All Classes and Interfaces
Class
Description
Raise a exception to abort the process when called.
Optional attributes for
AbortComputes the absolute value of a tensor.
Helper base class for custom gradient adapters INTERNAL USE ONLY
Returns the element-wise sum of a list of tensors.
Applies a gradient to a given accumulator.
Returns the number of gradients aggregated in the given accumulators.
Updates the accumulator with a new value for global_step.
Extracts the average gradient in the given ConditionalAccumulator.
Computes acos of x element-wise.
Computes inverse hyperbolic cosine of x element-wise.
Returns x + y element-wise.
Add an
N-minibatch SparseTensor to a SparseTensorsMap, return N handles.Optional attributes for
AddManySparseToTensorsMapAdd all input tensors element wise.
Add a
SparseTensor to a SparseTensorsMap return its handle.Optional attributes for
AddSparseToTensorsMapAdjust the contrast of one or more images.
Adjust the hue of one or more images.
Adjust the saturation of one or more images.
Computes the "logical and" of elements across dimensions of a tensor.
Optional attributes for
AllGenerates labels for candidate sampling with a learned unigram distribution.
Optional attributes for
AllCandidateSamplerAn Op to exchange data across TPU replicas.
Returns the argument of a complex number.
Creates a uninitialized anonymous hash table.
A container for an iterator resource.
The AnonymousMemoryCache operation
A container for a multi device iterator resource.
Creates an empty anonymous mutable hash table that uses tensors as the backing store.
Optional attributes for
AnonymousMutableDenseHashTableCreates an empty anonymous mutable hash table.
Creates an empty anonymous mutable hash table of vector values.
Optional attributes for
AnonymousMutableHashTableOfTensorsThe AnonymousRandomSeedGenerator operation
The AnonymousSeedGenerator operation
Computes the "logical or" of elements across dimensions of a tensor.
Optional attributes for
AnyUpdate '*var' according to the adadelta scheme.
Optional attributes for
ApplyAdadeltaUpdate '*var' according to the adagrad scheme.
Optional attributes for
ApplyAdagradUpdate '*var' according to the proximal adagrad scheme.
Optional attributes for
ApplyAdagradDaUpdate '*var' according to the adagrad scheme.
Optional attributes for
ApplyAdagradV2Update '*var' according to the Adam algorithm.
Optional attributes for
ApplyAdamUpdate '*var' according to the AdaMax algorithm.
Optional attributes for
ApplyAdaMaxUpdate '*var' according to the AddSign update.
Optional attributes for
ApplyAddSignUpdate '*var' according to the centered RMSProp algorithm.
Optional attributes for
ApplyCenteredRmsPropUpdate '*var' according to the Ftrl-proximal scheme.
Optional attributes for
ApplyFtrlUpdate '*var' by subtracting 'alpha' * 'delta' from it.
Optional attributes for
ApplyGradientDescentUpdate '*var' according to the momentum scheme.
Optional attributes for
ApplyMomentumUpdate '*var' according to the AddSign update.
Optional attributes for
ApplyPowerSignUpdate '*var' and '*accum' according to FOBOS with Adagrad learning rate.
Optional attributes for
ApplyProximalAdagradUpdate '*var' as FOBOS algorithm with fixed learning rate.
Optional attributes for
ApplyProximalGradientDescentUpdate '*var' according to the RMSProp algorithm.
Optional attributes for
ApplyRmsPropReturns the truth value of abs(x-y) < tolerance element-wise.
Optional attributes for
ApproximateEqualReturns min/max k values and their indices of the input operand in an approximate manner.
Optional attributes for
ApproxTopKReturns the index with the largest value across dimensions of a tensor.
Returns the index with the smallest value across dimensions of a tensor.
Computes the trignometric inverse sine of x element-wise.
Computes inverse hyperbolic sine of x element-wise.
The AssertCardinalityDataset operation
A transformation that asserts which transformations happen next.
The ExperimentalAssertNextDataset operation
A transformation that asserts which transformations happened previously.
Asserts that the given condition is true.
Optional attributes for
AssertThatUpdate 'ref' by assigning 'value' to it.
Optional attributes for
AssignUpdate 'ref' by adding 'value' to it.
Optional attributes for
AssignAddAdds a value to the current value of a variable.
Update 'ref' by subtracting 'value' from it.
Optional attributes for
AssignSubSubtracts a value from the current value of a variable.
Concats input tensor across all dimensions.
Optional attributes for
AssignVariableConcatNDAssigns a new value to a variable.
Optional attributes for
AssignVariableOpConverts each entry in the given tensor to strings.
Optional attributes for
AsStringComputes the trignometric inverse tangent of x element-wise.
Computes arctangent of
y/x element-wise, respecting signs of the arguments.Computes inverse hyperbolic tangent of x element-wise.
Metadata of an op's attribute.
An API for building
audio operations as OpsProduces a visualization of audio data over time.
Optional attributes for
AudioSpectrogramOutputs a
Summary protocol buffer with audio.Optional attributes for
AudioSummaryCreates a dataset that shards the input dataset.
Creates a dataset that shards the input dataset.
Optional attributes for
AutoShardDatasetOptional attributes for
AutoShardDatasetPerforms average pooling on the input.
Optional attributes for
AvgPoolPerforms 3D average pooling on the input.
Optional attributes for
AvgPool3dComputes gradients of average pooling function.
Optional attributes for
AvgPool3dGradComputes gradients of the average pooling function.
Optional attributes for
AvgPoolGradThe BandedTriangularSolve operation
Optional attributes for
BandedTriangularSolveCopy a tensor setting everything outside a central band in each innermost matrix to zero.
Defines a barrier that persists across different graph executions.
Optional attributes for
BarrierCloses the given barrier.
Optional attributes for
BarrierCloseComputes the number of incomplete elements in the given barrier.
For each key, assigns the respective value to the specified component.
Computes the number of complete elements in the given barrier.
Takes the given number of completed elements from a barrier.
Optional attributes for
BarrierTakeManyBatches all input tensors nondeterministically.
Optional attributes for
BatchThe BatchCholesky operation
The BatchCholeskyGrad operation
Creates a dataset that batches
batch_size elements from input_dataset.Optional attributes for
BatchDatasetThe BatchFFT operation
The BatchFFT2D operation
The BatchFFT3D operation
Batches all the inputs tensors to the computation done by the function.
Optional attributes for
BatchFunctionThe BatchIFFT operation
The BatchIFFT2D operation
The BatchIFFT3D operation
Multiplies slices of two tensors in batches.
Optional attributes for
BatchMatMulThe BatchMatrixBandPart operation
The BatchMatrixDeterminant operation
The BatchMatrixDiag operation
The BatchMatrixDiagPart operation
The BatchMatrixInverse operation
Optional attributes for
BatchMatrixInverseThe BatchMatrixSetDiag operation
The BatchMatrixSolve operation
Optional attributes for
BatchMatrixSolveThe BatchMatrixSolveLs operation
Optional attributes for
BatchMatrixSolveLsThe BatchMatrixTriangularSolve operation
Optional attributes for
BatchMatrixTriangularSolveBatch normalization.
Gradients for batch normalization.
The BatchSelfAdjointEigV2 operation
Optional attributes for
BatchSelfAdjointEigThe BatchSvd operation
Optional attributes for
BatchSvdBatchToSpace for 4-D tensors of type T.
BatchToSpace for N-D tensors of type T.
The BesselI0 operation
The BesselI0e operation
The BesselI1 operation
The BesselI1e operation
The BesselJ0 operation
The BesselJ1 operation
The BesselK0 operation
The BesselK0e operation
The BesselK1 operation
The BesselK1e operation
The BesselY0 operation
The BesselY1 operation
Compute the regularized incomplete beta integral \(I_x(a, b)\).
Adds
bias to value.Optional attributes for
BiasAddThe backward operation for "BiasAdd" on the "bias" tensor.
Optional attributes for
BiasAddGradCounts the number of occurrences of each value in an integer array.
Bitcasts a tensor from one type to another without copying data.
Elementwise computes the bitwise AND of
x and y.An API for building
bitwise operations as OpsElementwise computes the bitwise OR of
x and y.Elementwise computes the bitwise XOR of
x and y.Computes the LSTM cell forward propagation for all the time steps.
Optional attributes for
BlockLSTMComputes the LSTM cell backward propagation for the entire time sequence.
Optional attributes for
BooleanMaskOptional attributes for
BooleanMaskUpdateReturn the shape of s0 op s1 with broadcast.
Return the reduction indices for computing gradients of s0 op s1 with broadcast.
Broadcast an array for a compatible shape.
Bucketizes 'input' based on 'boundaries'.
Records the bytes size of each element of
input_dataset in a StatsAggregator.Records the bytes size of each element of
input_dataset in a StatsAggregator.The CacheDatasetV2 operation
Optional attributes for
CacheDatasetAn n-way switch statement which calls a single branch function.
Optional attributes for
CaseCast x of type SrcT to y of DstT.
Optional attributes for
CastReturns element-wise smallest integer not less than x.
Checks a tensor for NaN, -Inf and +Inf values.
Computes the Cholesky decomposition of one or more square matrices.
Computes the reverse mode backpropagated gradient of the Cholesky algorithm.
The ChooseFastestBranchDataset operation
The ChooseFastestDataset operation
The ExperimentalChooseFastestDataset operation
Clips tensor values to a specified min and max.
The CloseSummaryWriter operation
An API for building
cluster operations as OpsAn op that merges the string-encoded memory config protos from all hosts.
Mutually exchanges multiple tensors of identical type and shape.
Optional attributes for
CollectiveAllToAllAssign group keys based on group assignment.
Receives a tensor value broadcast from another device.
Optional attributes for
CollectiveBcastRecvBroadcasts a tensor value to one or more other devices.
Optional attributes for
CollectiveBcastSendMutually accumulates multiple tensors of identical type and shape.
Optional attributes for
CollectiveGatherInitializes a group for collective operations.
Optional attributes for
CollectiveInitializeCommunicatorAn API for building
collective operations as OpsAn Op to permute tensors across replicated TPU instances.
Mutually reduces multiple tensors of identical type and shape.
Optional attributes for
CollectiveReduceMutually reduces multiple tensors of identical type and shape and scatters the result.
Optional attributes for
CollectiveReduceScatterGreedily selects a subset of bounding boxes in descending order of score,
This operation performs non_max_suppression on the inputs per batch, across
all classes.
Optional attributes for
CombinedNonMaxSuppressionReturns the result of a TPU compilation.
Compiles a computations for execution on one or more TPU devices.
Asserts that compilation succeeded.
Converts two real numbers to a complex number.
Computes the complex absolute value of a tensor.
Encodes an
ExtensionType value into a variant scalar Tensor.Decodes a
variant scalar Tensor into an ExtensionType value.Compresses a dataset element.
Computes the ids of the positions in sampled_candidates that match true_labels.
Optional attributes for
ComputeAccidentalHitsComputes the static batch size of a dataset sans partial batches.
An op computes the size of the deduplication data from embedding core and returns the updated config.
An op computes tuple mask of deduplication data from embedding core.
Concatenates tensors along one dimension.
Creates a dataset that concatenates
input_dataset with another_dataset.Optional attributes for
ConcatenateDatasetConcats input tensor across all dimensions.
Optional attributes for
ConcatNDComputes offsets of concat inputs within its output.
A graph that can be invoked as a single function, with an input and output signature.
A conditional accumulator for aggregating gradients.
Optional attributes for
ConditionalAccumulatorAn op that sets up the centralized structures for a distributed TPU system.
Optional attributes for
ConfigureAndInitializeGlobalTPUSets up the centralized structures for a distributed TPU system.
Optional attributes for
ConfigureDistributedTPUSets up TPUEmbedding in a distributed TPU system.
An op that configures the TPUEmbedding software on a host.
An op that configures the TPUEmbedding software on a host.
Returns the complex conjugate of a complex number.
Shuffle dimensions of x according to a permutation and conjugate the result.
An op that sets up communication between TPUEmbedding host software instances
after ConfigureTPUEmbeddingHost has been called on each host.
An operator producing a constant value.
This op consumes a lock created by
MutexLock.Does nothing.
Computes a N-D convolution given (N+1+batch_dims)-D
input and (N+2)-D filter tensors.Optional attributes for
ConvComputes a 2-D convolution given 4-D
input and filter tensors.Optional attributes for
Conv2dComputes the gradients of convolution with respect to the filter.
Optional attributes for
Conv2dBackpropFilterComputes the gradients of convolution with respect to the filter.
Optional attributes for
Conv2dBackpropFilterV2Computes the gradients of convolution with respect to the input.
Optional attributes for
Conv2dBackpropInputComputes the gradients of convolution with respect to the input.
Optional attributes for
Conv2dBackpropInputV2Computes a 3-D convolution given 5-D
input and filter tensors.Optional attributes for
Conv3dComputes the gradients of 3-D convolution with respect to the filter.
Optional attributes for
Conv3dBackpropFilterComputes the gradients of 3-D convolution with respect to the input.
Optional attributes for
Conv3dBackpropInputThe ConvertToCooTensor operation
Copy a tensor from CPU-to-CPU or GPU-to-GPU.
Optional attributes for
CopyCopy a tensor to host.
Optional attributes for
CopyHostThe CopyToMesh operation
The CopyToMeshGrad operation
Computes cos of x element-wise.
Computes hyperbolic cosine of x element-wise.
Increments 'ref' until it reaches 'limit'.
The CreateSummaryDbWriter operation
The CreateSummaryFileWriter operation
Extracts crops from the input image tensor and resizes them.
Optional attributes for
CropAndResizeComputes the gradient of the crop_and_resize op wrt the input boxes tensor.
Optional attributes for
CropAndResizeGradBoxesComputes the gradient of the crop_and_resize op wrt the input image tensor.
Optional attributes for
CropAndResizeGradImageCompute the pairwise cross product.
An Op to sum inputs across replicated TPU instances.
Reads out the CSR components at batch
index.Convert a (possibly batched) CSRSparseMatrix to dense.
Converts a (possibly batched) CSRSparesMatrix to a SparseTensor.
The CSVDatasetV2 operation
The ExperimentalCSVDataset operation
Performs beam search decoding on the logits given in input.
Optional attributes for
CtcBeamSearchDecoderPerforms greedy decoding on the logits given in inputs.
Optional attributes for
CtcGreedyDecoderCalculates the CTC Loss (log probability) for each batch entry.
Optional attributes for
CtcLossCalculates the CTC Loss (log probability) for each batch entry.
Optional attributes for
CTCLossV2A RNN backed by cuDNN.
Optional attributes for
CudnnRNNBackprop step of CudnnRNNV3.
Optional attributes for
CudnnRNNBackpropConverts CudnnRNN params from canonical form to usable form.
Optional attributes for
CudnnRNNCanonicalToParamsComputes size of weights that can be used by a Cudnn RNN model.
Optional attributes for
CudnnRnnParamsSizeRetrieves CudnnRNN params in canonical form.
Optional attributes for
CudnnRNNParamsToCanonicalCompute the cumulative product of the tensor
x along axis.Optional attributes for
CumprodCompute the cumulative sum of the tensor
x along axis.Optional attributes for
CumsumCompute the cumulative product of the tensor
x along axis.Optional attributes for
CumulativeLogsumexpA custom gradient for ops of type
CustomGradient.An API for building
data.experimental operations as OpsReturns the dimension index in the destination data format given the one in
the source data format.
Optional attributes for
DataFormatDimMapPermute input tensor from
src_format to dst_format.Optional attributes for
DataFormatVecPermuteAn API for building
data operations as OpsCreates a dataset that reads data from the tf.data service.
Optional attributes for
DataServiceDatasetReturns the cardinality of
input_dataset.Returns the cardinality of
input_dataset.Optional attributes for
DatasetCardinalityReturns the fingerprint of
input_dataset.Creates a dataset from the given
graph_def.Returns a serialized GraphDef representing
input_dataset.Optional attributes for
DatasetToGraphOutputs the single element from the given dataset.
Optional attributes for
DatasetToSingleElementWrites the given dataset to the given file using the TFRecord format.
Writes the given dataset to the given file using the TFRecord format.
The Dawsn operation
An API for building
debugging operations as OpsIdentity op for gradient debugging.
Identity op for gradient debugging.
Provides an identity mapping of the non-Ref type input tensor for debugging.
Optional attributes for
DebugIdentityDebug NaN Value Counter Op.
Optional attributes for
DebugNanCountDebug Numeric Summary V2 Op.
Optional attributes for
DebugNumericsSummaryDecode and Crop a JPEG-encoded image to a uint8 tensor.
Optional attributes for
DecodeAndCropJpegDecode web-safe base64-encoded strings.
Decode the first frame of a BMP-encoded image to a uint8 tensor.
Optional attributes for
DecodeBmpDecompress strings.
Optional attributes for
DecodeCompressedConvert CSV records to tensors.
Optional attributes for
DecodeCsvDecode the frame(s) of a GIF-encoded image to a uint8 tensor.
Function for decode_bmp, decode_gif, decode_jpeg, and decode_png.
Optional attributes for
DecodeImageDecode a JPEG-encoded image to a uint8 tensor.
Optional attributes for
DecodeJpegConvert JSON-encoded Example records to binary protocol buffer strings.
Reinterpret the bytes of a string as a vector of numbers.
Optional attributes for
DecodePaddedRawDecode a PNG-encoded image to a uint8 or uint16 tensor.
Optional attributes for
DecodePngThe op extracts fields from a serialized protocol buffers message into tensors.
Optional attributes for
DecodeProtoReinterpret the bytes of a string as a vector of numbers.
Optional attributes for
DecodeRawDecode a 16-bit PCM WAV file to a float tensor.
Optional attributes for
DecodeWavMakes a copy of
x.A container for an iterator resource.
The DeleteMemoryCache operation
A container for an iterator resource.
The DeleteRandomSeedGenerator operation
The DeleteSeedGenerator operation
Delete the tensor specified by its handle in the session.
Counts the number of occurrences of each value in an integer array.
Optional attributes for
DenseBincountPerforms sparse-output bin counting for a tf.tensor input.
Optional attributes for
DenseCountSparseOutputConverts a dense tensor to a (possibly batched) CSRSparseMatrix.
Applies set operation along last dimension of 2
Tensor inputs.Optional attributes for
DenseToDenseSetOperationCreates a dataset that batches input elements into a SparseTensor.
Creates a dataset that batches input elements into a SparseTensor.
Applies set operation along last dimension of
Tensor and SparseTensor.Optional attributes for
DenseToSparseSetOperationDepthToSpace for tensors of type T.
Optional attributes for
DepthToSpaceComputes a 2-D depthwise convolution given 4-D
input and filter tensors.Optional attributes for
DepthwiseConv2dNativeComputes the gradients of depthwise convolution with respect to the filter.
Optional attributes for
DepthwiseConv2dNativeBackpropFilterComputes the gradients of depthwise convolution with respect to the input.
Optional attributes for
DepthwiseConv2dNativeBackpropInputDequantize the 'input' tensor into a float or bfloat16 Tensor.
Optional attributes for
DequantizeConverts the given variant tensor to an iterator and stores it in the given resource.
Deserialize and concatenate
SparseTensors from a serialized minibatch.Deserialize
SparseTensor objects.Deletes the resource specified by the handle.
Optional attributes for
DestroyResourceOpDestroys the temporary variable and returns its final value.
Computes the determinant of one or more square matrices.
Return the index of device the op runs.
Represents a (possibly partial) specification for a TensorFlow device.
A Builder class for building
DeviceSpec class.Computes Psi, the derivative of Lgamma (the log of the absolute value of
Gamma(x)), element-wise.Computes the grayscale dilation of 4-D
input and 3-D filter tensors.Computes the gradient of morphological 2-D dilation with respect to the filter.
Computes the gradient of morphological 2-D dilation with respect to the input.
A substitute for
InterleaveDataset on a fixed list of N datasets.A substitute for
InterleaveDataset on a fixed list of N datasets.Optional attributes for
DirectedInterleaveDatasetTurns off the copy-on-read mode.
The DistributedSave operation
Optional attributes for
DistributedSaveAn API for building
distribute operations as OpsReturns x / y element-wise.
Returns 0 if the denominator is zero.
Draw bounding boxes on a batch of images.
The DTensorRestoreV2 operation
An API for building
dtypes operations as OpsThe DummyIterationCounter operation
The DummyMemoryCache operation
The DummySeedGenerator operation
Eases the porting of code that uses tf.nn.embedding_lookup_sparse().
Optional attributes for
DynamicEnqueueTPUEmbeddingArbitraryTensorBatchThe DynamicEnqueueTPUEmbeddingRaggedTensorBatch operation
Optional attributes for
DynamicEnqueueTPUEmbeddingRaggedTensorBatchPartitions
data into num_partitions tensors using indices from partitions.Interleave the values from the
data tensors into a single tensor.An environment for executing TensorFlow operations eagerly.
Controls how to act when we try to run an operation on a given device but some input tensors
are not on that device.
Computes the (possibly normalized) Levenshtein Edit Distance.
Optional attributes for
EditDistanceComputes the eigen decomposition of one or more square matrices.
Optional attributes for
EigTensor contraction according to Einstein summation convention.
Computes the exponential linear function.
Computes gradients for the exponential linear (Elu) operation.
An op enabling differentiation of TPU Embeddings.
Creates a tensor with the given shape.
Optional attributes for
EmptyCreates and returns an empty tensor list.
Creates and returns an empty tensor map.
Encode strings into web-safe base64 format.
Optional attributes for
EncodeBase64JPEG-encode an image.
Optional attributes for
EncodeJpegJPEG encode input image with provided compression quality.
PNG-encode an image.
Optional attributes for
EncodePngThe op serializes protobuf messages provided in the input tensors.
Optional attributes for
EncodeProtoEncode audio data using the WAV file format.
Eases the porting of code that uses tf.nn.embedding_lookup_sparse().
Optional attributes for
EnqueueTPUEmbeddingArbitraryTensorBatchAn op that enqueues a list of input batch tensors to TPUEmbedding.
Optional attributes for
EnqueueTPUEmbeddingBatchAn op that enqueues a list of input batch tensors to TPUEmbedding.
Optional attributes for
EnqueueTPUEmbeddingIntegerBatchEases the porting of code that uses tf.nn.embedding_lookup().
Optional attributes for
EnqueueTPUEmbeddingRaggedTensorBatchAn op that enqueues TPUEmbedding input indices from a SparseTensor.
Optional attributes for
EnqueueTPUEmbeddingSparseBatchEases the porting of code that uses tf.nn.embedding_lookup_sparse().
Optional attributes for
EnqueueTPUEmbeddingSparseTensorBatchEnsures that the tensor's shape matches the expected shape.
Creates or finds a child frame, and makes
data available to the child frame.Optional attributes for
EnterReturns the truth value of (x == y) element-wise.
Optional attributes for
EqualComputes the Gauss error function of
x element-wise.Computes the complementary error function of
x element-wise.The Erfinv operation
Computes the euclidean norm of elements across dimensions of a tensor.
Optional attributes for
EuclideanNormOp that loads and executes a TPU program on a TPU device.
Op that executes a program with optional in-place variable updates.
An op that executes the TPUEmbedding partitioner on the central configuration
device and computes the HBM size (in bytes) required for TPUEmbedding operation.
Defines an environment for creating and executing TensorFlow
Operations.Exits the current frame to its parent frame.
Computes exponential of x element-wise.
Inserts a dimension of 1 into a tensor's shape.
The Expint operation
Computes
exp(x) - 1 element-wise.Extracts a glimpse from the input tensor.
Optional attributes for
ExtractGlimpseExtract
patches from images and put them in the "depth" output dimension.Extract the shape information of a JPEG-encoded image.
Extract
patches from input and put them in the "depth" output dimension.Output a fact about factorials.
This op is used as a placeholder in If branch functions.
Fake-quantize the 'inputs' tensor, type float to 'outputs' tensor of same shape and type.
Optional attributes for
FakeQuantWithMinMaxArgsCompute gradients for a FakeQuantWithMinMaxArgs operation.
Optional attributes for
FakeQuantWithMinMaxArgsGradientFake-quantize the 'inputs' tensor of type float via global float scalars
Fake-quantize the
inputs tensor of type float via global float scalars
min and max to outputs tensor of same shape as inputs.Optional attributes for
FakeQuantWithMinMaxVarsCompute gradients for a FakeQuantWithMinMaxVars operation.
Optional attributes for
FakeQuantWithMinMaxVarsGradientFake-quantize the 'inputs' tensor of type float via per-channel floats
Fake-quantize the
inputs tensor of type float per-channel and one of the
shapes: [d], [b, d] [b, h, w, d] via per-channel floats min and max
of shape [d] to outputs tensor of same shape as inputs.Optional attributes for
FakeQuantWithMinMaxVarsPerChannelCompute gradients for a FakeQuantWithMinMaxVarsPerChannel operation.
Optional attributes for
FakeQuantWithMinMaxVarsPerChannelGradientDeprecated.
Fast Fourier transform.
2D fast Fourier transform.
3D fast Fourier transform.
ND fast Fourier transform.
A queue that produces elements in first-in first-out order.
Optional attributes for
FifoQueueSet configuration of the file system.
Creates a tensor filled with a scalar value.
Creates a dataset containing elements of first component of
input_dataset having true in the last component.Creates a dataset containing elements of
input_dataset matching predicate.Optional attributes for
FilterDatasetCreates a dataset by applying
tf.data.Options to input_dataset.Optional attributes for
FinalizeDatasetAn op that finalizes the TPUEmbedding configuration.
Generates fingerprint values.
The FixedLengthRecordDatasetV2 operation
Optional attributes for
FixedLengthRecordDatasetA Reader that outputs fixed-length records from a file.
Optional attributes for
FixedLengthRecordReaderGenerates labels for candidate sampling with a learned unigram distribution.
Optional attributes for
FixedUnigramCandidateSamplerCreates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
FlatMapDatasetReturns element-wise largest integer not greater than x.
Returns x // y element-wise.
Returns element-wise remainder of division.
The FlushSummaryWriter operation
Applies a for loop.
Performs fractional average pooling on the input.
Optional attributes for
FractionalAvgPoolComputes gradient of the FractionalAvgPool function.
Optional attributes for
FractionalAvgPoolGradPerforms fractional max pooling on the input.
Optional attributes for
FractionalMaxPoolComputes gradient of the FractionalMaxPool function.
Optional attributes for
FractionalMaxPoolGradThe FresnelCos operation
The FresnelSin operation
Ops for calling
ConcreteFunction.Batch normalization.
Optional attributes for
FusedBatchNormGradient for batch normalization.
Optional attributes for
FusedBatchNormGradPerforms a padding as a preprocess during a convolution.
Performs a resize and padding as a preprocess during a convolution.
Optional attributes for
FusedResizeAndPadConv2dGather slices from
params axis axis according to indices.Optional attributes for
GatherGather slices from
params into a Tensor with shape specified by indices.This op produces Region of Interests from given bounding boxes(bbox_deltas) encoded wrt anchors according to eq.2 in arXiv:1506.01497
Optional attributes for
GenerateBoundingBoxProposalsGiven a path to new and old vocabulary files, returns a remapping Tensor of
length
num_new_vocab, where remapping[i] contains the row number in the old
vocabulary that corresponds to row i in the new vocabulary (starting at line
new_vocab_offset and up to num_new_vocab entities), or -1 if entry i
in the new vocabulary is not in the old vocabulary.Optional attributes for
GenerateVocabRemappingCreates a dataset that invokes a function to generate elements.
Optional attributes for
GeneratorDatasetGets the element at the specified index in a dataset.
The GetMinibatchesInCsrWithPhysicalReplica operation
The GetMinibatchSplitsWithPhysicalReplica operation
Returns the
tf.data.Options attached to input_dataset.Store the input tensor in the state of the current session.
Get the value of the tensor specified by its handle.
The GlobalIterId operation
Adds operations to compute the partial derivatives of sum of
ys w.r.t xs, i.e.,
d(y_1 + y_2 + ...)/dx_1, d(y_1 + y_2 + ...)/dx_2...Optional attributes for
GradientsA data flow graph representing a TensorFlow computation.
Used to instantiate an abstract class which overrides the buildSubgraph method to build a
conditional or body subgraph for a while loop.
Returns the truth value of (x > y) element-wise.
Returns the truth value of (x >= y) element-wise.
Creates a dataset that computes a group-by on
input_dataset.Creates a dataset that computes a group-by on
input_dataset.Creates a dataset that computes a windowed group-by on
input_dataset.Creates a dataset that computes a windowed group-by on
input_dataset.Optional attributes for
GroupByWindowDatasetComputes the GRU cell forward propagation for 1 time step.
Computes the GRU cell back-propagation for 1 time step.
Gives a guarantee to the TF runtime that the input tensor is a constant.
Creates a non-initialized hash table.
Optional attributes for
HashTableContainer class for core methods which add or perform several operations and return one of them.
Return histogram of values.
Outputs a
Summary protocol buffer with a histogram.Returns a constant tensor on the host.
Convert one or more images from HSV to RGB.
Return a tensor with the same shape and contents as the input tensor or value.
Returns a list of tensors with the same shapes and contents as the input
tensors.
A Reader that outputs the queued work as both the key and value.
Optional attributes for
IdentityReaderoutput = cond ? then_branch(input) : else_branch(input)
Optional attributes for
IfInverse fast Fourier transform.
Inverse 2D fast Fourier transform.
Inverse 3D fast Fourier transform.
ND inverse fast Fourier transform.
Compute the lower regularized incomplete Gamma function
P(a, x).Compute the upper regularized incomplete Gamma function
Q(a, x).Computes the gradient of
igamma(a, x) wrt a.Creates a dataset that contains the elements of
input_dataset ignoring errors.Creates a dataset that contains the elements of
input_dataset ignoring errors.Optional attributes for
IgnoreErrorsDatasetOptional attributes for
IgnoreErrorsDatasetReturns the imaginary part of a complex number.
An API for building
image operations as OpsApplies the given transform to each of the images.
Optional attributes for
ImageProjectiveTransformV2Applies the given transform to each of the images.
Optional attributes for
ImageProjectiveTransformV3Outputs a
Summary protocol buffer with images.Optional attributes for
ImageSummaryReturns immutable tensor from memory region.
The ImportEvent operation
A placeholder op for a value that will be fed into the computation.
Fetches multiple values from infeed as an XLA tuple.
An op which feeds a single Tensor value into the computation.
Optional attributes for
InfeedEnqueueAn op which enqueues prelinearized buffer into TPU infeed.
Optional attributes for
InfeedEnqueuePrelinearizedBufferFeeds multiple Tensor values into the computation as an XLA tuple.
Optional attributes for
InfeedEnqueueTupleTable initializer that takes two tensors for keys and values respectively.
The InitializeTableFromDataset operation
Initializes a table from a text file.
Optional attributes for
InitializeTableFromTextFileAdds v into specified rows of x.
Subtracts `v` into specified rows of `x`.
Updates specified rows 'i' with values 'v'.
Creates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
InterleaveDatasetSays whether the targets are in the top
K predictions.Computes the inverse of one or more square invertible matrices or their adjoints (conjugate transposes).
Optional attributes for
InvInvert (flip) each bit of supported types; for example, type
uint8 value 01010101 becomes 10101010.Computes the inverse permutation of a tensor.
Computes the gradient for the inverse of
x wrt its input.An API for building
io operations as OpsInverse real-valued fast Fourier transform.
Inverse 2D real-valued fast Fourier transform.
Inverse 3D real-valued fast Fourier transform.
ND inverse real fast Fourier transform.
Returns which elements of x are finite.
Returns which elements of x are Inf.
Returns which elements of x are NaN.
Solves a batch of isotonic regression problems.
Whether TPU Embedding is initialized in a distributed TPU system.
Optional attributes for
IsTPUEmbeddingInitializedChecks whether a tensor has been initialized.
The IteratorV2 operation
The IteratorFromStringHandleV2 operation
Optional attributes for
IteratorFromStringHandleReturns the name of the device on which
resource has been placed.Returns the name of the device on which
resource has been placed.Gets the next output from the given iterator .
Gets the next output from the given iterator as an Optional variant.
Gets the next output from the given iterator.
Converts the given
resource_handle representing an iterator to a string.Joins the strings in the given list of string tensors into one tensor;
with the given separator (default is an empty separator).
Optional attributes for
JoinReturns the index of a data point that should be added to the seed set.
Selects num_to_sample rows of input using the KMeans++ criterion.
Computes the Kth order statistic of a data set.
L2 Loss.
Records the latency of producing
input_dataset elements in a StatsAggregator.Records the latency of producing
input_dataset elements in a StatsAggregator.Computes rectified linear:
max(features, features * alpha).Optional attributes for
LeakyReluComputes rectified linear gradients for a LeakyRelu operation.
Optional attributes for
LeakyReluGradGenerates labels for candidate sampling with a learned unigram distribution.
Optional attributes for
LearnedUnigramCandidateSamplerElementwise computes the bitwise left-shift of
x and y.Creates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
LegacyParallelInterleaveDatasetReturns the truth value of (x < y) element-wise.
Returns the truth value of (x <= y) element-wise.
Computes the log of the absolute value of
Gamma(x) element-wise.An API for building
linalg operations as OpsAn API for building
linalg.sparse operations as OpsGenerates values in an interval.
Creates a dataset that emits each of
tensors once.Optional attributes for
ListDatasetThe ListSnapshotChunksDataset operation
The ExperimentalLMDBDataset operation
Creates a dataset that emits the key-value pairs in one or more LMDB files.
A Reader that outputs the records from a LMDB file.
Optional attributes for
LmdbReaderAn op that loads optimization parameters into embedding memory.
Loads a 2-D (matrix)
Tensor with name old_tensor_name from the checkpoint
at ckpt_path and potentially reorders its rows and columns using the
specified remappings.Optional attributes for
LoadAndRemapMatrixThe LoadDataset operation
Optional attributes for
LoadDatasetLoad Adadelta embedding parameters.
Optional attributes for
LoadTPUEmbeddingAdadeltaParametersLoad Adagrad Momentum embedding parameters.
Optional attributes for
LoadTPUEmbeddingAdagradMomentumParametersLoad Adagrad embedding parameters.
Optional attributes for
LoadTPUEmbeddingAdagradParametersLoad ADAM embedding parameters.
Optional attributes for
LoadTPUEmbeddingADAMParametersLoad centered RMSProp embedding parameters.
Optional attributes for
LoadTPUEmbeddingCenteredRMSPropParametersLoad frequency estimator embedding parameters.
Optional attributes for
LoadTPUEmbeddingFrequencyEstimatorParametersLoad FTRL embedding parameters.
Optional attributes for
LoadTPUEmbeddingFTRLParametersLoad MDL Adagrad Light embedding parameters.
Optional attributes for
LoadTPUEmbeddingMDLAdagradLightParametersLoad Momentum embedding parameters.
Optional attributes for
LoadTPUEmbeddingMomentumParametersLoad proximal Adagrad embedding parameters.
Optional attributes for
LoadTPUEmbeddingProximalAdagradParametersThe LoadTPUEmbeddingProximalYogiParameters operation
Optional attributes for
LoadTPUEmbeddingProximalYogiParametersLoad RMSProp embedding parameters.
Optional attributes for
LoadTPUEmbeddingRMSPropParametersLoad SGD embedding parameters.
Optional attributes for
LoadTPUEmbeddingStochasticGradientDescentParametersLocal Response Normalization.
Optional attributes for
LocalResponseNormalizationGradients for Local Response Normalization.
Optional attributes for
LocalResponseNormalizationGradComputes natural logarithm of x element-wise.
Computes natural logarithm of (1 + x) element-wise.
Returns the truth value of x AND y element-wise.
Returns the truth value of
NOT x element-wise.Returns the truth value of x OR y element-wise.
Computes the sign and the log of the absolute value of the determinant of
one or more square matrices.
Computes log softmax activations.
Generates labels for candidate sampling with a log-uniform distribution.
Optional attributes for
LogUniformCandidateSamplerOutputs all keys and values in the table.
Looks up keys in a table, outputs the corresponding values.
Replaces the contents of the table with the specified keys and values.
Updates the table to associates keys with values.
Removes keys and its associated values from a table.
Computes the number of elements in the given table.
Forwards the input to the output.
Converts all uppercase characters into their respective lowercase replacements.
Optional attributes for
LowerApplies lower_bound(sorted_search_values, values) along each row.
Computes the LSTM cell forward propagation for 1 time step.
Optional attributes for
LSTMBlockCellComputes the LSTM cell backward propagation for 1 timestep.
Computes the LU decomposition of one or more square matrices.
Makes a new iterator from the given
dataset and stores it in iterator.Make all elements in the non-Batch dimension unique, but "close" to
their initial value.
Creates a dataset that fuses mapping with batching.
Creates a dataset that fuses mapping with batching.
Optional attributes for
MapAndBatchDatasetOptional attributes for
MapAndBatchDatasetOp removes all elements in the underlying container.
Optional attributes for
MapClearCreates a dataset that applies
f to the outputs of input_dataset.Creates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
MapDatasetOptional attributes for
MapDatasetMaps a function on the list of tensors unpacked from arguments on dimension 0.
Optional attributes for
MapDefunOp returns the number of incomplete elements in the underlying container.
Optional attributes for
MapIncompleteSizeOp peeks at the values at the specified key.
Optional attributes for
MapPeekOp returns the number of elements in the underlying container.
Optional attributes for
MapSizeStage (key, values) in the underlying container which behaves like a hashtable.
Optional attributes for
MapStageOp removes and returns the values associated with the key
from the underlying container.
Optional attributes for
MapUnstageOp removes and returns a random (key, value)
from the underlying container.
Optional attributes for
MapUnstageNoKeyReturns the set of files matching one or more glob patterns.
The ExperimentalMatchingFilesDataset operation
The MatchingFilesDataset operation
An API for building
math operations as OpsAn API for building
math.special operations as OpsMultiply the matrix "a" by the matrix "b".
Optional attributes for
MatMulReturns a batched diagonal tensor with given batched diagonal values.
Returns the batched diagonal part of a batched tensor.
Returns the batched diagonal part of a batched tensor.
Optional attributes for
MatrixDiagPartV3Returns a batched diagonal tensor with given batched diagonal values.
Optional attributes for
MatrixDiagV3Deprecated, use python implementation tf.linalg.matrix_exponential.
Computes the matrix logarithm of one or more square matrices:
\(log(exp(A)) = A\)
Returns a batched matrix tensor with new batched diagonal values.
Optional attributes for
MatrixSetDiagSolves one or more linear least-squares problems.
Optional attributes for
MatrixSolveLsComputes the maximum of elements across dimensions of a tensor.
Optional attributes for
MaxReturns the max of x and y (i.e.
Creates a dataset that overrides the maximum intra-op parallelism.
Creates a dataset that overrides the maximum intra-op parallelism.
Performs max pooling on the input.
Optional attributes for
MaxPoolPerforms 3D max pooling on the input.
Optional attributes for
MaxPool3dComputes gradients of 3D max pooling function.
Optional attributes for
MaxPool3dGradComputes second-order gradients of the maxpooling function.
Optional attributes for
MaxPool3dGradGradComputes gradients of the maxpooling function.
Optional attributes for
MaxPoolGradComputes second-order gradients of the maxpooling function.
Optional attributes for
MaxPoolGradGradComputes second-order gradients of the maxpooling function.
Optional attributes for
MaxPoolGradGradWithArgmaxComputes gradients of the maxpooling function.
Optional attributes for
MaxPoolGradWithArgmaxPerforms max pooling on the input and outputs both max values and indices.
Optional attributes for
MaxPoolWithArgmaxComputes the mean of elements across dimensions of a tensor.
Optional attributes for
MeanForwards the value of an available tensor from
inputs to output.An op merges elements of integer and float tensors into deduplication data as
XLA tuple.
Optional attributes for
MergeDedupDataMerges summaries.
V2 format specific: merges the metadata files of sharded checkpoints.
Optional attributes for
MergeV2CheckpointsTransforms a spectrogram into a form that's useful for speech recognition.
Optional attributes for
MfccComputes the minimum of elements across dimensions of a tensor.
Optional attributes for
MinReturns the min of x and y (i.e.
Pads a tensor with mirrored values.
Gradient op for
MirrorPad op.Wraps an arbitrary MLIR computation expressed as a module with a main() function.
Returns element-wise remainder of division.
Identity transformation that models performance.
Optional attributes for
ModelDatasetReturns x * y element-wise.
Returns x * y element-wise.
Creates a MultiDeviceIterator resource.
Generates a MultiDeviceIterator resource from its provided string handle.
Optional attributes for
MultiDeviceIteratorFromStringHandleGets next element for the provided shard number.
Initializes the multi device iterator with the given dataset.
Produces a string handle for the given MultiDeviceIterator.
Draws samples from a multinomial distribution.
Optional attributes for
MultinomialCreates an empty hash table that uses tensors as the backing store.
Optional attributes for
MutableDenseHashTableCreates an empty hash table.
Optional attributes for
MutableHashTableCreates an empty hash table.
Optional attributes for
MutableHashTableOfTensorsCreates a Mutex resource that can be locked by
MutexLock.Optional attributes for
MutexLocks a mutex resource.
A
Scope implementation backed by a native scope.Deprecated.
Outputs a tensor containing the reduction across all input tensors.
Deprecated.
use
NcclBroadcast insteadSends
input to all devices that are connected to the output.Deprecated.
use
NcclReduce insteadReduces
input from num_devices using reduction to a single device.The Ndtri operation
Selects the k nearest centers for each point.
Computes numerical negative value element-wise.
Training via negative sampling.
Returns the next representable value of
x1 in the direction of x2, element-wise.Makes its input available to the next iteration.
An API for building
nn operations as OpsNon-deterministically generates some integers.
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.
Optional attributes for
NonMaxSuppressionGreedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high overlaps
with previously selected boxes.
The ExperimentalNonSerializableDataset operation
The NonSerializableDataset operation
Does nothing.
Returns the truth value of (x != y) element-wise.
Optional attributes for
NotEqualFinds values of the
n-th order statistic for the last dimension.Optional attributes for
NthElementReturns a one-hot tensor.
Optional attributes for
OneHotAn operator creating a constant initialized with ones of the shape given by `dims`.
Makes a "one-shot" iterator that can be iterated only once.
Optional attributes for
OneShotIteratorReturns a tensor of ones with the same shape and type as x.
A logical unit of computation.
Interface implemented by operands of a TensorFlow operation.
Utilities for manipulating operand related types and lists.
Performs computation on Tensors.
Helper type for attribute getters, so we don't clutter the operation classes too much.
A builder for
Operations.Annotation used by classes to make TensorFlow operations conveniently accessible via
org.tensorflow.op.Ops or one of its groups.An annotation to provide metadata about an op inputs accessor class.
An annotation to provide metadata about an op.
An API for building operations as
OpsA Java implementation of
Scope.Creates a dataset by applying related optimizations to
input_dataset.Optional attributes for
OptimizeDatasetConstructs an Optional variant from a tuple of tensors.
Returns the value stored in an Optional variant or raises an error if none exists.
Returns true if and only if the given Optional variant has a value.
Creates an Optional variant with no value.
Creates a dataset by attaching tf.data.Options to
input_dataset.Optional attributes for
OptionsDatasetOp removes all elements in the underlying container.
Optional attributes for
OrderedMapClearOp returns the number of incomplete elements in the underlying container.
Optional attributes for
OrderedMapIncompleteSizeOp peeks at the values at the specified key.
Optional attributes for
OrderedMapPeekOp returns the number of elements in the underlying container.
Optional attributes for
OrderedMapSizeStage (key, values) in the underlying container which behaves like a ordered
associative container.
Optional attributes for
OrderedMapStageOp removes and returns the values associated with the key
from the underlying container.
Optional attributes for
OrderedMapUnstageOp removes and returns the (key, value) element with the smallest
key from the underlying container.
Optional attributes for
OrderedMapUnstageNoKeyA TPU core selector Op.
Retrieves a single tensor from the computation outfeed.
Optional attributes for
OutfeedDequeueRetrieve multiple values from the computation outfeed.
Optional attributes for
OutfeedDequeueTupleRetrieve multiple values from the computation outfeed.
Retrieves a single tensor from the computation outfeed.
Enqueue a Tensor on the computation outfeed.
Enqueue multiple Tensor values on the computation outfeed.
A symbolic handle to a tensor produced by an
Operation.Pads a tensor.
Creates a dataset that batches and pads
batch_size elements from the input.Optional attributes for
PaddedBatchDatasetA queue that produces elements in first-in first-out order.
Optional attributes for
PaddingFifoQueueThe ParallelBatchDataset operation
Optional attributes for
ParallelBatchDatasetConcatenates a list of
N tensors along the first dimension.Interleave the values from the
data tensors into a single tensor.Creates a dataset containing elements of
input_dataset matching predicate.Optional attributes for
ParallelFilterDatasetCreates a dataset that applies
f to the outputs of input_dataset.Creates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
ParallelInterleaveDatasetCreates a dataset that applies
f to the outputs of input_dataset.Optional attributes for
ParallelMapDatasetOutputs random values from a normal distribution.
Optional attributes for
ParameterizedTruncatedNormalTransforms a vector of tf.Example protos (as strings) into typed tensors.
Transforms
input_dataset containing Example protos as vectors of DT_STRING into a dataset of Tensor or SparseTensor objects representing the parsed features.Transforms
input_dataset containing Example protos as vectors of DT_STRING into a dataset of Tensor or SparseTensor objects representing the parsed features.Optional attributes for
ParseExampleDatasetOptional attributes for
ParseExampleDatasetTransforms a vector of tf.io.SequenceExample protos (as strings) into
typed tensors.
Optional attributes for
ParseSequenceExampleTransforms a tf.Example proto (as a string) into typed tensors.
Transforms a scalar brain.SequenceExample proto (as strings) into typed tensors.
Optional attributes for
ParseSingleSequenceExampleTransforms a serialized tensorflow.TensorProto proto into a Tensor.
returns
f(inputs), where f's body is placed and partitioned.Calls a function placed on a specified TPU device.
Optional attributes for
PartitionedCallOptional attributes for
PartitionedCallAn op that groups a list of partitioned inputs together.
Optional attributes for
PartitionedInputAn op that demultiplexes a tensor to be sharded by XLA to a list of partitioned
outputs outside the XLA computation.
A placeholder op for a value that will be fed into the computation.
Optional attributes for
PlaceholderA placeholder op that passes through
input when its output is not fed.Compute the polygamma function \(\psi^{(n)}(x)\).
Computes element-wise population count (a.k.a.
Computes the power of one value to another.
Creates a dataset that asynchronously prefetches elements from
input_dataset.Optional attributes for
PrefetchDatasetAn op which linearizes one Tensor value to an opaque variant tensor.
Optional attributes for
PrelinearizeAn op which linearizes multiple Tensor values to an opaque variant tensor.
Optional attributes for
PrelinearizeTupleAn identity op that triggers an error if a gradient is requested.
Optional attributes for
PreventGradientPrints a string scalar.
Optional attributes for
PrintA queue that produces elements sorted by the first component value.
Optional attributes for
PriorityQueueCreates a dataset that uses a custom thread pool to compute
input_dataset.Creates a dataset that uses a custom thread pool to compute
input_dataset.Computes the product of elements across dimensions of a tensor.
Optional attributes for
ProdComputes the QR decompositions of one or more matrices.
Optional attributes for
QrAn API for building
quantization operations as OpsQuantize the 'input' tensor of type float to 'output' tensor of type 'T'.
Optional attributes for
QuantizeQuantizes then dequantizes a tensor.
Optional attributes for
QuantizeAndDequantizeQuantizes then dequantizes a tensor.
Optional attributes for
QuantizeAndDequantizeV3Quantizes then dequantizes a tensor.
Optional attributes for
QuantizeAndDequantizeV4Returns the gradient of
QuantizeAndDequantizeV4.Optional attributes for
QuantizeAndDequantizeV4GradReturns x + y element-wise, working on quantized buffers.
Produces the average pool of the input tensor for quantized types.
Quantized Batch normalization.
Adds Tensor 'bias' to Tensor 'input' for Quantized types.
Concatenates quantized tensors along one dimension.
Computes a 2D convolution given quantized 4D input and filter tensors.
Optional attributes for
QuantizedConv2dThe QuantizedConv2DAndRelu operation
Optional attributes for
QuantizedConv2DAndReluThe QuantizedConv2DAndReluAndRequantize operation
Optional attributes for
QuantizedConv2DAndReluAndRequantizeThe QuantizedConv2DAndRequantize operation
Optional attributes for
QuantizedConv2DAndRequantizeComputes QuantizedConv2D per channel.
Optional attributes for
QuantizedConv2DPerChannelThe QuantizedConv2DWithBias operation
Optional attributes for
QuantizedConv2DWithBiasThe QuantizedConv2DWithBiasAndRelu operation
Optional attributes for
QuantizedConv2DWithBiasAndReluThe QuantizedConv2DWithBiasAndReluAndRequantize operation
Optional attributes for
QuantizedConv2DWithBiasAndReluAndRequantizeThe QuantizedConv2DWithBiasAndRequantize operation
Optional attributes for
QuantizedConv2DWithBiasAndRequantizeThe QuantizedConv2DWithBiasSignedSumAndReluAndRequantize operation
Optional attributes for
QuantizedConv2DWithBiasSignedSumAndReluAndRequantizeThe QuantizedConv2DWithBiasSumAndRelu operation
Optional attributes for
QuantizedConv2DWithBiasSumAndReluThe QuantizedConv2DWithBiasSumAndReluAndRequantize operation
Optional attributes for
QuantizedConv2DWithBiasSumAndReluAndRequantizeComputes quantized depthwise Conv2D.
Optional attributes for
QuantizedDepthwiseConv2DComputes quantized depthwise Conv2D with Bias.
Optional attributes for
QuantizedDepthwiseConv2DWithBiasComputes quantized depthwise Conv2D with Bias and Relu.
Optional attributes for
QuantizedDepthwiseConv2DWithBiasAndReluComputes quantized depthwise Conv2D with Bias, Relu and Requantize.
Optional attributes for
QuantizedDepthwiseConv2DWithBiasAndReluAndRequantizeQuantized Instance normalization.
Optional attributes for
QuantizedInstanceNormPerform a quantized matrix multiplication of
a by the matrix b.Optional attributes for
QuantizedMatMulPerforms a quantized matrix multiplication of
a by the matrix b with bias
add.Optional attributes for
QuantizedMatMulWithBiasThe QuantizedMatMulWithBiasAndDequantize operation
Optional attributes for
QuantizedMatMulWithBiasAndDequantizePerform a quantized matrix multiplication of
a by the matrix b with bias
add and relu fusion.Optional attributes for
QuantizedMatMulWithBiasAndReluPerform a quantized matrix multiplication of
a by the matrix b with bias
add and relu and requantize fusion.Optional attributes for
QuantizedMatMulWithBiasAndReluAndRequantizeThe QuantizedMatMulWithBiasAndRequantize operation
Optional attributes for
QuantizedMatMulWithBiasAndRequantizeProduces the max pool of the input tensor for quantized types.
Returns x * y element-wise, working on quantized buffers.
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.
Computes Quantized Rectified Linear:
max(features, 0)Computes Quantized Rectified Linear 6:
min(max(features, 0), 6)Computes Quantized Rectified Linear X:
min(max(features, 0), max_value)Reshapes a quantized tensor as per the Reshape op.
Resize quantized
images to size using quantized bilinear interpolation.Optional attributes for
QuantizedResizeBilinearCloses the given queue.
Optional attributes for
QueueCloseDequeues a tuple of one or more tensors from the given queue.
Optional attributes for
QueueDequeueDequeues
n tuples of one or more tensors from the given queue.Optional attributes for
QueueDequeueManyDequeues
n tuples of one or more tensors from the given queue.Optional attributes for
QueueDequeueUpToEnqueues a tuple of one or more tensors in the given queue.
Optional attributes for
QueueEnqueueEnqueues zero or more tuples of one or more tensors in the given queue.
Optional attributes for
QueueEnqueueManyReturns true if queue is closed.
Computes the number of elements in the given queue.
Counts the number of occurrences of each value in an integer array.
Optional attributes for
RaggedBincountPerforms sparse-output bin counting for a ragged tensor input.
Optional attributes for
RaggedCountSparseOutputGenerates a feature cross from a list of tensors, and returns it as a
RaggedTensor.
The RaggedFillEmptyRows operation
The RaggedFillEmptyRowsGrad operation
Gather ragged slices from
params axis 0 according to indices.An API for building
ragged operations as OpsReturns a
RaggedTensor containing the specified sequences of numbers.Decodes a
variant Tensor into a RaggedTensor.Converts a
RaggedTensor into a SparseTensor with the same values.Create a dense tensor from a ragged tensor, possibly altering its shape.
Encodes a
RaggedTensor into a variant Tensor.Helper used to compute the gradient for
RaggedTensorToVariant.Randomly crop
image.Optional attributes for
RandomCropCreates a Dataset that returns pseudorandom numbers.
Creates a Dataset that returns pseudorandom numbers.
Optional attributes for
RandomDatasetAn API for building
random.experimental operations as OpsOutputs random values from the Gamma distribution(s) described by alpha.
Optional attributes for
RandomGammaComputes the derivative of a Gamma random sample w.r.t.
Outputs the position of
value in a permutation of [0, ..., max_index].Optional attributes for
RandomIndexShuffleAn API for building
random operations as OpsOutputs random values from the Poisson distribution(s) described by rate.
Optional attributes for
RandomPoissonRandomly shuffles a tensor along its first dimension.
Optional attributes for
RandomShuffleA queue that randomizes the order of elements.
Optional attributes for
RandomShuffleQueueOutputs random values from a normal distribution.
Optional attributes for
RandomStandardNormalOutputs random values from a uniform distribution.
Optional attributes for
RandomUniformOutputs random integers from a uniform distribution.
Optional attributes for
RandomUniformIntCreates a sequence of numbers.
Creates a dataset with a range of values.
Optional attributes for
RangeDatasetReturns the rank of a tensor.
A custom gradient for an op of unspecified type.
A base class for operation input accessors.
A tensor which memory has not been mapped to a data space directly accessible from the JVM.
Returns the number of records this Reader has produced.
Returns the number of work units this Reader has finished processing.
Returns the next record (key, value pair) produced by a Reader.
Returns up to
num_records (key, value) pairs produced by a Reader.Restore a Reader to its initial clean state.
Restore a reader to a previously saved state.
Produce a string tensor that encodes the state of a Reader.
Reads and outputs the entire contents of the input filename.
Reads the value of a variable.
Splits resource variable input tensor across all dimensions.
Optional attributes for
ReadVariableSplitNDReturns the real part of a complex number.
Returns x / y element-wise for real types.
Creates a dataset that changes the batch size.
Optional attributes for
RebatchDatasetCreates a dataset that changes the batch size.
Computes the reciprocal of x element-wise.
Computes the gradient for the inverse of
x wrt its input.Emits randomized records.
Optional attributes for
RecordInputReceives the named tensor from send_device on recv_device.
Optional attributes for
RecvAn op that receives embedding activations on the TPU.
Computes the "logical and" of elements across dimensions of a tensor.
Optional attributes for
ReduceAllComputes the "logical or" of elements across dimensions of a tensor.
Optional attributes for
ReduceAnyReduces the input dataset to a singleton using a reduce function.
Optional attributes for
ReduceDatasetJoins a string Tensor across the given dimensions.
Optional attributes for
ReduceJoinComputes the maximum of elements across dimensions of a tensor.
Optional attributes for
ReduceMaxComputes the minimum of elements across dimensions of a tensor.
Optional attributes for
ReduceMinComputes the product of elements across dimensions of a tensor.
Optional attributes for
ReduceProdComputes the sum of elements across dimensions of a tensor.
Optional attributes for
ReduceSumCreates or finds a child frame, and makes
data available to the child frame.Optional attributes for
RefEnterExits the current frame to its parent frame.
Return the same ref tensor as the input ref tensor.
Forwards the value of an available tensor from
inputs to output.Makes its input available to the next iteration.
Forwards the
indexth element of inputs to output.Forwards the ref tensor
data to the output port determined by pred.Check if the input matches the regex pattern.
Replaces matches of the
pattern regular expression in input with the
replacement string provided in rewrite.Optional attributes for
RegexReplaceRegisters a dataset with the tf.data service.
Optional attributes for
RegisterDatasetThe Relayout operation
The RelayoutLike operation
Computes rectified linear:
max(features, 0).Computes rectified linear 6:
min(max(features, 0), 6).Computes rectified linear 6 gradients for a Relu6 operation.
Computes rectified linear gradients for a Relu operation.
Runs function
f on a remote device indicated by target.Creates a dataset that emits the outputs of
input_dataset count times.Optional attributes for
RepeatDatasetConnects N inputs to an N-way replicated TPU computation.
Optional attributes for
ReplicatedInputConnects N outputs from an N-way replicated TPU computation.
Metadata indicating how the TPU computation should be replicated.
Optional attributes for
ReplicateMetadataComputes a range that covers the actual values present in a quantized tensor.
Computes requantization range per channel.
Converts the quantized
input tensor into a lower-precision output.Requantizes input with min and max values known per channel.
Reshapes a tensor.
Resize
images to size using area interpolation.Optional attributes for
ResizeAreaResize
images to size using bicubic interpolation.Optional attributes for
ResizeBicubicComputes the gradient of bicubic interpolation.
Optional attributes for
ResizeBicubicGradResize
images to size using bilinear interpolation.Optional attributes for
ResizeBilinearComputes the gradient of bilinear interpolation.
Optional attributes for
ResizeBilinearGradResize
images to size using nearest neighbor interpolation.Optional attributes for
ResizeNearestNeighborComputes the gradient of nearest neighbor interpolation.
Optional attributes for
ResizeNearestNeighborGradApplies a gradient to a given accumulator.
Returns the number of gradients aggregated in the given accumulators.
Updates the accumulator with a new value for global_step.
Extracts the average gradient in the given ConditionalAccumulator.
Update '*var' according to the adadelta scheme.
Optional attributes for
ResourceApplyAdadeltaUpdate '*var' according to the adagrad scheme.
Optional attributes for
ResourceApplyAdagradUpdate '*var' according to the proximal adagrad scheme.
Optional attributes for
ResourceApplyAdagradDaUpdate '*var' according to the Adam algorithm.
Optional attributes for
ResourceApplyAdamUpdate '*var' according to the AdaMax algorithm.
Optional attributes for
ResourceApplyAdaMaxUpdate '*var' according to the Adam algorithm.
Optional attributes for
ResourceApplyAdamWithAmsgradUpdate '*var' according to the AddSign update.
Optional attributes for
ResourceApplyAddSignUpdate '*var' according to the centered RMSProp algorithm.
Optional attributes for
ResourceApplyCenteredRmsPropUpdate '*var' according to the Ftrl-proximal scheme.
Optional attributes for
ResourceApplyFtrlUpdate '*var' by subtracting 'alpha' * 'delta' from it.
Optional attributes for
ResourceApplyGradientDescentUpdate '*var' according to the momentum scheme.
Optional attributes for
ResourceApplyKerasMomentumUpdate '*var' according to the momentum scheme.
Optional attributes for
ResourceApplyMomentumUpdate '*var' according to the AddSign update.
Optional attributes for
ResourceApplyPowerSignUpdate '*var' and '*accum' according to FOBOS with Adagrad learning rate.
Optional attributes for
ResourceApplyProximalAdagradUpdate '*var' as FOBOS algorithm with fixed learning rate.
Optional attributes for
ResourceApplyProximalGradientDescentUpdate '*var' according to the RMSProp algorithm.
Optional attributes for
ResourceApplyRmsPropA conditional accumulator for aggregating gradients.
Optional attributes for
ResourceConditionalAccumulatorIncrements variable pointed to by 'resource' until it reaches 'limit'.
Gather slices from the variable pointed to by
resource according to indices.Optional attributes for
ResourceGatherThe ResourceGatherNd operation
Adds sparse updates to the variable referenced by
resource.Divides sparse updates into the variable referenced by
resource.Reduces sparse updates into the variable referenced by
resource using the max operation.Reduces sparse updates into the variable referenced by
resource using the min operation.Multiplies sparse updates into the variable referenced by
resource.Applies sparse addition to individual values or slices in a Variable.
Optional attributes for
ResourceScatterNdAddThe ResourceScatterNdMax operation
Optional attributes for
ResourceScatterNdMaxThe ResourceScatterNdMin operation
Optional attributes for
ResourceScatterNdMinApplies sparse subtraction to individual values or slices in a Variable.
Optional attributes for
ResourceScatterNdSubApplies sparse
updates to individual values or slices within a given
variable according to indices.Optional attributes for
ResourceScatterNdUpdateSubtracts sparse updates from the variable referenced by
resource.Assigns sparse updates to the variable referenced by
resource.var: Should be from a Variable().
Optional attributes for
ResourceSparseApplyAdadeltaUpdate relevant entries in '*var' and '*accum' according to the adagrad scheme.
Optional attributes for
ResourceSparseApplyAdagradUpdate entries in '*var' and '*accum' according to the proximal adagrad scheme.
Optional attributes for
ResourceSparseApplyAdagradDaUpdate relevant entries in '*var' and '*accum' according to the adagrad scheme.
Optional attributes for
ResourceSparseApplyAdagradV2Update '*var' according to the centered RMSProp algorithm.
Optional attributes for
ResourceSparseApplyCenteredRmsPropUpdate relevant entries in '*var' according to the Ftrl-proximal scheme.
Optional attributes for
ResourceSparseApplyFtrlUpdate relevant entries in '*var' and '*accum' according to the momentum scheme.
Optional attributes for
ResourceSparseApplyKerasMomentumUpdate relevant entries in '*var' and '*accum' according to the momentum scheme.
Optional attributes for
ResourceSparseApplyMomentumSparse update entries in '*var' and '*accum' according to FOBOS algorithm.
Optional attributes for
ResourceSparseApplyProximalAdagradSparse update '*var' as FOBOS algorithm with fixed learning rate.
Optional attributes for
ResourceSparseApplyProximalGradientDescentUpdate '*var' according to the RMSProp algorithm.
Optional attributes for
ResourceSparseApplyRmsPropAssign
value to the sliced l-value reference of ref.Optional attributes for
ResourceStridedSliceAssignRestores tensors from a V2 checkpoint.
Restores a tensor from checkpoint files.
Optional attributes for
RestoreSliceAn op that retrieves optimization parameters from embedding to host memory.
Retrieve Adadelta embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingAdadeltaParametersRetrieve Adagrad Momentum embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingAdagradMomentumParametersRetrieve Adagrad embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingAdagradParametersRetrieve ADAM embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingADAMParametersRetrieve centered RMSProp embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingCenteredRMSPropParametersRetrieve frequency estimator embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingFrequencyEstimatorParametersRetrieve FTRL embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingFTRLParametersRetrieve MDL Adagrad Light embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingMDLAdagradLightParametersRetrieve Momentum embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingMomentumParametersRetrieve proximal Adagrad embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingProximalAdagradParametersThe RetrieveTPUEmbeddingProximalYogiParameters operation
Optional attributes for
RetrieveTPUEmbeddingProximalYogiParametersRetrieve RMSProp embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingRMSPropParametersRetrieve SGD embedding parameters.
Optional attributes for
RetrieveTPUEmbeddingStochasticGradientDescentParametersReverses specific dimensions of a tensor.
Reverses variable length slices.
Optional attributes for
ReverseSequenceThe RewriteDataset operation
Real-valued fast Fourier transform.
2D real-valued fast Fourier transform.
3D real-valued fast Fourier transform.
ND fast real Fourier transform.
Converts one or more images from RGB to HSV.
Elementwise computes the bitwise right-shift of
x and y.Returns element-wise integer closest to x.
Advance the counter of a counter-based RNG.
Advance the counter of a counter-based RNG.
Rolls the elements of a tensor along an axis.
Rounds the values of a tensor to the nearest integer, element-wise.
Computes reciprocal of square root of x element-wise.
Computes the gradient for the rsqrt of
x wrt its input.Generate a single randomly distorted bounding box for an image.
Optional attributes for
SampleDistortedBoundingBoxCreates a dataset that takes a Bernoulli sample of the contents of another dataset.
Saves tensors in V2 checkpoint format.
The SaveDatasetV2 operation
Optional attributes for
SaveDatasetSavedModelBundle represents a model loaded from storage.
Options for exporting a SavedModel.
Options for loading a SavedModel.
Saves input tensors slices to disk.
Outputs a
Summary protocol buffer with scalar values.The ScaleAndTranslate operation
Optional attributes for
ScaleAndTranslateThe ScaleAndTranslateGrad operation
Optional attributes for
ScaleAndTranslateGradCreates a dataset successively reduces
f over the elements of input_dataset.Creates a dataset successively reduces
f over the elements of input_dataset.Optional attributes for
ScanDatasetOptional attributes for
ScanDatasetAdds sparse updates to a variable reference.
Optional attributes for
ScatterAddDivides a variable reference by sparse updates.
Optional attributes for
ScatterDivReduces sparse updates into a variable reference using the
max operation.Optional attributes for
ScatterMaxReduces sparse updates into a variable reference using the
min operation.Optional attributes for
ScatterMinMultiplies sparse updates into a variable reference.
Optional attributes for
ScatterMulScatters
updates into a tensor of shape shape according to indices.Applies sparse addition to individual values or slices in a Variable.
Optional attributes for
ScatterNdAddComputes element-wise maximum.
Optional attributes for
ScatterNdMaxComputes element-wise minimum.
Optional attributes for
ScatterNdMinApplies sparse addition to
input using individual values or slices
from updates according to indices indices.Applies sparse subtraction to individual values or slices in a Variable.
Optional attributes for
ScatterNdSubApplies sparse
updates to individual values or slices within a given
variable according to indices.Optional attributes for
ScatterNdUpdateSubtracts sparse updates to a variable reference.
Optional attributes for
ScatterSubApplies sparse updates to a variable reference.
Optional attributes for
ScatterUpdateManages groups of related properties when creating Tensorflow Operations, such as a common name
prefix.
Computes fingerprints of the input strings.
Distributed version of Stochastic Dual Coordinate Ascent (SDCA) optimizer for
linear models with L1 + L2 regularization.
Optional attributes for
SdcaOptimizerApplies L1 regularization shrink step on the parameters.
Computes the maximum along segments of a tensor.
Computes the mean along segments of a tensor.
Computes the minimum along segments of a tensor.
Computes the product along segments of a tensor.
Computes the sum along segments of a tensor.
The SelectV2 operation
Computes the eigen decomposition of one or more square self-adjoint matrices.
Optional attributes for
SelfAdjointEigComputes scaled exponential linear:
scale * alpha * (exp(features) - 1)
if < 0, scale * features otherwise.Computes gradients for the scaled exponential linear (Selu) operation.
Sends the named tensor from send_device to recv_device.
Optional attributes for
SendPerforms gradient updates of embedding tables.
Optional attributes for
SendTPUEmbeddingGradientsConverts the given
resource_handle representing an iterator to a variant tensor.Optional attributes for
SerializeIteratorSerialize an
N-minibatch SparseTensor into an [N, 3] Tensor object.Serialize a
SparseTensor into a [3] Tensor object.Transforms a Tensor into a serialized TensorProto proto.
An in-process TensorFlow server, for use in distributed training.
Driver for
Graph execution.A callable function backed by a session.
Computes the difference between two lists of numbers or strings.
Number of unique elements along last dimension of input
set.Optional attributes for
SetSizeThe ExperimentalSetStatsAggregatorDataset operation
The SetStatsAggregatorDataset operation
Returns the shape of a tensor.
Returns shape of tensors.
An API for building
shape operations as OpsAn operator providing methods on org.tensorflow.op.core.Shape tensors and 1d operands that
represent the dimensions of a shape.
Creates a
Dataset that includes only 1/num_shards of this dataset.Optional attributes for
ShardDatasetGenerate a sharded filename.
Generate a glob pattern matching all sharded file names.
The ShuffleAndRepeatDatasetV2 operation
Optional attributes for
ShuffleAndRepeatDatasetThe ShuffleDatasetV3 operation
Optional attributes for
ShuffleDatasetShuts down a running distributed TPU system.
An op that shuts down the TPU system.
Computes sigmoid of
x element-wise.Computes the gradient of the sigmoid of
x wrt its input.Returns an element-wise indication of the sign of a number.
An API for building
signal operations as OpsDescribe the inputs and outputs of an executable entity, such as a
ConcreteFunction,
among other useful metadata.Builds a new function signature.
Computes sine of x element-wise.
Computes hyperbolic sine of x element-wise.
Returns the size of a tensor.
Creates a dataset that skips
count elements from the input_dataset.Optional attributes for
SkipDatasetParses a text file and creates a batch of examples.
Optional attributes for
SkipgramThe ExperimentalSleepDataset operation
The SleepDataset operation
Return a slice from 'input'.
Creates a dataset that passes a sliding window over
input_dataset.Creates a dataset that passes a sliding window over
input_dataset.Optional attributes for
SlidingWindowDatasetReturns a copy of the input tensor.
The SnapshotChunkDataset operation
Optional attributes for
SnapshotChunkDatasetCreates a dataset that will write to / read from a snapshot.
Optional attributes for
SnapshotDatasetThe SnapshotDatasetReader operation
Optional attributes for
SnapshotDatasetReaderThe SnapshotNestedDatasetReader operation
Generates points from the Sobol sequence.
Computes softmax activations.
Computes softmax cross entropy cost and gradients to backpropagate.
The Softplus operation
Computes softplus gradients for a softplus operation.
Computes softsign:
features / (abs(features) + 1).Computes softsign gradients for a softsign operation.
Solves systems of linear equations.
Optional attributes for
SolveSpaceToBatch for 4-D tensors of type T.
SpaceToBatch for N-D tensors of type T.
SpaceToDepth for tensors of type T.
Optional attributes for
SpaceToDepthApplies a sparse gradient to a given accumulator.
Extracts the average sparse gradient in a SparseConditionalAccumulator.
Adds two
SparseTensor objects to produce another SparseTensor.The gradient operator for the SparseAdd op.
var: Should be from a Variable().
Optional attributes for
SparseApplyAdadeltaUpdate relevant entries in '*var' and '*accum' according to the adagrad scheme.
Optional attributes for
SparseApplyAdagradUpdate entries in '*var' and '*accum' according to the proximal adagrad scheme.
Optional attributes for
SparseApplyAdagradDaUpdate '*var' according to the centered RMSProp algorithm.
Optional attributes for
SparseApplyCenteredRmsPropUpdate relevant entries in '*var' according to the Ftrl-proximal scheme.
Optional attributes for
SparseApplyFtrlUpdate relevant entries in '*var' and '*accum' according to the momentum scheme.
Optional attributes for
SparseApplyMomentumSparse update entries in '*var' and '*accum' according to FOBOS algorithm.
Optional attributes for
SparseApplyProximalAdagradSparse update '*var' as FOBOS algorithm with fixed learning rate.
Optional attributes for
SparseApplyProximalGradientDescentUpdate '*var' according to the RMSProp algorithm.
Optional attributes for
SparseApplyRmsPropCounts the number of occurrences of each value in an integer array.
Optional attributes for
SparseBincountConcatenates a list of
SparseTensor along the specified dimension.A conditional accumulator for aggregating sparse gradients.
Optional attributes for
SparseConditionalAccumulatorPerforms sparse-output bin counting for a sparse tensor input.
Optional attributes for
SparseCountSparseOutputGenerates sparse cross from a list of sparse and dense tensors.
Generates sparse cross from a list of sparse and dense tensors.
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.
Component-wise divides a SparseTensor by a dense Tensor.
Component-wise multiplies a SparseTensor by a dense Tensor.
Fills empty rows in the input 2-D
SparseTensor with a default value.The gradient of SparseFillEmptyRows.
Multiply matrix "a" by matrix "b".
Optional attributes for
SparseMatMulSparse addition of two CSR matrices, C = alpha * A + beta * B.
Matrix-multiplies a sparse matrix with a dense matrix.
Optional attributes for
SparseMatrixMatMulElement-wise multiplication of a sparse matrix with a dense tensor.
Returns the number of nonzeroes of
sparse_matrix.Computes the Approximate Minimum Degree (AMD) ordering of
input.Calculates the softmax of a CSRSparseMatrix.
Calculates the gradient of the SparseMatrixSoftmax op.
Computes the sparse Cholesky decomposition of
input.Sparse-matrix-multiplies two CSR matrices
a and b.Optional attributes for
SparseMatrixSparseMatMulTransposes the inner (matrix) dimensions of a CSRSparseMatrix.
Optional attributes for
SparseMatrixTransposeCreates an all-zeros CSRSparseMatrix with shape
dense_shape.An API for building
sparse operations as OpsComputes the max of elements across dimensions of a SparseTensor.
Optional attributes for
SparseReduceMaxComputes the max of elements across dimensions of a SparseTensor.
Optional attributes for
SparseReduceMaxSparseComputes the sum of elements across dimensions of a SparseTensor.
Optional attributes for
SparseReduceSumComputes the sum of elements across dimensions of a SparseTensor.
Optional attributes for
SparseReduceSumSparseReorders a SparseTensor into the canonical, row-major ordering.
Reshapes a SparseTensor to represent values in a new dense shape.
Computes the mean along sparse segments of a tensor.
Optional attributes for
SparseSegmentMeanComputes gradients for SparseSegmentMean.
Computes the mean along sparse segments of a tensor.
Optional attributes for
SparseSegmentMeanWithNumSegmentsComputes the sum along sparse segments of a tensor divided by the sqrt of N.
Optional attributes for
SparseSegmentSqrtNComputes gradients for SparseSegmentSqrtN.
Computes the sum along sparse segments of a tensor divided by the sqrt of N.
Optional attributes for
SparseSegmentSqrtNWithNumSegmentsComputes the sum along sparse segments of a tensor.
Optional attributes for
SparseSegmentSumComputes gradients for SparseSegmentSum.
Computes the sum along sparse segments of a tensor.
Optional attributes for
SparseSegmentSumWithNumSegmentsSlice a
SparseTensor based on the start and size.The gradient operator for the SparseSlice op.
Applies softmax to a batched N-D
SparseTensor.Computes softmax cross entropy cost and gradients to backpropagate.
Returns the element-wise max of two SparseTensors.
Returns the element-wise min of two SparseTensors.
Split a
SparseTensor into num_split tensors along one dimension.A virtual type of
Tensor composed of three dense tensors (indices, values and dimensions)
used to represent the sparse data into a multi-dimensional dense space.Adds up a
SparseTensor and a dense Tensor, producing a dense Tensor.Multiply SparseTensor (of rank 2) "A" by dense matrix "B".
Optional attributes for
SparseTensorDenseMatMulCreates a dataset that splits a SparseTensor into elements row-wise.
Converts a SparseTensor to a (possibly batched) CSRSparseMatrix.
Converts a sparse representation into a dense tensor.
Optional attributes for
SparseToDenseApplies set operation along last dimension of 2
SparseTensor inputs.Optional attributes for
SparseToSparseSetOperationThe Spence operation
Splits a tensor into
num_split tensors along one dimension.An op splits input deduplication data XLA tuple into integer and floating point
tensors.
Optional attributes for
SplitDedupDataSplits input tensor across all dimensions.
Optional attributes for
SplitNDSplits a tensor into
num_split tensors along one dimension.Creates a dataset that executes a SQL query and emits rows of the result set.
Creates a dataset that executes a SQL query and emits rows of the result set.
Computes square root of x element-wise.
Computes the gradient for the sqrt of
x wrt its input.Computes the matrix square root of one or more square matrices:
matmul(sqrtm(A), sqrtm(A)) = A
Computes square of x element-wise.
Returns conj(x - y)(x - y) element-wise.
Removes dimensions of size 1 from the shape of a tensor.
Optional attributes for
SqueezePacks a list of
N rank-R tensors into one rank-(R+1) tensor.Optional attributes for
StackDelete the stack from its resource container.
A stack that produces elements in first-in last-out order.
Optional attributes for
StackCreatePop the element at the top of the stack.
Push an element onto the stack.
Optional attributes for
StackPushStage values similar to a lightweight Enqueue.
Optional attributes for
StageOp removes all elements in the underlying container.
Optional attributes for
StageClearOp peeks at the values at the specified index.
Optional attributes for
StagePeekOp returns the number of elements in the underlying container.
Optional attributes for
StageSizeAn n-way switch statement which calls a single branch function.
output = cond ? then_branch(input) : else_branch(input)
returns
f(inputs), where f's body is placed and partitioned.Optional attributes for
StatefulPartitionedCallThe StatefulRandomBinomial operation
Outputs random values from a normal distribution.
Outputs random values from a truncated normal distribution.
Outputs random values from a uniform distribution.
Outputs random integers from a uniform distribution.
Outputs random integers from a uniform distribution.
output = input; While (Cond(output)) { output = Body(output) }
An n-way switch statement which calls a single branch function.
output = cond ? then_branch(input) : else_branch(input)
Draws samples from a multinomial distribution.
The StatelessParameterizedTruncatedNormal operation
Outputs deterministic pseudorandom random numbers from a binomial distribution.
Outputs deterministic pseudorandom random numbers from a gamma distribution.
Picks the best counter-based RNG algorithm based on device.
Scrambles seed into key and counter, using the best algorithm based on device.
Picks the best algorithm based on device, and scrambles seed into key and counter.
Outputs deterministic pseudorandom values from a normal distribution.
Outputs deterministic pseudorandom values from a normal distribution.
Outputs deterministic pseudorandom random numbers from a Poisson distribution.
Outputs deterministic pseudorandom random values from a uniform distribution.
Outputs deterministic pseudorandom random integers from a uniform distribution.
Outputs deterministic pseudorandom random integers from a uniform distribution.
Outputs deterministic pseudorandom random integers from a uniform distribution.
Outputs deterministic pseudorandom random integers from a uniform distribution.
Outputs deterministic pseudorandom random values from a uniform distribution.
Generate a randomly distorted bounding box for an image deterministically.
Optional attributes for
StatelessSampleDistortedBoundingBoxRandomly and deterministically shuffles a tensor along its first dimension.
Outputs deterministic pseudorandom values from a truncated normal distribution.
Outputs deterministic pseudorandom values from a truncated normal distribution.
output = input; While (Cond(output)) { output = Body(output) }
Check if the input matches the regex pattern.
Replaces the match of pattern in input with rewrite.
Optional attributes for
StaticRegexReplaceCreates a statistics manager resource.
The StatsAggregatorHandleV2 operation
Optional attributes for
StatsAggregatorHandleOptional attributes for
StatsAggregatorHandleSet a summary_writer_interface to record statistics using given stats_aggregator.
Produces a summary of any statistics recorded by the given statistics manager.
Produces a summary of any statistics recorded by the given statistics manager.
Stochastically cast a given tensor from floats to ints.
Stops gradient computation.
The StoreMinibatchStatisticsInFdo operation
Return a strided slice from
input.Optional attributes for
StridedSliceAssign
value to the sliced l-value reference of ref.Optional attributes for
StridedSliceAssignReturns the gradient of
StridedSlice.Optional attributes for
StridedSliceGradHelper endpoint methods for Python like indexing.
Formats a string template using a list of tensors.
Optional attributes for
StringFormatString lengths of
input.Optional attributes for
StringLengthCreates ngrams from ragged string data.
An API for building
strings operations as OpsSplit elements of
source based on sep into a SparseTensor.Optional attributes for
StringSplitStrip leading and trailing whitespaces from the Tensor.
Returns x - y element-wise.
Return substrings from
Tensor of strings.Optional attributes for
SubstrComputes the sum of elements across dimensions of a tensor.
Optional attributes for
SumAn API for building
summary operations as OpsThe SummaryWriter operation
Optional attributes for
SummaryWriterComputes the singular value decompositions of one or more matrices.
Optional attributes for
SvdForwards
data to the output port determined by pred.Computes the gradient function for function f via backpropagation.
Synchronizes the device this op is run on.
Creates a dataset that contains
count elements from the input_dataset.Optional attributes for
TakeDatasetRead
SparseTensors from a SparseTensorsMap and concatenate them.Optional attributes for
TakeManySparseFromTensorsMapCreates a dataset that stops iteration when predicate` is false.
Creates a dataset that stops iteration when predicate` is false.
Optional attributes for
TakeWhileDatasetComputes tan of x element-wise.
Computes hyperbolic tangent of
x element-wise.Computes the gradient for the tanh of
x wrt its input.Brain 16-bit float tensor type.
Boolean tensor type.
Returns a tensor that may be mutated, but only persists within a single step.
Optional attributes for
TemporaryVariableA statically typed multi-dimensional array.
An array of Tensors of given size.
Optional attributes for
TensorArrayDelete the TensorArray from its resource container.
Concat the elements from the TensorArray into value
value.Optional attributes for
TensorArrayConcatGather specific elements from the TensorArray into output
value.Optional attributes for
TensorArrayGatherCreates a TensorArray for storing the gradients of values in the given handle.
Creates a TensorArray for storing multiple gradients of values in the given handle.
The TensorArrayPack operation
Optional attributes for
TensorArrayPackRead an element from the TensorArray into output
value.Scatter the data from the input value into specific TensorArray elements.
Get the current size of the TensorArray.
Split the data from the input value into TensorArray elements.
The TensorArrayUnpack operation
Push an element onto the tensor_array.
Creates a dataset that emits
components as a tuple of tensors once.Optional attributes for
TensorDatasetReturns a diagonal tensor with a given diagonal values.
Returns the diagonal part of the tensor.
Static utility methods describing the TensorFlow runtime.
A function that can be called with tensors.
Concats all tensors in the list along the 0th dimension.
The TensorListConcatLists operation
The shape of the elements of the given list, as a tensor.
Creates a TensorList which, when stacked, has the value of
tensor.Creates a Tensor by indexing into the TensorList.
Returns the item in the list with the given index.
Returns the number of tensors in the input tensor list.
Returns the last element of the input list as well as a list with all but that element.
Returns a list which has the passed-in
Tensor as last element and the other elements of the given list in input_handle.The TensorListPushBackBatch operation
List of the given size with empty elements.
Resizes the list.
Creates a TensorList by indexing into a Tensor.
Scatters tensor at indices in an input list.
Sets the index-th position of the list to contain the given tensor.
Optional attributes for
TensorListSetItemSplits a tensor into a list.
Stacks all tensors in the list.
Optional attributes for
TensorListStackReturns a tensor map with item from given key erased.
Returns whether the given key exists in the map.
Returns a map that is the 'input_handle' with the given key-value pair inserted.
Returns the value from a given key in a tensor map.
Maps the native memory of a
RawTensor to a n-dimensional typed data space accessible from
the JVM.Returns the number of tensors in the input tensor map.
Returns a Tensor stack of all keys in a tensor map.
Adds sparse
updates to an existing tensor according to indices.Apply a sparse update to a tensor taking the element-wise maximum.
The TensorScatterMin operation
Subtracts sparse
updates from an existing tensor according to indices.Scatter
updates into an existing tensor according to indices.Creates a dataset that emits each dim-0 slice of
components once.Optional attributes for
TensorSliceDatasetAssign
value to the sliced l-value reference of input.Optional attributes for
TensorStridedSliceUpdateOutputs a
Summary protocol buffer with a tensor and per-plugin data.Annotation for all tensor types.
Creates a dataset that emits the lines of one or more text files.
Optional attributes for
TextLineDatasetA Reader that outputs the lines of a file delimited by '\n'.
Optional attributes for
TextLineReaderIEEE-754 half-precision 16-bit float tensor type.
IEEE-754 single-precision 32-bit float tensor type.
IEEE-754 double-precision 64-bit float tensor type.
Common interface for all floating point tensors.
Creates a dataset that emits the records from one or more TFRecord files.
Optional attributes for
TfRecordDatasetA Reader that outputs the records from a TensorFlow Records file.
Optional attributes for
TfRecordReaderCreates a dataset that uses a custom thread pool to compute
input_dataset.Creates a dataset that uses a custom thread pool to compute
input_dataset.Creates a dataset that uses a custom thread pool to compute
input_dataset.Creates a dataset that uses a custom thread pool to compute
input_dataset.Optional attributes for
ThreadPoolHandleOptional attributes for
ThreadPoolHandleGenerates labels for candidate sampling with a learned unigram distribution.
Optional attributes for
ThreadUnsafeUnigramCandidateSamplerConstructs a tensor by tiling a given tensor.
Returns the gradient of
Tile.Provides the time since epoch in seconds.
32-bit signed integer tensor type.
64-bit signed integer tensor type.
Common interface for all integral numeric tensors.
Common interface for all numeric tensors.
Converts a tensor to a scalar predicate.
Converts each string in the input Tensor to its hash mod by a number of buckets.
Converts each string in the input Tensor to its hash mod by a number of buckets.
Converts each string in the input Tensor to its hash mod by a number of buckets.
Converts each string in the input Tensor to the specified numeric type.
Finds values and indices of the
k largest elements for the last dimension.Optional attributes for
TopKReturns the TopK unique values in the array in sorted order.
Returns the TopK values in the array in sorted order.
The TPUAnnotateTensorsWithDynamicShape operation
Deprecated.
use
CompilationResult insteadOp that copies host tensor to device with dynamic shape support.
Deprecated.
use
EmbeddingActivations insteadConverts XRT's uid handles to TensorFlow-friendly input format.
An API for building
tpu operations as OpsDeprecated.
use
ReplicatedInput insteadOptional attributes for
TPUReplicatedInputDeprecated.
use
ReplicatedOutput insteadDeprecated.
use
ReplicateMetadata insteadOptional attributes for
TPUReplicateMetadataOp that reshards on-device TPU variables to specified state.
Round-robin load balancing on TPU cores.
An API for building
train operations as OpsShuffle dimensions of x according to a permutation.
Solves systems of linear equations with upper or lower triangular matrices by backsubstitution.
Optional attributes for
TriangularSolveCalculate product with tridiagonal matrix.
Solves tridiagonal systems of equations.
Optional attributes for
TridiagonalSolveReturns x / y element-wise, rounded towards zero.
Outputs random values from a truncated normal distribution.
Optional attributes for
TruncatedNormalReturns element-wise remainder of division.
String type.
Common interface for all typed tensors.
16-bit unsigned integer tensor type.
8-bit unsigned integer tensor type.
Reverses the operation of Batch for a single output Tensor.
Optional attributes for
UnbatchA dataset that splits the elements of its input into multiple elements.
A dataset that splits the elements of its input into multiple elements.
Optional attributes for
UnbatchDatasetGradient of Unbatch.
Optional attributes for
UnbatchGradUncompresses a compressed dataset element.
Decodes each string in
input into a sequence of Unicode code points.Optional attributes for
UnicodeDecodeDecodes each string in
input into a sequence of Unicode code points.Optional attributes for
UnicodeDecodeWithOffsetsEncode a tensor of ints into unicode strings.
Optional attributes for
UnicodeEncodeDetermine the script codes of a given tensor of Unicode integer code points.
Transcode the input text from a source encoding to a destination encoding.
Optional attributes for
UnicodeTranscodeGenerates labels for candidate sampling with a uniform distribution.
Optional attributes for
UniformCandidateSamplerPerform dequantization on the quantized Tensor
input.Optional attributes for
UniformDequantizePerform quantization on Tensor
input.Optional attributes for
UniformQuantizePerform quantized add of quantized Tensor
lhs and quantized Tensor rhs to make quantized output.Optional attributes for
UniformQuantizedAddPerform clip by value on the quantized Tensor
operand.Optional attributes for
UniformQuantizedClipByValuePerform quantized convolution of quantized Tensor
lhs and quantized Tensor rhs.Optional attributes for
UniformQuantizedConvolutionPerform hybrid quantized convolution of float Tensor
lhs and quantized Tensor rhs.Optional attributes for
UniformQuantizedConvolutionHybridPerform quantized dot of quantized Tensor
lhs and quantized Tensor rhs to make quantized output.Optional attributes for
UniformQuantizedDotPerform hybrid quantized dot of float Tensor
lhs and quantized Tensor rhs.Optional attributes for
UniformQuantizedDotHybridGiven quantized tensor
input, requantize it with new quantization parameters.Optional attributes for
UniformRequantizeFinds unique elements along an axis of a tensor.
Creates a dataset that contains the unique elements of
input_dataset.Creates a dataset that contains the unique elements of
input_dataset.Optional attributes for
UniqueDatasetFinds unique elements along an axis of a tensor.
Converts an array of flat indices into a tuple of coordinate arrays.
The UnsortedSegmentJoin operation
Optional attributes for
UnsortedSegmentJoinComputes the maximum along segments of a tensor.
Computes the minimum along segments of a tensor.
Computes the product along segments of a tensor.
Computes the sum along segments of a tensor.
Unpacks a given dimension of a rank-
R tensor into num rank-(R-1) tensors.Optional attributes for
UnstackOp is similar to a lightweight Dequeue.
Optional attributes for
UnstageThe UnwrapDatasetVariant operation
Converts all lowercase characters into their respective uppercase replacements.
Optional attributes for
UpperApplies upper_bound(sorted_search_values, values) along each row.
Creates a handle to a Variable resource.
Optional attributes for
VarHandleOpHolds state in the form of a tensor that persists across steps.
Optional attributes for
VariableReturns the shape of the variable pointed to by
resource.Checks whether a resource handle-based variable has been initialized.
Returns locations of nonzero / true values in a tensor.
output = input; While (Cond(output)) { output = Body(output) }
Optional attributes for
WhileA Reader that outputs the entire contents of a file as a value.
Optional attributes for
WholeFileReaderCombines (nests of) input elements into a dataset of (nests of) windows.
Optional attributes for
WindowDatasetThe WindowOp operation
Worker heartbeat op.
The WrapDatasetVariant operation
Writes an audio summary.
Optional attributes for
WriteAudioSummaryWrites
contents to the file at input filename.Writes a graph summary.
Writes a histogram summary.
Writes an image summary.
Optional attributes for
WriteImageSummaryWrites a serialized proto summary.
Writes a scalar summary.
Writes a tensor summary.
Returns 0 if x == 0, and x / y otherwise, elementwise.
A pseudo-op to represent host-side computation in an XLA program.
Optional attributes for
XlaHostComputeAn API for building
xla operations as OpsAn op to receive a tensor from the host.
An op that receives embedding activations on the TPU.
Receives deduplication data (indices and weights) from the embedding core.
An op to send a tensor to the host.
An op that performs gradient updates of embedding tables.
Optional attributes for
XlaSendTPUEmbeddingGradientsThe XlaSparseCoreAdagrad operation
The XlaSparseCoreAdagradMomentum operation
The XlaSparseCoreAdam operation
The XlaSparseCoreFtrl operation
The XlaSparseCoreSgd operation
The XlaSparseDenseMatmul operation
The XlaSparseDenseMatmulGradWithAdagradAndCsrInput operation
Optional attributes for
XlaSparseDenseMatmulGradWithAdagradAndCsrInputThe XlaSparseDenseMatmulGradWithAdagradMomentumAndCsrInput operation
Optional attributes for
XlaSparseDenseMatmulGradWithAdagradMomentumAndCsrInputThe XlaSparseDenseMatmulGradWithAdamAndCsrInput operation
Optional attributes for
XlaSparseDenseMatmulGradWithAdamAndCsrInputThe XlaSparseDenseMatmulGradWithFtrlAndCsrInput operation
Optional attributes for
XlaSparseDenseMatmulGradWithFtrlAndCsrInputThe XlaSparseDenseMatmulGradWithSgdAndCsrInput operation
Optional attributes for
XlaSparseDenseMatmulGradWithSgdAndCsrInputThe XlaSparseDenseMatmulWithCsrInput operation
Returns 0 if x == 0, and x * log1p(y) otherwise, elementwise.
Returns 0 if x == 0, and x * log(y) otherwise, elementwise.
An operator creating a constant initialized with zeros of the shape given by `dims`.
Returns a tensor of zeros with the same shape and type as x.
Compute the Hurwitz zeta function \(\zeta(x, q)\).
Creates a dataset that zips together
input_datasets.Optional attributes for
ZipDataset
NcclAllReduceinstead