-D `[batch, height, width, channels]` tensor. filter: 4-D with shape
`[filter_height, filter_width, in_channels, depthwise_multiplier]`.
out_backprop: 4-D with shape based on `data_format`.
For example, if `data_format` is 'NHWC' then out_backprop shape is `[batch, out_height, out_width, out_channels]`. Gradients w.r.t. the output of the convolution.
strides: The stride of the sliding window for each dimension of the input
of the convolution.
padding: The type of padding algorithm to use.
Returns 4-D with shape according to `data_format`. For example, if `data_format` is 'NHWC', output shape is `[batch, in_height, in_width, in_channels]`. Gradient w.r.t. the input of the convolution.
func Dequantize ¶
func Dequantize(scope *Scope, input tf.Output, min_range tf.Output, max_range tf.Output, optional ...DequantizeAttr) (output tf.Output)
Dequantize the 'input' tensor into a float 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:
“` number_of_steps = 1 << (# of bits in T) range_adjust = number_of_steps / (number_of_steps - 1) range = (range_max - range_min) * range_adjust range_scale = range / number_of_steps const double offset_input = static_cast<double>(input) - lowest_quantized; result = range_min + ((input - numeric_limits<T>::min()) * range_scale) “`
Arguments:
min_range: The minimum scalar value possibly produced for the input. max_range: The maximum scalar value possibly produced for the input.
func DeserializeManySparse ¶
func DeserializeManySparse(scope *Scope, serialized_sparse tf.Output, dtype tf.DataType) (sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output)
Deserialize and concatenate `SparseTensors` from a serialized minibatch.
The input `serialized_sparse` must be a string matrix of shape `[N x 3]` where `N` is the minibatch size and the rows correspond to packed outputs of `SerializeSparse`. The ranks of the original `SparseTensor` objects must all match. When the final `SparseTensor` is created, it has rank one higher than the ranks of the incoming `SparseTensor` objects (they have been concatenated along a new row dimension).
The output `SparseTensor` object's shape values for all dimensions but the first are the max across the input `SparseTensor` objects' shape values for the corresponding dimensions. Its first shape value is `N`, the minibatch size.
The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, after this step run `SparseReorder` to restore index ordering.
For example, if the serialized input is a `[2 x 3]` matrix representing two original `SparseTensor` objects:
index = [ 0]
[10]
[20]
values = [1, 2, 3]
shape = [50]
and
index = [ 2]
[10]
values = [4, 5]
shape = [30]
then the final deserialized `SparseTensor` will be:
index = [0 0]
[0 10]
[0 20]
[1 2]
[1 10]
values = [1, 2, 3, 4, 5]
shape = [2 50]
Arguments:
serialized_sparse: 2-D, The `N` serialized `SparseTensor` objects.
Must have 3 columns.
dtype: The `dtype` of the serialized `SparseTensor` objects.
func DestroyResourceOp ¶
func DestroyResourceOp(scope *Scope, resource tf.Output, optional ...DestroyResourceOpAttr) (o *tf.Operation)
Deletes the resource specified by the handle.
All subsequent operations using the resource will result in a NotFound error status.
Arguments:
resource: handle to the resource to delete.
Returns the created operation.
func Diag ¶
func Diag(scope *Scope, diagonal tf.Output) (output tf.Output)
Returns a diagonal tensor with a given diagonal values.
Given a `diagonal`, this operation returns a tensor with the `diagonal` and everything else padded with zeros. The diagonal is computed as follows:
Assume `diagonal` has dimensions [D1,..., Dk], then the output is a tensor of rank 2k with dimensions [D1,..., Dk, D1,..., Dk] where:
`output[i1,..., ik, i1,..., ik] = diagonal[i1, ..., ik]` and 0 everywhere else.
For example:
“`prettyprint # '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 3.
func DiagPart ¶
func DiagPart(scope *Scope, input tf.Output) (diagonal tf.Output)
Returns the diagonal part of the tensor.
This operation returns a tensor with the `diagonal` part of the `input`. The `diagonal` part is computed as follows:
Assume `input` has dimensions `[D1,..., Dk, D1,..., Dk]`, then the output is a tensor of rank `k` with dimensions `[D1,..., Dk]` where:
`diagonal[i1,..., ik] = input[i1, ..., ik, i1,..., ik]`.
For example:
“`prettyprint # '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 2, 4, or 6.
Returns The extracted diagonal.
func Digamma ¶
func Digamma(scope *Scope, x tf.Output) (y tf.Output)
Computes Psi, the derivative of Lgamma (the log of the absolute value of
`Gamma(x)`), element-wise.
func Dilation2D ¶
func Dilation2D(scope *Scope, input tf.Output, filter tf.Output, strides []int64, rates []int64, padding string) (output tf.Output)
Computes the grayscale dilation of 4-D `input` and 3-D `filter` tensors.
The `input` tensor has shape `[batch, in_height, in_width, depth]` and the `filter` tensor has shape `[filter_height, filter_width, depth]`, i.e., each input channel is processed independently of the others with its own structuring function. The `output` tensor has shape `[batch, out_height, out_width, depth]`. The spatial dimensions of the output tensor depend on the `padding` algorithm. We currently only support the default "NHWC" `data_format`.
In detail, the grayscale morphological 2-D dilation is the max-sum correlation (for consistency with `conv2d`, we use unmirrored filters):
output[b, y, x, c] =
max_{dy, dx} input[b,
strides[1] * y + rates[1] * dy,
strides[2] * x + rates[2] * dx,
c] +
filter[dy, dx, c]
Max-pooling is a special case when the filter has size equal to the pooling kernel size and contains all zeros.
Note on duality: The dilation of `input` by the `filter` is equal to the negation of the erosion of `-input` by the reflected `filter`.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, depth]`. filter: 3-D with shape `[filter_height, filter_width, depth]`. strides: The stride of the sliding window for each dimension of the input
tensor. Must be: `[1, stride_height, stride_width, 1]`.
rates: The input stride for atrous morphological dilation. Must be:
`[1, rate_height, rate_width, 1]`.
padding: The type of padding algorithm to use.
Returns 4-D with shape `[batch, out_height, out_width, depth]`.
func Dilation2DBackpropFilter ¶
func Dilation2DBackpropFilter(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, rates []int64, padding string) (filter_backprop tf.Output)
Computes the gradient of morphological 2-D dilation with respect to the filter.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, depth]`. filter: 3-D with shape `[filter_height, filter_width, depth]`. out_backprop: 4-D with shape `[batch, out_height, out_width, depth]`. strides: 1-D of length 4. The stride of the sliding window for each dimension of
the input tensor. Must be: `[1, stride_height, stride_width, 1]`.
rates: 1-D of length 4. The input stride for atrous morphological dilation.
Must be: `[1, rate_height, rate_width, 1]`.
padding: The type of padding algorithm to use.
Returns 3-D with shape `[filter_height, filter_width, depth]`.
func Dilation2DBackpropInput ¶
func Dilation2DBackpropInput(scope *Scope, input tf.Output, filter tf.Output, out_backprop tf.Output, strides []int64, rates []int64, padding string) (in_backprop tf.Output)
Computes the gradient of morphological 2-D dilation with respect to the input.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, depth]`. filter: 3-D with shape `[filter_height, filter_width, depth]`. out_backprop: 4-D with shape `[batch, out_height, out_width, depth]`. strides: 1-D of length 4. The stride of the sliding window for each dimension of
the input tensor. Must be: `[1, stride_height, stride_width, 1]`.
rates: 1-D of length 4. The input stride for atrous morphological dilation.
Must be: `[1, rate_height, rate_width, 1]`.
padding: The type of padding algorithm to use.
Returns 4-D with shape `[batch, in_height, in_width, depth]`.
func Div ¶
func Div(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x / y element-wise.
*NOTE*: `Div` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func DrawBoundingBoxes ¶
func DrawBoundingBoxes(scope *Scope, images tf.Output, boxes tf.Output) (output tf.Output)
Draw bounding boxes on a batch of images.
Outputs a copy of `images` but draws on top of the pixels zero or more bounding boxes specified by the locations in `boxes`. The coordinates of the each bounding box in `boxes` are encoded as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.
For example, if an image is 100 x 200 pixels and the bounding box is `[0.1, 0.2, 0.5, 0.9]`, the bottom-left and upper-right coordinates of the bounding box will be `(10, 40)` to `(50, 180)`.
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 DynamicPartition ¶
func DynamicPartition(scope *Scope, data tf.Output, partitions tf.Output, num_partitions int64) (outputs []tf.Output)
Partitions `data` into `num_partitions` tensors using indices from `partitions`.
For each index tuple `js` of size `partitions.ndim`, the slice `data[js, ...]` becomes part of `outputs[partitions[js]]`. The slices with `partitions[js] = i` are placed in `outputs[i]` in lexicographic order of `js`, and the first dimension of `outputs[i]` is the number of entries in `partitions` equal to `i`. In detail,
“`python
outputs[i].shape = [sum(partitions == i)] + data.shape[partitions.ndim:] outputs[i] = pack([data[js, ...] for js if partitions[js] == i])
“`
`data.shape` must start with `partitions.shape`.
For example:
“`python
# Scalar partitions. partitions = 1 num_partitions = 2 data = [10, 20] outputs[0] = [] # Empty with shape [0, 2] outputs[1] = [[10, 20]] # Vector partitions. partitions = [0, 0, 1, 1, 0] num_partitions = 2 data = [10, 20, 30, 40, 50] outputs[0] = [10, 20, 50] outputs[1] = [30, 40]
“`
<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="../../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 ¶
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.
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]]
“`
<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="../../images/DynamicStitch.png" alt> </div>
func EditDistance ¶
func EditDistance(scope *Scope, hypothesis_indices tf.Output, hypothesis_values tf.Output, hypothesis_shape tf.Output, truth_indices tf.Output, truth_values tf.Output, truth_shape tf.Output, optional ...EditDistanceAttr) (output tf.Output)
Computes the (possibly normalized) Levenshtein Edit Distance.
The inputs are variable-length sequences provided by SparseTensors
(hypothesis_indices, hypothesis_values, hypothesis_shape)
and
(truth_indices, truth_values, truth_shape).
The inputs are:
Arguments:
hypothesis_indices: The indices of the hypothesis list SparseTensor.
This is an N x R int64 matrix.
hypothesis_values: The values of the hypothesis list SparseTensor.
This is an N-length vector.
hypothesis_shape: The shape of the hypothesis list SparseTensor.
This is an R-length vector.
truth_indices: The indices of the truth list SparseTensor.
This is an M x R int64 matrix.
truth_values: The values of the truth list SparseTensor.
This is an M-length vector.
truth_shape: truth indices, vector.
Returns A dense float tensor with rank R - 1.
For the example input:
// hypothesis represents a 2x1 matrix with variable-length values:
// (0,0) = ["a"]
// (1,0) = ["b"]
hypothesis_indices = [[0, 0, 0],
[1, 0, 0]]
hypothesis_values = ["a", "b"]
hypothesis_shape = [2, 1, 1]
// truth represents a 2x2 matrix with variable-length values:
// (0,0) = []
// (0,1) = ["a"]
// (1,0) = ["b", "c"]
// (1,1) = ["a"]
truth_indices = [[0, 1, 0],
[1, 0, 0],
[1, 0, 1],
[1, 1, 0]]
truth_values = ["a", "b", "c", "a"]
truth_shape = [2, 2, 2]
normalize = true
The output will be:
// output is a 2x2 matrix with edit distances normalized by truth lengths.
output = [[inf, 1.0], // (0,0): no truth, (0,1): no hypothesis
[0.5, 1.0]] // (1,0): addition, (1,1): no hypothesis
func Elu ¶
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 ¶
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 EncodeBase64 ¶
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 ¶
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 EncodePng ¶
func EncodePng(scope *Scope, image tf.Output, optional ...EncodePngAttr) (contents tf.Output)
PNG-encode an image.
`image` is a 3-D uint8 or uint16 Tensor of shape `[height, width, channels]` where `channels` is:
* 1: for grayscale. * 2: for grayscale + alpha. * 3: for RGB. * 4: for RGBA.
The ZLIB compression level, `compression`, can be -1 for the PNG-encoder default or a value from 0 to 9. 9 is the highest compression level, generating the smallest output, but is slower.
Arguments:
image: 3-D with shape `[height, width, channels]`.
Returns 0-D. PNG-encoded image.
func Enter ¶
func Enter(scope *Scope, data tf.Output, frame_name string, optional ...EnterAttr) (output tf.Output)
Creates or finds a child frame, and makes `data` available to the child frame.
This op is used together with `Exit` to create loops in the graph. The unique `frame_name` is used by the `Executor` to identify frames. If `is_constant` is true, `output` is a constant in the child frame; otherwise it may be changed in the child frame. At most `parallel_iterations` iterations are run in parallel in the child frame.
Arguments:
data: The tensor to be made available to the child frame. frame_name: The name of the child frame.
Returns The same tensor as `data`.
func Equal ¶
func Equal(scope *Scope, x tf.Output, y tf.Output) (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)
func Erf ¶
func Erf(scope *Scope, x tf.Output) (y tf.Output)
Computes the Gauss error function of `x` element-wise.
func Erfc ¶
func Erfc(scope *Scope, x tf.Output) (y tf.Output)
Computes the complementary error function of `x` element-wise.
func Exit ¶
func Exit(scope *Scope, data tf.Output) (output tf.Output)
Exits the current frame to its parent frame.
Exit makes its input `data` available to the parent frame.
Arguments:
data: The tensor to be made available to the parent frame.
Returns The same tensor as `data`.
func Exp ¶
func Exp(scope *Scope, x tf.Output) (y tf.Output)
Computes exponential of x element-wise. \\(y = e^x\\).
func ExpandDims ¶
func ExpandDims(scope *Scope, input tf.Output, dim 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 `dim` of `input`'s shape. The dimension index `dim` starts at zero; if you specify a negative number for `dim` 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:
“`prettyprint # '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:
dim: 0-D (scalar). Specifies the dimension index at which to
expand the shape of `input`.
Returns Contains the same data as `input`, but its shape has an additional dimension of size 1 added.
func Expm1 ¶
func Expm1(scope *Scope, x tf.Output) (y tf.Output)
Computes exponential of x - 1 element-wise.
I.e., \\(y = (\exp x) - 1\\).
func ExtractGlimpse ¶
func ExtractGlimpse(scope *Scope, input tf.Output, size tf.Output, offsets tf.Output, optional ...ExtractGlimpseAttr) (glimpse tf.Output)
Extracts a glimpse from the input tensor.
Returns a set of windows called glimpses extracted at location `offsets` from the input tensor. If the windows only partially overlaps the inputs, the non overlapping areas will be filled with random noise.
The result is a 4-D tensor of shape `[batch_size, glimpse_height, glimpse_width, channels]`. The channels and batch dimensions are the same as that of the input tensor. The height and width of the output windows are specified in the `size` parameter.
The argument `normalized` and `centered` controls how the windows are built:
* If the coordinates are normalized but not centered, 0.0 and 1.0
correspond to the minimum and maximum of each height and width dimension.
* If the coordinates are both normalized and centered, they range from
-1.0 to 1.0. The coordinates (-1.0, -1.0) correspond to the upper left corner, the lower right corner is located at (1.0, 1.0) and the center is at (0, 0).
* If the coordinates are not normalized they are interpreted as
numbers of pixels.
Arguments:
input: A 4-D float tensor of shape `[batch_size, height, width, channels]`. size: A 1-D tensor of 2 elements containing the size of the glimpses
to extract. The glimpse height must be specified first, following by the glimpse width.
offsets: A 2-D integer tensor of shape `[batch_size, 2]` containing
the y, x locations of the center of each window.
Returns A tensor representing the glimpses `[batch_size, glimpse_height, glimpse_width, channels]`.
func ExtractImagePatches ¶
func ExtractImagePatches(scope *Scope, images tf.Output, ksizes []int64, strides []int64, rates []int64, padding string) (patches tf.Output)
Extract `patches` from `images` and put them in the "depth" output dimension.
Arguments:
images: 4-D Tensor with shape `[batch, in_rows, in_cols, depth]`. ksizes: The size of the sliding window for each dimension of `images`. strides: 1-D of length 4. How far the centers of two consecutive patches are in
the images. Must be: `[1, stride_rows, stride_cols, 1]`.
rates: 1-D of length 4. 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`.
padding: The type of padding algorithm to use.
We specify the size-related attributes as:
“`python
ksizes = [1, ksize_rows, ksize_cols, 1] strides = [1, strides_rows, strides_cols, 1] rates = [1, rates_rows, rates_cols, 1]
“`
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.
func FFT ¶
func FFT(scope *Scope, input tf.Output) (output tf.Output)
Compute the 1-dimensional discrete Fourier Transform over the inner-most
dimension of `input`.
Arguments:
input: A complex64 tensor.
Returns A complex64 tensor of the same shape as `input`. The inner-most
dimension of `input` is replaced with its 1D Fourier Transform.
@compatibility(numpy) Equivalent to np.fft.fft @end_compatibility
func FFT2D ¶
func FFT2D(scope *Scope, input tf.Output) (output tf.Output)
Compute the 2-dimensional discrete Fourier Transform over the inner-most
2 dimensions of `input`.
Arguments:
input: A complex64 tensor.
Returns A complex64 tensor of the same shape as `input`. The inner-most 2
dimensions of `input` are replaced with their 2D Fourier Transform.
@compatibility(numpy) Equivalent to np.fft.fft2 @end_compatibility
func FFT3D ¶
func FFT3D(scope *Scope, input tf.Output) (output tf.Output)
Compute 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 ¶
func FIFOQueueV2(scope *Scope, component_types []tf.DataType, optional ...FIFOQueueV2Attr) (handle tf.Output)
A queue that produces elements in first-in first-out order.
Arguments:
component_types: The type of each component in a value.
Returns The handle to the queue.
func Fact ¶
func Fact(scope *Scope) (fact tf.Output)
Output a fact about factorials.
func FakeQuantWithMinMaxArgs ¶
func FakeQuantWithMinMaxArgs(scope *Scope, inputs tf.Output, optional ...FakeQuantWithMinMaxArgsAttr) (outputs tf.Output)
Fake-quantize the 'inputs' tensor, type float to 'outputs' tensor of same type.
Attributes [min; max] define the clamping range for the 'inputs' data. Op divides this range into 255 steps (total of 256 values), then replaces each 'inputs' value with the closest of the quantized step values.
Quantization is called fake since the output is still in floating point.
func FakeQuantWithMinMaxArgsGradient ¶
func FakeQuantWithMinMaxArgsGradient(scope *Scope, gradients tf.Output, inputs tf.Output, optional ...FakeQuantWithMinMaxArgsGradientAttr) (backprops tf.Output)
Compute gradients for a FakeQuantWithMinMaxArgs operation.
Arguments:
gradients: Backpropagated gradients above the FakeQuantWithMinMaxArgs operation. inputs: Values passed as inputs to the FakeQuantWithMinMaxArgs operation.
Returns Backpropagated gradients below the FakeQuantWithMinMaxArgs operation: `gradients * (inputs >= min && inputs <= max)`.
func FakeQuantWithMinMaxVars ¶
func FakeQuantWithMinMaxVars(scope *Scope, inputs tf.Output, min tf.Output, max tf.Output) (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] is the clamping range for the 'inputs' data. Op divides this range into 255 steps (total of 256 values), then replaces each 'inputs' value with the closest of the quantized step values.
This operation has a gradient and thus allows for training `min` and `max` values.
func FakeQuantWithMinMaxVarsGradient ¶
func FakeQuantWithMinMaxVarsGradient(scope *Scope, gradients tf.Output, inputs tf.Output, min tf.Output, max tf.Output) (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 ¶
func FakeQuantWithMinMaxVarsPerChannel(scope *Scope, inputs tf.Output, min tf.Output, max tf.Output) (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] is the clamping range for the 'inputs' data in the corresponding depth channel. Op divides this range into 255 steps (total of 256 values), then replaces each 'inputs' value with the closest of the quantized step values.
This operation has a gradient and thus allows for training `min` and `max` values.
func FakeQuantWithMinMaxVarsPerChannelGradient ¶
func FakeQuantWithMinMaxVarsPerChannelGradient(scope *Scope, gradients tf.Output, inputs tf.Output, min tf.Output, max tf.Output) (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 ¶
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:
“`prettyprint # Output tensor has shape [2, 3]. fill([2, 3], 9) ==> [[9, 9, 9]
[9, 9, 9]]
“`
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 FixedLengthRecordReaderV2 ¶
func FixedLengthRecordReaderV2(scope *Scope, record_bytes int64, optional ...FixedLengthRecordReaderV2Attr) (reader_handle tf.Output)
A Reader that outputs fixed-length records from a file.
Returns The handle to reference the Reader.
func FixedUnigramCandidateSampler ¶
func FixedUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...FixedUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a learned unigram distribution.
A unigram sampler could use a fixed unigram distribution read from a file or passed in as an in-memory array instead of building up the distribution from data on the fly. There is also an option to skew the distribution by applying a distortion power to the weights.
The vocabulary file should be in CSV-like format, with the last field being the weight associated with the word.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to randomly sample per batch. 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 ¶
func Floor(scope *Scope, x tf.Output) (y tf.Output)
Returns element-wise largest integer not greater than x.
func FloorDiv ¶
func FloorDiv(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x // y element-wise.
*NOTE*: `FloorDiv` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func FloorMod ¶
func FloorMod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns element-wise remainder of division. 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 ¶
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 ¶
func FractionalAvgPoolGrad(scope *Scope, orig_input_tensor_shape tf.Output, out_backprop tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output, optional ...FractionalAvgPoolGradAttr) (output tf.Output)
Computes gradient of the FractionalAvgPool function.
Unlike FractionalMaxPoolGrad, we don't need to find arg_max for FractionalAvgPoolGrad, we just need to evenly back-propagate each element of out_backprop to those indices that form the same pooling cell. Therefore, we just need to know the shape of original input tensor, instead of the whole tensor.
Arguments:
orig_input_tensor_shape: Original input tensor shape for `fractional_avg_pool` out_backprop: 4-D with shape `[batch, height, width, channels]`. Gradients
w.r.t. the output of `fractional_avg_pool`.
row_pooling_sequence: row pooling sequence, form pooling region with
col_pooling_sequence.
col_pooling_sequence: column pooling sequence, form pooling region with
row_pooling sequence.
Returns 4-D. Gradients w.r.t. the input of `fractional_avg_pool`.
func FractionalMaxPool ¶
func FractionalMaxPool(scope *Scope, value tf.Output, pooling_ratio []float32, optional ...FractionalMaxPoolAttr) (output tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output)
Performs fractional max pooling on the input.
Fractional max pooling is slightly different than regular max pooling. In regular max pooling, you downsize an input set by taking the maximum value of smaller N x N subsections of the set (often 2x2), and try to reduce the set by a factor of N, where N is an integer. Fractional max pooling, as you might expect from the word "fractional", means that the overall reduction ratio N does not have to be an integer.
The sizes of the pooling regions are generated randomly but are fairly uniform. For example, let's look at the height dimension, and the constraints on the list of rows that will be pool boundaries.
First we define the following:
1. input_row_length : the number of rows from the input set 2. output_row_length : which will be smaller than the input 3. alpha = input_row_length / output_row_length : our reduction ratio 4. K = floor(alpha) 5. row_pooling_sequence : this is the result list of pool boundary rows
Then, row_pooling_sequence should satisfy:
1. a[0] = 0 : the first value of the sequence is 0 2. a[end] = input_row_length : the last value of the sequence is the size 3. K <= (a[i+1] - a[i]) <= K+1 : all intervals are K or K+1 size 4. length(row_pooling_sequence) = output_row_length+1
For more details on fractional max pooling, see this paper: [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071)
Arguments:
value: 4-D with shape `[batch, height, width, channels]`. pooling_ratio: Pooling ratio for each dimension of `value`, currently only
supports row and col dimension and should be >= 1.0. For example, a valid pooling ratio looks like [1.0, 1.44, 1.73, 1.0]. The first and last elements must be 1.0 because we don't allow pooling on batch and channels dimensions. 1.44 and 1.73 are pooling ratio on height and width dimensions respectively.
Returns output tensor after fractional max pooling.row pooling sequence, needed to calculate gradient.column pooling sequence, needed to calculate gradient.
func FractionalMaxPoolGrad ¶
func FractionalMaxPoolGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, out_backprop tf.Output, row_pooling_sequence tf.Output, col_pooling_sequence tf.Output, optional ...FractionalMaxPoolGradAttr) (output tf.Output)
Computes gradient of the FractionalMaxPool function.
Arguments:
orig_input: Original input for `fractional_max_pool` orig_output: Original output for `fractional_max_pool` out_backprop: 4-D with shape `[batch, height, width, channels]`. Gradients
w.r.t. the output of `fractional_max_pool`.
row_pooling_sequence: row pooling sequence, form pooling region with
col_pooling_sequence.
col_pooling_sequence: column pooling sequence, form pooling region with
row_pooling sequence.
Returns 4-D. Gradients w.r.t. the input of `fractional_max_pool`.
func FusedBatchNorm ¶
func FusedBatchNorm(scope *Scope, x tf.Output, scale tf.Output, offset tf.Output, mean tf.Output, variance tf.Output, optional ...FusedBatchNormAttr) (y tf.Output, batch_mean tf.Output, batch_variance tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output)
Batch normalization.
Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.
Arguments:
x: A 4D Tensor for input data. scale: A 1D Tensor for scaling factor, to scale the normalized x. offset: A 1D Tensor for offset, to shift to the normalized x. mean: A 1D Tensor for population mean. Used for inference only;
must be empty for training.
variance: A 1D Tensor for population variance. Used for inference only;
must be empty for training.
Returns 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 used in the gradient computation.
func FusedBatchNormGrad ¶
func FusedBatchNormGrad(scope *Scope, y_backprop tf.Output, x tf.Output, scale tf.Output, reserve_space_1 tf.Output, reserve_space_2 tf.Output, optional ...FusedBatchNormGradAttr) (x_backprop tf.Output, scale_backprop tf.Output, offset_backprop tf.Output, reserve_space_3 tf.Output, reserve_space_4 tf.Output)
Gradient for batch normalization.
Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW". The size of 1D Tensors matches the dimension C of the 4D Tensors.
Arguments:
y_backprop: A 4D Tensor for the gradient with respect to y. x: A 4D Tensor for input data. scale: A 1D Tensor for scaling factor, to scale the normalized x. reserve_space_1: A 1D Tensor for the computed batch mean, to be reused
in the gradient computation.
reserve_space_2: A 1D Tensor for the computed batch variance (inverted variance
in the cuDNN case), to be used in the 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 FusedPadConv2D ¶
func FusedPadConv2D(scope *Scope, input tf.Output, paddings tf.Output, filter tf.Output, mode string, strides []int64, padding string) (output tf.Output)
Performs a padding as a preprocess during a convolution.
Similar to FusedResizeAndPadConv2d, this op allows for an optimized implementation where the spatial padding transformation stage is fused with the im2col lookup, but in this case without the bilinear filtering required for resizing. Fusing the padding prevents the need to write out the intermediate results as whole tensors, reducing memory pressure, and we can get some latency gains by merging the transformation calculations. The data_format attribute for Conv2D isn't supported by this op, and 'NHWC' order is used instead. Internally this op uses a single per-graph scratch buffer, which means that it will block if multiple versions are being run in parallel. This is because this operator is primarily an optimization to minimize memory usage.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, in_channels]`. paddings: A two-column matrix specifying the padding sizes. The number of
rows must be the same as the rank of `input`.
filter: 4-D with shape
`[filter_height, filter_width, in_channels, out_channels]`.
strides: 1-D of length 4. The stride of the sliding window for each dimension
of `input`. Must be in the same order as the dimension specified with format.
padding: The type of padding algorithm to use.
func FusedResizeAndPadConv2D ¶
func FusedResizeAndPadConv2D(scope *Scope, input tf.Output, size tf.Output, paddings tf.Output, filter tf.Output, mode string, strides []int64, padding string, optional ...FusedResizeAndPadConv2DAttr) (output tf.Output)
Performs a resize and padding as a preprocess during a convolution.
It's often possible to do spatial transformations more efficiently as part of the packing stage of a convolution, so this op allows for an optimized implementation where these stages are fused together. This prevents the need to write out the intermediate results as whole tensors, reducing memory pressure, and we can get some latency gains by merging the transformation calculations. The data_format attribute for Conv2D isn't supported by this op, and defaults to 'NHWC' order. Internally this op uses a single per-graph scratch buffer, which means that it will block if multiple versions are being run in parallel. This is because this operator is primarily an optimization to minimize memory usage.
Arguments:
input: 4-D with shape `[batch, in_height, in_width, in_channels]`. size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
paddings: A two-column matrix specifying the padding sizes. The number of
rows must be the same as the rank of `input`.
filter: 4-D with shape
`[filter_height, filter_width, in_channels, out_channels]`.
strides: 1-D of length 4. The stride of the sliding window for each dimension
of `input`. Must be in the same order as the dimension specified with format.
padding: The type of padding algorithm to use.
func Gather ¶
func Gather(scope *Scope, params tf.Output, indices tf.Output, optional ...GatherAttr) (output tf.Output)
Gather slices from `params` according to `indices`.
`indices` must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape `indices.shape + params.shape[1:]` where:
“`python
# Scalar indices output[:, ..., :] = params[indices, :, ... :] # Vector indices output[i, :, ..., :] = params[indices[i], :, ... :] # Higher rank indices output[i, ..., j, :, ... :] = params[indices[i, ..., j], :, ..., :]
“`
If `indices` is a permutation and `len(indices) == params.shape[0]` then this operation will permute `params` accordingly.
<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="../../images/Gather.png" alt> </div>
func GatherNd ¶
func GatherNd(scope *Scope, params tf.Output, indices tf.Output) (output tf.Output)
Gather values or slices from `params` according to `indices`.
`params` is a Tensor of rank `P` and `indices` is a Tensor of rank `Q`.
`indices` must be integer tensor, containing indices into `params`. 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 `params`.
Produces an output tensor with shape
“` [d_0, ..., d_{Q-2}, params.shape[K], ..., params.shape[P-1]]. “`
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']]
“`
Arguments:
params: `P-D`. The tensor from which to gather values.
indices: `Q-D`. Index tensor having shape `[d_0, ..., d_{Q-2}, K]`.
Returns `(P+Q-K-1)-D`. Values from `params` gathered from indices given by `indices`.
func GetSessionHandleV2 ¶
func GetSessionHandleV2(scope *Scope, value tf.Output) (handle tf.Output)
Store the input tensor in the state of the current session.
Arguments:
value: The tensor to be stored.
Returns The handle for the tensor stored in the session state, represented as a ResourceHandle object.
func GetSessionTensor ¶
func GetSessionTensor(scope *Scope, handle tf.Output, dtype tf.DataType) (value tf.Output)
Get the value of the tensor specified by its handle.
Arguments:
handle: The handle for a tensor stored in the session state. dtype: The type of the output value.
Returns The tensor for the given handle.
func Greater ¶
func Greater(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of (x > y) element-wise.
*NOTE*: `Greater` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func GreaterEqual ¶
func GreaterEqual(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of (x >= y) element-wise.
*NOTE*: `GreaterEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func HSVToRGB ¶
func HSVToRGB(scope *Scope, images tf.Output) (output tf.Output)
Convert one or more images from HSV to RGB.
Outputs a tensor of the same shape as the `images` tensor, containing the RGB value of the pixels. The output is only well defined if the value in `images` are in `[0,1]`.
See `rgb_to_hsv` for a description of the HSV encoding.
Arguments:
images: 1-D or higher rank. HSV data to convert. Last dimension must be size 3.
Returns `images` converted to RGB.
func HistogramSummary ¶
func HistogramSummary(scope *Scope, tag tf.Output, values tf.Output) (summary tf.Output)
Outputs a `Summary` protocol buffer with a histogram.
The generated [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) has one summary value containing a histogram for `values`.
This op reports an `InvalidArgument` error if any value is not finite.
Arguments:
tag: Scalar. Tag to use for the `Summary.Value`. values: Any shape. Values to use to build the histogram.
Returns Scalar. Serialized `Summary` protocol buffer.
func IFFT ¶
func IFFT(scope *Scope, input tf.Output) (output tf.Output)
Compute the inverse 1-dimensional discrete Fourier Transform over the inner-most
dimension of `input`.
Arguments:
input: A complex64 tensor.
Returns A complex64 tensor of the same shape as `input`. The inner-most
dimension of `input` is replaced with its inverse 1D Fourier Transform.
@compatibility(numpy) Equivalent to np.fft.ifft @end_compatibility
func IFFT2D ¶
func IFFT2D(scope *Scope, input tf.Output) (output tf.Output)
Compute the inverse 2-dimensional discrete Fourier Transform over the inner-most
2 dimensions of `input`.
Arguments:
input: A complex64 tensor.
Returns A complex64 tensor of the same shape as `input`. The inner-most 2
dimensions of `input` are replaced with their inverse 2D Fourier Transform.
@compatibility(numpy) Equivalent to np.fft.ifft2 @end_compatibility
func IFFT3D ¶
func IFFT3D(scope *Scope, input tf.Output) (output tf.Output)
Compute 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 ¶
func IRFFT(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
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 ¶
func IRFFT2D(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
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 ¶
func IRFFT3D(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
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 ¶
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 IdentityReaderV2 ¶
func IdentityReaderV2(scope *Scope, optional ...IdentityReaderV2Attr) (reader_handle tf.Output)
A Reader that outputs the queued work as both the key and value.
To use, enqueue strings in a Queue. ReaderRead will take the front work string and output (work, work).
Returns The handle to reference the Reader.
func Igamma ¶
func Igamma(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)
Compute the lower regularized incomplete Gamma function `Q(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 Igammac ¶
func Igammac(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)
Compute the upper regularized incomplete Gamma function `Q(a, x)`.
The upper regularized incomplete Gamma function is defined as:
“` Q(a, x) = Gamma(a, x) / Gamma(a) = 1 - P(a, x) “` where “` Gamma(a, x) = int_{x}^{\infty} t^{a-1} exp(-t) dt “` is the upper incomplete Gama function.
Note, above `P(a, x)` (`Igamma`) is the lower regularized complete Gamma function.
func Imag ¶
func Imag(scope *Scope, input tf.Output, optional ...ImagAttr) (output tf.Output)
Returns the imaginary part of a complex number.
Given a tensor `input` of complex numbers, this operation returns a tensor of type `float` that is the imaginary part of each element in `input`. All elements in `input` must be complex numbers of the form \\(a + bj\\), where *a* is the real part and *b* is the imaginary part returned by this operation.
For example:
“` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.imag(input) ==> [4.75, 5.75] “`
func ImageSummary ¶
func ImageSummary(scope *Scope, tag tf.Output, tensor tf.Output, optional ...ImageSummaryAttr) (summary tf.Output)
Outputs a `Summary` protocol buffer with images.
The summary has up to `max_images` summary values containing images. The images are built from `tensor` which must be 4-D with shape `[batch_size, height, width, channels]` and where `channels` can be:
* 1: `tensor` is interpreted as Grayscale. * 3: `tensor` is interpreted as RGB. * 4: `tensor` is interpreted as RGBA.
The images have the same number of channels as the input tensor. For float input, the values are normalized one image at a time to fit in the range `[0, 255]`. `uint8` values are unchanged. The op uses two different normalization algorithms:
* If the input values are all positive, they are rescaled so the largest one
is 255.
* If any input value is negative, the values are shifted so input value 0.0
is at 127. They are then rescaled so that either the smallest value is 0, or the largest one is 255.
The `tag` argument is a scalar `Tensor` of type `string`. It is used to build the `tag` of the summary values:
* If `max_images` is 1, the summary value tag is '*tag*/image'. * If `max_images` is greater than 1, the summary value tags are
generated sequentially as '*tag*/image/0', '*tag*/image/1', etc.
The `bad_color` argument is the color to use in the generated images for non-finite input values. It is a `unit8` 1-D tensor of length `channels`. Each element must be in the range `[0, 255]` (It represents the value of a pixel in the output image). Non-finite values in the input tensor are replaced by this tensor in the output image. The default value is the color red.
Arguments:
tag: Scalar. Used to build the `tag` attribute of the summary values. tensor: 4-D of shape `[batch_size, height, width, channels]` where
`channels` is 1, 3, or 4.
Returns Scalar. Serialized `Summary` protocol buffer.
func ImmutableConst ¶
func ImmutableConst(scope *Scope, dtype tf.DataType, shape tf.Shape, memory_region_name string) (tensor tf.Output)
Returns immutable tensor from memory region.
The current implementation memmaps the tensor from a file.
Arguments:
dtype: Type of the returned tensor. shape: Shape of the returned tensor. memory_region_name: Name of readonly memory region used by the tensor, see
NewReadOnlyMemoryRegionFromFile in tensorflow::Env.
func InTopK ¶
func InTopK(scope *Scope, predictions tf.Output, targets tf.Output, k int64) (precision tf.Output)
Says whether the targets are in the top `K` predictions.
This outputs a `batch_size` bool array, an entry `out[i]` is `true` if the prediction for the target class is among the top `k` predictions among all predictions for example `i`. Note that the behavior of `InTopK` differs from the `TopK` op in its handling of ties; if multiple classes have the same prediction value and straddle the top-`k` boundary, all of those classes are considered to be in the top `k`.
More formally, let
\\(predictions_i\\) be the predictions for all classes for example `i`, \\(targets_i\\) be the target class for example `i`, \\(out_i\\) be the output for example `i`,
$$out_i = predictions_{i, targets_i} \in TopKIncludingTies(predictions_i)$$
Arguments:
predictions: A `batch_size` x `classes` tensor. targets: A `batch_size` vector of class ids. k: Number of top elements to look at for computing precision.
Returns Computed Precision at `k` as a `bool Tensor`.
func Inv ¶
func Inv(scope *Scope, x tf.Output) (y tf.Output)
Computes the reciprocal of x element-wise.
DEPRECATED at GraphDef version 17: Use Reciprocal
I.e., \\(y = 1 / x\\).
func InvGrad ¶
func InvGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient for the inverse of `x` wrt its input.
DEPRECATED at GraphDef version 17: Use ReciprocalGrad
Specifically, `grad = -dy * y*y`, where `y = 1/x`, and `dy` is the corresponding input gradient.
func InvertPermutation ¶
func InvertPermutation(scope *Scope, x tf.Output) (y tf.Output)
Computes the inverse permutation of a tensor.
This operation computes the inverse of an index permutation. It takes a 1-D integer tensor `x`, which represents the indices of a zero-based array, and swaps each value with its index position. In other words, for an output tensor `y` and an input tensor `x`, this operation computes the following:
`y[x[i]] = i for i in [0, 1, ..., len(x) - 1]`
The values must include 0. There can be no duplicate values or negative values.
For example:
“`prettyprint # tensor `x` is [3, 4, 0, 2, 1] invert_permutation(x) ==> [2, 4, 3, 0, 1] “`
Arguments:
x: 1-D.
Returns 1-D.
func IsFinite ¶
func IsFinite(scope *Scope, x tf.Output) (y tf.Output)
Returns which elements of x are finite.
@compatibility(numpy) Equivalent to np.isfinite @end_compatibility
func IsInf ¶
func IsInf(scope *Scope, x tf.Output) (y tf.Output)
Returns which elements of x are Inf.
@compatibility(numpy) Equivalent to np.isinf @end_compatibility
func IsNan ¶
func IsNan(scope *Scope, x tf.Output) (y tf.Output)
Returns which elements of x are NaN.
@compatibility(numpy) Equivalent to np.isnan @end_compatibility
func L2Loss ¶
func L2Loss(scope *Scope, t tf.Output) (output tf.Output)
L2 Loss.
Computes half the L2 norm of a tensor without the `sqrt`:
output = sum(t ** 2) / 2
Arguments:
t: Typically 2-D, but may have any dimensions.
Returns 0-D.
func LRN ¶
func LRN(scope *Scope, input tf.Output, optional ...LRNAttr) (output tf.Output)
Local Response Normalization.
The 4-D `input` tensor is treated as a 3-D array of 1-D vectors (along the last dimension), and each vector is normalized independently. Within a given vector, each component is divided by the weighted, squared sum of inputs within `depth_radius`. In detail,
sqr_sum[a, b, c, d] =
sum(input[a, b, c, d - depth_radius : d + depth_radius + 1] ** 2)
output = input / (bias + alpha * sqr_sum) ** beta
For details, see [Krizhevsky et al., ImageNet classification with deep convolutional neural networks (NIPS 2012)](http://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks).
Arguments:
input: 4-D.
func LRNGrad ¶
func LRNGrad(scope *Scope, input_grads tf.Output, input_image tf.Output, output_image tf.Output, optional ...LRNGradAttr) (output tf.Output)
Gradients for Local Response Normalization.
Arguments:
input_grads: 4-D with shape `[batch, height, width, channels]`. input_image: 4-D with shape `[batch, height, width, channels]`. output_image: 4-D with shape `[batch, height, width, channels]`.
Returns The gradients for LRN.
func LearnedUnigramCandidateSampler ¶
func LearnedUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...LearnedUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a learned unigram distribution.
See explanations of candidate sampling and the data formats at go/candidate-sampling.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to randomly sample per batch. 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 Less ¶
func Less(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of (x < y) element-wise.
*NOTE*: `Less` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func LessEqual ¶
func LessEqual(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of (x <= y) element-wise.
*NOTE*: `LessEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Lgamma ¶
func Lgamma(scope *Scope, x tf.Output) (y tf.Output)
Computes the log of the absolute value of `Gamma(x)` element-wise.
func LinSpace ¶
func LinSpace(scope *Scope, start tf.Output, stop tf.Output, num tf.Output) (output tf.Output)
Generates values in an interval.
A sequence of `num` evenly-spaced values are generated beginning at `start`. If `num > 1`, the values in the sequence increase by `stop - start / num - 1`, so that the last one is exactly `stop`.
For example:
“` tf.linspace(10.0, 12.0, 3, name="linspace") => [ 10.0 11.0 12.0] “`
Arguments:
start: First entry in the range. stop: Last entry in the range. num: Number of values to generate.
Returns 1-D. The generated values.
func ListDiff ¶
func ListDiff(scope *Scope, x tf.Output, y tf.Output, optional ...ListDiffAttr) (out tf.Output, idx tf.Output)
Computes the difference between two lists of numbers or strings.
Given a list `x` and a list `y`, this operation returns a list `out` that represents all values that are in `x` but not in `y`. The returned list `out` is sorted in the same order that the numbers appear in `x` (duplicates are preserved). This operation also returns a list `idx` that represents the position of each `out` element in `x`. In other words:
`out[i] = x[idx[i]] for i in [0, 1, ..., len(out) - 1]`
For example, given this input:
“`prettyprint x = [1, 2, 3, 4, 5, 6] y = [1, 3, 5] “`
This operation would return:
“`prettyprint 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 Log ¶
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 ¶
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 LogSoftmax ¶
func LogSoftmax(scope *Scope, logits tf.Output) (logsoftmax tf.Output)
Computes log softmax activations.
For each batch `i` and class `j` we have
logsoftmax[i, j] = logits[i, j] - log(sum(exp(logits[i])))
Arguments:
logits: 2-D with shape `[batch_size, num_classes]`.
Returns Same shape as `logits`.
func LogUniformCandidateSampler ¶
func LogUniformCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...LogUniformCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a log-uniform distribution.
See explanations of candidate sampling and the data formats at go/candidate-sampling.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to randomly sample per batch. 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 ¶
func LogicalAnd(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of x AND y element-wise.
*NOTE*: `LogicalAnd` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func LogicalNot ¶
func LogicalNot(scope *Scope, x tf.Output) (y tf.Output)
Returns the truth value of NOT x element-wise.
func LogicalOr ¶
func LogicalOr(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the truth value of x OR y element-wise.
*NOTE*: `LogicalOr` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func LoopCond ¶
func LoopCond(scope *Scope, input tf.Output) (output tf.Output)
Forwards the input to the output.
This operator represents the loop termination condition used by the "pivot" switches of a loop.
Arguments:
input: A boolean scalar, representing the branch predicate of the Switch op.
Returns The same tensor as `input`.
func MatMul ¶
func MatMul(scope *Scope, a tf.Output, b tf.Output, optional ...MatMulAttr) (product tf.Output)
Multiply the matrix "a" by the matrix "b".
The inputs must be two-dimensional matrices and the inner dimension of "a" (after being transposed if transpose_a is true) must match the outer dimension of "b" (after being transposed if transposed_b is true).
*Note*: The default kernel implementation for MatMul on GPUs uses cublas.
func MatchingFiles ¶
func MatchingFiles(scope *Scope, pattern tf.Output) (filenames tf.Output)
Returns the set of files matching one or more glob patterns.
Note that this routine only supports wildcard characters in the basename portion of the pattern, not in the directory portion.
Arguments:
pattern: Shell wildcard pattern(s). Scalar or vector of type string.
Returns A vector of matching filenames.
func MatrixBandPart ¶
func MatrixBandPart(scope *Scope, input tf.Output, num_lower tf.Output, num_upper tf.Output) (band tf.Output)
Copy a tensor setting everything outside a central band in each innermost matrix
to zero.
The `band` part is computed as follows: Assume `input` has `k` dimensions `[I, J, K, ..., M, N]`, then the output is a tensor with the same shape where
`band[i, j, k, ..., m, n] = in_band(m, n) * input[i, j, k, ..., m, n]`.
The indicator function
`in_band(m, n) = (num_lower < 0 || (m-n) <= num_lower)) &&
(num_upper < 0 || (n-m) <= num_upper)`.
For example:
“`prettyprint # 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:
“`prettyprint
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 ¶
func MatrixDeterminant(scope *Scope, input tf.Output) (output tf.Output)
Computes the determinant of one ore more square matrices.
The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor containing the determinants for all input submatrices `[..., :, :]`.
Arguments:
input: Shape is `[..., M, M]`.
Returns Shape is `[...]`.
func MatrixDiag ¶
func MatrixDiag(scope *Scope, diagonal tf.Output) (output tf.Output)
Returns a batched diagonal tensor with a given batched diagonal values.
Given a `diagonal`, this operation returns a tensor with the `diagonal` and everything else padded with zeros. The diagonal is computed as follows:
Assume `diagonal` has `k` dimensions `[I, J, K, ..., N]`, then the output is a tensor of rank `k+1` with dimensions [I, J, K, ..., N, N]` where:
`output[i, j, k, ..., m, n] = 1{m=n} * diagonal[i, j, k, ..., n]`.
For example:
“`prettyprint # 'diagonal' is [[1, 2, 3, 4], [5, 6, 7, 8]]
and diagonal.shape = (2, 4)
tf.matrix_diag(diagonal) ==> [[[1, 0, 0, 0]
[0, 2, 0, 0] [0, 0, 3, 0] [0, 0, 0, 4]], [[5, 0, 0, 0] [0, 6, 0, 0] [0, 0, 7, 0] [0, 0, 0, 8]]]
which has shape (2, 4, 4) “`
Arguments:
diagonal: Rank `k`, where `k >= 1`.
Returns Rank `k+1`, with `output.shape = diagonal.shape + [diagonal.shape[-1]]`.
func MatrixDiagPart ¶
func MatrixDiagPart(scope *Scope, input tf.Output) (diagonal tf.Output)
Returns the batched diagonal part of a batched tensor.
This operation returns a tensor with the `diagonal` part of the batched `input`. The `diagonal` part is computed as follows:
Assume `input` has `k` dimensions `[I, J, K, ..., M, N]`, then the output is a tensor of rank `k - 1` with dimensions `[I, J, K, ..., min(M, N)]` where:
`diagonal[i, j, k, ..., n] = input[i, j, k, ..., n, n]`.
The input must be at least a matrix.
For example:
“`prettyprint # '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 MatrixInverse ¶
func MatrixInverse(scope *Scope, input tf.Output, optional ...MatrixInverseAttr) (output tf.Output)
Computes the inverse of one or more square invertible matrices or their
adjoints (conjugate transposes).
The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. The output is a tensor of the same shape as the input containing the inverse for all input submatrices `[..., :, :]`.
The op uses LU decomposition with partial pivoting to compute the inverses.
If a matrix is not invertible there is no guarantee what the op does. It may detect the condition and raise an exception or it may simply return a garbage result.
Arguments:
input: Shape is `[..., M, M]`.
Returns Shape is `[..., M, M]`.
@compatibility(numpy) Equivalent to np.linalg.inv @end_compatibility
func MatrixSetDiag ¶
func MatrixSetDiag(scope *Scope, input tf.Output, diagonal tf.Output) (output tf.Output)
Returns a batched matrix tensor with new batched diagonal values.
Given `input` and `diagonal`, this operation returns a tensor with the same shape and values as `input`, except for the main diagonal of the innermost matrices. These will be overwritten by the values in `diagonal`.
The output is computed as follows:
Assume `input` has `k+1` dimensions `[I, J, K, ..., M, N]` and `diagonal` has `k` dimensions `[I, J, K, ..., min(M, N)]`. Then the output is a tensor of rank `k+1` with dimensions `[I, J, K, ..., M, N]` where:
* `output[i, j, k, ..., m, n] = diagonal[i, j, k, ..., n]` for `m == n`. * `output[i, j, k, ..., m, n] = input[i, j, k, ..., m, n]` for `m != n`.
Arguments:
input: Rank `k+1`, where `k >= 1`. diagonal: Rank `k`, where `k >= 1`.
Returns Rank `k+1`, with `output.shape = input.shape`.
func MatrixSolve ¶
func MatrixSolve(scope *Scope, matrix tf.Output, rhs tf.Output, optional ...MatrixSolveAttr) (output tf.Output)
Solves systems of linear equations.
`Matrix` is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. `Rhs` is a tensor of shape `[..., M, K]`. The `output` is a tensor shape `[..., M, K]`. If `adjoint` is `False` then each output matrix satisfies `matrix[..., :, :] * output[..., :, :] = rhs[..., :, :]`. If `adjoint` is `True` then each output matrix satisfies `adjoint(matrix[..., :, :]) * output[..., :, :] = rhs[..., :, :]`.
Arguments:
matrix: Shape is `[..., M, M]`. rhs: Shape is `[..., M, K]`.
Returns Shape is `[..., M, K]`.
func MatrixSolveLs ¶
func MatrixSolveLs(scope *Scope, matrix tf.Output, rhs tf.Output, l2_regularizer tf.Output, optional ...MatrixSolveLsAttr) (output tf.Output)
Solves one or more linear least-squares problems.
`matrix` is a tensor of shape `[..., M, N]` whose inner-most 2 dimensions form matrices of size `[M, N]`. Rhs is a tensor of 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.
matrix and right-hand sides in the batch:
`matrix`=\\(A \in \Re^{m \times n}\\), `rhs`=\\(B \in \Re^{m \times k}\\), `output`=\\(X \in \Re^{n \times k}\\), `l2_regularizer`=\\(\lambda\\).
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^T A + \lambda I)^{-1} A^T 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^T (A A^T + \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 \Re^{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 MatrixTriangularSolve ¶
func MatrixTriangularSolve(scope *Scope, matrix tf.Output, rhs tf.Output, optional ...MatrixTriangularSolveAttr) (output tf.Output)
Solves systems of linear equations with upper or lower triangular matrices by
backsubstitution.
`matrix` is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices. If `lower` is `True` then the strictly upper triangular part of each inner-most matrix is assumed to be zero and not accessed. If `lower` is False then the strictly lower triangular part of each inner-most matrix is assumed to be zero and not accessed. `rhs` is a tensor of shape `[..., M, 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]`.
Arguments:
matrix: Shape is `[..., M, M]`. rhs: Shape is `[..., M, K]`.
Returns Shape is `[..., M, K]`.
func Max ¶
func Max(scope *Scope, input tf.Output, reduction_indices tf.Output, optional ...MaxAttr) (output tf.Output)
Computes the maximum of elements across dimensions of a tensor.
Reduces `input` along the dimensions given in `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func MaxPool ¶
func MaxPool(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolAttr) (output tf.Output)
Performs max pooling on the input.
Arguments:
input: 4-D input to pool over. ksize: The size of the window for each dimension of the input tensor. strides: The stride of the sliding window for each dimension of the
input tensor.
padding: The type of padding algorithm to use.
Returns The max pooled output tensor.
func MaxPool3D ¶
func MaxPool3D(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output)
Performs 3D max pooling on the input.
Arguments:
input: Shape `[batch, depth, rows, cols, channels]` tensor to pool over. ksize: 1-D tensor of length 5. The size of the window for each dimension of
the input tensor. Must have `ksize[0] = ksize[4] = 1`.
strides: 1-D tensor of length 5. The stride of the sliding window for each
dimension of `input`. Must have `strides[0] = strides[4] = 1`.
padding: The type of padding algorithm to use.
Returns The max pooled output tensor.
func MaxPool3DGrad ¶
func MaxPool3DGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string) (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 MaxPoolGrad ¶
func MaxPoolGrad(scope *Scope, orig_input tf.Output, orig_output tf.Output, grad tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolGradAttr) (output tf.Output)
Computes gradients of the maxpooling function.
Arguments:
orig_input: The original input tensor. orig_output: The original output tensor. grad: 4-D. Gradients w.r.t. the output of `max_pool`. ksize: The size of the window for each dimension of the input tensor. strides: The stride of the sliding window for each dimension of the
input tensor.
padding: The type of padding algorithm to use.
Returns Gradients w.r.t. the input to `max_pool`.
func MaxPoolGradWithArgmax ¶
func MaxPoolGradWithArgmax(scope *Scope, input tf.Output, grad tf.Output, argmax tf.Output, ksize []int64, strides []int64, padding string) (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 MaxPoolWithArgmax ¶
func MaxPoolWithArgmax(scope *Scope, input tf.Output, ksize []int64, strides []int64, padding string, optional ...MaxPoolWithArgmaxAttr) (output tf.Output, argmax tf.Output)
Performs max pooling on the input and outputs both max values and indices.
The indices in `argmax` are flattened, so that a maximum value at position `[b, y, x, c]` becomes flattened index `((b * height + y) * width + x) * channels + c`.
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 ¶
func Maximum(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the max of x and y (i.e. x > y ? x : y) element-wise.
*NOTE*: `Maximum` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Mean ¶
func Mean(scope *Scope, input tf.Output, reduction_indices tf.Output, optional ...MeanAttr) (output tf.Output)
Computes the mean of elements across dimensions of a tensor.
Reduces `input` along the dimensions given in `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func Merge ¶
func Merge(scope *Scope, inputs []tf.Output) (output tf.Output, value_index tf.Output)
Forwards the value of an available tensor from `inputs` to `output`.
`Merge` waits for at least one of the tensors in `inputs` to become available. It is usually combined with `Switch` to implement branching.
`Merge` forwards the first tensor for 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 ¶
func MergeSummary(scope *Scope, inputs []tf.Output) (summary tf.Output)
Merges summaries.
This op creates a [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) protocol buffer that contains the union of all the values in the input summaries.
When the Op is run, it reports an `InvalidArgument` error if multiple values in the summaries to merge use the same tag.
Arguments:
inputs: Can be of any shape. Each must contain serialized `Summary` protocol
buffers.
Returns Scalar. Serialized `Summary` protocol buffer.
func MergeV2Checkpoints ¶
func MergeV2Checkpoints(scope *Scope, checkpoint_prefixes tf.Output, destination_prefix tf.Output, optional ...MergeV2CheckpointsAttr) (o *tf.Operation)
V2 format specific: merges the metadata files of sharded checkpoints. The
result is one logical checkpoint, with one physical metadata file and renamed data files.
Intended for "grouping" multiple checkpoints in a sharded checkpoint setup.
If delete_old_dirs is true, attempts to delete recursively the dirname of each path in the input checkpoint_prefixes. This is useful when those paths are non user-facing temporary locations.
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 Min ¶
func Min(scope *Scope, input tf.Output, reduction_indices tf.Output, optional ...MinAttr) (output tf.Output)
Computes the minimum of elements across dimensions of a tensor.
Reduces `input` along the dimensions given in `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func Minimum ¶
func Minimum(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns the min of x and y (i.e. x < y ? x : y) element-wise.
*NOTE*: `Minimum` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func MirrorPad ¶
func MirrorPad(scope *Scope, input tf.Output, paddings tf.Output, mode string) (output tf.Output)
Pads a tensor with mirrored values.
This operation pads a `input` with mirrored values according to the `paddings` you specify. `paddings` is an integer tensor with shape `[n, 2]`, where n is the rank of `input`. For each dimension D of `input`, `paddings[D, 0]` indicates how many values to add before the contents of `input` in that dimension, and `paddings[D, 1]` indicates how many values to add after the contents of `input` in that dimension. Both `paddings[D, 0]` and `paddings[D, 1]` must be no greater than `input.dim_size(D)` (or `input.dim_size(D) - 1`) if `copy_border` is true (if false, respectively).
The padded size of each dimension D of the output is:
`paddings(D, 0) + input.dim_size(D) + paddings(D, 1)`
For example:
“`prettyprint # 't' is [[1, 2, 3], [4, 5, 6]]. # 'paddings' is [[1, 1]], [2, 2]]. # 'mode' is SYMMETRIC. # rank of 't' is 2. pad(t, paddings) ==> [[2, 1, 1, 2, 3, 3, 2]
[2, 1, 1, 2, 3, 3, 2] [5, 4, 4, 5, 6, 6, 5] [5, 4, 4, 5, 6, 6, 5]]
“`
Arguments:
input: The input tensor to be padded. paddings: A two-column matrix specifying the padding sizes. The number of
rows must be the same as the rank of `input`.
mode: Either `REFLECT` or `SYMMETRIC`. In reflect mode the padded regions
do not include the borders, while in symmetric mode the padded regions do include the borders. For example, if `input` is `[1, 2, 3]` and `paddings` is `[0, 2]`, then the output is `[1, 2, 3, 2, 1]` in reflect mode, and it is `[1, 2, 3, 3, 2]` in symmetric mode.
Returns The padded tensor.
func MirrorPadGrad ¶
func MirrorPadGrad(scope *Scope, input tf.Output, paddings tf.Output, mode string) (output tf.Output)
Gradient op for `MirrorPad` op. This op folds a mirror-padded tensor.
This operation folds the padded areas of `input` by `MirrorPad` according to the `paddings` you specify. `paddings` must be the same as `paddings` argument given to the corresponding `MirrorPad` op.
The folded size of each dimension D of the output is:
`input.dim_size(D) - paddings(D, 0) - paddings(D, 1)`
For example:
“`prettyprint # '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 ¶
func Mod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns element-wise remainder of division.
*NOTE*: `Mod` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Mul ¶
func Mul(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x * y element-wise.
*NOTE*: `Mul` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Multinomial ¶
func Multinomial(scope *Scope, logits tf.Output, num_samples tf.Output, optional ...MultinomialAttr) (output tf.Output)
Draws samples from a multinomial distribution.
Arguments:
logits: 2-D Tensor with shape `[batch_size, num_classes]`. Each slice `[i, :]`
represents the unnormalized log probabilities for all classes.
num_samples: 0-D. Number of independent samples to draw for each row slice.
Returns 2-D Tensor with shape `[batch_size, num_samples]`. Each slice `[i, :]` contains the drawn class labels with range `[0, num_classes)`.
func Neg ¶
func Neg(scope *Scope, x tf.Output) (y tf.Output)
Computes numerical negative value element-wise.
I.e., \\(y = -x\\).
func NextIteration ¶
func NextIteration(scope *Scope, data tf.Output) (output tf.Output)
Makes its input available to the next iteration.
Arguments:
data: The tensor to be made available to the next iteration.
Returns The same tensor as `data`.
func NoOp ¶
func NoOp(scope *Scope) (o *tf.Operation)
Does nothing. Only useful as a placeholder for control edges.
Returns the created operation.
func NonMaxSuppression ¶
func NonMaxSuppression(scope *Scope, boxes tf.Output, scores tf.Output, max_output_size tf.Output, optional ...NonMaxSuppressionAttr) (selected_indices tf.Output)
Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes are supplied as [y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any diagonal pair of box corners and the coordinates can be provided as normalized (i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm is agnostic to where the origin is in the coordinate system. Note that this algorithm is invariant to orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input collection of bounding boxes representing the selected boxes. The bounding box coordinates corresponding to the selected indices can then be obtained using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression(
boxes, scores, max_output_size, iou_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Arguments:
boxes: A 2-D float tensor of shape `[num_boxes, 4]`. scores: A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
Returns A 1-D integer tensor of shape `[M]` representing the selected indices from the boxes tensor, where `M <= max_output_size`.
func NotEqual ¶
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 OneHot ¶
func OneHot(scope *Scope, indices tf.Output, depth tf.Output, on_value tf.Output, off_value tf.Output, optional ...OneHotAttr) (output tf.Output)
Returns a one-hot tensor.
The locations represented by indices in `indices` take value `on_value`, while all other locations take value `off_value`.
If the input `indices` is rank `N`, the output will have rank `N+1`, The new axis is created at dimension `axis` (default: the new axis is appended at the end).
If `indices` is a scalar the output shape will be a vector of length `depth`.
If `indices` is a vector of length `features`, the output shape will be: “`
features x depth if axis == -1 depth x features if axis == 0
“`
If `indices` is a matrix (batch) with shape `[batch, features]`, the output shape will be: “`
batch x features x depth if axis == -1 batch x depth x features if axis == 1 depth x batch x features if axis == 0
“`
Examples =========
Suppose that
“`
indices = [0, 2, -1, 1] depth = 3 on_value = 5.0 off_value = 0.0 axis = -1
“`
Then output is `[4 x 3]`:
```output = [5.0 0.0 0.0] // one_hot(0) [0.0 0.0 5.0] // one_hot(2) [0.0 0.0 0.0] // one_hot(-1) [0.0 5.0 0.0] // one_hot(1) ```
Suppose that
“`
indices = [0, 2, -1, 1] depth = 3 on_value = 0.0 off_value = 3.0 axis = 0
“`
Then output is `[3 x 4]`:
```output = [0.0 3.0 3.0 3.0] [3.0 3.0 3.0 0.0] [3.0 3.0 3.0 3.0] [3.0 0.0 3.0 3.0] // ^ one_hot(0) // ^ one_hot(2) // ^ one_hot(-1) // ^ one_hot(1) ```
Suppose that
“`
indices = [[0, 2], [1, -1]] depth = 3 on_value = 1.0 off_value = 0.0 axis = -1
“`
Then output is `[2 x 2 x 3]`:
```output =
[
[1.0, 0.0, 0.0] // one_hot(0)
[0.0, 0.0, 1.0] // one_hot(2)
][
[0.0, 1.0, 0.0] // one_hot(1)
[0.0, 0.0, 0.0] // one_hot(-1)
]```
Arguments:
indices: A tensor of indices. depth: A scalar defining the depth of the one hot dimension. on_value: A scalar defining the value to fill in output when `indices[j] = i`. off_value: A scalar defining the value to fill in output when `indices[j] != i`.
Returns The one-hot tensor.
func Pack ¶
func Pack(scope *Scope, values []tf.Output, optional ...PackAttr) (output tf.Output)
Packs a list of `N` rank-`R` tensors into one rank-`(R+1)` tensor.
Packs the `N` tensors in `values` into a tensor with rank one higher than each tensor in `values`, by packing them along the `axis` dimension. Given a list of tensors of shape `(A, B, C)`;
if `axis == 0` then the `output` tensor will have the shape `(N, A, B, C)`. if `axis == 1` then the `output` tensor will have the shape `(A, N, B, C)`. Etc.
For example:
“`prettyprint # 'x' is [1, 4] # 'y' is [2, 5] # 'z' is [3, 6] pack([x, y, z]) => [[1, 4], [2, 5], [3, 6]] # Pack along first dim. pack([x, y, z], axis=1) => [[1, 2, 3], [4, 5, 6]] “`
This is the opposite of `unpack`.
Arguments:
values: Must be of same shape and type.
Returns The packed tensor.
func Pad ¶
func Pad(scope *Scope, input tf.Output, paddings tf.Output) (output tf.Output)
Pads a tensor with zeros.
This operation pads a `input` with zeros according to the `paddings` you specify. `paddings` is an integer tensor with shape `[Dn, 2]`, where n is the rank of `input`. For each dimension D of `input`, `paddings[D, 0]` indicates how many zeros to add before the contents of `input` in that dimension, and `paddings[D, 1]` indicates how many zeros to add after the contents of `input` in that dimension.
The padded size of each dimension D of the output is:
`paddings(D, 0) + input.dim_size(D) + paddings(D, 1)`
For example:
“`prettyprint # '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 PaddingFIFOQueueV2 ¶
func PaddingFIFOQueueV2(scope *Scope, component_types []tf.DataType, optional ...PaddingFIFOQueueV2Attr) (handle tf.Output)
A queue that produces elements in first-in first-out order.
Variable-size shapes are allowed by setting the corresponding shape dimensions to 0 in the shape attr. In this case DequeueMany will pad up to the maximum size of any given element in the minibatch. See below for details.
Arguments:
component_types: The type of each component in a value.
Returns The handle to the queue.
func ParallelConcat ¶
func ParallelConcat(scope *Scope, values []tf.Output, shape tf.Shape) (output tf.Output)
Concatenates a list of `N` tensors along the first dimension.
The input tensors are all required to have size 1 in the first dimension.
For example:
“`prettyprint # '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 ParameterizedTruncatedNormal ¶
func ParameterizedTruncatedNormal(scope *Scope, shape tf.Output, means tf.Output, stdevs tf.Output, minvals tf.Output, maxvals tf.Output, optional ...ParameterizedTruncatedNormalAttr) (output tf.Output)
Outputs random values from a normal distribution. The parameters may each be a
scalar which applies to the entire output, or a vector of length shape[0] which stores the parameters for each batch.
Arguments:
shape: The shape of the output tensor. Batches are indexed by the 0th dimension. means: The mean parameter of each batch. stdevs: The standard deviation parameter of each batch. Must be greater than 0. minvals: The minimum cutoff. May be -infinity. maxvals: The maximum cutoff. May be +infinity, and must be more than the minval
for each batch.
Returns A matrix of shape num_batches x samples_per_batch, filled with random truncated normal values using the parameters for each row.
func ParseExample ¶
func ParseExample(scope *Scope, serialized tf.Output, names tf.Output, sparse_keys []tf.Output, dense_keys []tf.Output, dense_defaults []tf.Output, sparse_types []tf.DataType, dense_shapes []tf.Shape) (sparse_indices []tf.Output, sparse_values []tf.Output, sparse_shapes []tf.Output, dense_values []tf.Output)
Transforms a vector of brain.Example protos (as strings) into typed tensors.
Arguments:
serialized: A vector containing a batch of binary serialized Example protos. names: A vector containing the names of the serialized protos.
May contain, for example, table key (descriptive) names for the corresponding serialized protos. These are purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty vector if no names are available. If non-empty, this vector must be the same length as "serialized".
sparse_keys: A list of Nsparse string Tensors (scalars).
The keys expected in the Examples' features associated with sparse values.
dense_keys: A list of Ndense string Tensors (scalars).
The keys expected in the Examples' features associated with dense values.
dense_defaults: A list of Ndense Tensors (some may be empty).
dense_defaults[j] provides default values when the example's feature_map lacks dense_key[j]. If an empty Tensor is provided for dense_defaults[j], then the Feature dense_keys[j] is required. The input type is inferred from dense_defaults[j], even when it's empty. If dense_defaults[j] is not empty, and dense_shapes[j] is fully defined, then the shape of dense_defaults[j] must match that of dense_shapes[j]. If dense_shapes[j] has an undefined major dimension (variable strides dense feature), dense_defaults[j] must contain a single element: the padding element.
sparse_types: A list of Nsparse types; the data types of data in each Feature
given in sparse_keys. Currently the ParseExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList).
dense_shapes: A list of Ndense shapes; the shapes of data in each Feature
given in dense_keys. The number of elements in the Feature corresponding to dense_key[j] must always equal dense_shapes[j].NumEntries(). If dense_shapes[j] == (D0, D1, ..., DN) then the shape of output Tensor dense_values[j] will be (|serialized|, D0, D1, ..., DN): The dense outputs are just the inputs row-stacked by batch. This works for dense_shapes[j] = (-1, D1, ..., DN). In this case the shape of the output Tensor dense_values[j] will be (|serialized|, M, D1, .., DN), where M is the maximum number of blocks of elements of length D1 * .... * DN, across all minibatch entries in the input. Any minibatch entry with less than M blocks of elements of length D1 * ... * DN will be padded with the corresponding default_value scalar element along the second dimension.
func ParseSingleSequenceExample ¶
func ParseSingleSequenceExample(scope *Scope, serialized tf.Output, feature_list_dense_missing_assumed_empty tf.Output, context_sparse_keys []tf.Output, context_dense_keys []tf.Output, feature_list_sparse_keys []tf.Output, feature_list_dense_keys []tf.Output, context_dense_defaults []tf.Output, debug_name tf.Output, optional ...ParseSingleSequenceExampleAttr) (context_sparse_indices []tf.Output, context_sparse_values []tf.Output, context_sparse_shapes []tf.Output, context_dense_values []tf.Output, feature_list_sparse_indices []tf.Output, feature_list_sparse_values []tf.Output, feature_list_sparse_shapes []tf.Output, feature_list_dense_values []tf.Output)
Transforms a scalar brain.SequenceExample proto (as strings) into typed tensors.
Arguments:
serialized: A scalar containing a binary serialized SequenceExample proto. feature_list_dense_missing_assumed_empty: A vector listing the
FeatureList keys which may be missing from the SequenceExample. If the associated FeatureList is missing, it is treated as empty. By default, any FeatureList not listed in this vector must exist in the SequenceExample.
context_sparse_keys: A list of Ncontext_sparse string Tensors (scalars).
The keys expected in the Examples' features associated with context_sparse values.
context_dense_keys: A list of Ncontext_dense string Tensors (scalars).
The keys expected in the SequenceExamples' context features associated with dense values.
feature_list_sparse_keys: A list of Nfeature_list_sparse string Tensors
(scalars). The keys expected in the FeatureLists associated with sparse values.
feature_list_dense_keys: A list of Nfeature_list_dense string Tensors (scalars).
The keys expected in the SequenceExamples' feature_lists associated with lists of dense values.
context_dense_defaults: A list of Ncontext_dense Tensors (some may be empty).
context_dense_defaults[j] provides default values when the SequenceExample's context map lacks context_dense_key[j]. If an empty Tensor is provided for context_dense_defaults[j], then the Feature context_dense_keys[j] is required. The input type is inferred from context_dense_defaults[j], even when it's empty. If context_dense_defaults[j] is not empty, its shape must match context_dense_shapes[j].
debug_name: A scalar containing the name of the serialized proto.
May contain, for example, table key (descriptive) name for the corresponding serialized proto. This is purely useful for debugging purposes, and the presence of values here has no effect on the output. May also be an empty scalar if no name is available.
func ParseTensor ¶
func ParseTensor(scope *Scope, serialized tf.Output, out_type tf.DataType) (output tf.Output)
Transforms a serialized tensorflow.TensorProto proto into a Tensor.
Arguments:
serialized: A scalar string containing a serialized TensorProto proto. out_type: The type of the serialized tensor. The provided type must match the
type of the serialized tensor and no implicit conversion will take place.
Returns A Tensor of type `out_type`.
func Placeholder ¶
func Placeholder(scope *Scope, dtype tf.DataType, optional ...PlaceholderAttr) (output tf.Output)
A placeholder op for a value that will be fed into the computation.
N.B. This operation will fail with an error if it is executed. It is intended as a way to represent a value that will always be fed, and to provide attrs that enable the fed value to be checked at runtime.
Arguments:
dtype: The type of elements in the tensor.
Returns A placeholder tensor that must be replaced using the feed mechanism.
func PlaceholderV2 ¶
func PlaceholderV2(scope *Scope, dtype tf.DataType, shape tf.Shape) (output tf.Output)
A placeholder op for a value that will be fed into the computation.
N.B. This operation will fail with an error if it is executed. It is intended as a way to represent a value that will always be fed, and to provide attrs that enable the fed value to be checked at runtime.
Arguments:
dtype: The type of elements in the tensor. shape: The shape of the tensor. The shape can be any partially-specified
shape. To be unconstrained, pass in a shape with unknown rank.
Returns A placeholder tensor that must be replaced using the feed mechanism.
func PlaceholderWithDefault ¶
func PlaceholderWithDefault(scope *Scope, input tf.Output, shape tf.Shape) (output tf.Output)
A placeholder op that passes through `input` when its output is not fed.
Arguments:
input: The default value to produce when `output` is not fed. shape: The (possibly partial) shape of the tensor.
Returns A placeholder tensor that defaults to `input` if it is not fed.
func Polygamma ¶
func Polygamma(scope *Scope, a tf.Output, x tf.Output) (z tf.Output)
Compute the polygamma function \\(\psi^{(n)}(x)\\).
The polygamma function is defined as:
“` \psi^{(n)}(x) = \frac{d^n}{dx^n} \psi(x) “` where \\(\psi(x)\\) is the digamma function.
func Pow ¶
func Pow(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the power of one value to another.
Given a tensor `x` and a tensor `y`, this operation computes \\(x^y\\) for corresponding elements in `x` and `y`. For example:
“` # tensor 'x' is [[2, 2]], [3, 3]] # tensor 'y' is [[8, 16], [2, 3]] tf.pow(x, y) ==> [[256, 65536], [9, 27]] “`
func PreventGradient ¶
func PreventGradient(scope *Scope, input tf.Output, optional ...PreventGradientAttr) (output tf.Output)
An identity op that triggers an error if a gradient is requested.
When executed in a graph, this op outputs its input tensor as-is.
When building ops to compute gradients, the TensorFlow gradient system will return an error when trying to lookup the gradient of this op, because no gradient must ever be registered for this function. This op exists to prevent subtle bugs from silently returning unimplemented gradients in some corner cases.
Arguments:
input: any tensor.
Returns the same input tensor.
func Print ¶
func Print(scope *Scope, input tf.Output, data []tf.Output, optional ...PrintAttr) (output tf.Output)
Prints a list of tensors.
Passes `input` through to `output` and prints `data` when evaluating.
Arguments:
input: The tensor passed to `output` data: A list of tensors to print out when op is evaluated.
Returns = The unmodified `input` tensor
func PriorityQueueV2 ¶
func PriorityQueueV2(scope *Scope, shapes []tf.Shape, optional ...PriorityQueueV2Attr) (handle tf.Output)
A queue that produces elements sorted by the first component value.
Note that the PriorityQueue requires the first component of any element to be a scalar int64, in addition to the other elements declared by component_types. Therefore calls to Enqueue and EnqueueMany (resp. Dequeue and DequeueMany) on a PriorityQueue will all require (resp. output) one extra entry in their input (resp. output) lists.
Arguments:
shapes: The shape of each component in a value. The length of this attr must
be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time.
Returns The handle to the queue.
func Prod ¶
func Prod(scope *Scope, input tf.Output, reduction_indices tf.Output, optional ...ProdAttr) (output tf.Output)
Computes the product of elements across dimensions of a tensor.
Reduces `input` along the dimensions given in `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func Qr ¶
func Qr(scope *Scope, input tf.Output, optional ...QrAttr) (q tf.Output, r tf.Output)
Computes the QR decompositions of one or more matrices.
Computes the QR decomposition of each inner matrix in `tensor` such that `tensor[..., :, :] = q[..., :, :] * r[..., :,:])`
“`prettyprint # 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 ¶
func QuantizeAndDequantize(scope *Scope, input tf.Output, optional ...QuantizeAndDequantizeAttr) (output tf.Output)
Use QuantizeAndDequantizeV2 instead.
DEPRECATED at GraphDef version 22: Replaced by QuantizeAndDequantizeV2
func QuantizeAndDequantizeV2 ¶
func QuantizeAndDequantizeV2(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, optional ...QuantizeAndDequantizeV2Attr) (output tf.Output)
Quantizes then dequantizes a tensor.
This op simulates the precision loss from the quantized forward pass by: 1. Quantizing the tensor to fixed point numbers, which should match the target
quantization method when it is used in inference.
2. Dequantizing it back to floating point numbers for the following ops, most
likely matmul.
There are different ways to quantize. This version does 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.
To perform this op, 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
1. m = max(abs(input_min), abs(input_max)) if range_given is true, 2. m = max(abs(min_elem(input)), abs(max_elem(input))) otherwise.
Our input tensor range is then [-m, m].
Next, we choose our fixed-point quantization buckets, [min_fixed, max_fixed]. If signed_input is true, this is
[min_fixed, max_fixed ] =
[-(1 << (num_bits - 1) - 1), (1 << (num_bits - 1)) - 1].
Otherwise, if signed_input is false, the fixed-point range is
[min_fixed, max_fixed] = [0, (1 << num_bits) - 1].
From this we compute our scaling factor, s:
s = (max_fixed - min_fixed) / (2 * m).
Now we can quantize and dequantize the elements of our tensor. An element e is transformed into e':
e' = (e * s).round_to_nearest() / s.
Note that we have a different number of buckets in the signed vs. unsigned cases. For example, if num_bits == 8, we get 254 buckets in the signed case vs. 255 in the unsigned case.
For example, suppose num_bits = 8 and m = 1. Then
[min_fixed, max_fixed] = [-127, 127], and s = (127 + 127) / 2 = 127.
Given the vector {-1, -0.5, 0, 0.3}, this is quantized to {-127, -63, 0, 38}, and dequantized to {-1, -63.0/127, 0, 38.0/127}.
Arguments:
input: Tensor to quantize and then dequantize. input_min: If range_given, this is the min of the range, otherwise this input
will be ignored.
input_max: If range_given, this is the max of the range, otherwise this input
will be ignored.
func QuantizeDownAndShrinkRange ¶
func QuantizeDownAndShrinkRange(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, out_type tf.DataType) (output tf.Output, output_min tf.Output, output_max tf.Output)
Convert the quantized 'input' tensor into a lower-precision 'output', using the
actual distribution of the values to maximize the usage of the lower bit depth and adjusting the output min and max ranges accordingly.
[input_min, input_max] are scalar floats that specify the range for the float interpretation of the 'input' data. For example, if input_min is -1.0f and input_max is 1.0f, and we are dealing with quint16 quantized data, then a 0 value in the 16-bit data should be interpreted as -1.0f, and a 65535 means 1.0f.
This operator tries to squeeze as much precision as possible into an output with a lower bit depth by calculating the actual min and max values found in the data. For example, maybe that quint16 input has no values lower than 16,384 and none higher than 49,152. That means only half the range is actually needed, all the float interpretations are between -0.5f and 0.5f, so if we want to compress the data into a quint8 output, we can use that range rather than the theoretical -1.0f to 1.0f that is suggested by the input min and max.
In practice, this is most useful for taking output from operations like QuantizedMatMul that can produce higher bit-depth outputs than their inputs and may have large potential output ranges, but in practice have a distribution of input values that only uses a small fraction of the possible range. By feeding that output into this operator, we can reduce it from 32 bits down to 8 with minimal loss of accuracy.
Arguments:
input_min: The float value that the minimum quantized input value represents. input_max: The float value that the maximum quantized input value represents. out_type: The type of the output. Should be a lower bit depth than Tinput.
Returns The float value that the minimum quantized output value represents.The float value that the maximum quantized output value represents.
func QuantizeV2 ¶
func QuantizeV2(scope *Scope, input tf.Output, min_range tf.Output, max_range tf.Output, T tf.DataType, optional ...QuantizeV2Attr) (output tf.Output, output_min tf.Output, output_max tf.Output)
Quantize the 'input' tensor of type float to 'output' tensor of type 'T'.
[min_range, max_range] are scalar floats that specify the range for the 'input' data. The 'mode' attribute controls exactly which calculations are used to convert the float values to their quantized equivalents.
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:
“` number_of_steps = 1 << (# of bits in T) range_adjust = number_of_steps / (number_of_steps - 1) range = (range_max - range_min) * range_adjust range_scale = number_of_steps / 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.
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 QuantizedAvgPool ¶
func QuantizedAvgPool(scope *Scope, input tf.Output, min_input tf.Output, max_input tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output, min_output tf.Output, max_output tf.Output)
Produces the average pool of the input tensor for quantized types.
Arguments:
input: 4-D with shape `[batch, height, width, channels]`. min_input: The float value that the lowest quantized input value represents. max_input: The float value that the highest quantized input value represents. ksize: The size of the window for each dimension of the input tensor.
The length must be 4 to match the number of dimensions of the input.
strides: The stride of the sliding window for each dimension of the input
tensor. The length must be 4 to match the number of dimensions of the input.
padding: The type of padding algorithm to use.
Returns The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.
func QuantizedBatchNormWithGlobalNormalization ¶
func QuantizedBatchNormWithGlobalNormalization(scope *Scope, t tf.Output, t_min tf.Output, t_max tf.Output, m tf.Output, m_min tf.Output, m_max tf.Output, v tf.Output, v_min tf.Output, v_max tf.Output, beta tf.Output, beta_min tf.Output, beta_max tf.Output, gamma tf.Output, gamma_min tf.Output, gamma_max tf.Output, out_type tf.DataType, variance_epsilon float32, scale_after_normalization bool) (result tf.Output, result_min tf.Output, result_max tf.Output)
Quantized Batch normalization.
This op is deprecated and will be removed in the future. Prefer `tf.nn.batch_normalization`.
Arguments:
t: A 4D input Tensor. t_min: The value represented by the lowest quantized input. t_max: The value represented by the highest quantized input. m: A 1D mean Tensor with size matching the last dimension of t.
This is the first output from tf.nn.moments, or a saved moving average thereof.
m_min: The value represented by the lowest quantized mean. m_max: The value represented by the highest quantized mean. v: A 1D variance Tensor with size matching the last dimension of t.
This is the second output from tf.nn.moments, or a saved moving average thereof.
v_min: The value represented by the lowest quantized variance. v_max: The value represented by the highest quantized variance. beta: A 1D beta Tensor with size matching the last dimension of t.
An offset to be added to the normalized tensor.
beta_min: The value represented by the lowest quantized offset. beta_max: The value represented by the highest quantized offset. gamma: A 1D gamma Tensor with size matching the last dimension of t.
If "scale_after_normalization" is true, this tensor will be multiplied with the normalized tensor.
gamma_min: The value represented by the lowest quantized gamma. gamma_max: The value represented by the highest quantized gamma. variance_epsilon: A small float number to avoid dividing by 0. scale_after_normalization: A bool indicating whether the resulted tensor
needs to be multiplied with gamma.
func QuantizedBiasAdd ¶
func QuantizedBiasAdd(scope *Scope, input tf.Output, bias tf.Output, min_input tf.Output, max_input tf.Output, min_bias tf.Output, max_bias tf.Output, out_type tf.DataType) (output tf.Output, min_out tf.Output, max_out tf.Output)
Adds Tensor 'bias' to Tensor 'input' for Quantized types.
Broadcasts the values of bias on dimensions 0..N-2 of 'input'.
Arguments:
bias: A 1D bias Tensor with size matching the last dimension of 'input'. min_input: The float value that the lowest quantized input value represents. max_input: The float value that the highest quantized input value represents. min_bias: The float value that the lowest quantized bias value represents. max_bias: The float value that the highest quantized bias value represents.
Returns The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.
func QuantizedConcat ¶
func QuantizedConcat(scope *Scope, concat_dim tf.Output, values []tf.Output, input_mins []tf.Output, input_maxes []tf.Output) (output tf.Output, output_min tf.Output, output_max tf.Output)
Concatenates quantized tensors along one dimension.
Arguments:
concat_dim: 0-D. The dimension along which to concatenate. Must be in the
range [0, rank(values)).
values: The `N` Tensors to concatenate. Their ranks and types must match,
and their sizes must match in all dimensions except `concat_dim`.
input_mins: The minimum scalar values for each of the input tensors. input_maxes: The maximum scalar values for each of the input tensors.
Returns 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 ¶
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 QuantizedInstanceNorm ¶
func QuantizedInstanceNorm(scope *Scope, x tf.Output, x_min tf.Output, x_max tf.Output, optional ...QuantizedInstanceNormAttr) (y tf.Output, y_min tf.Output, y_max tf.Output)
Quantized Instance normalization.
Arguments:
x: A 4D input Tensor. x_min: The value represented by the lowest quantized input. x_max: The value represented by the highest quantized input.
Returns A 4D Tensor.The value represented by the lowest quantized output.The value represented by the highest quantized output.
func QuantizedMatMul ¶
func QuantizedMatMul(scope *Scope, a tf.Output, b tf.Output, min_a tf.Output, max_a tf.Output, min_b tf.Output, max_b tf.Output, optional ...QuantizedMatMulAttr) (out tf.Output, min_out tf.Output, max_out tf.Output)
Perform a quantized matrix multiplication of `a` by the matrix `b`.
The inputs must be two-dimensional matrices and the inner dimension of `a` (after being transposed if `transpose_a` is non-zero) must match the outer dimension of `b` (after being transposed if `transposed_b` is non-zero).
Arguments:
a: Must be a two-dimensional tensor. b: Must be a two-dimensional tensor. min_a: The float value that the lowest quantized `a` value represents. max_a: The float value that the highest quantized `a` value represents. min_b: The float value that the lowest quantized `b` value represents. max_b: The float value that the highest quantized `b` value represents.
Returns The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.
func QuantizedMaxPool ¶
func QuantizedMaxPool(scope *Scope, input tf.Output, min_input tf.Output, max_input tf.Output, ksize []int64, strides []int64, padding string) (output tf.Output, min_output tf.Output, max_output tf.Output)
Produces the max pool of the input tensor for quantized types.
Arguments:
input: The 4D (batch x rows x cols x depth) Tensor to MaxReduce over. min_input: The float value that the lowest quantized input value represents. max_input: The float value that the highest quantized input value represents. ksize: The size of the window for each dimension of the input tensor.
The length must be 4 to match the number of dimensions of the input.
strides: The stride of the sliding window for each dimension of the input
tensor. The length must be 4 to match the number of dimensions of the input.
padding: The type of padding algorithm to use.
Returns The float value that the lowest quantized output value represents.The float value that the highest quantized output value represents.
func QuantizedMul ¶
func QuantizedMul(scope *Scope, x tf.Output, y tf.Output, min_x tf.Output, max_x tf.Output, min_y tf.Output, max_y tf.Output, optional ...QuantizedMulAttr) (z tf.Output, min_z tf.Output, max_z tf.Output)
Returns x * y element-wise, working on quantized buffers.
Arguments:
min_x: The float value that the lowest quantized `x` value represents. max_x: The float value that the highest quantized `x` value represents. min_y: The float value that the lowest quantized `y` value represents. max_y: The float value that the highest quantized `y` value represents.
Returns 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 ¶
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 ¶
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 ¶
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 ¶
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 QueueCloseV2 ¶
func QueueCloseV2(scope *Scope, handle tf.Output, optional ...QueueCloseV2Attr) (o *tf.Operation)
Closes the given queue.
This operation signals that no more elements will be enqueued in the given queue. Subsequent Enqueue(Many) operations will fail. Subsequent Dequeue(Many) operations will continue to succeed if sufficient elements remain in the queue. Subsequent Dequeue(Many) operations that would block will fail immediately.
Arguments:
handle: The handle to a queue.
Returns the created operation.
func QueueDequeueManyV2 ¶
func QueueDequeueManyV2(scope *Scope, handle tf.Output, n tf.Output, component_types []tf.DataType, optional ...QueueDequeueManyV2Attr) (components []tf.Output)
Dequeues n tuples of one or more tensors from the given queue.
If the queue is closed and there are fewer than n elements, then an OutOfRange error is returned.
This operation concatenates queue-element component tensors along the 0th dimension to make a single component tensor. All of the components in the dequeued tuple will have size n in the 0th dimension.
This operation has k outputs, where k is the number of components in the tuples stored in the given queue, and output i is the ith component of the dequeued tuple.
N.B. If the queue is empty, this operation will block until n elements have been dequeued (or 'timeout_ms' elapses, if specified).
Arguments:
handle: The handle to a queue. n: The number of tuples to dequeue. component_types: The type of each component in a tuple.
Returns One or more tensors that were dequeued as a tuple.
func QueueDequeueUpToV2 ¶
func QueueDequeueUpToV2(scope *Scope, handle tf.Output, n tf.Output, component_types []tf.DataType, optional ...QueueDequeueUpToV2Attr) (components []tf.Output)
Dequeues n tuples of one or more tensors from the given queue.
This operation is not supported by all queues. If a queue does not support DequeueUpTo, then an Unimplemented error is returned.
If the queue is closed and there are more than 0 but less than n elements remaining, then instead of returning an OutOfRange error like QueueDequeueMany, less than `n` elements are returned immediately. If the queue is closed and there are 0 elements left in the queue, then an OutOfRange error is returned just like in QueueDequeueMany. Otherwise the behavior is identical to QueueDequeueMany:
This operation concatenates queue-element component tensors along the 0th dimension to make a single component tensor. All of the components in the dequeued tuple will have size n in the 0th dimension.
This operation has k outputs, where k is the number of components in the tuples stored in the given queue, and output i is the ith component of the dequeued tuple.
Arguments:
handle: The handle to a queue. n: The number of tuples to dequeue. component_types: The type of each component in a tuple.
Returns One or more tensors that were dequeued as a tuple.
func QueueDequeueV2 ¶
func QueueDequeueV2(scope *Scope, handle tf.Output, component_types []tf.DataType, optional ...QueueDequeueV2Attr) (components []tf.Output)
Dequeues a tuple of one or more tensors from the given queue.
This operation has k outputs, where k is the number of components in the tuples stored in the given queue, and output i is the ith component of the dequeued tuple.
N.B. If the queue is empty, this operation will block until an element has been dequeued (or 'timeout_ms' elapses, if specified).
Arguments:
handle: The handle to a queue. component_types: The type of each component in a tuple.
Returns One or more tensors that were dequeued as a tuple.
func QueueEnqueueManyV2 ¶
func QueueEnqueueManyV2(scope *Scope, handle tf.Output, components []tf.Output, optional ...QueueEnqueueManyV2Attr) (o *tf.Operation)
Enqueues zero or more tuples of one or more tensors in the given queue.
This operation slices each component tensor along the 0th dimension to make multiple queue elements. All of the tuple components must have the same size in the 0th dimension.
The components input has k elements, which correspond to the components of tuples stored in the given queue.
N.B. If the queue is full, this operation will block until the given elements have been enqueued (or 'timeout_ms' elapses, if specified).
Arguments:
handle: The handle to a queue. components: One or more tensors from which the enqueued tensors should
be taken.
Returns the created operation.
func QueueEnqueueV2 ¶
func QueueEnqueueV2(scope *Scope, handle tf.Output, components []tf.Output, optional ...QueueEnqueueV2Attr) (o *tf.Operation)
Enqueues a tuple of one or more tensors in the given queue.
The components input has k elements, which correspond to the components of tuples stored in the given queue.
N.B. If the queue is full, this operation will block until the given element has been enqueued (or 'timeout_ms' elapses, if specified).
Arguments:
handle: The handle to a queue. components: One or more tensors from which the enqueued tensors should be taken.
Returns the created operation.
func QueueSizeV2 ¶
func QueueSizeV2(scope *Scope, handle tf.Output) (size tf.Output)
Computes the number of elements in the given queue.
Arguments:
handle: The handle to a queue.
Returns The number of elements in the given queue.
func RFFT ¶
func RFFT(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
Arguments:
input: A float32 tensor. fft_length: An int32 tensor of shape [1]. The FFT length.
Returns A complex64 tensor of the same rank as `input`. The inner-most
dimension of `input` is replaced with the `fft_length / 2 + 1` unique frequency components of its 1D Fourier Transform.
@compatibility(numpy) Equivalent to np.fft.rfft @end_compatibility
func RFFT2D ¶
func RFFT2D(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
Arguments:
input: A float32 tensor. fft_length: An int32 tensor of shape [2]. The FFT length for each dimension.
Returns A complex64 tensor of the same rank as `input`. The inner-most 2
dimensions of `input` are replaced with their 2D Fourier Transform. The inner-most dimension contains `fft_length / 2 + 1` unique frequency components.
@compatibility(numpy) Equivalent to np.fft.rfft2 @end_compatibility
func RFFT3D ¶
func RFFT3D(scope *Scope, input tf.Output, fft_length tf.Output) (output tf.Output)
Compute 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.
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 ¶
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 RandomCrop ¶
func RandomCrop(scope *Scope, image tf.Output, size tf.Output, optional ...RandomCropAttr) (output tf.Output)
Randomly crop `image`.
DEPRECATED at GraphDef version 8: Random crop is now pure Python
`size` is a 1-D int64 tensor with 2 elements representing the crop height and width. The values must be non negative.
This Op picks a random location in `image` and crops a `height` by `width` rectangle from that location. The random location is picked so the cropped area will fit inside the original image.
Arguments:
image: 3-D of shape `[height, width, channels]`. size: 1-D of length 2 containing: `crop_height`, `crop_width`..
Returns 3-D of shape `[crop_height, crop_width, channels].`
func RandomGamma ¶
func RandomGamma(scope *Scope, shape tf.Output, alpha tf.Output, optional ...RandomGammaAttr) (output tf.Output)
Outputs random values from the Gamma distribution(s) described by alpha.
This op uses the algorithm by Marsaglia et al. to acquire samples via transformation-rejection from pairs of uniform and normal random variables. See http://dl.acm.org/citation.cfm?id=358414
Arguments:
shape: 1-D integer tensor. Shape of independent samples to draw from each
distribution described by the shape parameters given in alpha.
alpha: A tensor in which each scalar is a "shape" parameter describing the
associated gamma distribution.
Returns A tensor with shape `shape + shape(alpha)`. Each slice `[:, ..., :, i0, i1, ...iN]` contains the samples drawn for `alpha[i0, i1, ...iN]`. The dtype of the output matches the dtype of alpha.
func RandomPoisson ¶
func RandomPoisson(scope *Scope, shape tf.Output, rate tf.Output, optional ...RandomPoissonAttr) (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]`. The dtype of the output matches the dtype of rate.
func RandomShuffle ¶
func RandomShuffle(scope *Scope, value tf.Output, optional ...RandomShuffleAttr) (output tf.Output)
Randomly shuffles a tensor along its first dimension.
The tensor is shuffled along dimension 0, such that each `value[j]` is mapped to one and only one `output[i]`. For example, a mapping that might occur for a 3x2 tensor is:
“`prettyprint [[1, 2], [[5, 6],
[3, 4], ==> [1, 2], [5, 6]] [3, 4]]
“`
Arguments:
value: The tensor to be shuffled.
Returns A tensor of same shape and type as `value`, shuffled along its first dimension.
func RandomShuffleQueueV2 ¶
func RandomShuffleQueueV2(scope *Scope, component_types []tf.DataType, optional ...RandomShuffleQueueV2Attr) (handle tf.Output)
A queue that randomizes the order of elements.
Arguments:
component_types: The type of each component in a value.
Returns The handle to the queue.
func RandomStandardNormal ¶
func RandomStandardNormal(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...RandomStandardNormalAttr) (output tf.Output)
Outputs random values from a normal distribution.
The generated values will have mean 0 and standard deviation 1.
Arguments:
shape: The shape of the output tensor. dtype: The type of the output.
Returns A tensor of the specified shape filled with random normal values.
func RandomUniform ¶
func RandomUniform(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...RandomUniformAttr) (output tf.Output)
Outputs random values from a uniform distribution.
The generated values follow a uniform distribution in the range `[0, 1)`. The lower bound 0 is included in the range, while the upper bound 1 is excluded.
Arguments:
shape: The shape of the output tensor. dtype: The type of the output.
Returns A tensor of the specified shape filled with uniform random values.
func RandomUniformInt ¶
func RandomUniformInt(scope *Scope, shape tf.Output, minval tf.Output, maxval tf.Output, optional ...RandomUniformIntAttr) (output tf.Output)
Outputs random integers from a uniform distribution.
The generated values are uniform integers in the range `[minval, maxval)`. The lower bound `minval` is included in the range, while the upper bound `maxval` is excluded.
The random integers are slightly biased unless `maxval - minval` is an exact power of two. The bias is small for values of `maxval - minval` significantly smaller than the range of the output (either `2^32` or `2^64`).
Arguments:
shape: The shape of the output tensor. minval: 0-D. Inclusive lower bound on the generated integers. maxval: 0-D. Exclusive upper bound on the generated integers.
Returns A tensor of the specified shape filled with uniform random integers.
func Range ¶
func Range(scope *Scope, start tf.Output, limit tf.Output, delta tf.Output) (output tf.Output)
Creates a sequence of numbers.
This operation creates a sequence of numbers that begins at `start` and extends by increments of `delta` up to but not including `limit`.
For example:
“` # 'start' is 3 # 'limit' is 18 # 'delta' is 3 tf.range(start, limit, delta) ==> [3, 6, 9, 12, 15] “`
Arguments:
start: 0-D (scalar). First entry in the sequence. limit: 0-D (scalar). Upper limit of sequence, exclusive. delta: 0-D (scalar). Optional. Default is 1. Number that increments `start`.
Returns 1-D.
func Rank ¶
func Rank(scope *Scope, input tf.Output) (output tf.Output)
Returns the rank of a tensor.
This operation returns an integer representing the rank of `input`.
For example:
“`prettyprint # 't' is [[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]] # shape of tensor 't' is [2, 2, 3] rank(t) ==> 3 “`
**Note**: The rank of a tensor is not the same as the rank of a matrix. The rank of a tensor is the number of indices required to uniquely select each element of the tensor. Rank is also known as "order", "degree", or "ndims."
func ReadFile ¶
func ReadFile(scope *Scope, filename tf.Output) (contents tf.Output)
Reads and outputs the entire contents of the input filename.
func ReadVariableOp ¶
func ReadVariableOp(scope *Scope, resource tf.Output, dtype tf.DataType) (value tf.Output)
Reads the value of a variable.
The tensor returned by this operation is immutable.
The value returned by this operation is guaranteed to be influenced by all the writes on which this operation depends directly or indirectly, and to not be influenced by any of the writes which depend directly or indirectly on this operation.
Arguments:
resource: handle to the resource in which to store the variable. dtype: the dtype of the value.
func ReaderNumRecordsProducedV2 ¶
func ReaderNumRecordsProducedV2(scope *Scope, reader_handle tf.Output) (records_produced tf.Output)
Returns the number of records this Reader has produced.
This is the same as the number of ReaderRead executions that have succeeded.
Arguments:
reader_handle: Handle to a Reader.
func ReaderNumWorkUnitsCompletedV2 ¶
func ReaderNumWorkUnitsCompletedV2(scope *Scope, reader_handle tf.Output) (units_completed tf.Output)
Returns the number of work units this Reader has finished processing.
Arguments:
reader_handle: Handle to a Reader.
func ReaderReadUpToV2 ¶
func ReaderReadUpToV2(scope *Scope, reader_handle tf.Output, queue_handle tf.Output, num_records tf.Output) (keys tf.Output, values tf.Output)
Returns up to `num_records` (key, value) pairs produced by a Reader.
Will dequeue from the input queue if necessary (e.g. when the Reader needs to start reading from a new file since it has finished with the previous file). It may return less than `num_records` even before the last batch.
Arguments:
reader_handle: Handle to a `Reader`. queue_handle: Handle to a `Queue`, with string work items. num_records: number of records to read from `Reader`.
Returns A 1-D tensor.A 1-D tensor.
func ReaderReadV2 ¶
func ReaderReadV2(scope *Scope, reader_handle tf.Output, queue_handle tf.Output) (key tf.Output, value tf.Output)
Returns the next record (key, value pair) produced by a Reader.
Will dequeue from the input queue if necessary (e.g. when the Reader needs to start reading from a new file since it has finished with the previous file).
Arguments:
reader_handle: Handle to a Reader. queue_handle: Handle to a Queue, with string work items.
Returns A scalar.A scalar.
func ReaderResetV2 ¶
func ReaderResetV2(scope *Scope, reader_handle tf.Output) (o *tf.Operation)
Restore a Reader to its initial clean state.
Arguments:
reader_handle: Handle to a Reader.
Returns the created operation.
func ReaderRestoreStateV2 ¶
func ReaderRestoreStateV2(scope *Scope, reader_handle tf.Output, state tf.Output) (o *tf.Operation)
Restore a reader to a previously saved state.
Not all Readers support being restored, so this can produce an Unimplemented error.
Arguments:
reader_handle: Handle to a Reader. state: Result of a ReaderSerializeState of a Reader with type
matching reader_handle.
Returns the created operation.
func ReaderSerializeStateV2 ¶
func ReaderSerializeStateV2(scope *Scope, reader_handle tf.Output) (state tf.Output)
Produce a string tensor that encodes the state of a Reader.
Not all Readers support being serialized, so this can produce an Unimplemented error.
Arguments:
reader_handle: Handle to a Reader.
func Real ¶
func Real(scope *Scope, input tf.Output, optional ...RealAttr) (output tf.Output)
Returns the real part of a complex number.
Given a tensor `input` of complex numbers, this operation returns a tensor of type `float` that is the real part of each element in `input`. All elements in `input` must be complex numbers of the form \\(a + bj\\), where *a* is the real
part returned by this operation and *b* is the imaginary part.
For example:
“` # tensor 'input' is [-2.25 + 4.75j, 3.25 + 5.75j] tf.real(input) ==> [-2.25, 3.25] “`
func RealDiv ¶
func RealDiv(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x / y element-wise for real types.
If `x` and `y` are reals, this will return the floating-point division.
*NOTE*: `Div` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Reciprocal ¶
func Reciprocal(scope *Scope, x tf.Output) (y tf.Output)
Computes the reciprocal of x element-wise.
I.e., \\(y = 1 / x\\).
func ReciprocalGrad ¶
func ReciprocalGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient for the inverse of `x` wrt its input.
Specifically, `grad = -dy * y*y`, where `y = 1/x`, and `dy` is the corresponding input gradient.
func RecordInput ¶
func RecordInput(scope *Scope, file_pattern string, optional ...RecordInputAttr) (records tf.Output)
Emits randomized records.
Arguments:
file_pattern: Glob pattern for the data files.
Returns A tensor of shape [batch_size].
func ReduceJoin ¶
func ReduceJoin(scope *Scope, inputs tf.Output, reduction_indices tf.Output, optional ...ReduceJoinAttr) (output tf.Output)
Joins a string Tensor across the given dimensions.
Computes the string join across dimensions in the given string Tensor of shape `[d_0, d_1, ..., d_n-1]`. Returns a new Tensor created by joining the input strings with the given separator (default: empty string). Negative indices are counted backwards from the end, with `-1` being equivalent to `n - 1`.
For example:
“` # 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, []) ==> ["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 Relu ¶
func Relu(scope *Scope, features tf.Output) (activations tf.Output)
Computes rectified linear: `max(features, 0)`.
func Relu6 ¶
func Relu6(scope *Scope, features tf.Output) (activations tf.Output)
Computes rectified linear 6: `min(max(features, 0), 6)`.
func Relu6Grad ¶
func Relu6Grad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)
Computes rectified linear 6 gradients for a Relu6 operation.
Arguments:
gradients: The backpropagated gradients to the corresponding Relu6 operation. features: The features passed as input to the corresponding Relu6 operation.
Returns The gradients: `gradients * (features > 0) * (features < 6)`.
func ReluGrad ¶
func ReluGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)
Computes rectified linear gradients for a Relu operation.
Arguments:
gradients: The backpropagated gradients to the corresponding Relu operation. features: The features passed as input to the corresponding Relu operation, OR
the outputs of that operation (both work equivalently).
Returns `gradients * (features > 0)`.
func RemoteFusedGraphExecute ¶
func RemoteFusedGraphExecute(scope *Scope, values []tf.Output, N int64, U tf.DataType, serialized_graph_transfer_info string) (output []tf.Output)
Execute a sub graph on a remote processor transferred by GraphTransferer.
The graph specifications are serialized by protobuf as graph_transfer_info. The implementation / limitations may differ for each platform and each available peripheral.
func RequantizationRange ¶
func RequantizationRange(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output) (output_min tf.Output, output_max tf.Output)
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 Requantize ¶
func Requantize(scope *Scope, input tf.Output, input_min tf.Output, input_max tf.Output, requested_output_min tf.Output, requested_output_max tf.Output, out_type tf.DataType) (output tf.Output, output_min tf.Output, output_max tf.Output)
Convert 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 Reshape ¶
func Reshape(scope *Scope, tensor tf.Output, shape tf.Output) (output tf.Output)
Reshapes a tensor.
Given `tensor`, this operation returns a tensor that has the same values as `tensor` with shape `shape`.
If one component of `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:
“`prettyprint # tensor 't' is [1, 2, 3, 4, 5, 6, 7, 8, 9] # tensor 't' has shape [9] reshape(t, [3, 3]) ==> [[1, 2, 3],
[4, 5, 6], [7, 8, 9]]
# tensor 't' is [[[1, 1], [2, 2]], # [[3, 3], [4, 4]]] # tensor 't' has shape [2, 2, 2] reshape(t, [2, 4]) ==> [[1, 1, 2, 2],
[3, 3, 4, 4]]
# tensor 't' is [[[1, 1, 1], # [2, 2, 2]], # [[3, 3, 3], # [4, 4, 4]], # [[5, 5, 5], # [6, 6, 6]]] # tensor 't' has shape [3, 2, 3] # pass '[-1]' to flatten 't' reshape(t, [-1]) ==> [1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6]
# -1 can also be used to infer the shape
# -1 is inferred to be 9: reshape(t, [2, -1]) ==> [[1, 1, 1, 2, 2, 2, 3, 3, 3],
[4, 4, 4, 5, 5, 5, 6, 6, 6]]
# -1 is inferred to be 2: reshape(t, [-1, 9]) ==> [[1, 1, 1, 2, 2, 2, 3, 3, 3],
[4, 4, 4, 5, 5, 5, 6, 6, 6]]
# -1 is inferred to be 3: reshape(t, [ 2, -1, 3]) ==> [[[1, 1, 1],
[2, 2, 2], [3, 3, 3]], [[4, 4, 4], [5, 5, 5], [6, 6, 6]]]
# tensor 't' is [7] # shape `[]` reshapes to a scalar reshape(t, []) ==> 7 “`
Arguments:
shape: Defines the shape of the output tensor.
func ResizeArea ¶
func ResizeArea(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeAreaAttr) (resized_images tf.Output)
Resize `images` to `size` using area interpolation.
Input images can be of different types but output images are always float.
Arguments:
images: 4-D with shape `[batch, height, width, channels]`. size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
Returns 4-D with shape `[batch, new_height, new_width, channels]`.
func ResizeBicubic ¶
func ResizeBicubic(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeBicubicAttr) (resized_images tf.Output)
Resize `images` to `size` using bicubic interpolation.
Input images can be of different types but output images are always float.
Arguments:
images: 4-D with shape `[batch, height, width, channels]`. size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
Returns 4-D with shape `[batch, new_height, new_width, channels]`.
func ResizeBilinear ¶
func ResizeBilinear(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeBilinearAttr) (resized_images tf.Output)
Resize `images` to `size` using bilinear interpolation.
Input images can be of different types but output images are always float.
Arguments:
images: 4-D with shape `[batch, height, width, channels]`. size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
Returns 4-D with shape `[batch, new_height, new_width, channels]`.
func ResizeBilinearGrad ¶
func ResizeBilinearGrad(scope *Scope, grads tf.Output, original_image tf.Output, optional ...ResizeBilinearGradAttr) (output tf.Output)
Computes the gradient of bilinear interpolation.
Arguments:
grads: 4-D with shape `[batch, height, width, channels]`. original_image: 4-D with shape `[batch, orig_height, orig_width, channels]`,
The image tensor that was resized.
Returns 4-D with shape `[batch, orig_height, orig_width, channels]`. Gradients with respect to the input image. Input image must have been float or double.
func ResizeNearestNeighbor ¶
func ResizeNearestNeighbor(scope *Scope, images tf.Output, size tf.Output, optional ...ResizeNearestNeighborAttr) (resized_images tf.Output)
Resize `images` to `size` using nearest neighbor interpolation.
Arguments:
images: 4-D with shape `[batch, height, width, channels]`. size: = A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
Returns 4-D with shape `[batch, new_height, new_width, channels]`.
func ResizeNearestNeighborGrad ¶
func ResizeNearestNeighborGrad(scope *Scope, grads tf.Output, size tf.Output, optional ...ResizeNearestNeighborGradAttr) (output tf.Output)
Computes the gradient of nearest neighbor interpolation.
Arguments:
grads: 4-D with shape `[batch, height, width, channels]`. size: = A 1-D int32 Tensor of 2 elements: `orig_height, orig_width`. The
original input size.
Returns 4-D with shape `[batch, orig_height, orig_width, channels]`. Gradients with respect to the input image.
func ResourceApplyAdadelta ¶
func ResourceApplyAdadelta(scope *Scope, var_ tf.Output, accum tf.Output, accum_update tf.Output, lr tf.Output, rho tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdadeltaAttr) (o *tf.Operation)
Update '*var' according to the adadelta scheme.
accum = rho() * accum + (1 - rho()) * grad.square(); update = (update_accum + epsilon).sqrt() * (accum + epsilon()).rsqrt() * grad; update_accum = rho() * update_accum + (1 - rho()) * update.square(); var -= update;
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). accum_update: Should be from a Variable(). lr: Scaling factor. Must be a scalar. rho: Decay factor. Must be a scalar. epsilon: Constant factor. Must be a scalar. grad: The gradient.
Returns the created operation.
func ResourceApplyAdagrad ¶
func ResourceApplyAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, optional ...ResourceApplyAdagradAttr) (o *tf.Operation)
Update '*var' according to the adagrad scheme.
accum += grad * grad var -= lr * grad * (1 / sqrt(accum))
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Scaling factor. Must be a scalar. grad: The gradient.
Returns the created operation.
func ResourceApplyAdagradDA ¶
func ResourceApplyAdagradDA(scope *Scope, var_ tf.Output, gradient_accumulator tf.Output, gradient_squared_accumulator tf.Output, grad tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, global_step tf.Output, optional ...ResourceApplyAdagradDAAttr) (o *tf.Operation)
Update '*var' according to the proximal adagrad scheme.
Arguments:
var_: Should be from a Variable(). gradient_accumulator: Should be from a Variable(). gradient_squared_accumulator: Should be from a Variable(). grad: The gradient. lr: Scaling factor. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. global_step: Training step number. Must be a scalar.
Returns the created operation.
func ResourceApplyAdam ¶
func ResourceApplyAdam(scope *Scope, var_ tf.Output, m tf.Output, v tf.Output, beta1_power tf.Output, beta2_power tf.Output, lr tf.Output, beta1 tf.Output, beta2 tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyAdamAttr) (o *tf.Operation)
Update '*var' according to the Adam algorithm.
lr_t <- learning_rate * sqrt(1 - beta2^t) / (1 - beta1^t) m_t <- beta1 * m_{t-1} + (1 - beta1) * g_t v_t <- beta2 * v_{t-1} + (1 - beta2) * g_t * g_t 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 ResourceApplyCenteredRMSProp ¶
func ResourceApplyCenteredRMSProp(scope *Scope, var_ tf.Output, mg tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyCenteredRMSPropAttr) (o *tf.Operation)
Update '*var' according to the centered RMSProp algorithm.
The centered RMSProp algorithm uses an estimate of the centered second moment (i.e., the variance) for normalization, as opposed to regular RMSProp, which uses the (uncentered) second moment. This often helps with training, but is slightly more expensive in terms of computation and memory.
Note that in dense implementation of this algorithm, mg, ms, and mom will update even if the grad is zero, but in this sparse implementation, mg, ms, and mom will not update in iterations during which the grad is zero.
mean_square = decay * mean_square + (1-decay) * gradient ** 2 mean_grad = decay * mean_grad + (1-decay) * gradient
Delta = learning_rate * gradient / sqrt(mean_square + epsilon - mean_grad ** 2)
mg <- rho * mg_{t-1} + (1-rho) * grad ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms - mg * mg + epsilon) var <- var - mom
Arguments:
var_: Should be from a Variable(). mg: Should be from a Variable(). ms: Should be from a Variable(). mom: Should be from a Variable(). lr: Scaling factor. Must be a scalar. rho: Decay rate. Must be a scalar. epsilon: Ridge term. Must be a scalar. grad: The gradient.
Returns the created operation.
func ResourceApplyFtrl ¶
func ResourceApplyFtrl(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, lr_power tf.Output, optional ...ResourceApplyFtrlAttr) (o *tf.Operation)
Update '*var' according to the Ftrl-proximal scheme.
accum_new = accum + grad * grad linear += grad + (accum_new^(-lr_power) - accum^(-lr_power)) / lr * var quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). linear: Should be from a Variable(). grad: The gradient. lr: Scaling factor. Must be a scalar. l1: L1 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 ResourceApplyGradientDescent ¶
func ResourceApplyGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, delta tf.Output, optional ...ResourceApplyGradientDescentAttr) (o *tf.Operation)
Update '*var' by subtracting 'alpha' * 'delta' from it.
Arguments:
var_: Should be from a Variable(). alpha: Scaling factor. Must be a scalar. delta: The change.
Returns the created operation.
func ResourceApplyMomentum ¶
func ResourceApplyMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, momentum tf.Output, optional ...ResourceApplyMomentumAttr) (o *tf.Operation)
Update '*var' according to the momentum scheme. Set use_nesterov = True if you
want to use Nesterov momentum.
accum = accum * momentum + grad var -= lr * accum
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Scaling factor. Must be a scalar. grad: The gradient. momentum: Momentum. Must be a scalar.
Returns the created operation.
func ResourceApplyProximalAdagrad ¶
func ResourceApplyProximalAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, optional ...ResourceApplyProximalAdagradAttr) (o *tf.Operation)
Update '*var' and '*accum' according to FOBOS with Adagrad learning rate.
accum += grad * grad prox_v = var - lr * grad * (1 / sqrt(accum)) var = sign(prox_v)/(1+lr*l2) * max{|prox_v|-lr*l1,0}
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Scaling factor. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. grad: The gradient.
Returns the created operation.
func ResourceApplyProximalGradientDescent ¶
func ResourceApplyProximalGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, l1 tf.Output, l2 tf.Output, delta tf.Output, optional ...ResourceApplyProximalGradientDescentAttr) (o *tf.Operation)
Update '*var' as FOBOS algorithm with fixed learning rate.
prox_v = var - alpha * delta var = sign(prox_v)/(1+alpha*l2) * max{|prox_v|-alpha*l1,0}
Arguments:
var_: Should be from a Variable(). alpha: Scaling factor. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. delta: The change.
Returns the created operation.
func ResourceApplyRMSProp ¶
func ResourceApplyRMSProp(scope *Scope, var_ tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, optional ...ResourceApplyRMSPropAttr) (o *tf.Operation)
Update '*var' according to the RMSProp algorithm.
Note that in dense implementation of this algorithm, ms and mom will update even if the grad is zero, but in this sparse implementation, ms and mom will not update in iterations during which the grad is zero.
mean_square = decay * mean_square + (1-decay) * gradient ** 2 Delta = learning_rate * gradient / sqrt(mean_square + epsilon)
ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom
Arguments:
var_: Should be from a Variable(). ms: Should be from a Variable(). mom: Should be from a Variable(). lr: Scaling factor. Must be a scalar. rho: Decay rate. Must be a scalar. epsilon: Ridge term. Must be a scalar. grad: The gradient.
Returns the created operation.
func ResourceGather ¶
func ResourceGather(scope *Scope, resource tf.Output, indices tf.Output, dtype tf.DataType, optional ...ResourceGatherAttr) (output tf.Output)
Gather slices from the variable pointed to by `resource` according to `indices`.
`indices` must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape `indices.shape + params.shape[1:]` where:
“`python
# Scalar indices output[:, ..., :] = params[indices, :, ... :] # Vector indices output[i, :, ..., :] = params[indices[i], :, ... :] # Higher rank indices output[i, ..., j, :, ... :] = params[indices[i, ..., j], :, ..., :]
“`
func ResourceScatterAdd ¶
func ResourceScatterAdd(scope *Scope, resource tf.Output, indices tf.Output, updates tf.Output) (o *tf.Operation)
Adds sparse updates to the variable referenced by `resource`.
This operation computes
# Scalar indices ref[indices, ...] += updates[...] # Vector indices (for each i) ref[indices[i], ...] += updates[i, ...] # High rank indices (for each i, ..., j) ref[indices[i, ..., j], ...] += updates[i, ..., j, ...]
Duplicate entries are handled correctly: if multiple `indices` reference the same location, their contributions add.
Requires `updates.shape = indices.shape + ref.shape[1:]`.
<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="../../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 ResourceSparseApplyAdadelta ¶
func ResourceSparseApplyAdadelta(scope *Scope, var_ tf.Output, accum tf.Output, accum_update tf.Output, lr tf.Output, rho tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyAdadeltaAttr) (o *tf.Operation)
var: Should be from a Variable().
Arguments:
accum: Should be from a Variable(). accum_update: : Should be from a Variable(). lr: Learning rate. Must be a scalar. rho: Decay factor. Must be a scalar. epsilon: Constant factor. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var and accum.
Returns the created operation.
func ResourceSparseApplyAdagrad ¶
func ResourceSparseApplyAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyAdagradAttr) (o *tf.Operation)
Update relevant entries in '*var' and '*accum' according to the adagrad scheme.
That is for rows we have grad for, we update var and accum as follows: accum += grad * grad var -= lr * grad * (1 / sqrt(accum))
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Learning rate. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var and accum.
Returns the created operation.
func ResourceSparseApplyAdagradDA ¶
func ResourceSparseApplyAdagradDA(scope *Scope, var_ tf.Output, gradient_accumulator tf.Output, gradient_squared_accumulator tf.Output, grad tf.Output, indices tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, global_step tf.Output, optional ...ResourceSparseApplyAdagradDAAttr) (o *tf.Operation)
Update entries in '*var' and '*accum' according to the proximal adagrad scheme.
Arguments:
var_: Should be from a Variable(). gradient_accumulator: Should be from a Variable(). gradient_squared_accumulator: Should be from a Variable(). grad: The gradient. indices: A vector of indices into the first dimension of var and accum. lr: Learning rate. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. global_step: Training step number. Must be a scalar.
Returns the created operation.
func ResourceSparseApplyCenteredRMSProp ¶
func ResourceSparseApplyCenteredRMSProp(scope *Scope, var_ tf.Output, mg tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyCenteredRMSPropAttr) (o *tf.Operation)
Update '*var' according to the centered RMSProp algorithm.
The centered RMSProp algorithm uses an estimate of the centered second moment (i.e., the variance) for normalization, as opposed to regular RMSProp, which uses the (uncentered) second moment. This often helps with training, but is slightly more expensive in terms of computation and memory.
Note that in dense implementation of this algorithm, mg, ms, and mom will update even if the grad is zero, but in this sparse implementation, mg, ms, and mom will not update in iterations during which the grad is zero.
mean_square = decay * mean_square + (1-decay) * gradient ** 2 mean_grad = decay * mean_grad + (1-decay) * gradient Delta = learning_rate * gradient / sqrt(mean_square + epsilon - mean_grad ** 2)
ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom
Arguments:
var_: Should be from a Variable(). mg: Should be from a Variable(). ms: Should be from a Variable(). mom: Should be from a Variable(). lr: Scaling factor. Must be a scalar. rho: Decay rate. Must be a scalar. epsilon: Ridge term. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var, ms and mom.
Returns the created operation.
func ResourceSparseApplyFtrl ¶
func ResourceSparseApplyFtrl(scope *Scope, var_ tf.Output, accum tf.Output, linear tf.Output, grad tf.Output, indices tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, lr_power tf.Output, optional ...ResourceSparseApplyFtrlAttr) (o *tf.Operation)
Update relevant entries in '*var' according to the Ftrl-proximal scheme.
That is for rows we have grad for, we update var, accum and linear as follows: accum_new = accum + grad * grad linear += grad + (accum_new^(-lr_power) - accum^(-lr_power)) / lr * var quadratic = 1.0 / (accum_new^(lr_power) * lr) + 2 * l2 var = (sign(linear) * l1 - linear) / quadratic if |linear| > l1 else 0.0 accum = accum_new
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). linear: Should be from a Variable(). grad: The gradient. indices: A vector of indices into the first dimension of var and accum. lr: Scaling factor. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. lr_power: Scaling factor. Must be a scalar.
Returns the created operation.
func ResourceSparseApplyMomentum ¶
func ResourceSparseApplyMomentum(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, grad tf.Output, indices tf.Output, momentum tf.Output, optional ...ResourceSparseApplyMomentumAttr) (o *tf.Operation)
Update relevant entries in '*var' and '*accum' according to the momentum scheme.
Set use_nesterov = True if you want to use Nesterov momentum.
That is for rows we have grad for, we update var and accum as follows:
accum = accum * momentum + grad var -= lr * accum
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Learning rate. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var and accum. momentum: Momentum. Must be a scalar.
Returns the created operation.
func ResourceSparseApplyProximalAdagrad ¶
func ResourceSparseApplyProximalAdagrad(scope *Scope, var_ tf.Output, accum tf.Output, lr tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyProximalAdagradAttr) (o *tf.Operation)
Sparse update entries in '*var' and '*accum' according to FOBOS algorithm.
That is for rows we have grad for, we update var and accum as follows: accum += grad * grad prox_v = var prox_v -= lr * grad * (1 / sqrt(accum)) var = sign(prox_v)/(1+lr*l2) * max{|prox_v|-lr*l1,0}
Arguments:
var_: Should be from a Variable(). accum: Should be from a Variable(). lr: Learning rate. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var and accum.
Returns the created operation.
func ResourceSparseApplyProximalGradientDescent ¶
func ResourceSparseApplyProximalGradientDescent(scope *Scope, var_ tf.Output, alpha tf.Output, l1 tf.Output, l2 tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyProximalGradientDescentAttr) (o *tf.Operation)
Sparse update '*var' as FOBOS algorithm with fixed learning rate.
That is for rows we have grad for, we update var as follows: prox_v = var - alpha * grad var = sign(prox_v)/(1+alpha*l2) * max{|prox_v|-alpha*l1,0}
Arguments:
var_: Should be from a Variable(). alpha: Scaling factor. Must be a scalar. l1: L1 regularization. Must be a scalar. l2: L2 regularization. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var and accum.
Returns the created operation.
func ResourceSparseApplyRMSProp ¶
func ResourceSparseApplyRMSProp(scope *Scope, var_ tf.Output, ms tf.Output, mom tf.Output, lr tf.Output, rho tf.Output, momentum tf.Output, epsilon tf.Output, grad tf.Output, indices tf.Output, optional ...ResourceSparseApplyRMSPropAttr) (o *tf.Operation)
Update '*var' according to the RMSProp algorithm.
Note that in dense implementation of this algorithm, ms and mom will update even if the grad is zero, but in this sparse implementation, ms and mom will not update in iterations during which the grad is zero.
mean_square = decay * mean_square + (1-decay) * gradient ** 2 Delta = learning_rate * gradient / sqrt(mean_square + epsilon)
ms <- rho * ms_{t-1} + (1-rho) * grad * grad mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon) var <- var - mom
Arguments:
var_: Should be from a Variable(). ms: Should be from a Variable(). mom: Should be from a Variable(). lr: Scaling factor. Must be a scalar. rho: Decay rate. Must be a scalar. epsilon: Ridge term. Must be a scalar. grad: The gradient. indices: A vector of indices into the first dimension of var, ms and mom.
Returns the created operation.
func Restore ¶
func Restore(scope *Scope, file_pattern tf.Output, tensor_name tf.Output, dt tf.DataType, optional ...RestoreAttr) (tensor tf.Output)
Restores a tensor from checkpoint files.
Reads a tensor stored in one or several files. If there are several files (for instance because a tensor was saved as slices), `file_pattern` may contain wildcard symbols (`*` and `?`) in the filename portion only, not in the directory portion.
If a `file_pattern` matches several files, `preferred_shard` can be used to hint in which file the requested tensor is likely to be found. This op will first open the file at index `preferred_shard` in the list of matching files and try to restore tensors from that file. Only if some tensors or tensor slices are not found in that first file, then the Op opens all the files. Setting `preferred_shard` to match the value passed as the `shard` input of a matching `Save` Op may speed up Restore. This attribute only affects performance, not correctness. The default value -1 means files are processed in order.
See also `RestoreSlice`.
Arguments:
file_pattern: Must have a single element. The pattern of the files from
which we read the tensor.
tensor_name: Must have a single element. The name of the tensor to be
restored.
dt: The type of the tensor to be restored.
Returns The restored tensor.
func RestoreSlice ¶
func RestoreSlice(scope *Scope, file_pattern tf.Output, tensor_name tf.Output, shape_and_slice tf.Output, dt tf.DataType, optional ...RestoreSliceAttr) (tensor tf.Output)
Restores a tensor from checkpoint files.
This is like `Restore` except that restored tensor can be listed as filling only a slice of a larger tensor. `shape_and_slice` specifies the shape of the larger tensor and the slice that the restored tensor covers.
The `shape_and_slice` input has the same format as the elements of the `shapes_and_slices` input of the `SaveSlices` op.
Arguments:
file_pattern: Must have a single element. The pattern of the files from
which we read the tensor.
tensor_name: Must have a single element. The name of the tensor to be
restored.
shape_and_slice: Scalar. The shapes and slice specifications to use when
restoring a tensors.
dt: The type of the tensor to be restored.
Returns The restored tensor.
func RestoreV2 ¶
func RestoreV2(scope *Scope, prefix tf.Output, tensor_names tf.Output, shape_and_slices tf.Output, dtypes []tf.DataType) (tensors []tf.Output)
Restores tensors from a V2 checkpoint.
For backward compatibility with the V1 format, this Op currently allows restoring from a V1 checkpoint as well:
- This Op first attempts to find the V2 index file pointed to by "prefix", and if found proceed to read it as a V2 checkpoint; - Otherwise the V1 read path is invoked.
Relying on this behavior is not recommended, as the ability to fall back to read V1 might be deprecated and eventually removed.
By default, restores the named tensors in full. If the caller wishes to restore specific slices of stored tensors, "shape_and_slices" should be non-empty strings and correspondingly well-formed.
Callers must ensure all the named tensors are indeed stored in the checkpoint.
Arguments:
prefix: Must have a single element. The prefix of a V2 checkpoint.
tensor_names: shape {N}. The names of the tensors to be restored.
shape_and_slices: shape {N}. The slice specs of the tensors to be restored.
Empty strings indicate that they are non-partitioned tensors.
dtypes: shape {N}. The list of expected dtype for the tensors. Must match
those stored in the checkpoint.
Returns shape {N}. The restored tensors, whose shapes are read from the checkpoint directly.
func Reverse ¶
func Reverse(scope *Scope, tensor tf.Output, dims tf.Output) (output tf.Output)
Reverses specific dimensions of a tensor.
Given a `tensor`, and a `bool` tensor `dims` representing the dimensions of `tensor`, this operation reverses each dimension i of `tensor` where `dims[i]` is `True`.
`tensor` can have up to 8 dimensions. The number of dimensions of `tensor` must equal the number of elements in `dims`. In other words:
`rank(tensor) = size(dims)`
For example:
“`prettyprint # tensor 't' is [[[[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]], # [[12, 13, 14, 15], # [16, 17, 18, 19], # [20, 21, 22, 23]]]] # tensor 't' shape is [1, 2, 3, 4]
# 'dims' is [False, False, False, True] reverse(t, dims) ==> [[[[ 3, 2, 1, 0],
[ 7, 6, 5, 4], [ 11, 10, 9, 8]], [[15, 14, 13, 12], [19, 18, 17, 16], [23, 22, 21, 20]]]]
# 'dims' is [False, True, False, False] reverse(t, dims) ==> [[[[12, 13, 14, 15],
[16, 17, 18, 19], [20, 21, 22, 23] [[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]]]]
# 'dims' is [False, False, True, False] reverse(t, dims) ==> [[[[8, 9, 10, 11],
[4, 5, 6, 7], [0, 1, 2, 3]] [[20, 21, 22, 23], [16, 17, 18, 19], [12, 13, 14, 15]]]]
“`
Arguments:
tensor: Up to 8-D. dims: 1-D. The dimensions to reverse.
Returns The same shape as `tensor`.
func ReverseSequence ¶
func ReverseSequence(scope *Scope, input tf.Output, seq_lengths tf.Output, seq_dim int64, optional ...ReverseSequenceAttr) (output tf.Output)
Reverses variable length slices.
This op first slices `input` along the dimension `batch_dim`, and for each slice `i`, reverses the first `seq_lengths[i]` elements along the dimension `seq_dim`.
The elements of `seq_lengths` must obey `seq_lengths[i] <= input.dims[seq_dim]`, and `seq_lengths` must be a vector of length `input.dims[batch_dim]`.
The output slice `i` along dimension `batch_dim` is then given by input slice `i`, with the first `seq_lengths[i]` slices along dimension `seq_dim` reversed.
For example:
“`prettyprint # 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:
“`prettyprint # Given this: batch_dim = 2 seq_dim = 0 input.dims = (8, ?, 4, ...) seq_lengths = [7, 2, 3, 5]
# then slices of input are reversed on seq_dim, but only up to seq_lengths: output[0:7, :, 0, :, ...] = input[7:0:-1, :, 0, :, ...] output[0:2, :, 1, :, ...] = input[2:0:-1, :, 1, :, ...] output[0:3, :, 2, :, ...] = input[3:0:-1, :, 2, :, ...] output[0:5, :, 3, :, ...] = input[5:0:-1, :, 3, :, ...]
# while entries past seq_lens are copied through: output[7:, :, 0, :, ...] = input[7:, :, 0, :, ...] output[2:, :, 1, :, ...] = input[2:, :, 1, :, ...] output[3:, :, 2, :, ...] = input[3:, :, 2, :, ...] output[2:, :, 3, :, ...] = input[2:, :, 3, :, ...] “`
Arguments:
input: The input to reverse. seq_lengths: 1-D with length `input.dims(batch_dim)` and
`max(seq_lengths) <= input.dims(seq_dim)`
seq_dim: The dimension which is partially reversed.
Returns The partially reversed input. It has the same shape as `input`.
func ReverseV2 ¶
func ReverseV2(scope *Scope, tensor tf.Output, axis tf.Output) (output tf.Output)
Reverses specific dimensions of a tensor.
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:
“`prettyprint # 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.
Returns The same shape as `tensor`.
func Rint ¶
func Rint(scope *Scope, x tf.Output) (y tf.Output)
Returns element-wise integer closest to x.
If the result is midway between two representable values, the even representable is chosen. For example:
“` rint(-1.5) ==> -2.0 rint(0.5000001) ==> 1.0 rint([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0]) ==> [-2., -2., -0., 0., 2., 2., 2.] “`
func Round ¶
func Round(scope *Scope, x tf.Output) (y tf.Output)
Rounds the values of a tensor to the nearest integer, element-wise.
Rounds half to even. Also known as bankers rounding. If you want to round according to the current system rounding mode use std::cint.
func Rsqrt ¶
func Rsqrt(scope *Scope, x tf.Output) (y tf.Output)
Computes reciprocal of square root of x element-wise.
I.e., \\(y = 1 / \sqrt{x}\\).
func RsqrtGrad ¶
func RsqrtGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient for the rsqrt of `x` wrt its input.
Specifically, `grad = dy * -0.5 * y^3`, where `y = rsqrt(x)`, and `dy` is the corresponding input gradient.
func SampleDistortedBoundingBox ¶
func SampleDistortedBoundingBox(scope *Scope, image_size tf.Output, bounding_boxes tf.Output, optional ...SampleDistortedBoundingBoxAttr) (begin tf.Output, size tf.Output, bboxes tf.Output)
Generate a single randomly distorted bounding box for an image.
Bounding box annotations are often supplied in addition to ground-truth labels in image recognition or object localization tasks. A common technique for training such a system is to randomly distort an image while preserving its content, i.e. *data augmentation*. This Op outputs a randomly distorted localization of an object, i.e. bounding box, given an `image_size`, `bounding_boxes` and a series of constraints.
The output of this Op is a single bounding box that may be used to crop the original image. The output is returned as 3 tensors: `begin`, `size` and `bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize what the bounding box looks like.
Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and height of the underlying image.
For example,
“`python
# Generate a single distorted bounding box.
begin, size, bbox_for_draw = tf.image.sample_distorted_bounding_box(
tf.shape(image),
bounding_boxes=bounding_boxes)
# Draw the bounding box in an image summary.
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
bbox_for_draw)
tf.image_summary('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 Save ¶
func Save(scope *Scope, filename tf.Output, tensor_names tf.Output, data []tf.Output) (o *tf.Operation)
Saves the input tensors to disk.
The size of `tensor_names` must match the number of tensors in `data`. `data[i]` is written to `filename` with name `tensor_names[i]`.
See also `SaveSlices`.
Arguments:
filename: Must have a single element. The name of the file to which we write
the tensor.
tensor_names: Shape `[N]`. The names of the tensors to be saved. data: `N` tensors to save.
Returns the created operation.
func SaveSlices ¶
func SaveSlices(scope *Scope, filename tf.Output, tensor_names tf.Output, shapes_and_slices tf.Output, data []tf.Output) (o *tf.Operation)
Saves input tensors slices to disk.
This is like `Save` except that tensors can be listed in the saved file as being a slice of a larger tensor. `shapes_and_slices` specifies the shape of the larger tensor and the slice that this tensor covers. `shapes_and_slices` must have as many elements as `tensor_names`.
Elements of the `shapes_and_slices` input must either be:
* The empty string, in which case the corresponding tensor is
saved normally.
* A string of the form `dim0 dim1 ... dimN-1 slice-spec` where the
`dimI` are the dimensions of the larger tensor and `slice-spec` specifies what part is covered by the tensor to save.
`slice-spec` itself is a `:`-separated list: `slice0:slice1:...:sliceN-1` where each `sliceI` is either:
* The string `-` meaning that the slice covers all indices of this dimension * `start,length` where `start` and `length` are integers. In that
case the slice covers `length` indices starting at `start`.
See also `Save`.
Arguments:
filename: Must have a single element. The name of the file to which we write the
tensor.
tensor_names: Shape `[N]`. The names of the tensors to be saved. shapes_and_slices: Shape `[N]`. The shapes and slice specifications to use when
saving the tensors.
data: `N` tensors to save.
Returns the created operation.
func SaveV2 ¶
func SaveV2(scope *Scope, prefix tf.Output, tensor_names tf.Output, shape_and_slices tf.Output, tensors []tf.Output) (o *tf.Operation)
Saves tensors in V2 checkpoint format.
By default, saves the named tensors in full. If the caller wishes to save specific slices of full tensors, "shape_and_slices" should be non-empty strings and correspondingly well-formed.
Arguments:
prefix: Must have a single element. The prefix of the V2 checkpoint to which we
write the tensors.
tensor_names: shape {N}. The names of the tensors to be saved.
shape_and_slices: shape {N}. The slice specs of the tensors to be saved.
Empty strings indicate that they are non-partitioned tensors.
tensors: `N` tensors to save.
Returns the created operation.
func ScalarSummary ¶
func ScalarSummary(scope *Scope, tags tf.Output, values tf.Output) (summary tf.Output)
Outputs a `Summary` protocol buffer with scalar values.
The input `tags` and `values` must have the same shape. The generated summary has a summary value for each tag-value pair in `tags` and `values`.
Arguments:
tags: Tags for the summary. values: Same shape as `tags. Values for the summary.
Returns Scalar. Serialized `Summary` protocol buffer.
func ScatterNd ¶
func ScatterNd(scope *Scope, indices tf.Output, updates tf.Output, shape tf.Output) (output tf.Output)
Creates a new tensor by applying sparse `updates` to individual
values or slices within a zero tensor of the given `shape` tensor according to indices. This operator is the inverse of the [tf.gather_nd](#gather_nd) operator which extracts values or slices from a given tensor.
TODO(simister): Add a link to Variable.__getitem__ documentation on slice syntax.
`shape` is a `TensorShape` with rank `P` and `indices` is a `Tensor` of rank `Q`.
`indices` must be integer tensor, containing indices into `shape`. 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 `shape`.
`updates` is Tensor of rank `Q-1+P-K` with shape:
“` [d_0, ..., d_{Q-2}, shape[K], ..., shape[P-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="../../images/ScatterNd1.png" alt> </div>
In Python, this scatter operation would look like this:
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="../../images/ScatterNd2.png" alt> </div>
In Python, this scatter operation would look like this:
indices = tf.constant([[0], [2]])
updates = tf.constant([[[5, 5, 5, 5], [6, 6, 6, 6],
[7, 7, 7, 7], [8, 8, 8, 8]],
[[5, 5, 5, 5], [6, 6, 6, 6],
[7, 7, 7, 7], [8, 8, 8, 8]]])
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]]]
Arguments:
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 tensor. A tensor of updated values
to store in ref.
shape: A vector. The shape of the resulting tensor.
Returns A new tensor with the given shape and updates applied according to the indices.
func SdcaFprint ¶
func SdcaFprint(scope *Scope, input tf.Output) (output tf.Output)
Computes fingerprints of the input strings.
Arguments:
input: vector of strings to compute fingerprints on.
Returns a (N,2) shaped matrix where N is the number of elements in the input vector. Each row contains the low and high parts of the fingerprint.
func SdcaOptimizer ¶
func SdcaOptimizer(scope *Scope, sparse_example_indices []tf.Output, sparse_feature_indices []tf.Output, sparse_feature_values []tf.Output, dense_features []tf.Output, example_weights tf.Output, example_labels tf.Output, sparse_indices []tf.Output, sparse_weights []tf.Output, dense_weights []tf.Output, example_state_data tf.Output, loss_type string, l1 float32, l2 float32, num_loss_partitions int64, num_inner_iterations int64, optional ...SdcaOptimizerAttr) (out_example_state_data tf.Output, out_delta_sparse_weights []tf.Output, out_delta_dense_weights []tf.Output)
Distributed version of Stochastic Dual Coordinate Ascent (SDCA) optimizer for
linear models with L1 + L2 regularization. As global optimization objective is strongly-convex, the optimizer optimizes the dual objective at each step. The optimizer applies each update one example at a time. Examples are sampled uniformly, and the optimizer is learning rate free and enjoys linear convergence rate.
Proximal Stochastic Dual Coordinate Ascent, Shalev-Shwartz, Shai; Zhang, Tong. 2012 arXiv1211.2717S: http://arxiv.org/pdf/1211.2717v1.pdf
Loss objective = \sum f_{i}(wx_{i}) + (l2 / 2) * |w|^2 + l1 * |w|
Adding vs. Averaging in Distributed Primal-Dual Optimization. Chenxin Ma, Virginia Smith, Martin Jaggi, Michael I. Jordan, Peter Richtarik, Martin Takac http://arxiv.org/abs/1502.03508
Stochastic Dual Coordinate Ascent with Adaptive Probabilities Dominik Csiba, Zheng Qu, Peter Richtarik https://arxiv.org/abs/1502.08053
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 ommitted 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 ¶
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](../../api_docs/python/math_ops.md#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="../../images/SegmentMax.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank 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 ¶
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](../../api_docs/python/math_ops.md#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="../../images/SegmentMean.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank 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 ¶
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](../../api_docs/python/math_ops.md#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="../../images/SegmentMin.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank 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 ¶
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](../../api_docs/python/math_ops.md#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="../../images/SegmentProd.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank 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 ¶
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](../../api_docs/python/math_ops.md#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="../../images/SegmentSum.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank 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 ¶
func Select(scope *Scope, condition tf.Output, t tf.Output, e tf.Output) (output tf.Output)
Selects elements from `t` or `e`, depending on `condition`.
The `t`, and `e` tensors must all have the same shape, and the output will also have that shape.
The `condition` tensor must be a scalar if `t` and `e` are scalars. If `t` and `e` are vectors or higher rank, then `condition` must be either a scalar, a vector with size matching the first dimension of `t`, or must have the same shape as `t`.
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 `t` (if true) or `e` (if false).
If `condition` is a vector and `t` and `e` are higher rank matrices, then it chooses which row (outer dimension) to copy from `t` and `e`. If `condition` has the same shape as `t` and `e`, then it chooses which element to copy from `t` and `e`.
For example:
“`prettyprint # '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:
t: = A `Tensor` which may have the same shape as `condition`.
If `condition` is rank 1, `t` may have higher rank, but its first dimension must match the size of `condition`.
e: = A `Tensor` with the same type and shape as `t`.
Returns = A `Tensor` with the same type and shape as `t` and `e`.
func SelfAdjointEig ¶
func SelfAdjointEig(scope *Scope, input tf.Output) (output tf.Output)
Computes the Eigen Decomposition of a batch of square self-adjoint matrices.
DEPRECATED at GraphDef version 11: Use SelfAdjointEigV2 instead.
The input is a tensor of shape `[..., M, M]` whose inner-most 2 dimensions form square matrices, with the same constraints as the single matrix SelfAdjointEig.
The result is a [..., M+1, M] matrix with [..., 0,:] containing the eigenvalues, and subsequent [...,1:, :] containing the eigenvectors.
Arguments:
input: Shape is `[..., M, M]`.
Returns Shape is `[..., M+1, M]`.
func SelfAdjointEigV2 ¶
func SelfAdjointEigV2(scope *Scope, input tf.Output, optional ...SelfAdjointEigV2Attr) (e tf.Output, v tf.Output)
Computes the eigen decomposition of one or more square self-adjoint matrices.
Computes the eigenvalues and (optionally) eigenvectors of each inner matrix in `input` such that `input[..., :, :] = v[..., :, :] * diag(e[..., :])`.
“`prettyprint # 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 SerializeManySparse ¶
func SerializeManySparse(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output) (serialized_sparse tf.Output)
Serialize an `N`-minibatch `SparseTensor` into an `[N, 3]` string `Tensor`.
The `SparseTensor` must have rank `R` greater than 1, and the first dimension is treated as the minibatch dimension. Elements of the `SparseTensor` must be sorted in increasing order of this first dimension. The serialized `SparseTensor` objects going into each row of `serialized_sparse` will have rank `R-1`.
The minibatch size `N` is extracted from `sparse_shape[0]`.
Arguments:
sparse_indices: 2-D. The `indices` of the minibatch `SparseTensor`. sparse_values: 1-D. The `values` of the minibatch `SparseTensor`. sparse_shape: 1-D. The `shape` of the minibatch `SparseTensor`.
func SerializeSparse ¶
func SerializeSparse(scope *Scope, sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output) (serialized_sparse tf.Output)
Serialize a `SparseTensor` into a string 3-vector (1-D `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 SetSize ¶
func SetSize(scope *Scope, set_indices tf.Output, set_values tf.Output, set_shape tf.Output, optional ...SetSizeAttr) (size tf.Output)
Number of unique elements along last dimension of input `set`.
Input `set` is a `SparseTensor` represented by `set_indices`, `set_values`, and `set_shape`. The last dimension contains values in a set, duplicates are allowed but ignored.
If `validate_indices` is `True`, this op validates the order and range of `set` indices.
Arguments:
set_indices: 2D `Tensor`, indices of a `SparseTensor`. set_values: 1D `Tensor`, values of a `SparseTensor`. set_shape: 1D `Tensor`, shape of a `SparseTensor`.
Returns For `set` ranked `n`, this is a `Tensor` with rank `n-1`, and the same 1st `n-1` dimensions as `set`. Each value is the number of unique elements in the corresponding `[0...n-1]` dimension of `set`.
func Shape ¶
func Shape(scope *Scope, input tf.Output, optional ...ShapeAttr) (output tf.Output)
Returns the shape of a tensor.
This operation returns a 1-D integer tensor representing the shape of `input`.
For example:
“`prettyprint # 't' is [[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]] shape(t) ==> [2, 2, 3] “`
func ShapeN ¶
func ShapeN(scope *Scope, input []tf.Output, optional ...ShapeNAttr) (output []tf.Output)
Returns shape of tensors.
This operation returns N 1-D integer tensors representing shape of `input[i]s`.
func ShardedFilename ¶
func ShardedFilename(scope *Scope, basename tf.Output, shard tf.Output, num_shards tf.Output) (filename tf.Output)
Generate a sharded filename. The filename is printf formatted as
%s-%05d-of-%05d, basename, shard, num_shards.
func ShardedFilespec ¶
func ShardedFilespec(scope *Scope, basename tf.Output, num_shards tf.Output) (filename tf.Output)
Generate a glob pattern matching all sharded file names.
func Sigmoid ¶
func Sigmoid(scope *Scope, x tf.Output) (y tf.Output)
Computes sigmoid of `x` element-wise.
Specifically, `y = 1 / (1 + exp(-x))`.
func SigmoidGrad ¶
func SigmoidGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient of the sigmoid of `x` wrt its input.
Specifically, `grad = dy * y * (1 - y)`, where `y = sigmoid(x)`, and `dy` is the corresponding input gradient.
func Sign ¶
func Sign(scope *Scope, x tf.Output) (y tf.Output)
Returns an element-wise indication of the sign of a number.
`y = sign(x) = -1` if `x < 0`; 0 if `x == 0`; 1 if `x > 0`.
For complex numbers, `y = sign(x) = x / |x|` if `x != 0`, otherwise `y = 0`.
func Sin ¶
func Sin(scope *Scope, x tf.Output) (y tf.Output)
Computes sin of x element-wise.
func Size ¶
func Size(scope *Scope, input tf.Output, optional ...SizeAttr) (output tf.Output)
Returns the size of a tensor.
This operation returns an integer representing the number of elements in `input`.
For example:
“`prettyprint # 't' is [[[1, 1,, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]]] size(t) ==> 12 “`
func Skipgram ¶
func Skipgram(scope *Scope, filename string, batch_size int64, optional ...SkipgramAttr) (vocab_word tf.Output, vocab_freq tf.Output, words_per_epoch tf.Output, current_epoch tf.Output, total_words_processed tf.Output, examples tf.Output, labels tf.Output)
Parses a text file and creates a batch of examples.
DEPRECATED at GraphDef version 19: Moving word2vec into tensorflow_models/tutorials and deprecating its ops here as a result
Arguments:
filename: The corpus's text file name. batch_size: The size of produced batch.
Returns 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 ¶
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 Softmax ¶
func Softmax(scope *Scope, logits tf.Output) (softmax tf.Output)
Computes softmax activations.
For each batch `i` and class `j` we have
softmax[i, j] = exp(logits[i, j]) / sum_j(exp(logits[i, j]))
Arguments:
logits: 2-D with shape `[batch_size, num_classes]`.
Returns Same shape as `logits`.
func SoftmaxCrossEntropyWithLogits ¶
func SoftmaxCrossEntropyWithLogits(scope *Scope, features tf.Output, labels tf.Output) (loss tf.Output, backprop tf.Output)
Computes softmax cross entropy cost and gradients to backpropagate.
Inputs are the logits, not probabilities.
Arguments:
features: batch_size x num_classes matrix labels: batch_size x num_classes matrix
The caller must ensure that each batch of labels represents a valid probability distribution.
Returns Per example loss (batch_size vector).backpropagated gradients (batch_size x num_classes matrix).
func Softplus ¶
func Softplus(scope *Scope, features tf.Output) (activations tf.Output)
Computes softplus: `log(exp(features) + 1)`.
func SoftplusGrad ¶
func SoftplusGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)
Computes softplus gradients for a softplus operation.
Arguments:
gradients: The backpropagated gradients to the corresponding softplus operation. features: The features passed as input to the corresponding softplus operation.
Returns The gradients: `gradients / (1 + exp(-features))`.
func Softsign ¶
func Softsign(scope *Scope, features tf.Output) (activations tf.Output)
Computes softsign: `features / (abs(features) + 1)`.
func SoftsignGrad ¶
func SoftsignGrad(scope *Scope, gradients tf.Output, features tf.Output) (backprops tf.Output)
Computes softsign gradients for a softsign operation.
Arguments:
gradients: The backpropagated gradients to the corresponding softsign operation. features: The features passed as input to the corresponding softsign operation.
Returns The gradients: `gradients / (1 + abs(-features)) ** 2`.
func SpaceToBatch ¶
func SpaceToBatch(scope *Scope, input tf.Output, paddings tf.Output, block_size int64) (output tf.Output)
SpaceToBatch for 4-D tensors of type T.
This is a legacy version of the more general SpaceToBatchND.
Zero-pads and then rearranges (permutes) blocks of spatial data into batch. More specifically, this op outputs a copy of the input tensor where values from the `height` and `width` dimensions are moved to the `batch` dimension. After the zero-padding, both `height` and `width` of the input must be divisible by the block size.
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:
“`prettyprint x = [[[[1], [2]], [[3], [4]]]] “`
The output tensor has shape `[4, 1, 1, 1]` and value:
“`prettyprint [[[[1]]], [[[2]]], [[[3]]], [[[4]]]] “`
(2) For the following input of shape `[1, 2, 2, 3]` and block_size of 2:
“`prettyprint x = [[[[1, 2, 3], [4, 5, 6]],
[[7, 8, 9], [10, 11, 12]]]]
“`
The output tensor has shape `[4, 1, 1, 3]` and value:
“`prettyprint [[[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:
“`prettyprint 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:
“`prettyprint 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:
“`prettyprint 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:
“`prettyprint 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 ¶
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
3. Permute dimensions of `reshaped_padded` to produce
`permuted_reshaped_padded` of shape: block_shape + [batch] + [padded_shape[1] / block_shape[0], ..., padded_shape[M] / block_shape[M-1]] + remaining_shape
4. Reshape `permuted_reshaped_padded` to flatten `block_shape` into the batch
dimension, producing an output tensor of shape: [batch * prod(block_shape)] + [padded_shape[1] / block_shape[0], ..., padded_shape[M] / block_shape[M-1]] + remaining_shape
Some examples:
(1) For the following input of shape `[1, 2, 2, 1]`, `block_shape = [2, 2]`, and
`paddings = [[0, 0], [0, 0]]`:
“`prettyprint x = [[[[1], [2]], [[3], [4]]]] “`
The output tensor has shape `[4, 1, 1, 1]` and value:
“`prettyprint [[[[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]]`:
“`prettyprint x = [[[[1, 2, 3], [4, 5, 6]],
[[7, 8, 9], [10, 11, 12]]]]
“`
The output tensor has shape `[4, 1, 1, 3]` and value:
“`prettyprint [[[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]]`:
“`prettyprint 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:
“`prettyprint 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]]`:
“`prettyprint 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:
“`prettyprint 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 ¶
func SpaceToDepth(scope *Scope, input tf.Output, block_size int64) (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 and how the data is moved.
* Non-overlapping blocks of size `block_size x block size` are rearranged into depth at each location. * The depth of the output tensor is `input_depth * block_size * block_size`. * The input tensor's height and width must be divisible by block_size.
That is, assuming the input is in the shape: `[batch, height, width, depth]`, the shape of the output will be: `[batch, height/block_size, width/block_size, depth*block_size*block_size]`
This operation requires that the input tensor be of rank 4, and that `block_size` be >=1 and a divisor of both the input `height` and `width`.
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 this input of shape `[1, 2, 2, 1]`, and block_size of 2:
“`prettyprint x = [[[[1], [2]],
[[3], [4]]]]
“`
This operation will output a tensor of shape `[1, 1, 1, 4]`:
“`prettyprint [[[[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.
“`prettyprint 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]`
“`prettyprint [[[[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:
“`prettyprint 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]`:
“`prettyprint x = [[[[1, 2, 3, 4],
[5, 6, 7, 8]], [[9, 10, 11, 12], [13, 14, 15, 16]]]]
“`
Arguments:
block_size: The size of the spatial block.
func SparseAdd ¶
func SparseAdd(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output, thresh tf.Output) (sum_indices tf.Output, sum_values tf.Output, sum_shape tf.Output)
Adds two `SparseTensor` objects to produce another `SparseTensor`.
The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, before this step run `SparseReorder` to restore index ordering.
By default, if two values sum to zero at some index, the output `SparseTensor` would still include that particular location in its index, storing a zero in the corresponding value slot. To override this, callers can specify `thresh`, indicating that if the sum has a magnitude strictly smaller than `thresh`, its corresponding value and index would then not be included. In particular, `thresh == 0` (default) means everything is kept and actual thresholding happens only for a positive value.
In the following shapes, `nnz` is the count after taking `thresh` into account.
Arguments:
a_indices: 2-D. The `indices` of the first `SparseTensor`, size `[nnz, ndims]` Matrix. a_values: 1-D. The `values` of the first `SparseTensor`, size `[nnz]` Vector. a_shape: 1-D. The `shape` of the first `SparseTensor`, size `[ndims]` Vector. b_indices: 2-D. The `indices` of the second `SparseTensor`, size `[nnz, ndims]` Matrix. b_values: 1-D. The `values` of the second `SparseTensor`, size `[nnz]` Vector. b_shape: 1-D. The `shape` of the second `SparseTensor`, size `[ndims]` Vector. thresh: 0-D. The magnitude threshold that determines if an output value/index
pair takes space.
func SparseAddGrad ¶
func SparseAddGrad(scope *Scope, backprop_val_grad tf.Output, a_indices tf.Output, b_indices tf.Output, sum_indices tf.Output) (a_val_grad tf.Output, b_val_grad tf.Output)
The gradient operator for the SparseAdd op.
The SparseAdd op calculates A + B, where A, B, and the sum are all represented as `SparseTensor` objects. This op takes in the upstream gradient w.r.t. non-empty values of the sum, and outputs the gradients w.r.t. the non-empty values of A and B.
Arguments:
backprop_val_grad: 1-D with shape `[nnz(sum)]`. The gradient with respect to
the non-empty values of the sum.
a_indices: 2-D. The `indices` of the `SparseTensor` A, size `[nnz(A), ndims]`. b_indices: 2-D. The `indices` of the `SparseTensor` B, size `[nnz(B), ndims]`. sum_indices: 2-D. The `indices` of the sum `SparseTensor`, size
`[nnz(sum), ndims]`.
Returns 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 ¶
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 SparseDenseCwiseAdd ¶
func SparseDenseCwiseAdd(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)
Adds up a SparseTensor and a dense Tensor, using these special rules:
(1) Broadcasts the dense side to have the same shape as the sparse side, if
eligible;
(2) Then, only the dense values pointed to by the indices of the SparseTensor
participate in the cwise addition.
By these rules, the result is a logical SparseTensor with exactly the same indices and shape, but possibly with different non-zero values. The output of this Op is the resultant non-zero values.
Arguments:
sp_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
sp_values: 1-D. `N` non-empty values corresponding to `sp_indices`. sp_shape: 1-D. Shape of the input SparseTensor. dense: `R`-D. The dense Tensor operand.
Returns 1-D. The `N` values that are operated on.
func SparseDenseCwiseDiv ¶
func SparseDenseCwiseDiv(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)
Component-wise divides a SparseTensor by a dense Tensor.
*Limitation*: this Op only broadcasts the dense side to the sparse side, but not the other direction.
Arguments:
sp_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
sp_values: 1-D. `N` non-empty values corresponding to `sp_indices`. sp_shape: 1-D. Shape of the input SparseTensor. dense: `R`-D. The dense Tensor operand.
Returns 1-D. The `N` values that are operated on.
func SparseDenseCwiseMul ¶
func SparseDenseCwiseMul(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output, dense tf.Output) (output tf.Output)
Component-wise multiplies a SparseTensor by a dense Tensor.
The output locations corresponding to the implicitly zero elements in the sparse tensor will be zero (i.e., will not take up storage space), regardless of the contents of the dense tensor (even if it's +/-INF and that INF*0 == NaN).
*Limitation*: this Op only broadcasts the dense side to the sparse side, but not the other direction.
Arguments:
sp_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
sp_values: 1-D. `N` non-empty values corresponding to `sp_indices`. sp_shape: 1-D. Shape of the input SparseTensor. dense: `R`-D. The dense Tensor operand.
Returns 1-D. The `N` values that are operated on.
func SparseMatMul ¶
func SparseMatMul(scope *Scope, a tf.Output, b tf.Output, optional ...SparseMatMulAttr) (product tf.Output)
Multiply matrix "a" by matrix "b".
The inputs must be two-dimensional matrices and the inner dimension of "a" must match the outer dimension of "b". This op is optimized for the case where at least one of "a" or "b" is sparse. The breakeven for using this versus a dense matrix multiply on one platform was 30% zero values in the sparse matrix.
func SparseReduceSum ¶
func SparseReduceSum(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceSumAttr) (output tf.Output)
Computes the sum of elements across dimensions of a SparseTensor.
This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_sum()`. In particular, this Op also returns a dense `Tensor` instead of a sparse one.
Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.
Arguments:
input_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
input_values: 1-D. `N` non-empty values corresponding to `input_indices`. input_shape: 1-D. Shape of the input SparseTensor. reduction_axes: 1-D. Length-`K` vector containing the reduction axes.
Returns `R-K`-D. The reduced Tensor.
func SparseReduceSumSparse ¶
func SparseReduceSumSparse(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output, reduction_axes tf.Output, optional ...SparseReduceSumSparseAttr) (output_indices tf.Output, output_values tf.Output, output_shape tf.Output)
Computes the sum of elements across dimensions of a SparseTensor.
This Op takes a SparseTensor and is the sparse counterpart to `tf.reduce_sum()`. In contrast to SparseReduceSum, this Op returns a SparseTensor.
Reduces `sp_input` along the dimensions given in `reduction_axes`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_axes`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
If `reduction_axes` has no entries, all dimensions are reduced, and a tensor with a single element is returned. Additionally, the axes can be negative, which are interpreted according to the indexing rules in Python.
Arguments:
input_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
input_values: 1-D. `N` non-empty values corresponding to `input_indices`. input_shape: 1-D. Shape of the input SparseTensor. reduction_axes: 1-D. Length-`K` vector containing the reduction axes.
func SparseReorder ¶
func SparseReorder(scope *Scope, input_indices tf.Output, input_values tf.Output, input_shape tf.Output) (output_indices tf.Output, output_values tf.Output)
Reorders a SparseTensor into the canonical, row-major ordering.
Note that by convention, all sparse ops preserve the canonical ordering along increasing dimension number. The only time ordering can be violated is during manual manipulation of the indices and values vectors to add entries.
Reordering does not affect the shape of the SparseTensor.
If the tensor has rank `R` and `N` non-empty values, `input_indices` has shape `[N, R]`, input_values has length `N`, and input_shape has length `R`.
Arguments:
input_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, possibly not in canonical ordering.
input_values: 1-D. `N` non-empty values corresponding to `input_indices`. input_shape: 1-D. Shape of the input SparseTensor.
Returns 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 ¶
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 ¶
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.
Read [the section on Segmentation](../../api_docs/python/math_ops.md#segmentation) for an explanation of segments.
Like `SegmentMean`, but `segment_ids` can have rank less than `data`'s first dimension, selecting a subset of dimension 0, specified by `indices`.
Arguments:
indices: A 1-D tensor. Has same rank as `segment_ids`. segment_ids: A 1-D tensor. Values should be sorted and can be repeated.
Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.
func SparseSegmentMeanGrad ¶
func SparseSegmentMeanGrad(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, output_dim0 tf.Output) (output tf.Output)
Computes gradients for SparseSegmentMean.
Returns tensor "output" with same shape as grad, except for dimension 0 whose value is output_dim0.
Arguments:
grad: gradient propagated to the SparseSegmentMean op. indices: indices passed to the corresponding SparseSegmentMean op. segment_ids: segment_ids passed to the corresponding SparseSegmentMean op. output_dim0: dimension 0 of "data" passed to SparseSegmentMean op.
func SparseSegmentSqrtN ¶
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.
Read [the section on Segmentation](../../api_docs/python/math_ops.md#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.
Returns Has same shape as data, except for dimension 0 which has size `k`, the number of segments.
func SparseSegmentSqrtNGrad ¶
func SparseSegmentSqrtNGrad(scope *Scope, grad tf.Output, indices tf.Output, segment_ids tf.Output, output_dim0 tf.Output) (output tf.Output)
Computes gradients for SparseSegmentSqrtN.
Returns tensor "output" with same shape as grad, except for dimension 0 whose value is output_dim0.
Arguments:
grad: gradient propagated to the SparseSegmentSqrtN op. indices: indices passed to the corresponding SparseSegmentSqrtN op. segment_ids: segment_ids passed to the corresponding SparseSegmentSqrtN op. output_dim0: dimension 0 of "data" passed to SparseSegmentSqrtN op.
func SparseSegmentSum ¶
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](../../api_docs/python/math_ops.md#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:
“`prettyprint 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 SparseSoftmax ¶
func SparseSoftmax(scope *Scope, sp_indices tf.Output, sp_values tf.Output, sp_shape tf.Output) (output tf.Output)
Applies softmax to a batched N-D `SparseTensor`.
The inputs represent an N-D SparseTensor with logical shape `[..., B, C]` (where `N >= 2`), and with indices sorted in the canonical lexicographic order.
This op is equivalent to applying the normal `tf.nn.softmax()` to each innermost logical submatrix with shape `[B, C]`, but with the catch that *the implicitly zero elements do not participate*. Specifically, the algorithm is equivalent to the following:
(1) Applies `tf.nn.softmax()` to a densified view of each innermost submatrix
with shape `[B, C]`, along the size-C dimension;
(2) Masks out the original implicitly-zero locations;
(3) Renormalizes the remaining elements.
Hence, the `SparseTensor` result has exactly the same non-zero indices and shape.
Arguments:
sp_indices: 2-D. `NNZ x R` matrix with the indices of non-empty values in a
SparseTensor, in canonical ordering.
sp_values: 1-D. `NNZ` non-empty values corresponding to `sp_indices`. sp_shape: 1-D. Shape of the input SparseTensor.
Returns 1-D. The `NNZ` values for the result `SparseTensor`.
func SparseSoftmaxCrossEntropyWithLogits ¶
func SparseSoftmaxCrossEntropyWithLogits(scope *Scope, features tf.Output, labels tf.Output) (loss tf.Output, backprop tf.Output)
Computes softmax cross entropy cost and gradients to backpropagate.
Unlike `SoftmaxCrossEntropyWithLogits`, this operation does not accept a matrix of label probabilities, but rather a single label per row of features. This label is considered to have probability 1.0 for the given row.
Inputs are the logits, not probabilities.
Arguments:
features: batch_size x num_classes matrix labels: batch_size vector with values in [0, num_classes).
This is the label for the given minibatch entry.
Returns Per example loss (batch_size vector).backpropagated gradients (batch_size x num_classes matrix).
func SparseSparseMaximum ¶
func SparseSparseMaximum(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output) (output_indices tf.Output, output_values tf.Output)
Returns the element-wise max of two SparseTensors.
Assumes the two SparseTensors have the same shape, i.e., no broadcasting.
Arguments:
a_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, in the canonical lexicographic ordering.
a_values: 1-D. `N` non-empty values corresponding to `a_indices`. a_shape: 1-D. Shape of the input SparseTensor. b_indices: counterpart to `a_indices` for the other operand. b_values: counterpart to `a_values` for the other operand; must be of the same dtype. b_shape: counterpart to `a_shape` for the other operand; the two shapes must be equal.
Returns 2-D. The indices of the output SparseTensor.1-D. The values of the output SparseTensor.
func SparseSparseMinimum ¶
func SparseSparseMinimum(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b_indices tf.Output, b_values tf.Output, b_shape tf.Output) (output_indices tf.Output, output_values tf.Output)
Returns the element-wise min of two SparseTensors.
Assumes the two SparseTensors have the same shape, i.e., no broadcasting.
Arguments:
a_indices: 2-D. `N x R` matrix with the indices of non-empty values in a
SparseTensor, in the canonical lexicographic ordering.
a_values: 1-D. `N` non-empty values corresponding to `a_indices`. a_shape: 1-D. Shape of the input SparseTensor. b_indices: counterpart to `a_indices` for the other operand. b_values: counterpart to `a_values` for the other operand; must be of the same dtype. b_shape: counterpart to `a_shape` for the other operand; the two shapes must be equal.
Returns 2-D. The indices of the output SparseTensor.1-D. The values of the output SparseTensor.
func SparseSplit ¶
func SparseSplit(scope *Scope, split_dim tf.Output, indices tf.Output, values tf.Output, shape tf.Output, num_split int64) (output_indices []tf.Output, output_values []tf.Output, output_shape []tf.Output)
Split a `SparseTensor` into `num_split` tensors along one dimension.
If the `shape[split_dim]` is not an integer multiple of `num_split`. Slices `[0 : shape[split_dim] % num_split]` gets one extra dimension. For example, if `split_dim = 1` and `num_split = 2` and the input is
input_tensor = shape = [2, 7] [ a d e ] [b c ]
Graphically the output tensors are:
output_tensor[0] = shape = [2, 4] [ a ] [b c ] output_tensor[1] = shape = [2, 3] [ d e ] [ ]
Arguments:
split_dim: 0-D. The dimension along which to split. Must be in the range
`[0, rank(shape))`.
indices: 2-D tensor represents the indices of the sparse tensor. values: 1-D tensor represents the values of the sparse tensor. shape: 1-D. tensor represents the shape of the sparse tensor.
output indices: A list of 1-D tensors represents the indices of the output sparse tensors.
num_split: The number of ways to split.
Returns 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 ¶
func SparseTensorDenseAdd(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b tf.Output) (output tf.Output)
Adds up a `SparseTensor` and a dense `Tensor`, producing a dense `Tensor`.
This Op does not require `a_indices` be sorted in standard lexicographic order.
Arguments:
a_indices: 2-D. The `indices` of the `SparseTensor`, with shape `[nnz, ndims]`. a_values: 1-D. The `values` of the `SparseTensor`, with shape `[nnz]`. a_shape: 1-D. The `shape` of the `SparseTensor`, with shape `[ndims]`. b: `ndims`-D Tensor. With shape `a_shape`.
func SparseTensorDenseMatMul ¶
func SparseTensorDenseMatMul(scope *Scope, a_indices tf.Output, a_values tf.Output, a_shape tf.Output, b tf.Output, optional ...SparseTensorDenseMatMulAttr) (product tf.Output)
Multiply SparseTensor (of rank 2) "A" by dense matrix "B".
No validity checking is performed on the indices of A. However, the following input format is recommended for optimal behavior:
if adjoint_a == false:
A should be sorted in lexicographically increasing order. Use SparseReorder if you're not sure.
if adjoint_a == true:
A should be sorted in order of increasing dimension 1 (i.e., "column major" order instead of "row major" order).
Arguments:
a_indices: 2-D. The `indices` of the `SparseTensor`, size `[nnz, 2]` Matrix. a_values: 1-D. The `values` of the `SparseTensor`, size `[nnz]` Vector. a_shape: 1-D. The `shape` of the `SparseTensor`, size `[2]` Vector. b: 2-D. A dense Matrix.
func SparseToDense ¶
func SparseToDense(scope *Scope, sparse_indices tf.Output, output_shape tf.Output, sparse_values tf.Output, default_value tf.Output, optional ...SparseToDenseAttr) (dense tf.Output)
Converts a sparse representation into a dense tensor.
Builds an array `dense` with shape `output_shape` such that
“`prettyprint # If sparse_indices is scalar dense[i] = (i == sparse_indices ? sparse_values : default_value)
# If sparse_indices is a vector, then for each i dense[sparse_indices[i]] = sparse_values[i]
# If sparse_indices is an n by d matrix, then for each i in [0, n) dense[sparse_indices[i][0], ..., sparse_indices[i][d-1]] = sparse_values[i] “`
All other values in `dense` are set to `default_value`. If `sparse_values` is a scalar, all sparse indices are set to this single value.
Indices should be sorted in lexicographic order, and indices must not contain any repeats. If `validate_indices` is true, these properties are checked during execution.
Arguments:
sparse_indices: 0-D, 1-D, or 2-D. `sparse_indices[i]` contains the complete
index where `sparse_values[i]` will be placed.
output_shape: 1-D. Shape of the dense output tensor. sparse_values: 1-D. Values corresponding to each row of `sparse_indices`,
or a scalar value to be used for all sparse indices.
default_value: Scalar value to set for indices not specified in
`sparse_indices`.
Returns Dense output tensor of shape `output_shape`.
func SparseToSparseSetOperation ¶
func SparseToSparseSetOperation(scope *Scope, set1_indices tf.Output, set1_values tf.Output, set1_shape tf.Output, set2_indices tf.Output, set2_values tf.Output, set2_shape tf.Output, set_operation string, optional ...SparseToSparseSetOperationAttr) (result_indices tf.Output, result_values tf.Output, result_shape tf.Output)
Applies set operation along last dimension of 2 `SparseTensor` inputs.
See SetOperationOp::SetOperationFromContext for values of `set_operation`.
If `validate_indices` is `True`, `SparseToSparseSetOperation` validates the order and range of `set1` and `set2` indices.
Input `set1` is a `SparseTensor` represented by `set1_indices`, `set1_values`, and `set1_shape`. For `set1` ranked `n`, 1st `n-1` dimensions must be the same as `set2`. Dimension `n` contains values in a set, duplicates are allowed but ignored.
Input `set2` is a `SparseTensor` represented by `set2_indices`, `set2_values`, and `set2_shape`. For `set2` ranked `n`, 1st `n-1` dimensions must be the same as `set1`. Dimension `n` contains values in a set, duplicates are allowed but ignored.
If `validate_indices` is `True`, this op validates the order and range of `set1` and `set2` indices.
Output `result` is a `SparseTensor` represented by `result_indices`, `result_values`, and `result_shape`. For `set1` and `set2` ranked `n`, this has rank `n` and the same 1st `n-1` dimensions as `set1` and `set2`. The `nth` dimension contains the result of `set_operation` applied to the corresponding `[0...n-1]` dimension of `set`.
Arguments:
set1_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major
order.
set1_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major
order.
set1_shape: 1D `Tensor`, shape of a `SparseTensor`. `set1_shape[0...n-1]` must
be the same as `set2_shape[0...n-1]`, `set1_shape[n]` is the max set size across `0...n-1` dimensions.
set2_indices: 2D `Tensor`, indices of a `SparseTensor`. Must be in row-major
order.
set2_values: 1D `Tensor`, values of a `SparseTensor`. Must be in row-major
order.
set2_shape: 1D `Tensor`, shape of a `SparseTensor`. `set2_shape[0...n-1]` must
be the same as `set1_shape[0...n-1]`, `set2_shape[n]` is the max set size across `0...n-1` dimensions.
Returns 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 ¶
func Split(scope *Scope, split_dim tf.Output, value tf.Output, num_split int64) (output []tf.Output)
Splits a tensor into `num_split` tensors along one dimension.
Arguments:
split_dim: 0-D. The dimension along which to split. Must be in the range
`[0, 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 `split_dim`, where their sizes are `values.shape[split_dim] / num_split`.
func SplitV ¶
func SplitV(scope *Scope, value tf.Output, size_splits tf.Output, split_dim 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.
split_dim: 0-D. The dimension along which to split. Must be in the range
`[0, rank(value))`.
Returns Tensors whose shape matches that of `value` except along `split_dim`, where their sizes are `size_splits[i]`.
func Sqrt ¶
func Sqrt(scope *Scope, x tf.Output) (y tf.Output)
Computes square root of x element-wise.
I.e., \\(y = \sqrt{x} = x^{1/2}\\).
func SqrtGrad ¶
func SqrtGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient for the sqrt of `x` wrt its input.
Specifically, `grad = dy * 0.5 / y`, where `y = sqrt(x)`, and `dy` is the corresponding input gradient.
func Square ¶
func Square(scope *Scope, x tf.Output) (y tf.Output)
Computes square of x element-wise.
I.e., \\(y = x * x = x^2\\).
func SquaredDifference ¶
func SquaredDifference(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns (x - y)(x - y) element-wise.
*NOTE*: `SquaredDifference` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Squeeze ¶
func Squeeze(scope *Scope, input tf.Output, optional ...SqueezeAttr) (output tf.Output)
Removes dimensions of size 1 from the shape of a tensor.
Given a tensor `input`, this operation returns a tensor of the same type with all dimensions of size 1 removed. If you don't want to remove all size 1 dimensions, you can remove specific size 1 dimensions by specifying `squeeze_dims`.
For example:
“`prettyprint # 't' is a tensor of shape [1, 2, 1, 3, 1, 1] shape(squeeze(t)) ==> [2, 3] “`
Or, to remove specific size 1 dimensions:
“`prettyprint # '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 Stage ¶
func Stage(scope *Scope, values []tf.Output, optional ...StageAttr) (o *tf.Operation)
Stage values similar to a lightweight Enqueue. The basic functionality of this
Op is similar to a queue with many fewer capabilities and options. This Op is optimized for performance.
Arguments:
values: a list of tensors
Returns the created operation.
func StopGradient ¶
func StopGradient(scope *Scope, input tf.Output) (output tf.Output)
Stops gradient computation.
When executed in a graph, this op outputs its input tensor as-is.
When building ops to compute gradients, this op prevents the contribution of its inputs to be taken into account. Normally, the gradient generator adds ops to a graph to compute the derivatives of a specified 'loss' by recursively finding out inputs that contributed to its computation. If you insert this op in the graph it inputs are masked from the gradient generator. They are not taken into account for computing gradients.
This is useful any time you want to compute a value with TensorFlow but need to pretend that the value was a constant. 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 ¶
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
“`prettyprint 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 ¶
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 StringJoin ¶
func StringJoin(scope *Scope, inputs []tf.Output, optional ...StringJoinAttr) (output tf.Output)
Joins the strings in the given list of string tensors into one tensor;
with the given separator (default is an empty separator).
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 StringSplit ¶
func StringSplit(scope *Scope, input tf.Output, delimiter tf.Output) (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 StringToHashBucket ¶
func StringToHashBucket(scope *Scope, string_tensor tf.Output, num_buckets int64) (output tf.Output)
Converts each string in the input Tensor to its hash mod by a number of buckets.
The hash function is deterministic on the content of the string within the process.
Note that the hash function may change from time to time. This functionality will be deprecated and it's recommended to use `tf.string_to_hash_bucket_fast()` or `tf.string_to_hash_bucket_strong()`.
Arguments:
num_buckets: The number of buckets.
Returns A Tensor of the same shape as the input `string_tensor`.
func StringToHashBucketFast ¶
func StringToHashBucketFast(scope *Scope, input tf.Output, num_buckets int64) (output tf.Output)
Converts each string in the input Tensor to its hash mod by a number of buckets.
The hash function is deterministic on the content of the string within the process and will never change. However, it is not suitable for cryptography. This function may be used when CPU time is scarce and inputs are trusted or unimportant. There is a risk of adversaries constructing inputs that all hash to the same bucket. To prevent this problem, use a strong hash function with `tf.string_to_hash_bucket_strong`.
Arguments:
input: The strings to assign a hash bucket. num_buckets: The number of buckets.
Returns A Tensor of the same shape as the input `string_tensor`.
func StringToHashBucketStrong ¶
func StringToHashBucketStrong(scope *Scope, input tf.Output, num_buckets int64, key []int64) (output tf.Output)
Converts each string in the input Tensor to its hash mod by a number of buckets.
The hash function is deterministic on the content of the string within the process. The hash function is a keyed hash function, where attribute `key` defines the key of the hash function. `key` is an array of 2 elements.
A strong hash is important when inputs may be malicious, e.g. URLs with additional components. Adversaries could try to make their inputs hash to the same bucket for a denial-of-service attack or to skew the results. A strong hash prevents this by making it difficult, if not infeasible, to compute inputs that hash to the same bucket. This 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 for the keyed hash function passed as a list of two uint64
elements.
Returns A Tensor of the same shape as the input `string_tensor`.
func StringToNumber ¶
func StringToNumber(scope *Scope, string_tensor tf.Output, optional ...StringToNumberAttr) (output tf.Output)
Converts each string in the input Tensor to the specified numeric type.
(Note that int32 overflow results in an error while float overflow results in a rounded value.)
Returns A Tensor of the same shape as the input `string_tensor`.
func Sub ¶
func Sub(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns x - y element-wise.
*NOTE*: `Sub` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func Substr ¶
func Substr(scope *Scope, input tf.Output, pos tf.Output, len tf.Output) (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.
If `pos` is negative or specifies a character index larger than 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`:
“` input = [b'Hello', b'World'] position = 1 length = 3
output = [b'ell', b'orl'] “`
Using `pos` and `len` with same shape as `input`:
“` 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 ¶
func Sum(scope *Scope, input tf.Output, reduction_indices tf.Output, optional ...SumAttr) (output tf.Output)
Computes the sum of elements across dimensions of a tensor.
Reduces `input` along the dimensions given in `reduction_indices`. Unless `keep_dims` is true, the rank of the tensor is reduced by 1 for each entry in `reduction_indices`. If `keep_dims` is true, the reduced dimensions are retained with length 1.
Arguments:
input: The tensor to reduce. reduction_indices: The dimensions to reduce.
Returns The reduced tensor.
func Svd ¶
func Svd(scope *Scope, input tf.Output, optional ...SvdAttr) (s tf.Output, u tf.Output, v tf.Output)
Computes the singular value decompositions of one or more matrices.
Computes the SVD of each inner matrix in `input` such that `input[..., :, :] = u[..., :, :] * diag(s[..., :, :]) * transpose(v[..., :, :])`
“`prettyprint # 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 ¶
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 TFRecordReaderV2 ¶
func TFRecordReaderV2(scope *Scope, optional ...TFRecordReaderV2Attr) (reader_handle tf.Output)
A Reader that outputs the records from a TensorFlow Records file.
Returns The handle to reference the Reader.
func TakeManySparseFromTensorsMap ¶
func TakeManySparseFromTensorsMap(scope *Scope, sparse_handles tf.Output, dtype tf.DataType, optional ...TakeManySparseFromTensorsMapAttr) (sparse_indices tf.Output, sparse_values tf.Output, sparse_shape tf.Output)
Read `SparseTensors` from a `SparseTensorsMap` and concatenate them.
The input `sparse_handles` must be an `int64` matrix of shape `[N, 1]` where `N` is the minibatch size and the rows correspond to the output handles of `AddSparseToTensorsMap` or `AddManySparseToTensorsMap`. The ranks of the original `SparseTensor` objects that went into the given input ops must all match. When the final `SparseTensor` is created, it has rank one higher than the ranks of the incoming `SparseTensor` objects (they have been concatenated along a new row dimension on the left).
The output `SparseTensor` object's shape values for all dimensions but the first are the max across the input `SparseTensor` objects' shape values for the corresponding dimensions. Its first shape value is `N`, the minibatch size.
The input `SparseTensor` objects' indices are assumed ordered in standard lexicographic order. If this is not the case, after this step run `SparseReorder` to restore index ordering.
For example, if the handles represent an input, which is a `[2, 3]` matrix representing two original `SparseTensor` objects:
“`
index = [ 0]
[10]
[20]
values = [1, 2, 3]
shape = [50]
“`
and
“`
index = [ 2]
[10]
values = [4, 5]
shape = [30]
“`
then the final `SparseTensor` will be:
“`
index = [0 0]
[0 10]
[0 20]
[1 2]
[1 10]
values = [1, 2, 3, 4, 5]
shape = [2 50]
“`
Arguments:
sparse_handles: 1-D, The `N` serialized `SparseTensor` objects.
Shape: `[N]`.
dtype: The `dtype` of the `SparseTensor` objects stored in the
`SparseTensorsMap`.
Returns 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 ¶
func Tan(scope *Scope, x tf.Output) (y tf.Output)
Computes tan of x element-wise.
func Tanh ¶
func Tanh(scope *Scope, x tf.Output) (y tf.Output)
Computes hyperbolic tangent of `x` element-wise.
func TanhGrad ¶
func TanhGrad(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Computes the gradient for the tanh of `x` wrt its input.
Specifically, `grad = dy * (1 - y*y)`, where `y = tanh(x)`, and `dy` is the corresponding input gradient.
func TensorArrayCloseV2 ¶
func TensorArrayCloseV2(scope *Scope, handle tf.Output) (o *tf.Operation)
Deprecated. Use TensorArrayCloseV3
Returns the created operation.
func TensorArrayCloseV3 ¶
func TensorArrayCloseV3(scope *Scope, handle tf.Output) (o *tf.Operation)
Delete the TensorArray from its resource container. This enables
the user to close and release the resource in the middle of a step/run.
Arguments:
handle: The handle to a TensorArray (output of TensorArray or TensorArrayGrad).
Returns the created operation.
func TensorArrayConcatV2 ¶
func TensorArrayConcatV2(scope *Scope, handle tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayConcatV2Attr) (value tf.Output, lengths tf.Output)
Deprecated. Use TensorArrayConcatV3
func TensorArrayConcatV3 ¶
func TensorArrayConcatV3(scope *Scope, handle tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayConcatV3Attr) (value tf.Output, lengths tf.Output)
Concat the elements from the TensorArray into value `value`.
Takes `T` elements of shapes
``` (n0 x d0 x d1 x ...), (n1 x d0 x d1 x ...), ..., (n(T-1) x d0 x d1 x ...) ```
and concatenates them into a Tensor of shape:
```(n0 + n1 + ... + n(T-1) x d0 x d1 x ...)```
All elements must have the same shape (excepting the first dimension).
Arguments:
handle: The handle to a TensorArray. flow_in: A float scalar that enforces proper chaining of operations. dtype: The type of the elem that is returned.
Returns 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 ¶
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
func TensorArrayGatherV3 ¶
func TensorArrayGatherV3(scope *Scope, handle tf.Output, indices tf.Output, flow_in tf.Output, dtype tf.DataType, optional ...TensorArrayGatherV3Attr) (value tf.Output)
Gather specific elements from the TensorArray into output `value`.
All elements selected by `indices` must have the same shape.
Arguments:
handle: The handle to a TensorArray. indices: The locations in the TensorArray from which to read tensor elements. flow_in: A float scalar that enforces proper chaining of operations. dtype: The type of the elem that is returned.
Returns All of the elements in the TensorArray, concatenated along a new axis (the new dimension 0).
func TensorArrayGradV2 ¶
func TensorArrayGradV2(scope *Scope, handle tf.Output, flow_in tf.Output, source string) (grad_handle tf.Output)
Deprecated. Use TensorArrayGradV3
func TensorArrayGradV3 ¶
func TensorArrayGradV3(scope *Scope, handle tf.Output, flow_in tf.Output, source string) (grad_handle tf.Output, flow_out tf.Output)
Creates a TensorArray for storing the gradients of values in the given handle.
If the given TensorArray gradient already exists, returns a reference to it.
Locks the size of the original TensorArray by disabling its dynamic size flag.
**A note about the input flow_in:**
The handle flow_in forces the execution of the gradient lookup to occur only after certain other operations have occurred. For example, when the forward TensorArray is dynamically sized, writes to this TensorArray may resize the object. The gradient TensorArray is statically sized based on the size of the forward TensorArray when this operation executes. Furthermore, the size of the forward TensorArray is frozen by this call. As a result, the flow is used to ensure that the call to generate the gradient TensorArray only happens after all writes are executed.
In the case of dynamically sized TensorArrays, gradient computation should only be performed on read operations that have themselves been chained via flow to occur only after all writes have executed. That way the final size of the forward TensorArray is known when this operation is called.
**A note about the source attribute:**
TensorArray gradient calls use an accumulator TensorArray object. If multiple gradients are calculated and run in the same session, the multiple gradient nodes may accidentally flow throuth 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 TensorArrayReadV2 ¶
func TensorArrayReadV2(scope *Scope, handle tf.Output, index tf.Output, flow_in tf.Output, dtype tf.DataType) (value tf.Output)
Deprecated. Use TensorArrayReadV3
func TensorArrayReadV3 ¶
func TensorArrayReadV3(scope *Scope, handle tf.Output, index tf.Output, flow_in tf.Output, dtype tf.DataType) (value tf.Output)
Read an element from the TensorArray into output `value`.
Arguments:
handle: The handle to a TensorArray. flow_in: A float scalar that enforces proper chaining of operations. dtype: The type of the elem that is returned.
Returns The tensor that is read from the TensorArray.
func TensorArrayScatterV2 ¶
func TensorArrayScatterV2(scope *Scope, handle tf.Output, indices tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)
Deprecated. Use TensorArrayScatterV3
func TensorArrayScatterV3 ¶
func TensorArrayScatterV3(scope *Scope, handle tf.Output, indices tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)
Scatter the data from the input value into specific TensorArray elements.
`indices` must be a vector, its length must match the first dim of `value`.
Arguments:
handle: The handle to a TensorArray. indices: The locations at which to write the tensor elements. value: The concatenated tensor to write to the TensorArray. flow_in: A float scalar that enforces proper chaining of operations.
Returns A float scalar that enforces proper chaining of operations.
func TensorArraySizeV2 ¶
func TensorArraySizeV2(scope *Scope, handle tf.Output, flow_in tf.Output) (size tf.Output)
Deprecated. Use TensorArraySizeV3
func TensorArraySizeV3 ¶
func TensorArraySizeV3(scope *Scope, handle tf.Output, flow_in tf.Output) (size tf.Output)
Get the current size of the TensorArray.
Arguments:
handle: The handle to a TensorArray (output of TensorArray or TensorArrayGrad). flow_in: A float scalar that enforces proper chaining of operations.
Returns The current size of the TensorArray.
func TensorArraySplitV2 ¶
func TensorArraySplitV2(scope *Scope, handle tf.Output, value tf.Output, lengths tf.Output, flow_in tf.Output) (flow_out tf.Output)
Deprecated. Use TensorArraySplitV3
func TensorArraySplitV3 ¶
func TensorArraySplitV3(scope *Scope, handle tf.Output, value tf.Output, lengths tf.Output, flow_in tf.Output) (flow_out tf.Output)
Split the data from the input value into TensorArray elements.
Assuming that `lengths` takes on values
```(n0, n1, ..., n(T-1))```
and that `value` has shape
```(n0 + n1 + ... + n(T-1) x d0 x d1 x ...)```,
this splits values into a TensorArray with T tensors.
TensorArray index t will be the subtensor of values with starting position
```(n0 + n1 + ... + n(t-1), 0, 0, ...)```
and having size
```nt x d0 x d1 x ...```
Arguments:
handle: The handle to a TensorArray. value: The concatenated tensor to write to the TensorArray. lengths: The vector of lengths, how to split the rows of value into the
TensorArray.
flow_in: A float scalar that enforces proper chaining of operations.
Returns A float scalar that enforces proper chaining of operations.
func TensorArrayV2 ¶
func TensorArrayV2(scope *Scope, size tf.Output, dtype tf.DataType, optional ...TensorArrayV2Attr) (handle tf.Output)
Deprecated. Use TensorArrayV3
func TensorArrayV3 ¶
func TensorArrayV3(scope *Scope, size tf.Output, dtype tf.DataType, optional ...TensorArrayV3Attr) (handle tf.Output, flow tf.Output)
An array of Tensors of given size, with data written 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 ¶
func TensorArrayWriteV2(scope *Scope, handle tf.Output, index tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)
Deprecated. Use TensorArrayGradV3
func TensorArrayWriteV3 ¶
func TensorArrayWriteV3(scope *Scope, handle tf.Output, index tf.Output, value tf.Output, flow_in tf.Output) (flow_out tf.Output)
Push an element onto the tensor_array.
Arguments:
handle: The handle to a TensorArray. index: The position to write to inside the TensorArray. value: The tensor to write to the TensorArray. flow_in: A float scalar that enforces proper chaining of operations.
Returns A float scalar that enforces proper chaining of operations.
func TensorSummary ¶
func TensorSummary(scope *Scope, tensor tf.Output, optional ...TensorSummaryAttr) (summary tf.Output)
Outputs a `Summary` protocol buffer with a tensor.
Arguments:
tensor: A tensor to serialize.
func TextLineReaderV2 ¶
func TextLineReaderV2(scope *Scope, optional ...TextLineReaderV2Attr) (reader_handle tf.Output)
A Reader that outputs the lines of a file delimited by '\n'.
Returns The handle to reference the Reader.
func ThreadUnsafeUnigramCandidateSampler ¶
func ThreadUnsafeUnigramCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...ThreadUnsafeUnigramCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a learned unigram distribution.
See explanations of candidate sampling and the data formats at go/candidate-sampling.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to randomly sample per batch. 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 ¶
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 ¶
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 TopK ¶
func TopK(scope *Scope, input tf.Output, k int64, optional ...TopKAttr) (values tf.Output, indices tf.Output)
Finds values and indices of the `k` largest elements for the last dimension.
DEPRECATED at GraphDef version 7: Use TopKV2 instead
If the input is a vector (rank-1), finds the `k` largest entries in the vector and outputs their values and indices as vectors. Thus `values[j]` is the `j`-th largest entry in `input`, and its index is `indices[j]`.
For matrices (resp. higher rank input), computes the top `k` entries in each row (resp. vector along the last dimension). Thus,
values.shape = indices.shape = input.shape[:-1] + [k]
If two elements are equal, the lower-index element appears first.
If `k` varies dynamically, use `TopKV2` below.
Arguments:
input: 1-D or higher with last dimension at least `k`. k: Number of top elements to look for along the last dimension (along each
row for matrices).
Returns The `k` largest elements along each last dimensional slice.The indices of `values` within the last dimension of `input`.
func TopKV2 ¶
func TopKV2(scope *Scope, input tf.Output, k tf.Output, optional ...TopKV2Attr) (values tf.Output, indices tf.Output)
Finds values and indices of the `k` largest elements for the last dimension.
If the input is a vector (rank-1), finds the `k` largest entries in the vector and outputs their values and indices as vectors. Thus `values[j]` is the `j`-th largest entry in `input`, and its index is `indices[j]`.
For matrices (resp. higher rank input), computes the top `k` entries in each row (resp. vector along the last dimension). Thus,
values.shape = indices.shape = input.shape[:-1] + [k]
If two elements are equal, the lower-index element appears first.
Arguments:
input: 1-D or higher with last dimension at least `k`. k: 0-D. Number of top elements to look for along the last dimension (along each
row for matrices).
Returns The `k` largest elements along each last dimensional slice.The indices of `values` within the last dimension of `input`.
func Transpose ¶
func Transpose(scope *Scope, x tf.Output, perm tf.Output) (y tf.Output)
Shuffle dimensions of x according to a permutation.
The output `y` has the same rank as `x`. The shapes of `x` and `y` satisfy:
`y.shape[i] == x.shape[perm[i]] for i in [0, 1, ..., rank(x) - 1]`
func TruncateDiv ¶
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 ¶
func TruncateMod(scope *Scope, x tf.Output, y tf.Output) (z tf.Output)
Returns element-wise remainder of division. This emulates C semantics where
true, this follows C semantics in that the result here is consistent with a flooring divide. E.g. `floor(x / y) * y + mod(x, y) = x`.
*NOTE*: `Mod` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html)
func TruncatedNormal ¶
func TruncatedNormal(scope *Scope, shape tf.Output, dtype tf.DataType, optional ...TruncatedNormalAttr) (output tf.Output)
Outputs random values from a truncated normal distribution.
The generated values follow a normal distribution with mean 0 and standard deviation 1, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.
Arguments:
shape: The shape of the output tensor. dtype: The type of the output.
Returns A tensor of the specified shape filled with random truncated normal values.
func UniformCandidateSampler ¶
func UniformCandidateSampler(scope *Scope, true_classes tf.Output, num_true int64, num_sampled int64, unique bool, range_max int64, optional ...UniformCandidateSamplerAttr) (sampled_candidates tf.Output, true_expected_count tf.Output, sampled_expected_count tf.Output)
Generates labels for candidate sampling with a uniform distribution.
See explanations of candidate sampling and the data formats at go/candidate-sampling.
For each batch, this op picks a single set of sampled candidate labels.
The advantages of sampling candidates per-batch are simplicity and the possibility of efficient dense matrix multiplication. The disadvantage is that the sampled candidates must be chosen independently of the context and of the true labels.
Arguments:
true_classes: A batch_size * num_true matrix, in which each row contains the
IDs of the num_true target_classes in the corresponding original label.
num_true: Number of true labels per context. num_sampled: Number of candidates to randomly sample per batch. 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 ¶
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:
“`prettyprint # 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 UniqueWithCounts ¶
func UniqueWithCounts(scope *Scope, x tf.Output, optional ...UniqueWithCountsAttr) (y tf.Output, idx tf.Output, count tf.Output)
Finds unique elements in a 1-D tensor.
This operation returns a tensor `y` containing all of the unique elements of `x` sorted in the same order that they occur in `x`. This operation also returns a tensor `idx` the same size as `x` that contains the index of each value of `x` in the unique output `y`. Finally, it returns a third tensor `count` that contains the count of each element of `y` in `x`. In other words:
`y[idx[i]] = x[i] for i in [0, 1,...,rank(x) - 1]`
For example:
“`prettyprint # 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 Unpack ¶
func Unpack(scope *Scope, value tf.Output, num int64, optional ...UnpackAttr) (output []tf.Output)
Unpacks a given dimension of a rank-`R` tensor into `num` rank-`(R-1)` tensors.
Unpacks `num` tensors from `value` by chipping it along the `axis` dimension. For example, given a tensor of shape `(A, B, C, D)`;
If `axis == 0` then the i'th tensor in `output` is the slice `value[i, :, :, :]`
and each tensor in `output` will have shape `(B, C, D)`. (Note that the dimension unpacked along is gone, unlike `split`).
If `axis == 1` then the i'th tensor in `output` is the slice `value[:, i, :, :]`
and each tensor in `output` will have shape `(A, C, D)`.
Etc.
This is the opposite of `pack`.
Arguments:
value: 1-D or higher, with `axis` dimension size equal to `num`.
Returns The list of tensors unpacked from `value`.
func UnsortedSegmentMax ¶
func UnsortedSegmentMax(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)
Computes the Max along segments of a tensor.
Read [the section on Segmentation](../../api_docs/python/math_ops.md#segmentation) for an explanation of segments.
This operator is similar to the [unsorted segment sum operator](../../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 `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 specific numeric type,
`output[i] = numeric_limits<T>::min()`.
<div style="width:70%; margin:auto; margin-bottom:10px; margin-top:20px;"> <img style="width:100%" src="../../images/UnsortedSegmentSum.png" alt> </div>
Arguments:
segment_ids: A 1-D tensor whose rank is equal to the rank of `data`'s
first dimension.
Returns Has same shape as data, except for dimension 0 which has size `num_segments`.
func UnsortedSegmentSum ¶
func UnsortedSegmentSum(scope *Scope, data tf.Output, segment_ids tf.Output, num_segments tf.Output) (output tf.Output)
Computes the sum along segments of a tensor.
Read [the section on Segmentation](../../api_docs/python/math_ops.md#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`.
`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="../../images/UnsortedSegmentSum.png" alt> </div>
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 ¶
func Unstage(scope *Scope, dtypes []tf.DataType, optional ...UnstageAttr) (values []tf.Output)
Op is similar to a lightweight Dequeue. The basic funtionality is similar to
dequeue with many fewer capabilities and options. This Op is optimized for performance.
func VarHandleOp ¶
func VarHandleOp(scope *Scope, dtype tf.DataType, shape tf.Shape, optional ...VarHandleOpAttr) (resource tf.Output)
Creates a handle to a Variable resource.
Arguments:
dtype: the type of this variable. Must agree with the dtypes
of all ops using this variable.
shape: The (possibly partially specified) shape of this variable.
func VarIsInitializedOp ¶
func VarIsInitializedOp(scope *Scope, resource tf.Output) (is_initialized tf.Output)
Checks whether a resource handle-based variable has been initialized.
Arguments:
resource: the input resource handle.
Returns a scalar boolean which is true if the variable has been initialized.
func Where ¶
func Where(scope *Scope, input tf.Output) (index tf.Output)
Returns locations of true values in a boolean tensor.
This operation returns the coordinates of true elements in `input`. 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 `input`. Indices are output in row-major order.
For example:
“`prettyprint # '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]]
# `input` 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]]
“`
func WholeFileReaderV2 ¶
func WholeFileReaderV2(scope *Scope, optional ...WholeFileReaderV2Attr) (reader_handle tf.Output)
A Reader that outputs the entire contents of a file as a value.
To use, enqueue filenames in a Queue. The output of ReaderRead will be a filename (key) and the contents of that file (value).
Returns The handle to reference the Reader.
func WriteFile ¶
func WriteFile(scope *Scope, filename tf.Output, contents tf.Output) (o *tf.Operation)
Writes contents to the file at input filename. Creates file 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 ZerosLike ¶
func ZerosLike(scope *Scope, x tf.Output) (y tf.Output)
Returns a tensor of zeros with the same shape and type as x.
Arguments:
x: a tensor of type T.
Returns a tensor of the same shape and type as x but filled with zeros.
func Zeta ¶
func Zeta(scope *Scope, x tf.Output, q tf.Output) (z tf.Output)
Compute the Hurwitz zeta function \\(\zeta(x, q)\\).
The Hurwitz zeta function is defined as:
“` \zeta(x, q) = \sum_{n=0}^{\infty} (q + n)^{-x} “`
type AbortAttr ¶
AbortAttr is an optional argument to Abort.
type AbortAttr func(optionalAttr)
func AbortErrorMsg ¶
func AbortErrorMsg(value string) AbortAttr
AbortErrorMsg sets the optional error_msg attribute to value.
value: A string which is the message associated with the exception. If not specified, defaults to ""
func AbortExitWithoutError ¶
func AbortExitWithoutError(value bool) AbortAttr
AbortExitWithoutError sets the optional exit_without_error attribute to value. If not specified, defaults to false
type AddManySparseToTensorsMapAttr ¶
AddManySparseToTensorsMapAttr is an optional argument to AddManySparseToTensorsMap.
type AddManySparseToTensorsMapAttr func(optionalAttr)
func AddManySparseToTensorsMapContainer ¶
func AddManySparseToTensorsMapContainer(value string) AddManySparseToTensorsMapAttr
AddManySparseToTensorsMapContainer sets the optional container attribute to value.
value: The container name for the `SparseTensorsMap` created by this op. If not specified, defaults to ""
func AddManySparseToTensorsMapSharedName ¶
func AddManySparseToTensorsMapSharedName(value string) AddManySparseToTensorsMapAttr
AddManySparseToTensorsMapSharedName sets the optional shared_name attribute to value.
value: The shared name for the `SparseTensorsMap` created by this op. If blank, the new Operation's unique name is used. If not specified, defaults to ""
type AddSparseToTensorsMapAttr ¶
AddSparseToTensorsMapAttr is an optional argument to AddSparseToTensorsMap.
type AddSparseToTensorsMapAttr func(optionalAttr)
func AddSparseToTensorsMapContainer ¶
func AddSparseToTensorsMapContainer(value string) AddSparseToTensorsMapAttr
AddSparseToTensorsMapContainer sets the optional container attribute to value.
value: The container name for the `SparseTensorsMap` created by this op. If not specified, defaults to ""
func AddSparseToTensorsMapSharedName ¶
func AddSparseToTensorsMapSharedName(value string) AddSparseToTensorsMapAttr
AddSparseToTensorsMapSharedName sets the optional shared_name attribute to value.
value: The shared name for the `SparseTensorsMap` created by this op. If blank, the new Operation's unique name is used. If not specified, defaults to ""
type AllAttr ¶
AllAttr is an optional argument to All.
type AllAttr func(optionalAttr)
func AllKeepDims ¶
func AllKeepDims(value bool) AllAttr
AllKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type AllCandidateSamplerAttr ¶
AllCandidateSamplerAttr is an optional argument to AllCandidateSampler.
type AllCandidateSamplerAttr func(optionalAttr)
func AllCandidateSamplerSeed ¶
func AllCandidateSamplerSeed(value int64) AllCandidateSamplerAttr
AllCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func AllCandidateSamplerSeed2 ¶
func AllCandidateSamplerSeed2(value int64) AllCandidateSamplerAttr
AllCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type AnyAttr ¶
AnyAttr is an optional argument to Any.
type AnyAttr func(optionalAttr)
func AnyKeepDims ¶
func AnyKeepDims(value bool) AnyAttr
AnyKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type ApproximateEqualAttr ¶
ApproximateEqualAttr is an optional argument to ApproximateEqual.
type ApproximateEqualAttr func(optionalAttr)
func ApproximateEqualTolerance ¶
func ApproximateEqualTolerance(value float32) ApproximateEqualAttr
ApproximateEqualTolerance sets the optional tolerance attribute to value. If not specified, defaults to 1e-05
type AsStringAttr ¶
AsStringAttr is an optional argument to AsString.
type AsStringAttr func(optionalAttr)
func AsStringFill ¶
func AsStringFill(value string) AsStringAttr
AsStringFill sets the optional fill attribute to value.
value: The value to pad if width > -1. If empty, pads with spaces. Another typical value is '0'. String cannot be longer than 1 character. If not specified, defaults to ""
func AsStringPrecision ¶
func AsStringPrecision(value int64) AsStringAttr
AsStringPrecision sets the optional precision attribute to value.
value: The post-decimal precision to use for floating point numbers. Only used if precision > -1. If not specified, defaults to -1
func AsStringScientific ¶
func AsStringScientific(value bool) AsStringAttr
AsStringScientific sets the optional scientific attribute to value.
value: Use scientific notation for floating point numbers. If not specified, defaults to false
func AsStringShortest ¶
func AsStringShortest(value bool) AsStringAttr
AsStringShortest sets the optional shortest attribute to value.
value: Use shortest representation (either scientific or standard) for floating point numbers. If not specified, defaults to false
func AsStringWidth ¶
func AsStringWidth(value int64) AsStringAttr
AsStringWidth sets the optional width attribute to value.
value: Pad pre-decimal numbers to this width. Applies to both floating point and integer numbers. Only used if width > -1. If not specified, defaults to -1
type AssertAttr ¶
AssertAttr is an optional argument to Assert.
type AssertAttr func(optionalAttr)
func AssertSummarize ¶
func AssertSummarize(value int64) AssertAttr
AssertSummarize sets the optional summarize attribute to value.
value: Print this many entries of each tensor. If not specified, defaults to 3
type AudioSummaryAttr ¶
AudioSummaryAttr is an optional argument to AudioSummary.
type AudioSummaryAttr func(optionalAttr)
func AudioSummaryMaxOutputs ¶
func AudioSummaryMaxOutputs(value int64) AudioSummaryAttr
AudioSummaryMaxOutputs sets the optional max_outputs attribute to value.
value: Max number of batch elements to generate audio for. If not specified, defaults to 3
REQUIRES: value >= 1
type AudioSummaryV2Attr ¶
AudioSummaryV2Attr is an optional argument to AudioSummaryV2.
type AudioSummaryV2Attr func(optionalAttr)
func AudioSummaryV2MaxOutputs ¶
func AudioSummaryV2MaxOutputs(value int64) AudioSummaryV2Attr
AudioSummaryV2MaxOutputs sets the optional max_outputs attribute to value.
value: Max number of batch elements to generate audio for. If not specified, defaults to 3
REQUIRES: value >= 1
type AvgPoolAttr ¶
AvgPoolAttr is an optional argument to AvgPool.
type AvgPoolAttr func(optionalAttr)
func AvgPoolDataFormat ¶
func AvgPoolDataFormat(value string) AvgPoolAttr
AvgPoolDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
type AvgPoolGradAttr ¶
AvgPoolGradAttr is an optional argument to AvgPoolGrad.
type AvgPoolGradAttr func(optionalAttr)
func AvgPoolGradDataFormat ¶
func AvgPoolGradDataFormat(value string) AvgPoolGradAttr
AvgPoolGradDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
type BatchMatMulAttr ¶
BatchMatMulAttr is an optional argument to BatchMatMul.
type BatchMatMulAttr func(optionalAttr)
func BatchMatMulAdjX ¶
func BatchMatMulAdjX(value bool) BatchMatMulAttr
BatchMatMulAdjX sets the optional adj_x attribute to value.
value: If `True`, adjoint the slices of `x`. Defaults to `False`. If not specified, defaults to false
func BatchMatMulAdjY ¶
func BatchMatMulAdjY(value bool) BatchMatMulAttr
BatchMatMulAdjY sets the optional adj_y attribute to value.
value: If `True`, adjoint the slices of `y`. Defaults to `False`. If not specified, defaults to false
type BiasAddAttr ¶
BiasAddAttr is an optional argument to BiasAdd.
type BiasAddAttr func(optionalAttr)
func BiasAddDataFormat ¶
func BiasAddDataFormat(value string) BiasAddAttr
BiasAddDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the bias tensor will be added to the last dimension of the value tensor. Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
The tensor will be added to "in_channels", the third-to-the-last
dimension.
If not specified, defaults to "NHWC"
type BiasAddGradAttr ¶
BiasAddGradAttr is an optional argument to BiasAddGrad.
type BiasAddGradAttr func(optionalAttr)
func BiasAddGradDataFormat ¶
func BiasAddGradDataFormat(value string) BiasAddGradAttr
BiasAddGradDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the bias tensor will be added to the last dimension of the value tensor. Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
The tensor will be added to "in_channels", the third-to-the-last
dimension.
If not specified, defaults to "NHWC"
type CTCBeamSearchDecoderAttr ¶
CTCBeamSearchDecoderAttr is an optional argument to CTCBeamSearchDecoder.
type CTCBeamSearchDecoderAttr func(optionalAttr)
func CTCBeamSearchDecoderMergeRepeated ¶
func CTCBeamSearchDecoderMergeRepeated(value bool) CTCBeamSearchDecoderAttr
CTCBeamSearchDecoderMergeRepeated sets the optional merge_repeated attribute to value.
value: If true, merge repeated classes in output. If not specified, defaults to true
type CTCGreedyDecoderAttr ¶
CTCGreedyDecoderAttr is an optional argument to CTCGreedyDecoder.
type CTCGreedyDecoderAttr func(optionalAttr)
func CTCGreedyDecoderMergeRepeated ¶
func CTCGreedyDecoderMergeRepeated(value bool) CTCGreedyDecoderAttr
CTCGreedyDecoderMergeRepeated sets the optional merge_repeated attribute to value.
value: If True, merge repeated classes in output. If not specified, defaults to false
type CTCLossAttr ¶
CTCLossAttr is an optional argument to CTCLoss.
type CTCLossAttr func(optionalAttr)
func CTCLossCtcMergeRepeated ¶
func CTCLossCtcMergeRepeated(value bool) CTCLossAttr
CTCLossCtcMergeRepeated sets the optional ctc_merge_repeated attribute to value.
value: Scalar. If set to false, *during* CTC calculation repeated non-blank labels will not be merged and are interpreted as individual labels. This is a simplified version of CTC. If not specified, defaults to true
func CTCLossPreprocessCollapseRepeated ¶
func CTCLossPreprocessCollapseRepeated(value bool) CTCLossAttr
CTCLossPreprocessCollapseRepeated sets the optional preprocess_collapse_repeated attribute to value.
value: Scalar, if true then repeated labels are collapsed prior to the CTC calculation. If not specified, defaults to false
type ComplexAbsAttr ¶
ComplexAbsAttr is an optional argument to ComplexAbs.
type ComplexAbsAttr func(optionalAttr)
func ComplexAbsTout ¶
func ComplexAbsTout(value tf.DataType) ComplexAbsAttr
ComplexAbsTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT
type ComplexAttr ¶
ComplexAttr is an optional argument to Complex.
type ComplexAttr func(optionalAttr)
func ComplexTout ¶
func ComplexTout(value tf.DataType) ComplexAttr
ComplexTout sets the optional Tout attribute to value. If not specified, defaults to DT_COMPLEX64
type ComputeAccidentalHitsAttr ¶
ComputeAccidentalHitsAttr is an optional argument to ComputeAccidentalHits.
type ComputeAccidentalHitsAttr func(optionalAttr)
func ComputeAccidentalHitsSeed ¶
func ComputeAccidentalHitsSeed(value int64) ComputeAccidentalHitsAttr
ComputeAccidentalHitsSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func ComputeAccidentalHitsSeed2 ¶
func ComputeAccidentalHitsSeed2(value int64) ComputeAccidentalHitsAttr
ComputeAccidentalHitsSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type Conv2DAttr ¶
Conv2DAttr is an optional argument to Conv2D.
type Conv2DAttr func(optionalAttr)
func Conv2DDataFormat ¶
func Conv2DDataFormat(value string) Conv2DAttr
Conv2DDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, height, width, channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, channels, height, width].
If not specified, defaults to "NHWC"
func Conv2DUseCudnnOnGpu ¶
func Conv2DUseCudnnOnGpu(value bool) Conv2DAttr
Conv2DUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true
type Conv2DBackpropFilterAttr ¶
Conv2DBackpropFilterAttr is an optional argument to Conv2DBackpropFilter.
type Conv2DBackpropFilterAttr func(optionalAttr)
func Conv2DBackpropFilterDataFormat ¶
func Conv2DBackpropFilterDataFormat(value string) Conv2DBackpropFilterAttr
Conv2DBackpropFilterDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
func Conv2DBackpropFilterUseCudnnOnGpu ¶
func Conv2DBackpropFilterUseCudnnOnGpu(value bool) Conv2DBackpropFilterAttr
Conv2DBackpropFilterUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true
type Conv2DBackpropInputAttr ¶
Conv2DBackpropInputAttr is an optional argument to Conv2DBackpropInput.
type Conv2DBackpropInputAttr func(optionalAttr)
func Conv2DBackpropInputDataFormat ¶
func Conv2DBackpropInputDataFormat(value string) Conv2DBackpropInputAttr
Conv2DBackpropInputDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
func Conv2DBackpropInputUseCudnnOnGpu ¶
func Conv2DBackpropInputUseCudnnOnGpu(value bool) Conv2DBackpropInputAttr
Conv2DBackpropInputUseCudnnOnGpu sets the optional use_cudnn_on_gpu attribute to value. If not specified, defaults to true
type CropAndResizeAttr ¶
CropAndResizeAttr is an optional argument to CropAndResize.
type CropAndResizeAttr func(optionalAttr)
func CropAndResizeExtrapolationValue ¶
func CropAndResizeExtrapolationValue(value float32) CropAndResizeAttr
CropAndResizeExtrapolationValue sets the optional extrapolation_value attribute to value.
value: Value used for extrapolation, when applicable. If not specified, defaults to 0
func CropAndResizeMethod ¶
func CropAndResizeMethod(value string) CropAndResizeAttr
CropAndResizeMethod sets the optional method attribute to value.
value: A string specifying the interpolation method. Only 'bilinear' is supported for now. If not specified, defaults to "bilinear"
type CropAndResizeGradBoxesAttr ¶
CropAndResizeGradBoxesAttr is an optional argument to CropAndResizeGradBoxes.
type CropAndResizeGradBoxesAttr func(optionalAttr)
func CropAndResizeGradBoxesMethod ¶
func CropAndResizeGradBoxesMethod(value string) CropAndResizeGradBoxesAttr
CropAndResizeGradBoxesMethod sets the optional method attribute to value.
value: A string specifying the interpolation method. Only 'bilinear' is supported for now. If not specified, defaults to "bilinear"
type CropAndResizeGradImageAttr ¶
CropAndResizeGradImageAttr is an optional argument to CropAndResizeGradImage.
type CropAndResizeGradImageAttr func(optionalAttr)
func CropAndResizeGradImageMethod ¶
func CropAndResizeGradImageMethod(value string) CropAndResizeGradImageAttr
CropAndResizeGradImageMethod sets the optional method attribute to value.
value: A string specifying the interpolation method. Only 'bilinear' is supported for now. If not specified, defaults to "bilinear"
type CumprodAttr ¶
CumprodAttr is an optional argument to Cumprod.
type CumprodAttr func(optionalAttr)
func CumprodExclusive ¶
func CumprodExclusive(value bool) CumprodAttr
CumprodExclusive sets the optional exclusive attribute to value. If not specified, defaults to false
func CumprodReverse ¶
func CumprodReverse(value bool) CumprodAttr
CumprodReverse sets the optional reverse attribute to value. If not specified, defaults to false
type CumsumAttr ¶
CumsumAttr is an optional argument to Cumsum.
type CumsumAttr func(optionalAttr)
func CumsumExclusive ¶
func CumsumExclusive(value bool) CumsumAttr
CumsumExclusive sets the optional exclusive attribute to value. If not specified, defaults to false
func CumsumReverse ¶
func CumsumReverse(value bool) CumsumAttr
CumsumReverse sets the optional reverse attribute to value. If not specified, defaults to false
type DecodeCSVAttr ¶
DecodeCSVAttr is an optional argument to DecodeCSV.
type DecodeCSVAttr func(optionalAttr)
func DecodeCSVFieldDelim ¶
func DecodeCSVFieldDelim(value string) DecodeCSVAttr
DecodeCSVFieldDelim sets the optional field_delim attribute to value.
value: delimiter to separate fields in a record. If not specified, defaults to ","
type DecodeJpegAttr ¶
DecodeJpegAttr is an optional argument to DecodeJpeg.
type DecodeJpegAttr func(optionalAttr)
func DecodeJpegAcceptableFraction ¶
func DecodeJpegAcceptableFraction(value float32) DecodeJpegAttr
DecodeJpegAcceptableFraction sets the optional acceptable_fraction attribute to value.
value: The minimum required fraction of lines before a truncated input is accepted. If not specified, defaults to 1
func DecodeJpegChannels ¶
func DecodeJpegChannels(value int64) DecodeJpegAttr
DecodeJpegChannels sets the optional channels attribute to value.
value: Number of color channels for the decoded image. If not specified, defaults to 0
func DecodeJpegDctMethod ¶
func DecodeJpegDctMethod(value string) DecodeJpegAttr
DecodeJpegDctMethod sets the optional dct_method attribute to value.
value: string specifying a hint about the algorithm used for decompression. Defaults to "" which maps to a system-specific default. Currently valid values are ["INTEGER_FAST", "INTEGER_ACCURATE"]. The hint may be ignored (e.g., the internal jpeg library changes to a version that does not have that specific option.) If not specified, defaults to ""
func DecodeJpegFancyUpscaling ¶
func DecodeJpegFancyUpscaling(value bool) DecodeJpegAttr
DecodeJpegFancyUpscaling sets the optional fancy_upscaling attribute to value.
value: If true use a slower but nicer upscaling of the chroma planes (yuv420/422 only). If not specified, defaults to true
func DecodeJpegRatio ¶
func DecodeJpegRatio(value int64) DecodeJpegAttr
DecodeJpegRatio sets the optional ratio attribute to value.
value: Downscaling ratio. If not specified, defaults to 1
func DecodeJpegTryRecoverTruncated ¶
func DecodeJpegTryRecoverTruncated(value bool) DecodeJpegAttr
DecodeJpegTryRecoverTruncated sets the optional try_recover_truncated attribute to value.
value: If true try to recover an image from truncated input. If not specified, defaults to false
type DecodePngAttr ¶
DecodePngAttr is an optional argument to DecodePng.
type DecodePngAttr func(optionalAttr)
func DecodePngChannels ¶
func DecodePngChannels(value int64) DecodePngAttr
DecodePngChannels sets the optional channels attribute to value.
value: Number of color channels for the decoded image. If not specified, defaults to 0
func DecodePngDtype ¶
func DecodePngDtype(value tf.DataType) DecodePngAttr
DecodePngDtype sets the optional dtype attribute to value. If not specified, defaults to DT_UINT8
type DecodeRawAttr ¶
DecodeRawAttr is an optional argument to DecodeRaw.
type DecodeRawAttr func(optionalAttr)
func DecodeRawLittleEndian ¶
func DecodeRawLittleEndian(value bool) DecodeRawAttr
DecodeRawLittleEndian sets the optional little_endian attribute to value.
value: Whether the input `bytes` are in little-endian order. Ignored for `out_type` values that are stored in a single byte like `uint8`. If not specified, defaults to true
type DenseToDenseSetOperationAttr ¶
DenseToDenseSetOperationAttr is an optional argument to DenseToDenseSetOperation.
type DenseToDenseSetOperationAttr func(optionalAttr)
func DenseToDenseSetOperationValidateIndices ¶
func DenseToDenseSetOperationValidateIndices(value bool) DenseToDenseSetOperationAttr
DenseToDenseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type DenseToSparseSetOperationAttr ¶
DenseToSparseSetOperationAttr is an optional argument to DenseToSparseSetOperation.
type DenseToSparseSetOperationAttr func(optionalAttr)
func DenseToSparseSetOperationValidateIndices ¶
func DenseToSparseSetOperationValidateIndices(value bool) DenseToSparseSetOperationAttr
DenseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type DepthwiseConv2dNativeAttr ¶
DepthwiseConv2dNativeAttr is an optional argument to DepthwiseConv2dNative.
type DepthwiseConv2dNativeAttr func(optionalAttr)
func DepthwiseConv2dNativeDataFormat ¶
func DepthwiseConv2dNativeDataFormat(value string) DepthwiseConv2dNativeAttr
DepthwiseConv2dNativeDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, height, width, channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, channels, height, width].
If not specified, defaults to "NHWC"
type DepthwiseConv2dNativeBackpropFilterAttr ¶
DepthwiseConv2dNativeBackpropFilterAttr is an optional argument to DepthwiseConv2dNativeBackpropFilter.
type DepthwiseConv2dNativeBackpropFilterAttr func(optionalAttr)
func DepthwiseConv2dNativeBackpropFilterDataFormat ¶
func DepthwiseConv2dNativeBackpropFilterDataFormat(value string) DepthwiseConv2dNativeBackpropFilterAttr
DepthwiseConv2dNativeBackpropFilterDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, height, width, channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, channels, height, width].
If not specified, defaults to "NHWC"
type DepthwiseConv2dNativeBackpropInputAttr ¶
DepthwiseConv2dNativeBackpropInputAttr is an optional argument to DepthwiseConv2dNativeBackpropInput.
type DepthwiseConv2dNativeBackpropInputAttr func(optionalAttr)
func DepthwiseConv2dNativeBackpropInputDataFormat ¶
func DepthwiseConv2dNativeBackpropInputDataFormat(value string) DepthwiseConv2dNativeBackpropInputAttr
DepthwiseConv2dNativeBackpropInputDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, height, width, channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, channels, height, width].
If not specified, defaults to "NHWC"
type DequantizeAttr ¶
DequantizeAttr is an optional argument to Dequantize.
type DequantizeAttr func(optionalAttr)
func DequantizeMode ¶
func DequantizeMode(value string) DequantizeAttr
DequantizeMode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"
type DestroyResourceOpAttr ¶
DestroyResourceOpAttr is an optional argument to DestroyResourceOp.
type DestroyResourceOpAttr func(optionalAttr)
func DestroyResourceOpIgnoreLookupError ¶
func DestroyResourceOpIgnoreLookupError(value bool) DestroyResourceOpAttr
DestroyResourceOpIgnoreLookupError sets the optional ignore_lookup_error attribute to value.
value: whether to ignore the error when the resource doesn't exist. If not specified, defaults to true
type EditDistanceAttr ¶
EditDistanceAttr is an optional argument to EditDistance.
type EditDistanceAttr func(optionalAttr)
func EditDistanceNormalize ¶
func EditDistanceNormalize(value bool) EditDistanceAttr
EditDistanceNormalize sets the optional normalize attribute to value.
value: boolean (if true, edit distances are normalized by length of truth).
The output is: If not specified, defaults to true
type EncodeBase64Attr ¶
EncodeBase64Attr is an optional argument to EncodeBase64.
type EncodeBase64Attr func(optionalAttr)
func EncodeBase64Pad ¶
func EncodeBase64Pad(value bool) EncodeBase64Attr
EncodeBase64Pad sets the optional pad attribute to value.
value: Bool whether padding is applied at the ends. If not specified, defaults to false
type EncodeJpegAttr ¶
EncodeJpegAttr is an optional argument to EncodeJpeg.
type EncodeJpegAttr func(optionalAttr)
func EncodeJpegChromaDownsampling ¶
func EncodeJpegChromaDownsampling(value bool) EncodeJpegAttr
EncodeJpegChromaDownsampling sets the optional chroma_downsampling attribute to value.
value: See http://en.wikipedia.org/wiki/Chroma_subsampling. If not specified, defaults to true
func EncodeJpegDensityUnit ¶
func EncodeJpegDensityUnit(value string) EncodeJpegAttr
EncodeJpegDensityUnit sets the optional density_unit attribute to value.
value: Unit used to specify `x_density` and `y_density`: pixels per inch (`'in'`) or centimeter (`'cm'`). If not specified, defaults to "in"
func EncodeJpegFormat ¶
func EncodeJpegFormat(value string) EncodeJpegAttr
EncodeJpegFormat sets the optional format attribute to value.
value: Per pixel image format. If not specified, defaults to ""
func EncodeJpegOptimizeSize ¶
func EncodeJpegOptimizeSize(value bool) EncodeJpegAttr
EncodeJpegOptimizeSize sets the optional optimize_size attribute to value.
value: If True, spend CPU/RAM to reduce size with no quality change. If not specified, defaults to false
func EncodeJpegProgressive ¶
func EncodeJpegProgressive(value bool) EncodeJpegAttr
EncodeJpegProgressive sets the optional progressive attribute to value.
value: If True, create a JPEG that loads progressively (coarse to fine). If not specified, defaults to false
func EncodeJpegQuality ¶
func EncodeJpegQuality(value int64) EncodeJpegAttr
EncodeJpegQuality sets the optional quality attribute to value.
value: Quality of the compression from 0 to 100 (higher is better and slower). If not specified, defaults to 95
func EncodeJpegXDensity ¶
func EncodeJpegXDensity(value int64) EncodeJpegAttr
EncodeJpegXDensity sets the optional x_density attribute to value.
value: Horizontal pixels per density unit. If not specified, defaults to 300
func EncodeJpegXmpMetadata ¶
func EncodeJpegXmpMetadata(value string) EncodeJpegAttr
EncodeJpegXmpMetadata sets the optional xmp_metadata attribute to value.
value: If not empty, embed this XMP metadata in the image header. If not specified, defaults to ""
func EncodeJpegYDensity ¶
func EncodeJpegYDensity(value int64) EncodeJpegAttr
EncodeJpegYDensity sets the optional y_density attribute to value.
value: Vertical pixels per density unit. If not specified, defaults to 300
type EncodePngAttr ¶
EncodePngAttr is an optional argument to EncodePng.
type EncodePngAttr func(optionalAttr)
func EncodePngCompression ¶
func EncodePngCompression(value int64) EncodePngAttr
EncodePngCompression sets the optional compression attribute to value.
value: Compression level. If not specified, defaults to -1
type EnterAttr ¶
EnterAttr is an optional argument to Enter.
type EnterAttr func(optionalAttr)
func EnterIsConstant ¶
func EnterIsConstant(value bool) EnterAttr
EnterIsConstant sets the optional is_constant attribute to value.
value: If true, the output is constant within the child frame. If not specified, defaults to false
func EnterParallelIterations ¶
func EnterParallelIterations(value int64) EnterAttr
EnterParallelIterations sets the optional parallel_iterations attribute to value.
value: The number of iterations allowed to run in parallel. If not specified, defaults to 10
type ExtractGlimpseAttr ¶
ExtractGlimpseAttr is an optional argument to ExtractGlimpse.
type ExtractGlimpseAttr func(optionalAttr)
func ExtractGlimpseCentered ¶
func ExtractGlimpseCentered(value bool) ExtractGlimpseAttr
ExtractGlimpseCentered sets the optional centered attribute to value.
value: indicates if the offset coordinates are centered relative to the image, in which case the (0, 0) offset is relative to the center of the input images. If false, the (0,0) offset corresponds to the upper left corner of the input images. If not specified, defaults to true
func ExtractGlimpseNormalized ¶
func ExtractGlimpseNormalized(value bool) ExtractGlimpseAttr
ExtractGlimpseNormalized sets the optional normalized attribute to value.
value: indicates if the offset coordinates are normalized. If not specified, defaults to true
func ExtractGlimpseUniformNoise ¶
func ExtractGlimpseUniformNoise(value bool) ExtractGlimpseAttr
ExtractGlimpseUniformNoise sets the optional uniform_noise attribute to value.
value: indicates if the noise should be generated using a uniform distribution or a Gaussian distribution. If not specified, defaults to true
type FIFOQueueV2Attr ¶
FIFOQueueV2Attr is an optional argument to FIFOQueueV2.
type FIFOQueueV2Attr func(optionalAttr)
func FIFOQueueV2Capacity ¶
func FIFOQueueV2Capacity(value int64) FIFOQueueV2Attr
FIFOQueueV2Capacity sets the optional capacity attribute to value.
value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1
func FIFOQueueV2Container ¶
func FIFOQueueV2Container(value string) FIFOQueueV2Attr
FIFOQueueV2Container sets the optional container attribute to value.
value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func FIFOQueueV2Shapes ¶
func FIFOQueueV2Shapes(value []tf.Shape) FIFOQueueV2Attr
FIFOQueueV2Shapes sets the optional shapes attribute to value.
value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func FIFOQueueV2SharedName ¶
func FIFOQueueV2SharedName(value string) FIFOQueueV2Attr
FIFOQueueV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""
type FakeQuantWithMinMaxArgsAttr ¶
FakeQuantWithMinMaxArgsAttr is an optional argument to FakeQuantWithMinMaxArgs.
type FakeQuantWithMinMaxArgsAttr func(optionalAttr)
func FakeQuantWithMinMaxArgsMax ¶
func FakeQuantWithMinMaxArgsMax(value float32) FakeQuantWithMinMaxArgsAttr
FakeQuantWithMinMaxArgsMax sets the optional max attribute to value. If not specified, defaults to 6
func FakeQuantWithMinMaxArgsMin ¶
func FakeQuantWithMinMaxArgsMin(value float32) FakeQuantWithMinMaxArgsAttr
FakeQuantWithMinMaxArgsMin sets the optional min attribute to value. If not specified, defaults to -6
type FakeQuantWithMinMaxArgsGradientAttr ¶
FakeQuantWithMinMaxArgsGradientAttr is an optional argument to FakeQuantWithMinMaxArgsGradient.
type FakeQuantWithMinMaxArgsGradientAttr func(optionalAttr)
func FakeQuantWithMinMaxArgsGradientMax ¶
func FakeQuantWithMinMaxArgsGradientMax(value float32) FakeQuantWithMinMaxArgsGradientAttr
FakeQuantWithMinMaxArgsGradientMax sets the optional max attribute to value. If not specified, defaults to 6
func FakeQuantWithMinMaxArgsGradientMin ¶
func FakeQuantWithMinMaxArgsGradientMin(value float32) FakeQuantWithMinMaxArgsGradientAttr
FakeQuantWithMinMaxArgsGradientMin sets the optional min attribute to value. If not specified, defaults to -6
type FixedLengthRecordReaderV2Attr ¶
FixedLengthRecordReaderV2Attr is an optional argument to FixedLengthRecordReaderV2.
type FixedLengthRecordReaderV2Attr func(optionalAttr)
func FixedLengthRecordReaderV2Container ¶
func FixedLengthRecordReaderV2Container(value string) FixedLengthRecordReaderV2Attr
FixedLengthRecordReaderV2Container sets the optional container attribute to value.
value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func FixedLengthRecordReaderV2FooterBytes ¶
func FixedLengthRecordReaderV2FooterBytes(value int64) FixedLengthRecordReaderV2Attr
FixedLengthRecordReaderV2FooterBytes sets the optional footer_bytes attribute to value. If not specified, defaults to 0
func FixedLengthRecordReaderV2HeaderBytes ¶
func FixedLengthRecordReaderV2HeaderBytes(value int64) FixedLengthRecordReaderV2Attr
FixedLengthRecordReaderV2HeaderBytes sets the optional header_bytes attribute to value. If not specified, defaults to 0
func FixedLengthRecordReaderV2SharedName ¶
func FixedLengthRecordReaderV2SharedName(value string) FixedLengthRecordReaderV2Attr
FixedLengthRecordReaderV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""
type FixedUnigramCandidateSamplerAttr ¶
FixedUnigramCandidateSamplerAttr is an optional argument to FixedUnigramCandidateSampler.
type FixedUnigramCandidateSamplerAttr func(optionalAttr)
func FixedUnigramCandidateSamplerDistortion ¶
func FixedUnigramCandidateSamplerDistortion(value float32) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerDistortion sets the optional distortion attribute to value.
value: The distortion is used to skew the unigram probability distribution. Each weight is first raised to the distortion's power before adding to the internal unigram distribution. As a result, distortion = 1.0 gives regular unigram sampling (as defined by the vocab file), and distortion = 0.0 gives a uniform distribution. If not specified, defaults to 1
func FixedUnigramCandidateSamplerNumReservedIds ¶
func FixedUnigramCandidateSamplerNumReservedIds(value int64) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerNumReservedIds sets the optional num_reserved_ids attribute to value.
value: Optionally some reserved IDs can be added in the range [0, ..., num_reserved_ids) by the users. One use case is that a special unknown word token is used as ID 0. These IDs will have a sampling probability of 0. If not specified, defaults to 0
func FixedUnigramCandidateSamplerNumShards ¶
func FixedUnigramCandidateSamplerNumShards(value int64) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerNumShards sets the optional num_shards attribute to value.
value: A sampler can be used to sample from a subset of the original range in order to speed up the whole computation through parallelism. This parameter (together with 'shard') indicates the number of partitions that are being used in the overall computation. If not specified, defaults to 1
REQUIRES: value >= 1
func FixedUnigramCandidateSamplerSeed ¶
func FixedUnigramCandidateSamplerSeed(value int64) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func FixedUnigramCandidateSamplerSeed2 ¶
func FixedUnigramCandidateSamplerSeed2(value int64) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
func FixedUnigramCandidateSamplerShard ¶
func FixedUnigramCandidateSamplerShard(value int64) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerShard sets the optional shard attribute to value.
value: A sampler can be used to sample from a subset of the original range in order to speed up the whole computation through parallelism. This parameter (together with 'num_shards') indicates the particular partition number of a sampler op, when partitioning is being used. If not specified, defaults to 0
REQUIRES: value >= 0
func FixedUnigramCandidateSamplerUnigrams ¶
func FixedUnigramCandidateSamplerUnigrams(value []float32) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerUnigrams sets the optional unigrams attribute to value.
value: A list of unigram counts or probabilities, one per ID in sequential order. Exactly one of vocab_file and unigrams should be passed to this op. If not specified, defaults to <>
func FixedUnigramCandidateSamplerVocabFile ¶
func FixedUnigramCandidateSamplerVocabFile(value string) FixedUnigramCandidateSamplerAttr
FixedUnigramCandidateSamplerVocabFile sets the optional vocab_file attribute to value.
value: Each valid line in this file (which should have a CSV-like format) corresponds to a valid word ID. IDs are in sequential order, starting from num_reserved_ids. The last entry in each line is expected to be a value corresponding to the count or relative probability. Exactly one of vocab_file and unigrams needs to be passed to this op. If not specified, defaults to ""
type FractionalAvgPoolAttr ¶
FractionalAvgPoolAttr is an optional argument to FractionalAvgPool.
type FractionalAvgPoolAttr func(optionalAttr)
func FractionalAvgPoolDeterministic ¶
func FractionalAvgPoolDeterministic(value bool) FractionalAvgPoolAttr
FractionalAvgPoolDeterministic sets the optional deterministic attribute to value.
value: When set to True, a fixed pooling region will be used when iterating over a FractionalAvgPool node in the computation graph. Mainly used in unit test to make FractionalAvgPool deterministic. If not specified, defaults to false
func FractionalAvgPoolOverlapping ¶
func FractionalAvgPoolOverlapping(value bool) FractionalAvgPoolAttr
FractionalAvgPoolOverlapping sets the optional overlapping attribute to value.
value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:
`index 0 1 2 3 4`
`value 20 5 16 3 7`
If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [41/3, 26/3] for fractional avg pooling. If not specified, defaults to false
func FractionalAvgPoolPseudoRandom ¶
func FractionalAvgPoolPseudoRandom(value bool) FractionalAvgPoolAttr
FractionalAvgPoolPseudoRandom sets the optional pseudo_random attribute to value.
value: When set to True, generates the pooling sequence in a pseudorandom fashion, otherwise, in a random fashion. Check paper [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071) for difference between pseudorandom and random. If not specified, defaults to false
func FractionalAvgPoolSeed ¶
func FractionalAvgPoolSeed(value int64) FractionalAvgPoolAttr
FractionalAvgPoolSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func FractionalAvgPoolSeed2 ¶
func FractionalAvgPoolSeed2(value int64) FractionalAvgPoolAttr
FractionalAvgPoolSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type FractionalAvgPoolGradAttr ¶
FractionalAvgPoolGradAttr is an optional argument to FractionalAvgPoolGrad.
type FractionalAvgPoolGradAttr func(optionalAttr)
func FractionalAvgPoolGradOverlapping ¶
func FractionalAvgPoolGradOverlapping(value bool) FractionalAvgPoolGradAttr
FractionalAvgPoolGradOverlapping sets the optional overlapping attribute to value.
value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:
`index 0 1 2 3 4`
`value 20 5 16 3 7`
If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [41/3, 26/3] for fractional avg pooling. If not specified, defaults to false
type FractionalMaxPoolAttr ¶
FractionalMaxPoolAttr is an optional argument to FractionalMaxPool.
type FractionalMaxPoolAttr func(optionalAttr)
func FractionalMaxPoolDeterministic ¶
func FractionalMaxPoolDeterministic(value bool) FractionalMaxPoolAttr
FractionalMaxPoolDeterministic sets the optional deterministic attribute to value.
value: When set to True, a fixed pooling region will be used when iterating over a FractionalMaxPool node in the computation graph. Mainly used in unit test to make FractionalMaxPool deterministic. If not specified, defaults to false
func FractionalMaxPoolOverlapping ¶
func FractionalMaxPoolOverlapping(value bool) FractionalMaxPoolAttr
FractionalMaxPoolOverlapping sets the optional overlapping attribute to value.
value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:
`index 0 1 2 3 4`
`value 20 5 16 3 7`
If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [20, 16] for fractional max pooling. If not specified, defaults to false
func FractionalMaxPoolPseudoRandom ¶
func FractionalMaxPoolPseudoRandom(value bool) FractionalMaxPoolAttr
FractionalMaxPoolPseudoRandom sets the optional pseudo_random attribute to value.
value: When set to True, generates the pooling sequence in a pseudorandom fashion, otherwise, in a random fashion. Check paper [Benjamin Graham, Fractional Max-Pooling](http://arxiv.org/abs/1412.6071) for difference between pseudorandom and random. If not specified, defaults to false
func FractionalMaxPoolSeed ¶
func FractionalMaxPoolSeed(value int64) FractionalMaxPoolAttr
FractionalMaxPoolSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func FractionalMaxPoolSeed2 ¶
func FractionalMaxPoolSeed2(value int64) FractionalMaxPoolAttr
FractionalMaxPoolSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type FractionalMaxPoolGradAttr ¶
FractionalMaxPoolGradAttr is an optional argument to FractionalMaxPoolGrad.
type FractionalMaxPoolGradAttr func(optionalAttr)
func FractionalMaxPoolGradOverlapping ¶
func FractionalMaxPoolGradOverlapping(value bool) FractionalMaxPoolGradAttr
FractionalMaxPoolGradOverlapping sets the optional overlapping attribute to value.
value: When set to True, it means when pooling, the values at the boundary of adjacent pooling cells are used by both cells. For example:
`index 0 1 2 3 4`
`value 20 5 16 3 7`
If the pooling sequence is [0, 2, 4], then 16, at index 2 will be used twice. The result would be [20, 16] for fractional max pooling. If not specified, defaults to false
type FusedBatchNormAttr ¶
FusedBatchNormAttr is an optional argument to FusedBatchNorm.
type FusedBatchNormAttr func(optionalAttr)
func FusedBatchNormDataFormat ¶
func FusedBatchNormDataFormat(value string) FusedBatchNormAttr
FusedBatchNormDataFormat sets the optional data_format attribute to value.
value: The data format for x and y. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"
func FusedBatchNormEpsilon ¶
func FusedBatchNormEpsilon(value float32) FusedBatchNormAttr
FusedBatchNormEpsilon sets the optional epsilon attribute to value.
value: A small float number added to the variance of x. If not specified, defaults to 0.0001
func FusedBatchNormIsTraining ¶
func FusedBatchNormIsTraining(value bool) FusedBatchNormAttr
FusedBatchNormIsTraining sets the optional is_training attribute to value.
value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true
type FusedBatchNormGradAttr ¶
FusedBatchNormGradAttr is an optional argument to FusedBatchNormGrad.
type FusedBatchNormGradAttr func(optionalAttr)
func FusedBatchNormGradDataFormat ¶
func FusedBatchNormGradDataFormat(value string) FusedBatchNormGradAttr
FusedBatchNormGradDataFormat sets the optional data_format attribute to value.
value: The data format for y_backprop, x, x_backprop. Either "NHWC" (default) or "NCHW". If not specified, defaults to "NHWC"
func FusedBatchNormGradEpsilon ¶
func FusedBatchNormGradEpsilon(value float32) FusedBatchNormGradAttr
FusedBatchNormGradEpsilon sets the optional epsilon attribute to value.
value: A small float number added to the variance of x. If not specified, defaults to 0.0001
func FusedBatchNormGradIsTraining ¶
func FusedBatchNormGradIsTraining(value bool) FusedBatchNormGradAttr
FusedBatchNormGradIsTraining sets the optional is_training attribute to value.
value: A bool value to indicate the operation is for training (default) or inference. If not specified, defaults to true
type FusedResizeAndPadConv2DAttr ¶
FusedResizeAndPadConv2DAttr is an optional argument to FusedResizeAndPadConv2D.
type FusedResizeAndPadConv2DAttr func(optionalAttr)
func FusedResizeAndPadConv2DResizeAlignCorners ¶
func FusedResizeAndPadConv2DResizeAlignCorners(value bool) FusedResizeAndPadConv2DAttr
FusedResizeAndPadConv2DResizeAlignCorners sets the optional resize_align_corners attribute to value.
value: If true, rescale input by (new_height - 1) / (height - 1), which exactly aligns the 4 corners of images and resized images. If false, rescale by new_height / height. Treat similarly the width dimension. If not specified, defaults to false
type GatherAttr ¶
GatherAttr is an optional argument to Gather.
type GatherAttr func(optionalAttr)
func GatherValidateIndices ¶
func GatherValidateIndices(value bool) GatherAttr
GatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type IdentityReaderV2Attr ¶
IdentityReaderV2Attr is an optional argument to IdentityReaderV2.
type IdentityReaderV2Attr func(optionalAttr)
func IdentityReaderV2Container ¶
func IdentityReaderV2Container(value string) IdentityReaderV2Attr
IdentityReaderV2Container sets the optional container attribute to value.
value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func IdentityReaderV2SharedName ¶
func IdentityReaderV2SharedName(value string) IdentityReaderV2Attr
IdentityReaderV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""
type ImagAttr ¶
ImagAttr is an optional argument to Imag.
type ImagAttr func(optionalAttr)
func ImagTout ¶
func ImagTout(value tf.DataType) ImagAttr
ImagTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT
type ImageSummaryAttr ¶
ImageSummaryAttr is an optional argument to ImageSummary.
type ImageSummaryAttr func(optionalAttr)
func ImageSummaryBadColor ¶
func ImageSummaryBadColor(value tf.Tensor) ImageSummaryAttr
ImageSummaryBadColor sets the optional bad_color attribute to value.
value: Color to use for pixels with non-finite values. If not specified, defaults to <dtype:DT_UINT8 tensor_shape:<dim:<size:4 > > int_val:255 int_val:0 int_val:0 int_val:255 >
func ImageSummaryMaxImages ¶
func ImageSummaryMaxImages(value int64) ImageSummaryAttr
ImageSummaryMaxImages sets the optional max_images attribute to value.
value: Max number of batch elements to generate images for. If not specified, defaults to 3
REQUIRES: value >= 1
type LRNAttr ¶
LRNAttr is an optional argument to LRN.
type LRNAttr func(optionalAttr)
func LRNAlpha ¶
func LRNAlpha(value float32) LRNAttr
LRNAlpha sets the optional alpha attribute to value.
value: A scale factor, usually positive. If not specified, defaults to 1
func LRNBeta ¶
func LRNBeta(value float32) LRNAttr
LRNBeta sets the optional beta attribute to value.
value: An exponent. If not specified, defaults to 0.5
func LRNBias ¶
func LRNBias(value float32) LRNAttr
LRNBias sets the optional bias attribute to value.
value: An offset (usually positive to avoid dividing by 0). If not specified, defaults to 1
func LRNDepthRadius ¶
func LRNDepthRadius(value int64) LRNAttr
LRNDepthRadius sets the optional depth_radius attribute to value.
value: 0-D. Half-width of the 1-D normalization window. If not specified, defaults to 5
type LRNGradAttr ¶
LRNGradAttr is an optional argument to LRNGrad.
type LRNGradAttr func(optionalAttr)
func LRNGradAlpha ¶
func LRNGradAlpha(value float32) LRNGradAttr
LRNGradAlpha sets the optional alpha attribute to value.
value: A scale factor, usually positive. If not specified, defaults to 1
func LRNGradBeta ¶
func LRNGradBeta(value float32) LRNGradAttr
LRNGradBeta sets the optional beta attribute to value.
value: An exponent. If not specified, defaults to 0.5
func LRNGradBias ¶
func LRNGradBias(value float32) LRNGradAttr
LRNGradBias sets the optional bias attribute to value.
value: An offset (usually > 0 to avoid dividing by 0). If not specified, defaults to 1
func LRNGradDepthRadius ¶
func LRNGradDepthRadius(value int64) LRNGradAttr
LRNGradDepthRadius sets the optional depth_radius attribute to value.
value: A depth radius. If not specified, defaults to 5
type LearnedUnigramCandidateSamplerAttr ¶
LearnedUnigramCandidateSamplerAttr is an optional argument to LearnedUnigramCandidateSampler.
type LearnedUnigramCandidateSamplerAttr func(optionalAttr)
func LearnedUnigramCandidateSamplerSeed ¶
func LearnedUnigramCandidateSamplerSeed(value int64) LearnedUnigramCandidateSamplerAttr
LearnedUnigramCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func LearnedUnigramCandidateSamplerSeed2 ¶
func LearnedUnigramCandidateSamplerSeed2(value int64) LearnedUnigramCandidateSamplerAttr
LearnedUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type ListDiffAttr ¶
ListDiffAttr is an optional argument to ListDiff.
type ListDiffAttr func(optionalAttr)
func ListDiffOutIdx ¶
func ListDiffOutIdx(value tf.DataType) ListDiffAttr
ListDiffOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32
type LogUniformCandidateSamplerAttr ¶
LogUniformCandidateSamplerAttr is an optional argument to LogUniformCandidateSampler.
type LogUniformCandidateSamplerAttr func(optionalAttr)
func LogUniformCandidateSamplerSeed ¶
func LogUniformCandidateSamplerSeed(value int64) LogUniformCandidateSamplerAttr
LogUniformCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func LogUniformCandidateSamplerSeed2 ¶
func LogUniformCandidateSamplerSeed2(value int64) LogUniformCandidateSamplerAttr
LogUniformCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type MatMulAttr ¶
MatMulAttr is an optional argument to MatMul.
type MatMulAttr func(optionalAttr)
func MatMulTransposeA ¶
func MatMulTransposeA(value bool) MatMulAttr
MatMulTransposeA sets the optional transpose_a attribute to value.
value: If true, "a" is transposed before multiplication. If not specified, defaults to false
func MatMulTransposeB ¶
func MatMulTransposeB(value bool) MatMulAttr
MatMulTransposeB sets the optional transpose_b attribute to value.
value: If true, "b" is transposed before multiplication. If not specified, defaults to false
type MatrixInverseAttr ¶
MatrixInverseAttr is an optional argument to MatrixInverse.
type MatrixInverseAttr func(optionalAttr)
func MatrixInverseAdjoint ¶
func MatrixInverseAdjoint(value bool) MatrixInverseAttr
MatrixInverseAdjoint sets the optional adjoint attribute to value. If not specified, defaults to false
type MatrixSolveAttr ¶
MatrixSolveAttr is an optional argument to MatrixSolve.
type MatrixSolveAttr func(optionalAttr)
func MatrixSolveAdjoint ¶
func MatrixSolveAdjoint(value bool) MatrixSolveAttr
MatrixSolveAdjoint sets the optional adjoint attribute to value.
value: Boolean indicating whether to solve with `matrix` or its (block-wise) adjoint. If not specified, defaults to false
type MatrixSolveLsAttr ¶
MatrixSolveLsAttr is an optional argument to MatrixSolveLs.
type MatrixSolveLsAttr func(optionalAttr)
func MatrixSolveLsFast ¶
func MatrixSolveLsFast(value bool) MatrixSolveLsAttr
MatrixSolveLsFast sets the optional fast attribute to value. If not specified, defaults to true
type MatrixTriangularSolveAttr ¶
MatrixTriangularSolveAttr is an optional argument to MatrixTriangularSolve.
type MatrixTriangularSolveAttr func(optionalAttr)
func MatrixTriangularSolveAdjoint ¶
func MatrixTriangularSolveAdjoint(value bool) MatrixTriangularSolveAttr
MatrixTriangularSolveAdjoint sets the optional adjoint attribute to value.
value: Boolean indicating whether to solve with `matrix` or its (block-wise)
adjoint.
@compatibility(numpy) Equivalent to np.linalg.triangular_solve @end_compatibility If not specified, defaults to false
func MatrixTriangularSolveLower ¶
func MatrixTriangularSolveLower(value bool) MatrixTriangularSolveAttr
MatrixTriangularSolveLower sets the optional lower attribute to value.
value: Boolean indicating whether the innermost matrices in `matrix` are lower or upper triangular. If not specified, defaults to true
type MaxAttr ¶
MaxAttr is an optional argument to Max.
type MaxAttr func(optionalAttr)
func MaxKeepDims ¶
func MaxKeepDims(value bool) MaxAttr
MaxKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type MaxPoolAttr ¶
MaxPoolAttr is an optional argument to MaxPool.
type MaxPoolAttr func(optionalAttr)
func MaxPoolDataFormat ¶
func MaxPoolDataFormat(value string) MaxPoolAttr
MaxPoolDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
type MaxPoolGradAttr ¶
MaxPoolGradAttr is an optional argument to MaxPoolGrad.
type MaxPoolGradAttr func(optionalAttr)
func MaxPoolGradDataFormat ¶
func MaxPoolGradDataFormat(value string) MaxPoolGradAttr
MaxPoolGradDataFormat sets the optional data_format attribute to value.
value: Specify the data format of the input and output data. With the default format "NHWC", the data is stored in the order of:
[batch, in_height, in_width, in_channels].
Alternatively, the format could be "NCHW", the data storage order of:
[batch, in_channels, in_height, in_width].
If not specified, defaults to "NHWC"
type MaxPoolWithArgmaxAttr ¶
MaxPoolWithArgmaxAttr is an optional argument to MaxPoolWithArgmax.
type MaxPoolWithArgmaxAttr func(optionalAttr)
func MaxPoolWithArgmaxTargmax ¶
func MaxPoolWithArgmaxTargmax(value tf.DataType) MaxPoolWithArgmaxAttr
MaxPoolWithArgmaxTargmax sets the optional Targmax attribute to value. If not specified, defaults to DT_INT64
type MeanAttr ¶
MeanAttr is an optional argument to Mean.
type MeanAttr func(optionalAttr)
func MeanKeepDims ¶
func MeanKeepDims(value bool) MeanAttr
MeanKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type MergeV2CheckpointsAttr ¶
MergeV2CheckpointsAttr is an optional argument to MergeV2Checkpoints.
type MergeV2CheckpointsAttr func(optionalAttr)
func MergeV2CheckpointsDeleteOldDirs ¶
func MergeV2CheckpointsDeleteOldDirs(value bool) MergeV2CheckpointsAttr
MergeV2CheckpointsDeleteOldDirs sets the optional delete_old_dirs attribute to value.
value: see above. If not specified, defaults to true
type MinAttr ¶
MinAttr is an optional argument to Min.
type MinAttr func(optionalAttr)
func MinKeepDims ¶
func MinKeepDims(value bool) MinAttr
MinKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type MultinomialAttr ¶
MultinomialAttr is an optional argument to Multinomial.
type MultinomialAttr func(optionalAttr)
func MultinomialSeed ¶
func MultinomialSeed(value int64) MultinomialAttr
MultinomialSeed sets the optional seed attribute to value.
value: If either seed or seed2 is set to be non-zero, the internal random number generator is seeded by the given seed. Otherwise, a random seed is used. If not specified, defaults to 0
func MultinomialSeed2 ¶
func MultinomialSeed2(value int64) MultinomialAttr
MultinomialSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type NonMaxSuppressionAttr ¶
NonMaxSuppressionAttr is an optional argument to NonMaxSuppression.
type NonMaxSuppressionAttr func(optionalAttr)
func NonMaxSuppressionIouThreshold ¶
func NonMaxSuppressionIouThreshold(value float32) NonMaxSuppressionAttr
NonMaxSuppressionIouThreshold sets the optional iou_threshold attribute to value.
value: A float representing the threshold for deciding whether boxes overlap too much with respect to IOU. If not specified, defaults to 0.5
type OneHotAttr ¶
OneHotAttr is an optional argument to OneHot.
type OneHotAttr func(optionalAttr)
func OneHotAxis ¶
func OneHotAxis(value int64) OneHotAttr
OneHotAxis sets the optional axis attribute to value.
value: The axis to fill (default: -1, a new inner-most axis). If not specified, defaults to -1
type PackAttr ¶
PackAttr is an optional argument to Pack.
type PackAttr func(optionalAttr)
func PackAxis ¶
func PackAxis(value int64) PackAttr
PackAxis sets the optional axis attribute to value.
value: Dimension along which to pack. Negative values wrap around, so the valid range is `[-(R+1), R+1)`. If not specified, defaults to 0
type PaddingFIFOQueueV2Attr ¶
PaddingFIFOQueueV2Attr is an optional argument to PaddingFIFOQueueV2.
type PaddingFIFOQueueV2Attr func(optionalAttr)
func PaddingFIFOQueueV2Capacity ¶
func PaddingFIFOQueueV2Capacity(value int64) PaddingFIFOQueueV2Attr
PaddingFIFOQueueV2Capacity sets the optional capacity attribute to value.
value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1
func PaddingFIFOQueueV2Container ¶
func PaddingFIFOQueueV2Container(value string) PaddingFIFOQueueV2Attr
PaddingFIFOQueueV2Container sets the optional container attribute to value.
value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func PaddingFIFOQueueV2Shapes ¶
func PaddingFIFOQueueV2Shapes(value []tf.Shape) PaddingFIFOQueueV2Attr
PaddingFIFOQueueV2Shapes sets the optional shapes attribute to value.
value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. Shapes of fixed rank but variable size are allowed by setting any shape dimension to -1. In this case, the inputs' shape may vary along the given dimension, and DequeueMany will pad the given dimension with zeros up to the maximum shape of all elements in the given batch. If the length of this attr is 0, different queue elements may have different ranks and shapes, but only one element may be dequeued at a time. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func PaddingFIFOQueueV2SharedName ¶
func PaddingFIFOQueueV2SharedName(value string) PaddingFIFOQueueV2Attr
PaddingFIFOQueueV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""
type ParameterizedTruncatedNormalAttr ¶
ParameterizedTruncatedNormalAttr is an optional argument to ParameterizedTruncatedNormal.
type ParameterizedTruncatedNormalAttr func(optionalAttr)
func ParameterizedTruncatedNormalSeed ¶
func ParameterizedTruncatedNormalSeed(value int64) ParameterizedTruncatedNormalAttr
ParameterizedTruncatedNormalSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func ParameterizedTruncatedNormalSeed2 ¶
func ParameterizedTruncatedNormalSeed2(value int64) ParameterizedTruncatedNormalAttr
ParameterizedTruncatedNormalSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type ParseSingleSequenceExampleAttr ¶
ParseSingleSequenceExampleAttr is an optional argument to ParseSingleSequenceExample.
type ParseSingleSequenceExampleAttr func(optionalAttr)
func ParseSingleSequenceExampleContextDenseShapes ¶
func ParseSingleSequenceExampleContextDenseShapes(value []tf.Shape) ParseSingleSequenceExampleAttr
ParseSingleSequenceExampleContextDenseShapes sets the optional context_dense_shapes attribute to value.
value: A list of Ncontext_dense shapes; the shapes of data in each context Feature given in context_dense_keys. The number of elements in the Feature corresponding to context_dense_key[j] must always equal context_dense_shapes[j].NumEntries(). The shape of context_dense_values[j] will match context_dense_shapes[j]. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func ParseSingleSequenceExampleContextSparseTypes ¶
func ParseSingleSequenceExampleContextSparseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr
ParseSingleSequenceExampleContextSparseTypes sets the optional context_sparse_types attribute to value.
value: A list of Ncontext_sparse types; the data types of data in each context Feature given in context_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to <>
REQUIRES: len(value) >= 0
func ParseSingleSequenceExampleFeatureListDenseShapes ¶
func ParseSingleSequenceExampleFeatureListDenseShapes(value []tf.Shape) ParseSingleSequenceExampleAttr
ParseSingleSequenceExampleFeatureListDenseShapes sets the optional feature_list_dense_shapes attribute to value.
value: A list of Nfeature_list_dense shapes; the shapes of data in each FeatureList given in feature_list_dense_keys. The shape of each Feature in the FeatureList corresponding to feature_list_dense_key[j] must always equal feature_list_dense_shapes[j].NumEntries(). If not specified, defaults to <>
REQUIRES: len(value) >= 0
func ParseSingleSequenceExampleFeatureListDenseTypes ¶
func ParseSingleSequenceExampleFeatureListDenseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr
ParseSingleSequenceExampleFeatureListDenseTypes sets the optional feature_list_dense_types attribute to value. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func ParseSingleSequenceExampleFeatureListSparseTypes ¶
func ParseSingleSequenceExampleFeatureListSparseTypes(value []tf.DataType) ParseSingleSequenceExampleAttr
ParseSingleSequenceExampleFeatureListSparseTypes sets the optional feature_list_sparse_types attribute to value.
value: A list of Nfeature_list_sparse types; the data types of data in each FeatureList given in feature_list_sparse_keys. Currently the ParseSingleSequenceExample supports DT_FLOAT (FloatList), DT_INT64 (Int64List), and DT_STRING (BytesList). If not specified, defaults to <>
REQUIRES: len(value) >= 0
type PlaceholderAttr ¶
PlaceholderAttr is an optional argument to Placeholder.
type PlaceholderAttr func(optionalAttr)
func PlaceholderShape ¶
func PlaceholderShape(value tf.Shape) PlaceholderAttr
PlaceholderShape sets the optional shape attribute to value.
value: (Optional) The shape of the tensor. If the shape has 0 dimensions, the shape is unconstrained. If not specified, defaults to <>
type PreventGradientAttr ¶
PreventGradientAttr is an optional argument to PreventGradient.
type PreventGradientAttr func(optionalAttr)
func PreventGradientMessage ¶
func PreventGradientMessage(value string) PreventGradientAttr
PreventGradientMessage sets the optional message attribute to value.
value: Will be printed in the error when anyone tries to differentiate this operation. If not specified, defaults to ""
type PrintAttr ¶
PrintAttr is an optional argument to Print.
type PrintAttr func(optionalAttr)
func PrintFirstN ¶
func PrintFirstN(value int64) PrintAttr
PrintFirstN sets the optional first_n attribute to value.
value: Only log `first_n` number of times. -1 disables logging. If not specified, defaults to -1
func PrintMessage ¶
func PrintMessage(value string) PrintAttr
PrintMessage sets the optional message attribute to value.
value: A string, prefix of the error message. If not specified, defaults to ""
func PrintSummarize ¶
func PrintSummarize(value int64) PrintAttr
PrintSummarize sets the optional summarize attribute to value.
value: Only print this many entries of each tensor. If not specified, defaults to 3
type PriorityQueueV2Attr ¶
PriorityQueueV2Attr is an optional argument to PriorityQueueV2.
type PriorityQueueV2Attr func(optionalAttr)
func PriorityQueueV2Capacity ¶
func PriorityQueueV2Capacity(value int64) PriorityQueueV2Attr
PriorityQueueV2Capacity sets the optional capacity attribute to value.
value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1
func PriorityQueueV2ComponentTypes ¶
func PriorityQueueV2ComponentTypes(value []tf.DataType) PriorityQueueV2Attr
PriorityQueueV2ComponentTypes sets the optional component_types attribute to value.
value: The type of each component in a value. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func PriorityQueueV2Container ¶
func PriorityQueueV2Container(value string) PriorityQueueV2Attr
PriorityQueueV2Container sets the optional container attribute to value.
value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func PriorityQueueV2SharedName ¶
func PriorityQueueV2SharedName(value string) PriorityQueueV2Attr
PriorityQueueV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""
type ProdAttr ¶
ProdAttr is an optional argument to Prod.
type ProdAttr func(optionalAttr)
func ProdKeepDims ¶
func ProdKeepDims(value bool) ProdAttr
ProdKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type QrAttr ¶
QrAttr is an optional argument to Qr.
type QrAttr func(optionalAttr)
func QrFullMatrices ¶
func QrFullMatrices(value bool) QrAttr
QrFullMatrices sets the optional full_matrices attribute to value.
value: If true, compute full-sized `q` and `r`. If false (the default), compute only the leading `P` columns of `q`. If not specified, defaults to false
type QuantizeAndDequantizeAttr ¶
QuantizeAndDequantizeAttr is an optional argument to QuantizeAndDequantize.
type QuantizeAndDequantizeAttr func(optionalAttr)
func QuantizeAndDequantizeInputMax ¶
func QuantizeAndDequantizeInputMax(value float32) QuantizeAndDequantizeAttr
QuantizeAndDequantizeInputMax sets the optional input_max attribute to value. If not specified, defaults to 0
func QuantizeAndDequantizeInputMin ¶
func QuantizeAndDequantizeInputMin(value float32) QuantizeAndDequantizeAttr
QuantizeAndDequantizeInputMin sets the optional input_min attribute to value. If not specified, defaults to 0
func QuantizeAndDequantizeNumBits ¶
func QuantizeAndDequantizeNumBits(value int64) QuantizeAndDequantizeAttr
QuantizeAndDequantizeNumBits sets the optional num_bits attribute to value. If not specified, defaults to 8
func QuantizeAndDequantizeRangeGiven ¶
func QuantizeAndDequantizeRangeGiven(value bool) QuantizeAndDequantizeAttr
QuantizeAndDequantizeRangeGiven sets the optional range_given attribute to value. If not specified, defaults to false
func QuantizeAndDequantizeSignedInput ¶
func QuantizeAndDequantizeSignedInput(value bool) QuantizeAndDequantizeAttr
QuantizeAndDequantizeSignedInput sets the optional signed_input attribute to value. If not specified, defaults to true
type QuantizeAndDequantizeV2Attr ¶
QuantizeAndDequantizeV2Attr is an optional argument to QuantizeAndDequantizeV2.
type QuantizeAndDequantizeV2Attr func(optionalAttr)
func QuantizeAndDequantizeV2NumBits ¶
func QuantizeAndDequantizeV2NumBits(value int64) QuantizeAndDequantizeV2Attr
QuantizeAndDequantizeV2NumBits sets the optional num_bits attribute to value.
value: The bitwidth of the quantization. If not specified, defaults to 8
func QuantizeAndDequantizeV2RangeGiven ¶
func QuantizeAndDequantizeV2RangeGiven(value bool) QuantizeAndDequantizeV2Attr
QuantizeAndDequantizeV2RangeGiven sets the optional range_given attribute to value.
value: If the range is given or should be computed from the tensor. If not specified, defaults to false
func QuantizeAndDequantizeV2SignedInput ¶
func QuantizeAndDequantizeV2SignedInput(value bool) QuantizeAndDequantizeV2Attr
QuantizeAndDequantizeV2SignedInput sets the optional signed_input attribute to value.
value: If the quantization is signed or unsigned. If not specified, defaults to true
type QuantizeV2Attr ¶
QuantizeV2Attr is an optional argument to QuantizeV2.
type QuantizeV2Attr func(optionalAttr)
func QuantizeV2Mode ¶
func QuantizeV2Mode(value string) QuantizeV2Attr
QuantizeV2Mode sets the optional mode attribute to value. If not specified, defaults to "MIN_COMBINED"
type QuantizedConv2DAttr ¶
QuantizedConv2DAttr is an optional argument to QuantizedConv2D.
type QuantizedConv2DAttr func(optionalAttr)
func QuantizedConv2DOutType ¶
func QuantizedConv2DOutType(value tf.DataType) QuantizedConv2DAttr
QuantizedConv2DOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QINT32
type QuantizedInstanceNormAttr ¶
QuantizedInstanceNormAttr is an optional argument to QuantizedInstanceNorm.
type QuantizedInstanceNormAttr func(optionalAttr)
func QuantizedInstanceNormGivenYMax ¶
func QuantizedInstanceNormGivenYMax(value float32) QuantizedInstanceNormAttr
QuantizedInstanceNormGivenYMax sets the optional given_y_max attribute to value.
value: Output in `y_max` if `output_range_given` is True. If not specified, defaults to 0
func QuantizedInstanceNormGivenYMin ¶
func QuantizedInstanceNormGivenYMin(value float32) QuantizedInstanceNormAttr
QuantizedInstanceNormGivenYMin sets the optional given_y_min attribute to value.
value: Output in `y_min` if `output_range_given` is True. If not specified, defaults to 0
func QuantizedInstanceNormMinSeparation ¶
func QuantizedInstanceNormMinSeparation(value float32) QuantizedInstanceNormAttr
QuantizedInstanceNormMinSeparation sets the optional min_separation attribute to value.
value: Minimum value of `y_max - y_min` If not specified, defaults to 0.001
func QuantizedInstanceNormOutputRangeGiven ¶
func QuantizedInstanceNormOutputRangeGiven(value bool) QuantizedInstanceNormAttr
QuantizedInstanceNormOutputRangeGiven sets the optional output_range_given attribute to value.
value: If True, `given_y_min` and `given_y_min` and `given_y_max` are used as the output range. Otherwise, the implementation computes the output range. If not specified, defaults to false
func QuantizedInstanceNormVarianceEpsilon ¶
func QuantizedInstanceNormVarianceEpsilon(value float32) QuantizedInstanceNormAttr
QuantizedInstanceNormVarianceEpsilon sets the optional variance_epsilon attribute to value.
value: A small float number to avoid dividing by 0. If not specified, defaults to 1e-05
type QuantizedMatMulAttr ¶
QuantizedMatMulAttr is an optional argument to QuantizedMatMul.
type QuantizedMatMulAttr func(optionalAttr)
func QuantizedMatMulTactivation ¶
func QuantizedMatMulTactivation(value tf.DataType) QuantizedMatMulAttr
QuantizedMatMulTactivation sets the optional Tactivation attribute to value.
value: The type of output produced by activation function following this operation. If not specified, defaults to DT_QUINT8
func QuantizedMatMulToutput ¶
func QuantizedMatMulToutput(value tf.DataType) QuantizedMatMulAttr
QuantizedMatMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32
func QuantizedMatMulTransposeA ¶
func QuantizedMatMulTransposeA(value bool) QuantizedMatMulAttr
QuantizedMatMulTransposeA sets the optional transpose_a attribute to value.
value: If true, `a` is transposed before multiplication. If not specified, defaults to false
func QuantizedMatMulTransposeB ¶
func QuantizedMatMulTransposeB(value bool) QuantizedMatMulAttr
QuantizedMatMulTransposeB sets the optional transpose_b attribute to value.
value: If true, `b` is transposed before multiplication. If not specified, defaults to false
type QuantizedMulAttr ¶
QuantizedMulAttr is an optional argument to QuantizedMul.
type QuantizedMulAttr func(optionalAttr)
func QuantizedMulToutput ¶
func QuantizedMulToutput(value tf.DataType) QuantizedMulAttr
QuantizedMulToutput sets the optional Toutput attribute to value. If not specified, defaults to DT_QINT32
type QuantizedRelu6Attr ¶
QuantizedRelu6Attr is an optional argument to QuantizedRelu6.
type QuantizedRelu6Attr func(optionalAttr)
func QuantizedRelu6OutType ¶
func QuantizedRelu6OutType(value tf.DataType) QuantizedRelu6Attr
QuantizedRelu6OutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8
type QuantizedReluAttr ¶
QuantizedReluAttr is an optional argument to QuantizedRelu.
type QuantizedReluAttr func(optionalAttr)
func QuantizedReluOutType ¶
func QuantizedReluOutType(value tf.DataType) QuantizedReluAttr
QuantizedReluOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8
type QuantizedReluXAttr ¶
QuantizedReluXAttr is an optional argument to QuantizedReluX.
type QuantizedReluXAttr func(optionalAttr)
func QuantizedReluXOutType ¶
func QuantizedReluXOutType(value tf.DataType) QuantizedReluXAttr
QuantizedReluXOutType sets the optional out_type attribute to value. If not specified, defaults to DT_QUINT8
type QueueCloseV2Attr ¶
QueueCloseV2Attr is an optional argument to QueueCloseV2.
type QueueCloseV2Attr func(optionalAttr)
func QueueCloseV2CancelPendingEnqueues ¶
func QueueCloseV2CancelPendingEnqueues(value bool) QueueCloseV2Attr
QueueCloseV2CancelPendingEnqueues sets the optional cancel_pending_enqueues attribute to value.
value: If true, all pending enqueue requests that are blocked on the given queue will be cancelled. If not specified, defaults to false
type QueueDequeueManyV2Attr ¶
QueueDequeueManyV2Attr is an optional argument to QueueDequeueManyV2.
type QueueDequeueManyV2Attr func(optionalAttr)
func QueueDequeueManyV2TimeoutMs ¶
func QueueDequeueManyV2TimeoutMs(value int64) QueueDequeueManyV2Attr
QueueDequeueManyV2TimeoutMs sets the optional timeout_ms attribute to value.
value: If the queue has fewer than n elements, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1
type QueueDequeueUpToV2Attr ¶
QueueDequeueUpToV2Attr is an optional argument to QueueDequeueUpToV2.
type QueueDequeueUpToV2Attr func(optionalAttr)
func QueueDequeueUpToV2TimeoutMs ¶
func QueueDequeueUpToV2TimeoutMs(value int64) QueueDequeueUpToV2Attr
QueueDequeueUpToV2TimeoutMs sets the optional timeout_ms attribute to value.
value: If the queue has fewer than n elements, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1
type QueueDequeueV2Attr ¶
QueueDequeueV2Attr is an optional argument to QueueDequeueV2.
type QueueDequeueV2Attr func(optionalAttr)
func QueueDequeueV2TimeoutMs ¶
func QueueDequeueV2TimeoutMs(value int64) QueueDequeueV2Attr
QueueDequeueV2TimeoutMs sets the optional timeout_ms attribute to value.
value: If the queue is empty, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1
type QueueEnqueueManyV2Attr ¶
QueueEnqueueManyV2Attr is an optional argument to QueueEnqueueManyV2.
type QueueEnqueueManyV2Attr func(optionalAttr)
func QueueEnqueueManyV2TimeoutMs ¶
func QueueEnqueueManyV2TimeoutMs(value int64) QueueEnqueueManyV2Attr
QueueEnqueueManyV2TimeoutMs sets the optional timeout_ms attribute to value.
value: If the queue is too full, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1
type QueueEnqueueV2Attr ¶
QueueEnqueueV2Attr is an optional argument to QueueEnqueueV2.
type QueueEnqueueV2Attr func(optionalAttr)
func QueueEnqueueV2TimeoutMs ¶
func QueueEnqueueV2TimeoutMs(value int64) QueueEnqueueV2Attr
QueueEnqueueV2TimeoutMs sets the optional timeout_ms attribute to value.
value: If the queue is full, this operation will block for up to timeout_ms milliseconds. Note: This option is not supported yet. If not specified, defaults to -1
type RandomCropAttr ¶
RandomCropAttr is an optional argument to RandomCrop.
type RandomCropAttr func(optionalAttr)
func RandomCropSeed ¶
func RandomCropSeed(value int64) RandomCropAttr
RandomCropSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomCropSeed2 ¶
func RandomCropSeed2(value int64) RandomCropAttr
RandomCropSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type RandomGammaAttr ¶
RandomGammaAttr is an optional argument to RandomGamma.
type RandomGammaAttr func(optionalAttr)
func RandomGammaSeed ¶
func RandomGammaSeed(value int64) RandomGammaAttr
RandomGammaSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomGammaSeed2 ¶
func RandomGammaSeed2(value int64) RandomGammaAttr
RandomGammaSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RandomPoissonAttr ¶
RandomPoissonAttr is an optional argument to RandomPoisson.
type RandomPoissonAttr func(optionalAttr)
func RandomPoissonSeed ¶
func RandomPoissonSeed(value int64) RandomPoissonAttr
RandomPoissonSeed 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 RandomPoissonSeed2 ¶
func RandomPoissonSeed2(value int64) RandomPoissonAttr
RandomPoissonSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RandomShuffleAttr ¶
RandomShuffleAttr is an optional argument to RandomShuffle.
type RandomShuffleAttr func(optionalAttr)
func RandomShuffleSeed ¶
func RandomShuffleSeed(value int64) RandomShuffleAttr
RandomShuffleSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomShuffleSeed2 ¶
func RandomShuffleSeed2(value int64) RandomShuffleAttr
RandomShuffleSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RandomShuffleQueueV2Attr ¶
RandomShuffleQueueV2Attr is an optional argument to RandomShuffleQueueV2.
type RandomShuffleQueueV2Attr func(optionalAttr)
func RandomShuffleQueueV2Capacity ¶
func RandomShuffleQueueV2Capacity(value int64) RandomShuffleQueueV2Attr
RandomShuffleQueueV2Capacity sets the optional capacity attribute to value.
value: The upper bound on the number of elements in this queue. Negative numbers mean no limit. If not specified, defaults to -1
func RandomShuffleQueueV2Container ¶
func RandomShuffleQueueV2Container(value string) RandomShuffleQueueV2Attr
RandomShuffleQueueV2Container sets the optional container attribute to value.
value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func RandomShuffleQueueV2MinAfterDequeue ¶
func RandomShuffleQueueV2MinAfterDequeue(value int64) RandomShuffleQueueV2Attr
RandomShuffleQueueV2MinAfterDequeue sets the optional min_after_dequeue attribute to value.
value: Dequeue will block unless there would be this many elements after the dequeue or the queue is closed. This ensures a minimum level of mixing of elements. If not specified, defaults to 0
func RandomShuffleQueueV2Seed ¶
func RandomShuffleQueueV2Seed(value int64) RandomShuffleQueueV2Attr
RandomShuffleQueueV2Seed sets the optional seed attribute to value.
value: If either seed or seed2 is set to be non-zero, the random number generator is seeded by the given seed. Otherwise, a random seed is used. If not specified, defaults to 0
func RandomShuffleQueueV2Seed2 ¶
func RandomShuffleQueueV2Seed2(value int64) RandomShuffleQueueV2Attr
RandomShuffleQueueV2Seed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
func RandomShuffleQueueV2Shapes ¶
func RandomShuffleQueueV2Shapes(value []tf.Shape) RandomShuffleQueueV2Attr
RandomShuffleQueueV2Shapes sets the optional shapes attribute to value.
value: The shape of each component in a value. The length of this attr must be either 0 or the same as the length of component_types. If the length of this attr is 0, the shapes of queue elements are not constrained, and only one element may be dequeued at a time. If not specified, defaults to <>
REQUIRES: len(value) >= 0
func RandomShuffleQueueV2SharedName ¶
func RandomShuffleQueueV2SharedName(value string) RandomShuffleQueueV2Attr
RandomShuffleQueueV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this queue will be shared under the given name across multiple sessions. If not specified, defaults to ""
type RandomStandardNormalAttr ¶
RandomStandardNormalAttr is an optional argument to RandomStandardNormal.
type RandomStandardNormalAttr func(optionalAttr)
func RandomStandardNormalSeed ¶
func RandomStandardNormalSeed(value int64) RandomStandardNormalAttr
RandomStandardNormalSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomStandardNormalSeed2 ¶
func RandomStandardNormalSeed2(value int64) RandomStandardNormalAttr
RandomStandardNormalSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RandomUniformAttr ¶
RandomUniformAttr is an optional argument to RandomUniform.
type RandomUniformAttr func(optionalAttr)
func RandomUniformSeed ¶
func RandomUniformSeed(value int64) RandomUniformAttr
RandomUniformSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomUniformSeed2 ¶
func RandomUniformSeed2(value int64) RandomUniformAttr
RandomUniformSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RandomUniformIntAttr ¶
RandomUniformIntAttr is an optional argument to RandomUniformInt.
type RandomUniformIntAttr func(optionalAttr)
func RandomUniformIntSeed ¶
func RandomUniformIntSeed(value int64) RandomUniformIntAttr
RandomUniformIntSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func RandomUniformIntSeed2 ¶
func RandomUniformIntSeed2(value int64) RandomUniformIntAttr
RandomUniformIntSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type RealAttr ¶
RealAttr is an optional argument to Real.
type RealAttr func(optionalAttr)
func RealTout ¶
func RealTout(value tf.DataType) RealAttr
RealTout sets the optional Tout attribute to value. If not specified, defaults to DT_FLOAT
type RecordInputAttr ¶
RecordInputAttr is an optional argument to RecordInput.
type RecordInputAttr func(optionalAttr)
func RecordInputBatchSize ¶
func RecordInputBatchSize(value int64) RecordInputAttr
RecordInputBatchSize sets the optional batch_size attribute to value.
value: The batch size. If not specified, defaults to 32
func RecordInputFileBufferSize ¶
func RecordInputFileBufferSize(value int64) RecordInputAttr
RecordInputFileBufferSize sets the optional file_buffer_size attribute to value.
value: The randomization shuffling buffer. If not specified, defaults to 10000
func RecordInputFileParallelism ¶
func RecordInputFileParallelism(value int64) RecordInputAttr
RecordInputFileParallelism sets the optional file_parallelism attribute to value.
value: How many sstables are opened and concurrently iterated over. If not specified, defaults to 16
func RecordInputFileRandomSeed ¶
func RecordInputFileRandomSeed(value int64) RecordInputAttr
RecordInputFileRandomSeed sets the optional file_random_seed attribute to value.
value: Random seeds used to produce randomized records. If not specified, defaults to 301
func RecordInputFileShuffleShiftRatio ¶
func RecordInputFileShuffleShiftRatio(value float32) RecordInputAttr
RecordInputFileShuffleShiftRatio sets the optional file_shuffle_shift_ratio attribute to value.
value: Shifts the list of files after the list is randomly shuffled. If not specified, defaults to 0
type ReduceJoinAttr ¶
ReduceJoinAttr is an optional argument to ReduceJoin.
type ReduceJoinAttr func(optionalAttr)
func ReduceJoinKeepDims ¶
func ReduceJoinKeepDims(value bool) ReduceJoinAttr
ReduceJoinKeepDims sets the optional keep_dims attribute to value.
value: If `True`, retain reduced dimensions with length `1`. If not specified, defaults to false
func ReduceJoinSeparator ¶
func ReduceJoinSeparator(value string) ReduceJoinAttr
ReduceJoinSeparator sets the optional separator attribute to value.
value: The separator to use when joining. If not specified, defaults to ""
type ResizeAreaAttr ¶
ResizeAreaAttr is an optional argument to ResizeArea.
type ResizeAreaAttr func(optionalAttr)
func ResizeAreaAlignCorners ¶
func ResizeAreaAlignCorners(value bool) ResizeAreaAttr
ResizeAreaAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale input by (new_height - 1) / (height - 1), which exactly aligns the 4 corners of images and resized images. If false, rescale by new_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResizeBicubicAttr ¶
ResizeBicubicAttr is an optional argument to ResizeBicubic.
type ResizeBicubicAttr func(optionalAttr)
func ResizeBicubicAlignCorners ¶
func ResizeBicubicAlignCorners(value bool) ResizeBicubicAttr
ResizeBicubicAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale input by (new_height - 1) / (height - 1), which exactly aligns the 4 corners of images and resized images. If false, rescale by new_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResizeBilinearAttr ¶
ResizeBilinearAttr is an optional argument to ResizeBilinear.
type ResizeBilinearAttr func(optionalAttr)
func ResizeBilinearAlignCorners ¶
func ResizeBilinearAlignCorners(value bool) ResizeBilinearAttr
ResizeBilinearAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale input by (new_height - 1) / (height - 1), which exactly aligns the 4 corners of images and resized images. If false, rescale by new_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResizeBilinearGradAttr ¶
ResizeBilinearGradAttr is an optional argument to ResizeBilinearGrad.
type ResizeBilinearGradAttr func(optionalAttr)
func ResizeBilinearGradAlignCorners ¶
func ResizeBilinearGradAlignCorners(value bool) ResizeBilinearGradAttr
ResizeBilinearGradAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale grads by (orig_height - 1) / (height - 1), which exactly aligns the 4 corners of grads and original_image. If false, rescale by orig_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResizeNearestNeighborAttr ¶
ResizeNearestNeighborAttr is an optional argument to ResizeNearestNeighbor.
type ResizeNearestNeighborAttr func(optionalAttr)
func ResizeNearestNeighborAlignCorners ¶
func ResizeNearestNeighborAlignCorners(value bool) ResizeNearestNeighborAttr
ResizeNearestNeighborAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale input by (new_height - 1) / (height - 1), which exactly aligns the 4 corners of images and resized images. If false, rescale by new_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResizeNearestNeighborGradAttr ¶
ResizeNearestNeighborGradAttr is an optional argument to ResizeNearestNeighborGrad.
type ResizeNearestNeighborGradAttr func(optionalAttr)
func ResizeNearestNeighborGradAlignCorners ¶
func ResizeNearestNeighborGradAlignCorners(value bool) ResizeNearestNeighborGradAttr
ResizeNearestNeighborGradAlignCorners sets the optional align_corners attribute to value.
value: If true, rescale grads by (orig_height - 1) / (height - 1), which exactly aligns the 4 corners of grads and original_image. If false, rescale by orig_height / height. Treat similarly the width dimension. If not specified, defaults to false
type ResourceApplyAdadeltaAttr ¶
ResourceApplyAdadeltaAttr is an optional argument to ResourceApplyAdadelta.
type ResourceApplyAdadeltaAttr func(optionalAttr)
func ResourceApplyAdadeltaUseLocking ¶
func ResourceApplyAdadeltaUseLocking(value bool) ResourceApplyAdadeltaAttr
ResourceApplyAdadeltaUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var, accum and update_accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyAdagradAttr ¶
ResourceApplyAdagradAttr is an optional argument to ResourceApplyAdagrad.
type ResourceApplyAdagradAttr func(optionalAttr)
func ResourceApplyAdagradUseLocking ¶
func ResourceApplyAdagradUseLocking(value bool) ResourceApplyAdagradAttr
ResourceApplyAdagradUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyAdagradDAAttr ¶
ResourceApplyAdagradDAAttr is an optional argument to ResourceApplyAdagradDA.
type ResourceApplyAdagradDAAttr func(optionalAttr)
func ResourceApplyAdagradDAUseLocking ¶
func ResourceApplyAdagradDAUseLocking(value bool) ResourceApplyAdagradDAAttr
ResourceApplyAdagradDAUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyAdamAttr ¶
ResourceApplyAdamAttr is an optional argument to ResourceApplyAdam.
type ResourceApplyAdamAttr func(optionalAttr)
func ResourceApplyAdamUseLocking ¶
func ResourceApplyAdamUseLocking(value bool) ResourceApplyAdamAttr
ResourceApplyAdamUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var, m, and v tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyCenteredRMSPropAttr ¶
ResourceApplyCenteredRMSPropAttr is an optional argument to ResourceApplyCenteredRMSProp.
type ResourceApplyCenteredRMSPropAttr func(optionalAttr)
func ResourceApplyCenteredRMSPropUseLocking ¶
func ResourceApplyCenteredRMSPropUseLocking(value bool) ResourceApplyCenteredRMSPropAttr
ResourceApplyCenteredRMSPropUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var, mg, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyFtrlAttr ¶
ResourceApplyFtrlAttr is an optional argument to ResourceApplyFtrl.
type ResourceApplyFtrlAttr func(optionalAttr)
func ResourceApplyFtrlUseLocking ¶
func ResourceApplyFtrlUseLocking(value bool) ResourceApplyFtrlAttr
ResourceApplyFtrlUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyGradientDescentAttr ¶
ResourceApplyGradientDescentAttr is an optional argument to ResourceApplyGradientDescent.
type ResourceApplyGradientDescentAttr func(optionalAttr)
func ResourceApplyGradientDescentUseLocking ¶
func ResourceApplyGradientDescentUseLocking(value bool) ResourceApplyGradientDescentAttr
ResourceApplyGradientDescentUseLocking sets the optional use_locking attribute to value.
value: If `True`, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyMomentumAttr ¶
ResourceApplyMomentumAttr is an optional argument to ResourceApplyMomentum.
type ResourceApplyMomentumAttr func(optionalAttr)
func ResourceApplyMomentumUseLocking ¶
func ResourceApplyMomentumUseLocking(value bool) ResourceApplyMomentumAttr
ResourceApplyMomentumUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
func ResourceApplyMomentumUseNesterov ¶
func ResourceApplyMomentumUseNesterov(value bool) ResourceApplyMomentumAttr
ResourceApplyMomentumUseNesterov sets the optional use_nesterov attribute to value.
value: If `True`, the tensor passed to compute grad will be var - lr * momentum * accum, so in the end, the var you get is actually var - lr * momentum * accum. If not specified, defaults to false
type ResourceApplyProximalAdagradAttr ¶
ResourceApplyProximalAdagradAttr is an optional argument to ResourceApplyProximalAdagrad.
type ResourceApplyProximalAdagradAttr func(optionalAttr)
func ResourceApplyProximalAdagradUseLocking ¶
func ResourceApplyProximalAdagradUseLocking(value bool) ResourceApplyProximalAdagradAttr
ResourceApplyProximalAdagradUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyProximalGradientDescentAttr ¶
ResourceApplyProximalGradientDescentAttr is an optional argument to ResourceApplyProximalGradientDescent.
type ResourceApplyProximalGradientDescentAttr func(optionalAttr)
func ResourceApplyProximalGradientDescentUseLocking ¶
func ResourceApplyProximalGradientDescentUseLocking(value bool) ResourceApplyProximalGradientDescentAttr
ResourceApplyProximalGradientDescentUseLocking sets the optional use_locking attribute to value.
value: If True, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceApplyRMSPropAttr ¶
ResourceApplyRMSPropAttr is an optional argument to ResourceApplyRMSProp.
type ResourceApplyRMSPropAttr func(optionalAttr)
func ResourceApplyRMSPropUseLocking ¶
func ResourceApplyRMSPropUseLocking(value bool) ResourceApplyRMSPropAttr
ResourceApplyRMSPropUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceGatherAttr ¶
ResourceGatherAttr is an optional argument to ResourceGather.
type ResourceGatherAttr func(optionalAttr)
func ResourceGatherValidateIndices ¶
func ResourceGatherValidateIndices(value bool) ResourceGatherAttr
ResourceGatherValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type ResourceSparseApplyAdadeltaAttr ¶
ResourceSparseApplyAdadeltaAttr is an optional argument to ResourceSparseApplyAdadelta.
type ResourceSparseApplyAdadeltaAttr func(optionalAttr)
func ResourceSparseApplyAdadeltaUseLocking ¶
func ResourceSparseApplyAdadeltaUseLocking(value bool) ResourceSparseApplyAdadeltaAttr
ResourceSparseApplyAdadeltaUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyAdagradAttr ¶
ResourceSparseApplyAdagradAttr is an optional argument to ResourceSparseApplyAdagrad.
type ResourceSparseApplyAdagradAttr func(optionalAttr)
func ResourceSparseApplyAdagradUseLocking ¶
func ResourceSparseApplyAdagradUseLocking(value bool) ResourceSparseApplyAdagradAttr
ResourceSparseApplyAdagradUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyAdagradDAAttr ¶
ResourceSparseApplyAdagradDAAttr is an optional argument to ResourceSparseApplyAdagradDA.
type ResourceSparseApplyAdagradDAAttr func(optionalAttr)
func ResourceSparseApplyAdagradDAUseLocking ¶
func ResourceSparseApplyAdagradDAUseLocking(value bool) ResourceSparseApplyAdagradDAAttr
ResourceSparseApplyAdagradDAUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyCenteredRMSPropAttr ¶
ResourceSparseApplyCenteredRMSPropAttr is an optional argument to ResourceSparseApplyCenteredRMSProp.
type ResourceSparseApplyCenteredRMSPropAttr func(optionalAttr)
func ResourceSparseApplyCenteredRMSPropUseLocking ¶
func ResourceSparseApplyCenteredRMSPropUseLocking(value bool) ResourceSparseApplyCenteredRMSPropAttr
ResourceSparseApplyCenteredRMSPropUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var, mg, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyFtrlAttr ¶
ResourceSparseApplyFtrlAttr is an optional argument to ResourceSparseApplyFtrl.
type ResourceSparseApplyFtrlAttr func(optionalAttr)
func ResourceSparseApplyFtrlUseLocking ¶
func ResourceSparseApplyFtrlUseLocking(value bool) ResourceSparseApplyFtrlAttr
ResourceSparseApplyFtrlUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyMomentumAttr ¶
ResourceSparseApplyMomentumAttr is an optional argument to ResourceSparseApplyMomentum.
type ResourceSparseApplyMomentumAttr func(optionalAttr)
func ResourceSparseApplyMomentumUseLocking ¶
func ResourceSparseApplyMomentumUseLocking(value bool) ResourceSparseApplyMomentumAttr
ResourceSparseApplyMomentumUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
func ResourceSparseApplyMomentumUseNesterov ¶
func ResourceSparseApplyMomentumUseNesterov(value bool) ResourceSparseApplyMomentumAttr
ResourceSparseApplyMomentumUseNesterov sets the optional use_nesterov attribute to value.
value: If `True`, the tensor passed to compute grad will be var - lr * momentum * accum, so in the end, the var you get is actually var - lr * momentum * accum. If not specified, defaults to false
type ResourceSparseApplyProximalAdagradAttr ¶
ResourceSparseApplyProximalAdagradAttr is an optional argument to ResourceSparseApplyProximalAdagrad.
type ResourceSparseApplyProximalAdagradAttr func(optionalAttr)
func ResourceSparseApplyProximalAdagradUseLocking ¶
func ResourceSparseApplyProximalAdagradUseLocking(value bool) ResourceSparseApplyProximalAdagradAttr
ResourceSparseApplyProximalAdagradUseLocking sets the optional use_locking attribute to value.
value: If True, updating of the var and accum tensors will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyProximalGradientDescentAttr ¶
ResourceSparseApplyProximalGradientDescentAttr is an optional argument to ResourceSparseApplyProximalGradientDescent.
type ResourceSparseApplyProximalGradientDescentAttr func(optionalAttr)
func ResourceSparseApplyProximalGradientDescentUseLocking ¶
func ResourceSparseApplyProximalGradientDescentUseLocking(value bool) ResourceSparseApplyProximalGradientDescentAttr
ResourceSparseApplyProximalGradientDescentUseLocking sets the optional use_locking attribute to value.
value: If True, the subtraction will be protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type ResourceSparseApplyRMSPropAttr ¶
ResourceSparseApplyRMSPropAttr is an optional argument to ResourceSparseApplyRMSProp.
type ResourceSparseApplyRMSPropAttr func(optionalAttr)
func ResourceSparseApplyRMSPropUseLocking ¶
func ResourceSparseApplyRMSPropUseLocking(value bool) ResourceSparseApplyRMSPropAttr
ResourceSparseApplyRMSPropUseLocking sets the optional use_locking attribute to value.
value: If `True`, updating of the var, ms, and mom tensors is protected by a lock; otherwise the behavior is undefined, but may exhibit less contention. If not specified, defaults to false
type RestoreAttr ¶
RestoreAttr is an optional argument to Restore.
type RestoreAttr func(optionalAttr)
func RestorePreferredShard ¶
func RestorePreferredShard(value int64) RestoreAttr
RestorePreferredShard sets the optional preferred_shard attribute to value.
value: Index of file to open first if multiple files match `file_pattern`. If not specified, defaults to -1
type RestoreSliceAttr ¶
RestoreSliceAttr is an optional argument to RestoreSlice.
type RestoreSliceAttr func(optionalAttr)
func RestoreSlicePreferredShard ¶
func RestoreSlicePreferredShard(value int64) RestoreSliceAttr
RestoreSlicePreferredShard sets the optional preferred_shard attribute to value.
value: Index of file to open first if multiple files match `file_pattern`. See the documentation for `Restore`. If not specified, defaults to -1
type ReverseSequenceAttr ¶
ReverseSequenceAttr is an optional argument to ReverseSequence.
type ReverseSequenceAttr func(optionalAttr)
func ReverseSequenceBatchDim ¶
func ReverseSequenceBatchDim(value int64) ReverseSequenceAttr
ReverseSequenceBatchDim sets the optional batch_dim attribute to value.
value: The dimension along which reversal is performed. If not specified, defaults to 0
type SampleDistortedBoundingBoxAttr ¶
SampleDistortedBoundingBoxAttr is an optional argument to SampleDistortedBoundingBox.
type SampleDistortedBoundingBoxAttr func(optionalAttr)
func SampleDistortedBoundingBoxAreaRange ¶
func SampleDistortedBoundingBoxAreaRange(value []float32) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxAreaRange sets the optional area_range attribute to value.
value: The cropped area of the image must contain a fraction of the supplied image within in this range. If not specified, defaults to <f:0.05 f:1 >
func SampleDistortedBoundingBoxAspectRatioRange ¶
func SampleDistortedBoundingBoxAspectRatioRange(value []float32) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxAspectRatioRange sets the optional aspect_ratio_range attribute to value.
value: The cropped area of the image must have an aspect ratio = width / height within this range. If not specified, defaults to <f:0.75 f:1.33 >
func SampleDistortedBoundingBoxMaxAttempts ¶
func SampleDistortedBoundingBoxMaxAttempts(value int64) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxMaxAttempts sets the optional max_attempts attribute to value.
value: Number of attempts at generating a cropped region of the image of the specified constraints. After `max_attempts` failures, return the entire image. If not specified, defaults to 100
func SampleDistortedBoundingBoxMinObjectCovered ¶
func SampleDistortedBoundingBoxMinObjectCovered(value float32) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxMinObjectCovered sets the optional min_object_covered attribute to value.
value: The cropped area of the image must contain at least this fraction of any bounding box supplied. The value of this parameter should be non-negative. In the case of 0, the cropped area does not need to overlap any of the bounding boxes supplied. If not specified, defaults to 0.1
func SampleDistortedBoundingBoxSeed ¶
func SampleDistortedBoundingBoxSeed(value int64) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to non-zero, the random number generator is seeded by the given `seed`. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func SampleDistortedBoundingBoxSeed2 ¶
func SampleDistortedBoundingBoxSeed2(value int64) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
func SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes ¶
func SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes(value bool) SampleDistortedBoundingBoxAttr
SampleDistortedBoundingBoxUseImageIfNoBoundingBoxes sets the optional use_image_if_no_bounding_boxes attribute to value.
value: Controls behavior if no bounding boxes supplied. If true, assume an implicit bounding box covering the whole input. If false, raise an error. If not specified, defaults to false
type Scope ¶
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.
type Scope struct {
// contains filtered or unexported fields
}
func NewScope ¶
func NewScope() *Scope
NewScope creates a Scope initialized with an empty Graph.
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
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.
type SdcaOptimizerAttr ¶
SdcaOptimizerAttr is an optional argument to SdcaOptimizer.
type SdcaOptimizerAttr func(optionalAttr)
func SdcaOptimizerAdaptative ¶
func SdcaOptimizerAdaptative(value bool) SdcaOptimizerAttr
SdcaOptimizerAdaptative sets the optional adaptative attribute to value.
value: Whether to use Adapative SDCA for the inner loop. If not specified, defaults to false
type SelfAdjointEigV2Attr ¶
SelfAdjointEigV2Attr is an optional argument to SelfAdjointEigV2.
type SelfAdjointEigV2Attr func(optionalAttr)
func SelfAdjointEigV2ComputeV ¶
func SelfAdjointEigV2ComputeV(value bool) SelfAdjointEigV2Attr
SelfAdjointEigV2ComputeV sets the optional compute_v attribute to value.
value: If `True` then eigenvectors will be computed and returned in `v`. Otherwise, only the eigenvalues will be computed. If not specified, defaults to true
type SetSizeAttr ¶
SetSizeAttr is an optional argument to SetSize.
type SetSizeAttr func(optionalAttr)
func SetSizeValidateIndices ¶
func SetSizeValidateIndices(value bool) SetSizeAttr
SetSizeValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type ShapeAttr ¶
ShapeAttr is an optional argument to Shape.
type ShapeAttr func(optionalAttr)
func ShapeOutType ¶
func ShapeOutType(value tf.DataType) ShapeAttr
ShapeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32
type ShapeNAttr ¶
ShapeNAttr is an optional argument to ShapeN.
type ShapeNAttr func(optionalAttr)
func ShapeNOutType ¶
func ShapeNOutType(value tf.DataType) ShapeNAttr
ShapeNOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32
type SizeAttr ¶
SizeAttr is an optional argument to Size.
type SizeAttr func(optionalAttr)
func SizeOutType ¶
func SizeOutType(value tf.DataType) SizeAttr
SizeOutType sets the optional out_type attribute to value. If not specified, defaults to DT_INT32
type SkipgramAttr ¶
SkipgramAttr is an optional argument to Skipgram.
type SkipgramAttr func(optionalAttr)
func SkipgramMinCount ¶
func SkipgramMinCount(value int64) SkipgramAttr
SkipgramMinCount sets the optional min_count attribute to value.
value: The minimum number of word occurrences for it to be included in the vocabulary. If not specified, defaults to 5
func SkipgramSubsample ¶
func SkipgramSubsample(value float32) SkipgramAttr
SkipgramSubsample sets the optional subsample attribute to value.
value: Threshold for word occurrence. Words that appear with higher frequency will be randomly down-sampled. Set to 0 to disable. If not specified, defaults to 0.001
func SkipgramWindowSize ¶
func SkipgramWindowSize(value int64) SkipgramAttr
SkipgramWindowSize sets the optional window_size attribute to value.
value: The number of words to predict to the left and right of the target. If not specified, defaults to 5
type SparseMatMulAttr ¶
SparseMatMulAttr is an optional argument to SparseMatMul.
type SparseMatMulAttr func(optionalAttr)
func SparseMatMulAIsSparse ¶
func SparseMatMulAIsSparse(value bool) SparseMatMulAttr
SparseMatMulAIsSparse sets the optional a_is_sparse attribute to value. If not specified, defaults to false
func SparseMatMulBIsSparse ¶
func SparseMatMulBIsSparse(value bool) SparseMatMulAttr
SparseMatMulBIsSparse sets the optional b_is_sparse attribute to value. If not specified, defaults to false
func SparseMatMulTransposeA ¶
func SparseMatMulTransposeA(value bool) SparseMatMulAttr
SparseMatMulTransposeA sets the optional transpose_a attribute to value. If not specified, defaults to false
func SparseMatMulTransposeB ¶
func SparseMatMulTransposeB(value bool) SparseMatMulAttr
SparseMatMulTransposeB sets the optional transpose_b attribute to value. If not specified, defaults to false
type SparseReduceSumAttr ¶
SparseReduceSumAttr is an optional argument to SparseReduceSum.
type SparseReduceSumAttr func(optionalAttr)
func SparseReduceSumKeepDims ¶
func SparseReduceSumKeepDims(value bool) SparseReduceSumAttr
SparseReduceSumKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type SparseReduceSumSparseAttr ¶
SparseReduceSumSparseAttr is an optional argument to SparseReduceSumSparse.
type SparseReduceSumSparseAttr func(optionalAttr)
func SparseReduceSumSparseKeepDims ¶
func SparseReduceSumSparseKeepDims(value bool) SparseReduceSumSparseAttr
SparseReduceSumSparseKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type SparseTensorDenseMatMulAttr ¶
SparseTensorDenseMatMulAttr is an optional argument to SparseTensorDenseMatMul.
type SparseTensorDenseMatMulAttr func(optionalAttr)
func SparseTensorDenseMatMulAdjointA ¶
func SparseTensorDenseMatMulAdjointA(value bool) SparseTensorDenseMatMulAttr
SparseTensorDenseMatMulAdjointA sets the optional adjoint_a attribute to value.
value: Use the adjoint of A in the matrix multiply. If A is complex, this is transpose(conj(A)). Otherwise it's transpose(A). If not specified, defaults to false
func SparseTensorDenseMatMulAdjointB ¶
func SparseTensorDenseMatMulAdjointB(value bool) SparseTensorDenseMatMulAttr
SparseTensorDenseMatMulAdjointB sets the optional adjoint_b attribute to value.
value: Use the adjoint of B in the matrix multiply. If B is complex, this is transpose(conj(B)). Otherwise it's transpose(B). If not specified, defaults to false
type SparseToDenseAttr ¶
SparseToDenseAttr is an optional argument to SparseToDense.
type SparseToDenseAttr func(optionalAttr)
func SparseToDenseValidateIndices ¶
func SparseToDenseValidateIndices(value bool) SparseToDenseAttr
SparseToDenseValidateIndices sets the optional validate_indices attribute to value.
value: If true, indices are checked to make sure they are sorted in lexicographic order and that there are no repeats. If not specified, defaults to true
type SparseToSparseSetOperationAttr ¶
SparseToSparseSetOperationAttr is an optional argument to SparseToSparseSetOperation.
type SparseToSparseSetOperationAttr func(optionalAttr)
func SparseToSparseSetOperationValidateIndices ¶
func SparseToSparseSetOperationValidateIndices(value bool) SparseToSparseSetOperationAttr
SparseToSparseSetOperationValidateIndices sets the optional validate_indices attribute to value. If not specified, defaults to true
type SqueezeAttr ¶
SqueezeAttr is an optional argument to Squeeze.
type SqueezeAttr func(optionalAttr)
func SqueezeSqueezeDims ¶
func SqueezeSqueezeDims(value []int64) SqueezeAttr
SqueezeSqueezeDims sets the optional squeeze_dims 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. If not specified, defaults to <>
REQUIRES: len(value) >= 0
type StageAttr ¶
StageAttr is an optional argument to Stage.
type StageAttr func(optionalAttr)
func StageContainer ¶
func StageContainer(value string) StageAttr
StageContainer sets the optional container attribute to value.
value: If non-empty, this queue is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func StageSharedName ¶
func StageSharedName(value string) StageAttr
StageSharedName sets the optional shared_name attribute to value.
value: It is necessary to match this name to the matching Unstage Op. If not specified, defaults to ""
type StridedSliceAttr ¶
StridedSliceAttr is an optional argument to StridedSlice.
type StridedSliceAttr func(optionalAttr)
func StridedSliceBeginMask ¶
func StridedSliceBeginMask(value int64) StridedSliceAttr
StridedSliceBeginMask sets the optional begin_mask attribute to value.
value: a bitmask where a bit i being 1 means to ignore the begin value and instead use the largest interval possible. At runtime begin[i] will be replaced with `[0, n-1) if `stride[i] > 0` or `[-1, n-1]` if `stride[i] < 0` If not specified, defaults to 0
func StridedSliceEllipsisMask ¶
func StridedSliceEllipsisMask(value int64) StridedSliceAttr
StridedSliceEllipsisMask sets the optional ellipsis_mask attribute to value.
value: a bitmask where bit `i` being 1 means the `i`th position is actually an ellipsis. One bit at most can be 1. If `ellipsis_mask == 0`, then an implicit ellipsis mask of `1 << (m+1)` is provided. This means that `foo[3:5] == foo[3:5, ...]`. An ellipsis implicitly creates as many range specifications as necessary to fully specify the sliced range for every dimension. For example for a 4-dimensional tensor `foo` the slice `foo[2, ..., 5:8]` implies `foo[2, :, :, 5:8]`. If not specified, defaults to 0
func StridedSliceEndMask ¶
func StridedSliceEndMask(value int64) StridedSliceAttr
StridedSliceEndMask sets the optional end_mask attribute to value.
value: analogous to `begin_mask` If not specified, defaults to 0
func StridedSliceNewAxisMask ¶
func StridedSliceNewAxisMask(value int64) StridedSliceAttr
StridedSliceNewAxisMask sets the optional new_axis_mask attribute to value.
value: a bitmask where bit `i` being 1 means the `i`th specification creates a new shape 1 dimension. For example `foo[:4, tf.newaxis, :2]` would produce a shape `(4, 1, 2)` tensor. If not specified, defaults to 0
func StridedSliceShrinkAxisMask ¶
func StridedSliceShrinkAxisMask(value int64) StridedSliceAttr
StridedSliceShrinkAxisMask sets the optional shrink_axis_mask attribute to value.
value: a bitmask where bit `i` implies that the `i`th specification should shrink the dimensionality. begin and end must imply a slice of size 1 in the dimension. For example in python one might do `foo[:, 3, :]` which would result in `shrink_axis_mask` being 2. If not specified, defaults to 0
type StridedSliceGradAttr ¶
StridedSliceGradAttr is an optional argument to StridedSliceGrad.
type StridedSliceGradAttr func(optionalAttr)
func StridedSliceGradBeginMask ¶
func StridedSliceGradBeginMask(value int64) StridedSliceGradAttr
StridedSliceGradBeginMask sets the optional begin_mask attribute to value. If not specified, defaults to 0
func StridedSliceGradEllipsisMask ¶
func StridedSliceGradEllipsisMask(value int64) StridedSliceGradAttr
StridedSliceGradEllipsisMask sets the optional ellipsis_mask attribute to value. If not specified, defaults to 0
func StridedSliceGradEndMask ¶
func StridedSliceGradEndMask(value int64) StridedSliceGradAttr
StridedSliceGradEndMask sets the optional end_mask attribute to value. If not specified, defaults to 0
func StridedSliceGradNewAxisMask ¶
func StridedSliceGradNewAxisMask(value int64) StridedSliceGradAttr
StridedSliceGradNewAxisMask sets the optional new_axis_mask attribute to value. If not specified, defaults to 0
func StridedSliceGradShrinkAxisMask ¶
func StridedSliceGradShrinkAxisMask(value int64) StridedSliceGradAttr
StridedSliceGradShrinkAxisMask sets the optional shrink_axis_mask attribute to value. If not specified, defaults to 0
type StringJoinAttr ¶
StringJoinAttr is an optional argument to StringJoin.
type StringJoinAttr func(optionalAttr)
func StringJoinSeparator ¶
func StringJoinSeparator(value string) StringJoinAttr
StringJoinSeparator sets the optional separator attribute to value.
value: string, an optional join separator. If not specified, defaults to ""
type StringToNumberAttr ¶
StringToNumberAttr is an optional argument to StringToNumber.
type StringToNumberAttr func(optionalAttr)
func StringToNumberOutType ¶
func StringToNumberOutType(value tf.DataType) StringToNumberAttr
StringToNumberOutType sets the optional out_type attribute to value.
value: The numeric type to interpret each string in `string_tensor` as. If not specified, defaults to DT_FLOAT
type SumAttr ¶
SumAttr is an optional argument to Sum.
type SumAttr func(optionalAttr)
func SumKeepDims ¶
func SumKeepDims(value bool) SumAttr
SumKeepDims sets the optional keep_dims attribute to value.
value: If true, retain reduced dimensions with length 1. If not specified, defaults to false
type SvdAttr ¶
SvdAttr is an optional argument to Svd.
type SvdAttr func(optionalAttr)
func SvdComputeUv ¶
func SvdComputeUv(value bool) SvdAttr
SvdComputeUv sets the optional compute_uv attribute to value.
value: If true, left and right singular vectors will be computed and returned in `u` and `v`, respectively. If false, `u` and `v` are not set and should never referenced. If not specified, defaults to true
func SvdFullMatrices ¶
func SvdFullMatrices(value bool) SvdAttr
SvdFullMatrices sets the optional full_matrices attribute to value.
value: If true, compute full-sized `u` and `v`. If false (the default), compute only the leading `P` singular vectors. Ignored if `compute_uv` is `False`. If not specified, defaults to false
type TFRecordReaderV2Attr ¶
TFRecordReaderV2Attr is an optional argument to TFRecordReaderV2.
type TFRecordReaderV2Attr func(optionalAttr)
func TFRecordReaderV2CompressionType ¶
func TFRecordReaderV2CompressionType(value string) TFRecordReaderV2Attr
TFRecordReaderV2CompressionType sets the optional compression_type attribute to value. If not specified, defaults to ""
func TFRecordReaderV2Container ¶
func TFRecordReaderV2Container(value string) TFRecordReaderV2Attr
TFRecordReaderV2Container sets the optional container attribute to value.
value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func TFRecordReaderV2SharedName ¶
func TFRecordReaderV2SharedName(value string) TFRecordReaderV2Attr
TFRecordReaderV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""
type TakeManySparseFromTensorsMapAttr ¶
TakeManySparseFromTensorsMapAttr is an optional argument to TakeManySparseFromTensorsMap.
type TakeManySparseFromTensorsMapAttr func(optionalAttr)
func TakeManySparseFromTensorsMapContainer ¶
func TakeManySparseFromTensorsMapContainer(value string) TakeManySparseFromTensorsMapAttr
TakeManySparseFromTensorsMapContainer sets the optional container attribute to value.
value: The container name for the `SparseTensorsMap` read by this op. If not specified, defaults to ""
func TakeManySparseFromTensorsMapSharedName ¶
func TakeManySparseFromTensorsMapSharedName(value string) TakeManySparseFromTensorsMapAttr
TakeManySparseFromTensorsMapSharedName sets the optional shared_name attribute to value.
value: The shared name for the `SparseTensorsMap` read by this op. It should not be blank; rather the `shared_name` or unique Operation name of the Op that created the original `SparseTensorsMap` should be used. If not specified, defaults to ""
type TensorArrayConcatV2Attr ¶
TensorArrayConcatV2Attr is an optional argument to TensorArrayConcatV2.
type TensorArrayConcatV2Attr func(optionalAttr)
func TensorArrayConcatV2ElementShapeExcept0 ¶
func TensorArrayConcatV2ElementShapeExcept0(value tf.Shape) TensorArrayConcatV2Attr
TensorArrayConcatV2ElementShapeExcept0 sets the optional element_shape_except0 attribute to value. If not specified, defaults to <unknown_rank:true >
type TensorArrayConcatV3Attr ¶
TensorArrayConcatV3Attr is an optional argument to TensorArrayConcatV3.
type TensorArrayConcatV3Attr func(optionalAttr)
func TensorArrayConcatV3ElementShapeExcept0 ¶
func TensorArrayConcatV3ElementShapeExcept0(value tf.Shape) TensorArrayConcatV3Attr
TensorArrayConcatV3ElementShapeExcept0 sets the optional element_shape_except0 attribute to value.
value: The expected shape of an element, if known, excluding the first dimension. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, concatenating zero-size TensorArrays is an error. If not specified, defaults to <unknown_rank:true >
type TensorArrayGatherV2Attr ¶
TensorArrayGatherV2Attr is an optional argument to TensorArrayGatherV2.
type TensorArrayGatherV2Attr func(optionalAttr)
func TensorArrayGatherV2ElementShape ¶
func TensorArrayGatherV2ElementShape(value tf.Shape) TensorArrayGatherV2Attr
TensorArrayGatherV2ElementShape sets the optional element_shape attribute to value. If not specified, defaults to <unknown_rank:true >
type TensorArrayGatherV3Attr ¶
TensorArrayGatherV3Attr is an optional argument to TensorArrayGatherV3.
type TensorArrayGatherV3Attr func(optionalAttr)
func TensorArrayGatherV3ElementShape ¶
func TensorArrayGatherV3ElementShape(value tf.Shape) TensorArrayGatherV3Attr
TensorArrayGatherV3ElementShape sets the optional element_shape attribute to value.
value: The expected shape of an element, if known. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, gathering zero-size TensorArrays is an error. If not specified, defaults to <unknown_rank:true >
type TensorArrayV2Attr ¶
TensorArrayV2Attr is an optional argument to TensorArrayV2.
type TensorArrayV2Attr func(optionalAttr)
func TensorArrayV2ClearAfterRead ¶
func TensorArrayV2ClearAfterRead(value bool) TensorArrayV2Attr
TensorArrayV2ClearAfterRead sets the optional clear_after_read attribute to value. If not specified, defaults to true
func TensorArrayV2DynamicSize ¶
func TensorArrayV2DynamicSize(value bool) TensorArrayV2Attr
TensorArrayV2DynamicSize sets the optional dynamic_size attribute to value. If not specified, defaults to false
func TensorArrayV2ElementShape ¶
func TensorArrayV2ElementShape(value tf.Shape) TensorArrayV2Attr
TensorArrayV2ElementShape sets the optional element_shape attribute to value. If not specified, defaults to <unknown_rank:true >
func TensorArrayV2TensorArrayName ¶
func TensorArrayV2TensorArrayName(value string) TensorArrayV2Attr
TensorArrayV2TensorArrayName sets the optional tensor_array_name attribute to value. If not specified, defaults to ""
type TensorArrayV3Attr ¶
TensorArrayV3Attr is an optional argument to TensorArrayV3.
type TensorArrayV3Attr func(optionalAttr)
func TensorArrayV3ClearAfterRead ¶
func TensorArrayV3ClearAfterRead(value bool) TensorArrayV3Attr
TensorArrayV3ClearAfterRead sets the optional clear_after_read attribute to value.
value: If true (default), Tensors in the TensorArray are cleared after being read. This disables multiple read semantics but allows early release of memory. If not specified, defaults to true
func TensorArrayV3DynamicSize ¶
func TensorArrayV3DynamicSize(value bool) TensorArrayV3Attr
TensorArrayV3DynamicSize sets the optional dynamic_size attribute to value.
value: A boolean that determines whether writes to the TensorArray are allowed to grow the size. By default, this is not allowed. If not specified, defaults to false
func TensorArrayV3ElementShape ¶
func TensorArrayV3ElementShape(value tf.Shape) TensorArrayV3Attr
TensorArrayV3ElementShape sets the optional element_shape attribute to value.
value: The expected shape of an element, if known. Used to validate the shapes of TensorArray elements. If this shape is not fully specified, gathering zero-size TensorArrays is an error. If not specified, defaults to <unknown_rank:true >
func TensorArrayV3TensorArrayName ¶
func TensorArrayV3TensorArrayName(value string) TensorArrayV3Attr
TensorArrayV3TensorArrayName sets the optional tensor_array_name attribute to value.
value: Overrides the name used for the temporary tensor_array resource. Default value is the name of the 'TensorArray' op (which is guaranteed unique). If not specified, defaults to ""
type TensorSummaryAttr ¶
TensorSummaryAttr is an optional argument to TensorSummary.
type TensorSummaryAttr func(optionalAttr)
func TensorSummaryDescription ¶
func TensorSummaryDescription(value string) TensorSummaryAttr
TensorSummaryDescription sets the optional description attribute to value.
value: A json-encoded SummaryDescription proto. If not specified, defaults to ""
func TensorSummaryDisplayName ¶
func TensorSummaryDisplayName(value string) TensorSummaryAttr
TensorSummaryDisplayName sets the optional display_name attribute to value.
value: An unused string. If not specified, defaults to ""
func TensorSummaryLabels ¶
func TensorSummaryLabels(value []string) TensorSummaryAttr
TensorSummaryLabels sets the optional labels attribute to value.
value: An unused list of strings. If not specified, defaults to <>
type TextLineReaderV2Attr ¶
TextLineReaderV2Attr is an optional argument to TextLineReaderV2.
type TextLineReaderV2Attr func(optionalAttr)
func TextLineReaderV2Container ¶
func TextLineReaderV2Container(value string) TextLineReaderV2Attr
TextLineReaderV2Container sets the optional container attribute to value.
value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func TextLineReaderV2SharedName ¶
func TextLineReaderV2SharedName(value string) TextLineReaderV2Attr
TextLineReaderV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""
func TextLineReaderV2SkipHeaderLines ¶
func TextLineReaderV2SkipHeaderLines(value int64) TextLineReaderV2Attr
TextLineReaderV2SkipHeaderLines sets the optional skip_header_lines attribute to value.
value: Number of lines to skip from the beginning of every file. If not specified, defaults to 0
type ThreadUnsafeUnigramCandidateSamplerAttr ¶
ThreadUnsafeUnigramCandidateSamplerAttr is an optional argument to ThreadUnsafeUnigramCandidateSampler.
type ThreadUnsafeUnigramCandidateSamplerAttr func(optionalAttr)
func ThreadUnsafeUnigramCandidateSamplerSeed ¶
func ThreadUnsafeUnigramCandidateSamplerSeed(value int64) ThreadUnsafeUnigramCandidateSamplerAttr
ThreadUnsafeUnigramCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func ThreadUnsafeUnigramCandidateSamplerSeed2 ¶
func ThreadUnsafeUnigramCandidateSamplerSeed2(value int64) ThreadUnsafeUnigramCandidateSamplerAttr
ThreadUnsafeUnigramCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type TopKAttr ¶
TopKAttr is an optional argument to TopK.
type TopKAttr func(optionalAttr)
func TopKSorted ¶
func TopKSorted(value bool) TopKAttr
TopKSorted sets the optional sorted attribute to value.
value: If true the resulting `k` elements will be sorted by the values in descending order. If not specified, defaults to true
type TopKV2Attr ¶
TopKV2Attr is an optional argument to TopKV2.
type TopKV2Attr func(optionalAttr)
func TopKV2Sorted ¶
func TopKV2Sorted(value bool) TopKV2Attr
TopKV2Sorted sets the optional sorted attribute to value.
value: If true the resulting `k` elements will be sorted by the values in descending order. If not specified, defaults to true
type TruncatedNormalAttr ¶
TruncatedNormalAttr is an optional argument to TruncatedNormal.
type TruncatedNormalAttr func(optionalAttr)
func TruncatedNormalSeed ¶
func TruncatedNormalSeed(value int64) TruncatedNormalAttr
TruncatedNormalSeed sets the optional seed attribute to value.
value: If either `seed` or `seed2` are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func TruncatedNormalSeed2 ¶
func TruncatedNormalSeed2(value int64) TruncatedNormalAttr
TruncatedNormalSeed2 sets the optional seed2 attribute to value.
value: A second seed to avoid seed collision. If not specified, defaults to 0
type UniformCandidateSamplerAttr ¶
UniformCandidateSamplerAttr is an optional argument to UniformCandidateSampler.
type UniformCandidateSamplerAttr func(optionalAttr)
func UniformCandidateSamplerSeed ¶
func UniformCandidateSamplerSeed(value int64) UniformCandidateSamplerAttr
UniformCandidateSamplerSeed sets the optional seed attribute to value.
value: If either seed or seed2 are set to be non-zero, the random number generator is seeded by the given seed. Otherwise, it is seeded by a random seed. If not specified, defaults to 0
func UniformCandidateSamplerSeed2 ¶
func UniformCandidateSamplerSeed2(value int64) UniformCandidateSamplerAttr
UniformCandidateSamplerSeed2 sets the optional seed2 attribute to value.
value: An second seed to avoid seed collision. If not specified, defaults to 0
type UniqueAttr ¶
UniqueAttr is an optional argument to Unique.
type UniqueAttr func(optionalAttr)
func UniqueOutIdx ¶
func UniqueOutIdx(value tf.DataType) UniqueAttr
UniqueOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32
type UniqueWithCountsAttr ¶
UniqueWithCountsAttr is an optional argument to UniqueWithCounts.
type UniqueWithCountsAttr func(optionalAttr)
func UniqueWithCountsOutIdx ¶
func UniqueWithCountsOutIdx(value tf.DataType) UniqueWithCountsAttr
UniqueWithCountsOutIdx sets the optional out_idx attribute to value. If not specified, defaults to DT_INT32
type UnpackAttr ¶
UnpackAttr is an optional argument to Unpack.
type UnpackAttr func(optionalAttr)
func UnpackAxis ¶
func UnpackAxis(value int64) UnpackAttr
UnpackAxis sets the optional axis attribute to value.
value: Dimension along which to unpack. Negative values wrap around, so the valid range is `[-R, R)`. If not specified, defaults to 0
type UnstageAttr ¶
UnstageAttr is an optional argument to Unstage.
type UnstageAttr func(optionalAttr)
func UnstageContainer ¶
func UnstageContainer(value string) UnstageAttr
UnstageContainer sets the optional container attribute to value. If not specified, defaults to ""
func UnstageSharedName ¶
func UnstageSharedName(value string) UnstageAttr
UnstageSharedName sets the optional shared_name attribute to value. If not specified, defaults to ""
type VarHandleOpAttr ¶
VarHandleOpAttr is an optional argument to VarHandleOp.
type VarHandleOpAttr func(optionalAttr)
func VarHandleOpContainer ¶
func VarHandleOpContainer(value string) VarHandleOpAttr
VarHandleOpContainer sets the optional container attribute to value.
value: the container this variable is placed in. If not specified, defaults to ""
func VarHandleOpSharedName ¶
func VarHandleOpSharedName(value string) VarHandleOpAttr
VarHandleOpSharedName sets the optional shared_name attribute to value.
value: the name by which this variable is referred to. If not specified, defaults to ""
type WholeFileReaderV2Attr ¶
WholeFileReaderV2Attr is an optional argument to WholeFileReaderV2.
type WholeFileReaderV2Attr func(optionalAttr)
func WholeFileReaderV2Container ¶
func WholeFileReaderV2Container(value string) WholeFileReaderV2Attr
WholeFileReaderV2Container sets the optional container attribute to value.
value: If non-empty, this reader is placed in the given container. Otherwise, a default container is used. If not specified, defaults to ""
func WholeFileReaderV2SharedName ¶
func WholeFileReaderV2SharedName(value string) WholeFileReaderV2Attr
WholeFileReaderV2SharedName sets the optional shared_name attribute to value.
value: If non-empty, this reader is named in the given bucket with this shared_name. Otherwise, the node name is used instead. If not specified, defaults to ""