Op

public interface Op

Known Indirect Subclasses Abort, All, AllToAll<T>, AnonymousHashTable, AnonymousIteratorV2, AnonymousIteratorV3, AnonymousMemoryCache, AnonymousMultiDeviceIterator, AnonymousMultiDeviceIteratorV3, AnonymousMutableDenseHashTable, AnonymousMutableHashTable, AnonymousMutableHashTableOfTensors, AnonymousRandomSeedGenerator, AnonymousSeedGenerator, Any, and 735 others. Abort Raise a exception to abort the process when called.  All Computes the "logical and" of elements across dimensions of a tensor.  AllToAll<T> An Op to exchange data across TPU replicas.  AnonymousHashTable Creates a uninitialized anonymous hash table.  AnonymousIteratorV2 A container for an iterator resource.  AnonymousIteratorV3 A container for an iterator resource.  AnonymousMemoryCache   AnonymousMultiDeviceIterator A container for a multi device iterator resource.  AnonymousMultiDeviceIteratorV3 A container for a multi device iterator resource.  AnonymousMutableDenseHashTable Creates an empty anonymous mutable hash table that uses tensors as the backing store.  AnonymousMutableHashTable Creates an empty anonymous mutable hash table.  AnonymousMutableHashTableOfTensors Creates an empty anonymous mutable hash table of vector values.  AnonymousRandomSeedGenerator   AnonymousSeedGenerator   Any Computes the "logical or" of elements across dimensions of a tensor.  ApplyAdagradV2<T> Update '*var' according to the adagrad scheme.  ApproxTopK<T extends Number> Returns min/max k values and their indices of the input operand in an approximate manner.  AssertCardinalityDataset   AssertNextDataset A transformation that asserts which transformations happen next.  AssertPrevDataset A transformation that asserts which transformations happened previously.  AssertThat Asserts that the given condition is true.  Assign<T> Update 'ref' by assigning 'value' to it.  AssignAdd<T> Update 'ref' by adding 'value' to it.  AssignAddVariableOp Adds a value to the current value of a variable.  AssignSub<T> Update 'ref' by subtracting 'value' from it.  AssignSubVariableOp Subtracts a value from the current value of a variable.  AssignVariableOp Assigns a new value to a variable.  AssignVariableXlaConcatND Concats input tensor across all dimensions.  AutoShardDataset Creates a dataset that shards the input dataset.  BandedTriangularSolve<T>   Barrier Defines a barrier that persists across different graph executions.  BarrierClose Closes the given barrier.  BarrierIncompleteSize Computes the number of incomplete elements in the given barrier.  BarrierInsertMany For each key, assigns the respective value to the specified component.  BarrierReadySize Computes the number of complete elements in the given barrier.  BarrierTakeMany Takes the given number of completed elements from a barrier.  Batch Batches all input tensors nondeterministically.  BatchMatMulV2<T> Multiplies slices of two tensors in batches.  BatchMatMulV3<V> Multiplies slices of two tensors in batches.  BatchToSpace<T> BatchToSpace for 4-D tensors of type T.  BatchToSpaceNd<T> BatchToSpace for N-D tensors of type T.  BesselI0<T extends Number>   BesselI1<T extends Number>   BesselJ0<T extends Number>   BesselJ1<T extends Number>   BesselK0<T extends Number>   BesselK0e<T extends Number>   BesselK1<T extends Number>   BesselK1e<T extends Number>   BesselY0<T extends Number>   BesselY1<T extends Number>   Bitcast<U> Bitcasts a tensor from one type to another without copying data.  BlockLSTM<T extends Number> Computes the LSTM cell forward propagation for all the time steps.  BlockLSTMGrad<T extends Number> Computes the LSTM cell backward propagation for the entire time sequence.  BlockLSTMGradV2<T extends Number> Computes the LSTM cell backward propagation for the entire time sequence.  BlockLSTMV2<T extends Number> Computes the LSTM cell forward propagation for all the time steps.  BoostedTreesAggregateStats Aggregates the summary of accumulated stats for the batch.  BoostedTreesBucketize Bucketize each feature based on bucket boundaries.  BoostedTreesCalculateBestFeatureSplit Calculates gains for each feature and returns the best possible split information for the feature.  BoostedTreesCalculateBestFeatureSplitV2 Calculates gains for each feature and returns the best possible split information for each node.  BoostedTreesCalculateBestGainsPerFeature Calculates gains for each feature and returns the best possible split information for the feature.  BoostedTreesCenterBias Calculates the prior from the training data (the bias) and fills in the first node with the logits' prior.  BoostedTreesCreateEnsemble Creates a tree ensemble model and returns a handle to it.  BoostedTreesCreateQuantileStreamResource Create the Resource for Quantile Streams.  BoostedTreesDeserializeEnsemble Deserializes a serialized tree ensemble config and replaces current tree ensemble.  BoostedTreesEnsembleResourceHandleOp Creates a handle to a BoostedTreesEnsembleResource   BoostedTreesExampleDebugOutputs Debugging/model interpretability outputs for each example.  BoostedTreesFlushQuantileSummaries Flush the quantile summaries from each quantile stream resource.  BoostedTreesGetEnsembleStates Retrieves the tree ensemble resource stamp token, number of trees and growing statistics.  BoostedTreesMakeQuantileSummaries Makes the summary of quantiles for the batch.  BoostedTreesMakeStatsSummary Makes the summary of accumulated stats for the batch.  BoostedTreesPredict Runs multiple additive regression ensemble predictors on input instances and computes the logits.  BoostedTreesQuantileStreamResourceAddSummaries Add the quantile summaries to each quantile stream resource.  BoostedTreesQuantileStreamResourceDeserialize Deserialize bucket boundaries and ready flag into current QuantileAccumulator.  BoostedTreesQuantileStreamResourceFlush Flush the summaries for a quantile stream resource.  BoostedTreesQuantileStreamResourceGetBucketBoundaries Generate the bucket boundaries for each feature based on accumulated summaries.  BoostedTreesQuantileStreamResourceHandleOp Creates a handle to a BoostedTreesQuantileStreamResource.  BoostedTreesSerializeEnsemble Serializes the tree ensemble to a proto.  BoostedTreesSparseAggregateStats Aggregates the summary of accumulated stats for the batch.  BoostedTreesSparseCalculateBestFeatureSplit Calculates gains for each feature and returns the best possible split information for the feature.  BoostedTreesTrainingPredict Runs multiple additive regression ensemble predictors on input instances and computes the update to cached logits.  BoostedTreesUpdateEnsemble Updates the tree ensemble by either adding a layer to the last tree being grown or by starting a new tree.  BoostedTreesUpdateEnsembleV2 Updates the tree ensemble by adding a layer to the last tree being grown or by starting a new tree.  BroadcastDynamicShape<T extends Number> Return the shape of s0 op s1 with broadcast.  BroadcastGradientArgs<T extends Number> Return the reduction indices for computing gradients of s0 op s1 with broadcast.  BroadcastTo<T> Broadcast an array for a compatible shape.  Bucketize Bucketizes 'input' based on 'boundaries'.  CSRSparseMatrixComponents<T> Reads out the CSR components at batch `index`.  CSRSparseMatrixToDense<T> Convert a (possibly batched) CSRSparseMatrix to dense.  CSRSparseMatrixToSparseTensor<T> Converts a (possibly batched) CSRSparesMatrix to a SparseTensor.  CSVDataset   CSVDatasetV2   CTCLossV2 Calculates the CTC Loss (log probability) for each batch entry.  CacheDatasetV2   CheckNumericsV2<T extends Number> Checks a tensor for NaN, -Inf and +Inf values.  ChooseFastestDataset   ClipByValue<T> Clips tensor values to a specified min and max.  CollateTPUEmbeddingMemory An op that merges the string-encoded memory config protos from all hosts.  CollectiveAllToAllV2<T extends Number> Mutually exchanges multiple tensors of identical type and shape.  CollectiveAllToAllV3<T extends Number> Mutually exchanges multiple tensors of identical type and shape.  CollectiveAssignGroupV2 Assign group keys based on group assignment.  CollectiveBcastRecvV2<U> Receives a tensor value broadcast from another device.  CollectiveBcastSendV2<T> Broadcasts a tensor value to one or more other devices.  CollectiveGather<T extends Number> Mutually accumulates multiple tensors of identical type and shape.  CollectiveGatherV2<T extends Number> Mutually accumulates multiple tensors of identical type and shape.  CollectiveInitializeCommunicator Initializes a group for collective operations.  CollectivePermute<T> An Op to permute tensors across replicated TPU instances.  CollectiveReduceScatterV2<T extends Number> Mutually reduces multiple tensors of identical type and shape and scatters the result.  CollectiveReduceV2<T extends Number> Mutually reduces multiple tensors of identical type and shape.  CollectiveReduceV3<T extends Number> Mutually reduces multiple tensors of identical type and shape.  CombinedNonMaxSuppression Greedily selects a subset of bounding boxes in descending order of score, This operation performs non_max_suppression on the inputs per batch, across all classes.  CompositeTensorVariantFromComponents Encodes an `ExtensionType` value into a `variant` scalar Tensor.  CompositeTensorVariantToComponents Decodes a `variant` scalar Tensor into an `ExtensionType` value.  CompressElement Compresses a dataset element.  ComputeBatchSize Computes the static batch size of a dataset sans partial batches.  ComputeDedupDataTupleMask An op computes tuple mask of deduplication data from embedding core.  Concat<T> Concatenates tensors along one dimension.  ConfigureAndInitializeGlobalTPU An op that sets up the centralized structures for a distributed TPU system.  ConfigureDistributedTPU Sets up the centralized structures for a distributed TPU system.  ConfigureTPUEmbedding Sets up TPUEmbedding in a distributed TPU system.  ConfigureTPUEmbeddingHost An op that configures the TPUEmbedding software on a host.  ConfigureTPUEmbeddingMemory An op that configures the TPUEmbedding software on a host.  ConnectTPUEmbeddingHosts An op that sets up communication between TPUEmbedding host software instances after ConfigureTPUEmbeddingHost has been called on each host.  Constant<T> An operator producing a constant value.  ConsumeMutexLock This op consumes a lock created by `MutexLock`.  ControlTrigger Does nothing.  Conv2DBackpropFilterV2<T extends Number> Computes the gradients of convolution with respect to the filter.  Conv2DBackpropInputV2<T extends Number> Computes the gradients of convolution with respect to the input.  Copy<T> Copy a tensor from CPU-to-CPU or GPU-to-GPU.  CopyHost<T> Copy a tensor to host.  CopyToMesh<T>   CopyToMeshGrad<T>   CountUpTo<T extends Number> Increments 'ref' until it reaches 'limit'.  CrossReplicaSum<T extends Number> An Op to sum inputs across replicated TPU instances.  CudnnRNNBackpropV3<T extends Number> Backprop step of CudnnRNNV3.  CudnnRNNCanonicalToParamsV2<T extends Number> Converts CudnnRNN params from canonical form to usable form.  CudnnRNNParamsToCanonicalV2<T extends Number> Retrieves CudnnRNN params in canonical form.  CudnnRNNV3<T extends Number> A RNN backed by cuDNN.  CumulativeLogsumexp<T extends Number> Compute the cumulative product of the tensor `x` along `axis`.  DTensorRestoreV2   DTensorSetGlobalTPUArray An op that informs a host of the global ids of all the of TPUs in the system.  DataServiceDataset Creates a dataset that reads data from the tf.data service.  DataServiceDatasetV2 Creates a dataset that reads data from the tf.data service.  DatasetCardinality Returns the cardinality of `input_dataset`.  DatasetFromGraph Creates a dataset from the given `graph_def`.  DatasetToGraphV2 Returns a serialized GraphDef representing `input_dataset`.  Dawsn<T extends Number>   DebugGradientIdentity<T> Identity op for gradient debugging.  DebugGradientRefIdentity<T> Identity op for gradient debugging.  DebugIdentity<T> Provides an identity mapping of the non-Ref type input tensor for debugging.  DebugIdentityV2<T> Debug Identity V2 Op.  DebugNanCount Debug NaN Value Counter Op.  DebugNumericSummary Debug Numeric Summary Op.  DebugNumericSummaryV2<U extends Number> Debug Numeric Summary V2 Op.  DecodeImage<T extends Number> Function for decode_bmp, decode_gif, decode_jpeg, and decode_png.  DecodePaddedRaw<T extends Number> Reinterpret the bytes of a string as a vector of numbers.  DecodeProto The op extracts fields from a serialized protocol buffers message into tensors.  DeepCopy<T> Makes a copy of `x`.  DeleteIterator A container for an iterator resource.  DeleteMemoryCache   DeleteMultiDeviceIterator A container for an iterator resource.  DeleteRandomSeedGenerator   DeleteSeedGenerator   DeleteSessionTensor Delete the tensor specified by its handle in the session.  DenseBincount<U extends Number> Counts the number of occurrences of each value in an integer array.  DenseCountSparseOutput<U extends Number> Performs sparse-output bin counting for a tf.tensor input.  DenseToCSRSparseMatrix Converts a dense tensor to a (possibly batched) CSRSparseMatrix.  DestroyResourceOp Deletes the resource specified by the handle.  DestroyTemporaryVariable<T> Destroys the temporary variable and returns its final value.  DeviceIndex Return the index of device the op runs.  DirectedInterleaveDataset A substitute for `InterleaveDataset` on a fixed list of `N` datasets.  DisableCopyOnRead Turns off the copy-on-read mode.  DistributedSave   DrawBoundingBoxesV2<T extends Number> Draw bounding boxes on a batch of images.  DummyIterationCounter   DummyMemoryCache   DummySeedGenerator   DynamicEnqueueTPUEmbeddingArbitraryTensorBatch Eases the porting of code that uses tf.nn.embedding_lookup_sparse().  DynamicPartition<T> Partitions `data` into `num_partitions` tensors using indices from `partitions`.  DynamicStitch<T> Interleave the values from the `data` tensors into a single tensor.  EditDistance Computes the (possibly normalized) Levenshtein Edit Distance.  Eig<U> Computes the eigen decomposition of one or more square matrices.  Einsum<T> Tensor contraction according to Einstein summation convention.  Empty<T> Creates a tensor with the given shape.  EmptyTensorList Creates and returns an empty tensor list.  EmptyTensorMap Creates and returns an empty tensor map.  EncodeProto The op serializes protobuf messages provided in the input tensors.  EnqueueTPUEmbeddingArbitraryTensorBatch Eases the porting of code that uses tf.nn.embedding_lookup_sparse().  EnqueueTPUEmbeddingBatch An op that enqueues a list of input batch tensors to TPUEmbedding.  EnqueueTPUEmbeddingIntegerBatch An op that enqueues a list of input batch tensors to TPUEmbedding.  EnqueueTPUEmbeddingRaggedTensorBatch Eases the porting of code that uses tf.nn.embedding_lookup().  EnqueueTPUEmbeddingSparseBatch An op that enqueues TPUEmbedding input indices from a SparseTensor.  EnqueueTPUEmbeddingSparseTensorBatch Eases the porting of code that uses tf.nn.embedding_lookup_sparse().  EnsureShape<T> Ensures that the tensor's shape matches the expected shape.  Enter<T> Creates or finds a child frame, and makes `data` available to the child frame.  Erfinv<T extends Number>   EuclideanNorm<T> Computes the euclidean norm of elements across dimensions of a tensor.  ExecuteTPUEmbeddingPartitioner An op that executes the TPUEmbedding partitioner on the central configuration device and computes the HBM size (in bytes) required for TPUEmbedding operation.  Exit<T> Exits the current frame to its parent frame.  ExpandDims<T> Inserts a dimension of 1 into a tensor's shape.  ExperimentalAutoShardDataset Creates a dataset that shards the input dataset.  ExperimentalBytesProducedStatsDataset Records the bytes size of each element of `input_dataset` in a StatsAggregator.  ExperimentalChooseFastestDataset   ExperimentalDatasetCardinality Returns the cardinality of `input_dataset`.  ExperimentalDatasetToTFRecord Writes the given dataset to the given file using the TFRecord format.  ExperimentalDenseToSparseBatchDataset Creates a dataset that batches input elements into a SparseTensor.  ExperimentalLatencyStatsDataset Records the latency of producing `input_dataset` elements in a StatsAggregator.  ExperimentalMatchingFilesDataset   ExperimentalMaxIntraOpParallelismDataset Creates a dataset that overrides the maximum intra-op parallelism.  ExperimentalParseExampleDataset Transforms `input_dataset` containing `Example` protos as vectors of DT_STRING into a dataset of `Tensor` or `SparseTensor` objects representing the parsed features.  ExperimentalPrivateThreadPoolDataset Creates a dataset that uses a custom thread pool to compute `input_dataset`.  ExperimentalRandomDataset Creates a Dataset that returns pseudorandom numbers.  ExperimentalRebatchDataset Creates a dataset that changes the batch size.  ExperimentalSetStatsAggregatorDataset   ExperimentalSlidingWindowDataset Creates a dataset that passes a sliding window over `input_dataset`.  ExperimentalSqlDataset Creates a dataset that executes a SQL query and emits rows of the result set.  ExperimentalStatsAggregatorHandle Creates a statistics manager resource.  ExperimentalStatsAggregatorSummary Produces a summary of any statistics recorded by the given statistics manager.  ExperimentalUnbatchDataset A dataset that splits the elements of its input into multiple elements.  Expint<T extends Number>   ExtractGlimpseV2 Extracts a glimpse from the input tensor.  ExtractVolumePatches<T extends Number> Extract `patches` from `input` and put them in the `"depth"` output dimension.  FileSystemSetConfiguration Set configuration of the file system.  Fill<U> Creates a tensor filled with a scalar value.  FinalizeDataset Creates a dataset by applying tf.data.Options to `input_dataset`.  FinalizeTPUEmbedding An op that finalizes the TPUEmbedding configuration.  Fingerprint Generates fingerprint values.  FresnelCos<T extends Number>   FresnelSin<T extends Number>   FusedBatchNormGradV3<T extends Number, U extends Number> Gradient for batch normalization.  FusedBatchNormV3<T extends Number, U extends Number> Batch normalization.  GRUBlockCell<T extends Number> Computes the GRU cell forward propagation for 1 time step.  GRUBlockCellGrad<T extends Number> Computes the GRU cell back-propagation for 1 time step.  Gather<T> Gather slices from `params` axis `axis` according to `indices`.  GatherNd<T> Gather slices from `params` into a Tensor with shape specified by `indices`.  GenerateBoundingBoxProposals This op produces Region of Interests from given bounding boxes(bbox_deltas) encoded wrt anchors according to eq.2 in arXiv:1506.01497 The op selects top `pre_nms_topn` scoring boxes, decodes them with respect to anchors, applies non-maximal suppression on overlapping boxes with higher than `nms_threshold` intersection-over-union (iou) value, discarding boxes where shorter side is less than `min_size`.  GetElementAtIndex Gets the element at the specified index in a dataset.  GetOptions Returns the tf.data.Options attached to `input_dataset`.  GetSessionHandle Store the input tensor in the state of the current session.  GetSessionTensor<T> Get the value of the tensor specified by its handle.  Gradients 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... If Options.dx() values are set, they are as the initial symbolic partial derivatives of some loss function L w.r.t.  GuaranteeConst<T> Gives a guarantee to the TF runtime that the input tensor is a constant.  HashTable Creates a non-initialized hash table.  HistogramFixedWidth<U extends Number> Return histogram of values.  Identity<T> Return a tensor with the same shape and contents as the input tensor or value.  IdentityN Returns a list of tensors with the same shapes and contents as the input tensors.  IgnoreErrorsDataset Creates a dataset that contains the elements of `input_dataset` ignoring errors.  ImageProjectiveTransformV2<T extends Number> Applies the given transform to each of the images.  ImageProjectiveTransformV3<T extends Number> Applies the given transform to each of the images.  ImmutableConst<T> Returns immutable tensor from memory region.  InfeedDequeue<T> A placeholder op for a value that will be fed into the computation.  InfeedDequeueTuple Fetches multiple values from infeed as an XLA tuple.  InfeedEnqueue An op which feeds a single Tensor value into the computation.  InfeedEnqueuePrelinearizedBuffer An op which enqueues prelinearized buffer into TPU infeed.  InfeedEnqueueTuple Feeds multiple Tensor values into the computation as an XLA tuple.  InitializeTable Table initializer that takes two tensors for keys and values respectively.  InitializeTableFromDataset   InitializeTableFromTextFile Initializes a table from a text file.  InplaceAdd<T> Adds v into specified rows of x.  InplaceSub<T> Subtracts `v` into specified rows of `x`.  InplaceUpdate<T> Updates specified rows 'i' with values 'v'.  IsBoostedTreesEnsembleInitialized Checks whether a tree ensemble has been initialized.  IsBoostedTreesQuantileStreamResourceInitialized Checks whether a quantile stream has been initialized.  IsTPUEmbeddingInitialized Whether TPU Embedding is initialized in a distributed TPU system.  IsVariableInitialized Checks whether a tensor has been initialized.  IsotonicRegression<U extends Number> Solves a batch of isotonic regression problems.  IteratorGetDevice Returns the name of the device on which `resource` has been placed.  KMC2ChainInitialization Returns the index of a data point that should be added to the seed set.  KmeansPlusPlusInitialization Selects num_to_sample rows of input using the KMeans++ criterion.  KthOrderStatistic Computes the Kth order statistic of a data set.  LMDBDataset Creates a dataset that emits the key-value pairs in one or more LMDB files.  LSTMBlockCell<T extends Number> Computes the LSTM cell forward propagation for 1 time step.  LSTMBlockCellGrad<T extends Number> Computes the LSTM cell backward propagation for 1 timestep.  LinSpace<T extends Number> Generates values in an interval.  ListDataset Creates a dataset that emits each of `tensors` once.  LoadAllTPUEmbeddingParameters An op that loads optimization parameters into embedding memory.  LoadTPUEmbeddingADAMParameters Load ADAM embedding parameters.  LoadTPUEmbeddingAdadeltaParameters Load Adadelta embedding parameters.  LoadTPUEmbeddingAdagradMomentumParameters Load Adagrad Momentum embedding parameters.  LoadTPUEmbeddingAdagradParameters Load Adagrad embedding parameters.  LoadTPUEmbeddingCenteredRMSPropParameters Load centered RMSProp embedding parameters.  LoadTPUEmbeddingFTRLParameters Load FTRL embedding parameters.  LoadTPUEmbeddingFrequencyEstimatorParameters Load frequency estimator embedding parameters.  LoadTPUEmbeddingMDLAdagradLightParameters Load MDL Adagrad Light embedding parameters.  LoadTPUEmbeddingMomentumParameters Load Momentum embedding parameters.  LoadTPUEmbeddingProximalAdagradParameters Load proximal Adagrad embedding parameters.  LoadTPUEmbeddingProximalYogiParameters   LoadTPUEmbeddingRMSPropParameters Load RMSProp embedding parameters.  LoadTPUEmbeddingStochasticGradientDescentParameters Load SGD embedding parameters.  LookupTableExport<T, U> Outputs all keys and values in the table.  LookupTableFind<U> Looks up keys in a table, outputs the corresponding values.  LookupTableImport Replaces the contents of the table with the specified keys and values.  LookupTableInsert Updates the table to associates keys with values.  LookupTableRemove Removes keys and its associated values from a table.  LookupTableSize Computes the number of elements in the given table.  LoopCond Forwards the input to the output.  LowerBound<U extends Number> Applies lower_bound(sorted_search_values, values) along each row.  Lu<T, U extends Number> Computes the LU decomposition of one or more square matrices.  MakeUnique Make all elements in the non-Batch dimension unique, but \"close\" to their initial value.  MapClear Op removes all elements in the underlying container.  MapIncompleteSize Op returns the number of incomplete elements in the underlying container.  MapPeek Op peeks at the values at the specified key.  MapSize Op returns the number of elements in the underlying container.  MapStage Stage (key, values) in the underlying container which behaves like a hashtable.  MapUnstage Op removes and returns the values associated with the key from the underlying container.  MapUnstageNoKey Op removes and returns a random (key, value) from the underlying container.  MatrixDiagPartV2<T> Returns the batched diagonal part of a batched tensor.  MatrixDiagPartV3<T> Returns the batched diagonal part of a batched tensor.  MatrixDiagV2<T> Returns a batched diagonal tensor with given batched diagonal values.  MatrixDiagV3<T> Returns a batched diagonal tensor with given batched diagonal values.  MatrixSetDiagV2<T> Returns a batched matrix tensor with new batched diagonal values.  MatrixSetDiagV3<T> Returns a batched matrix tensor with new batched diagonal values.  Max<T> Computes the maximum of elements across dimensions of a tensor.  MaxIntraOpParallelismDataset Creates a dataset that overrides the maximum intra-op parallelism.  Merge<T> Forwards the value of an available tensor from `inputs` to `output`.  MergeDedupData An op merges elements of integer and float tensors into deduplication data as XLA tuple.  Min<T> Computes the minimum of elements across dimensions of a tensor.  MirrorPad<T> Pads a tensor with mirrored values.  MirrorPadGrad<T> Gradient op for `MirrorPad` op.  MlirPassthroughOp Wraps an arbitrary MLIR computation expressed as a module with a main() function.  MulNoNan<T> Returns x * y element-wise.  MutableDenseHashTable Creates an empty hash table that uses tensors as the backing store.  MutableHashTable Creates an empty hash table.  MutableHashTableOfTensors Creates an empty hash table.  Mutex Creates a Mutex resource that can be locked by `MutexLock`.  MutexLock Locks a mutex resource.  NcclAllReduce<T extends Number> Outputs a tensor containing the reduction across all input tensors.  NcclBroadcast<T extends Number> Sends `input` to all devices that are connected to the output.  NcclReduce<T extends Number> Reduces `input` from `num_devices` using `reduction` to a single device.  Ndtri<T extends Number>   NearestNeighbors Selects the k nearest centers for each point.  NextAfter<T extends Number> Returns the next representable value of `x1` in the direction of `x2`, element-wise.  NextIteration<T> Makes its input available to the next iteration.  NoOp Does nothing.  NonDeterministicInts<U> Non-deterministically generates some integers.  NonMaxSuppressionV5<T extends Number> 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.  NonSerializableDataset   OneHot<U> Returns a one-hot tensor.  OnesLike<T> Returns a tensor of ones with the same shape and type as x.  OptimizeDatasetV2 Creates a dataset by applying related optimizations to `input_dataset`.  OptionsDataset Creates a dataset by attaching tf.data.Options to `input_dataset`.  OrderedMapClear Op removes all elements in the underlying container.  OrderedMapIncompleteSize Op returns the number of incomplete elements in the underlying container.  OrderedMapPeek Op peeks at the values at the specified key.  OrderedMapSize Op returns the number of elements in the underlying container.  OrderedMapStage Stage (key, values) in the underlying container which behaves like a ordered associative container.  OrderedMapUnstage Op removes and returns the values associated with the key from the underlying container.  OrderedMapUnstageNoKey Op removes and returns the (key, value) element with the smallest key from the underlying container.  OutfeedDequeue<T> Retrieves a single tensor from the computation outfeed.  OutfeedDequeueTuple Retrieve multiple values from the computation outfeed.  OutfeedDequeueTupleV2 Retrieve multiple values from the computation outfeed.  OutfeedDequeueV2<T> Retrieves a single tensor from the computation outfeed.  OutfeedEnqueue Enqueue a Tensor on the computation outfeed.  OutfeedEnqueueTuple Enqueue multiple Tensor values on the computation outfeed.  Pad<T> Pads a tensor.  ParallelBatchDataset   ParallelConcat<T> Concatenates a list of `N` tensors along the first dimension.  ParallelDynamicStitch<T> Interleave the values from the `data` tensors into a single tensor.  ParseExampleDatasetV2 Transforms `input_dataset` containing `Example` protos as vectors of DT_STRING into a dataset of `Tensor` or `SparseTensor` objects representing the parsed features.  ParseExampleV2 Transforms a vector of tf.Example protos (as strings) into typed tensors.  ParseSequenceExampleV2 Transforms a vector of tf.io.SequenceExample protos (as strings) into typed tensors.  Placeholder<T> A placeholder op for a value that will be fed into the computation.  PlaceholderWithDefault<T> A placeholder op that passes through `input` when its output is not fed.  Prelinearize An op which linearizes one Tensor value to an opaque variant tensor.  PrelinearizeTuple An op which linearizes multiple Tensor values to an opaque variant tensor.  PrimitiveOp A base class for Op implementations that are backed by a single Operation.  Print Prints a string scalar.  PrivateThreadPoolDataset Creates a dataset that uses a custom thread pool to compute `input_dataset`.  Prod<T> Computes the product of elements across dimensions of a tensor.  QuantizeAndDequantizeV4<T extends Number> Quantizes then dequantizes a tensor.  QuantizeAndDequantizeV4Grad<T extends Number> Returns the gradient of `QuantizeAndDequantizeV4`.  QuantizedConcat<T> Concatenates quantized tensors along one dimension.  QuantizedConcatV2<T>   QuantizedConv2DAndRelu<V>   QuantizedConv2DAndReluAndRequantize<V>   QuantizedConv2DAndRequantize<V>   QuantizedConv2DPerChannel<V> Computes QuantizedConv2D per channel.  QuantizedConv2DWithBias<V>   QuantizedConv2DWithBiasAndRelu<V>   QuantizedConv2DWithBiasAndReluAndRequantize<W>   QuantizedConv2DWithBiasAndRequantize<W>   QuantizedConv2DWithBiasSignedSumAndReluAndRequantize<X>   QuantizedConv2DWithBiasSumAndRelu<V>   QuantizedConv2DWithBiasSumAndReluAndRequantize<X>   QuantizedDepthwiseConv2D<V> Computes quantized depthwise Conv2D.  QuantizedDepthwiseConv2DWithBias<V> Computes quantized depthwise Conv2D with Bias.  QuantizedDepthwiseConv2DWithBiasAndRelu<V> Computes quantized depthwise Conv2D with Bias and Relu.  QuantizedDepthwiseConv2DWithBiasAndReluAndRequantize<W> Computes quantized depthwise Conv2D with Bias, Relu and Requantize.  QuantizedMatMulWithBias<W> Performs a quantized matrix multiplication of `a` by the matrix `b` with bias add.  QuantizedMatMulWithBiasAndDequantize<W extends Number>   QuantizedMatMulWithBiasAndRelu<V> Perform a quantized matrix multiplication of `a` by the matrix `b` with bias add and relu fusion.  QuantizedMatMulWithBiasAndReluAndRequantize<W> Perform a quantized matrix multiplication of `a` by the matrix `b` with bias add and relu and requantize fusion.  QuantizedMatMulWithBiasAndRequantize<W>   QuantizedReshape<T> Reshapes a quantized tensor as per the Reshape op.  RaggedBincount<U extends Number> Counts the number of occurrences of each value in an integer array.  RaggedCountSparseOutput<U extends Number> Performs sparse-output bin counting for a ragged tensor input.  RaggedCross<T, U extends Number> Generates a feature cross from a list of tensors, and returns it as a RaggedTensor.  RaggedFillEmptyRows<T>   RaggedFillEmptyRowsGrad<T>   RaggedGather<T extends Number, U> Gather ragged slices from `params` axis `0` according to `indices`.  RaggedRange<U extends Number, T extends Number> Returns a `RaggedTensor` containing the specified sequences of numbers.  RaggedTensorFromVariant<U extends Number, T> Decodes a `variant` Tensor into a `RaggedTensor`.  RaggedTensorToSparse<U> Converts a `RaggedTensor` into a `SparseTensor` with the same values.  RaggedTensorToTensor<U> Create a dense tensor from a ragged tensor, possibly altering its shape.  RaggedTensorToVariant Encodes a `RaggedTensor` into a `variant` Tensor.  RaggedTensorToVariantGradient<U> Helper used to compute the gradient for `RaggedTensorToVariant`.  RandomDatasetV2 Creates a Dataset that returns pseudorandom numbers.  RandomIndexShuffle<T extends Number> Outputs the position of `value` in a permutation of [0, ..., max_index].  Range<T extends Number> Creates a sequence of numbers.  Rank Returns the rank of a tensor.  ReadVariableOp<T> Reads the value of a variable.  ReadVariableXlaSplitND<T> Splits resource variable input tensor across all dimensions.  RebatchDataset Creates a dataset that changes the batch size.  RebatchDatasetV2 Creates a dataset that changes the batch size.  Recv<T> Receives the named tensor from send_device on recv_device.  RecvTPUEmbeddingActivations An op that receives embedding activations on the TPU.  ReduceAll Computes the "logical and" of elements across dimensions of a tensor.  ReduceAny Computes the "logical or" of elements across dimensions of a tensor.  ReduceMax<T> Computes the maximum of elements across dimensions of a tensor.  ReduceMin<T> Computes the minimum of elements across dimensions of a tensor.  ReduceProd<T> Computes the product of elements across dimensions of a tensor.  ReduceSum<T> Computes the sum of elements across dimensions of a tensor.  RefEnter<T> Creates or finds a child frame, and makes `data` available to the child frame.  RefExit<T> Exits the current frame to its parent frame.  RefIdentity<T> Return the same ref tensor as the input ref tensor.  RefMerge<T> Forwards the value of an available tensor from `inputs` to `output`.  RefNextIteration<T> Makes its input available to the next iteration.  RefSelect<T> Forwards the `index`th element of `inputs` to `output`.  RefSwitch<T> Forwards the ref tensor `data` to the output port determined by `pred`.  RegisterDataset Registers a dataset with the tf.data service.  RegisterDatasetV2 Registers a dataset with the tf.data service.  Relayout<T>   RelayoutGrad<T>   RequantizationRangePerChannel Computes requantization range per channel.  RequantizePerChannel<U> Requantizes input with min and max values known per channel.  Reshape<T> Reshapes a tensor.  ResourceAccumulatorApplyGradient Applies a gradient to a given accumulator.  ResourceAccumulatorNumAccumulated Returns the number of gradients aggregated in the given accumulators.  ResourceAccumulatorSetGlobalStep Updates the accumulator with a new value for global_step.  ResourceAccumulatorTakeGradient<T> Extracts the average gradient in the given ConditionalAccumulator.  ResourceApplyAdagradV2 Update '*var' according to the adagrad scheme.  ResourceApplyAdamWithAmsgrad Update '*var' according to the Adam algorithm.  ResourceApplyKerasMomentum Update '*var' according to the momentum scheme.  ResourceConditionalAccumulator A conditional accumulator for aggregating gradients.  ResourceCountUpTo<T extends Number> Increments variable pointed to by 'resource' until it reaches 'limit'.  ResourceGather<U> Gather slices from the variable pointed to by `resource` according to `indices`.  ResourceGatherNd<U>   ResourceScatterAdd Adds sparse updates to the variable referenced by `resource`.  ResourceScatterDiv Divides sparse updates into the variable referenced by `resource`.  ResourceScatterMax Reduces sparse updates into the variable referenced by `resource` using the `max` operation.  ResourceScatterMin Reduces sparse updates into the variable referenced by `resource` using the `min` operation.  ResourceScatterMul Multiplies sparse updates into the variable referenced by `resource`.  ResourceScatterNdAdd Applies sparse addition to individual values or slices in a Variable.  ResourceScatterNdMax   ResourceScatterNdMin   ResourceScatterNdSub Applies sparse subtraction to individual values or slices in a Variable.  ResourceScatterNdUpdate Applies sparse `updates` to individual values or slices within a given variable according to `indices`.  ResourceScatterSub Subtracts sparse updates from the variable referenced by `resource`.  ResourceScatterUpdate Assigns sparse updates to the variable referenced by `resource`.  ResourceSparseApplyAdagradV2 Update relevant entries in '*var' and '*accum' according to the adagrad scheme.  ResourceSparseApplyKerasMomentum Update relevant entries in '*var' and '*accum' according to the momentum scheme.  ResourceStridedSliceAssign Assign `value` to the sliced l-value reference of `ref`.  RetrieveAllTPUEmbeddingParameters An op that retrieves optimization parameters from embedding to host memory.  RetrieveTPUEmbeddingADAMParameters Retrieve ADAM embedding parameters.  RetrieveTPUEmbeddingAdadeltaParameters Retrieve Adadelta embedding parameters.  RetrieveTPUEmbeddingAdagradMomentumParameters Retrieve Adagrad Momentum embedding parameters.  RetrieveTPUEmbeddingAdagradParameters Retrieve Adagrad embedding parameters.  RetrieveTPUEmbeddingCenteredRMSPropParameters Retrieve centered RMSProp embedding parameters.  RetrieveTPUEmbeddingFTRLParameters Retrieve FTRL embedding parameters.  RetrieveTPUEmbeddingFrequencyEstimatorParameters Retrieve frequency estimator embedding parameters.  RetrieveTPUEmbeddingMDLAdagradLightParameters Retrieve MDL Adagrad Light embedding parameters.  RetrieveTPUEmbeddingMomentumParameters Retrieve Momentum embedding parameters.  RetrieveTPUEmbeddingProximalAdagradParameters Retrieve proximal Adagrad embedding parameters.  RetrieveTPUEmbeddingProximalYogiParameters   RetrieveTPUEmbeddingRMSPropParameters Retrieve RMSProp embedding parameters.  RetrieveTPUEmbeddingStochasticGradientDescentParameters Retrieve SGD embedding parameters.  Reverse<T> Reverses specific dimensions of a tensor.  ReverseSequence<T> Reverses variable length slices.  RewriteDataset   RiscAbs<T extends Number>   RiscAdd<T extends Number> Returns x + y element-wise.  RiscBinaryArithmetic<T extends Number>   RiscBinaryComparison   RiscBitcast<U>   RiscBroadcast<T>   RiscCast<U>   RiscCeil<T extends Number>   RiscCholesky<T extends Number>   RiscConcat<T>   RiscConv<T extends Number>   RiscCos<T extends Number>   RiscDiv<T extends Number>   RiscDot<T extends Number>   RiscExp<T extends Number>   RiscFft<T>   RiscFloor<T extends Number>   RiscGather<T>   RiscImag<U extends Number>   RiscIsFinite   RiscLog<T extends Number>   RiscLogicalAnd   RiscLogicalNot   RiscLogicalOr   RiscMax<T extends Number> Returns max(x, y) element-wise.  RiscMin<T extends Number>   RiscMul<T extends Number>   RiscNeg<T extends Number>   RiscPad<T extends Number>   RiscPool<T extends Number>   RiscPow<T extends Number>   RiscRandomUniform   RiscReal<U extends Number>   RiscReduce<T extends Number>   RiscRem<T extends Number>   RiscReshape<T extends Number>   RiscReverse<T extends Number>   RiscScatter<U extends Number>   RiscShape<U extends Number>   RiscSign<T extends Number>   RiscSlice<T extends Number>   RiscSort<T extends Number>   RiscSqueeze<T>   RiscSub<T extends Number>   RiscTranspose<T>   RiscTriangularSolve<T extends Number>   RiscUnary<T extends Number>   RngReadAndSkip Advance the counter of a counter-based RNG.  RngSkip Advance the counter of a counter-based RNG.  Roll<T> Rolls the elements of a tensor along an axis.  SamplingDataset Creates a dataset that takes a Bernoulli sample of the contents of another dataset.  ScaleAndTranslate   ScaleAndTranslateGrad<T extends Number>   ScatterAdd<T> Adds sparse updates to a variable reference.  ScatterDiv<T> Divides a variable reference by sparse updates.  ScatterMax<T extends Number> Reduces sparse updates into a variable reference using the `max` operation.  ScatterMin<T extends Number> Reduces sparse updates into a variable reference using the `min` operation.  ScatterMul<T> Multiplies sparse updates into a variable reference.  ScatterNd<U> Scatters `updates` into a tensor of shape `shape` according to `indices`.  ScatterNdAdd<T> Applies sparse addition to individual values or slices in a Variable.  ScatterNdMax<T> Computes element-wise maximum.  ScatterNdMin<T> Computes element-wise minimum.  ScatterNdNonAliasingAdd<T> Applies sparse addition to `input` using individual values or slices from `updates` according to indices `indices`.  ScatterNdSub<T> Applies sparse subtraction to individual values or slices in a Variable.  ScatterNdUpdate<T> Applies sparse `updates` to individual values or slices within a given variable according to `indices`.  ScatterSub<T> Subtracts sparse updates to a variable reference.  ScatterUpdate<T> Applies sparse updates to a variable reference.  SegmentMaxV2<T extends Number> Computes the maximum along segments of a tensor.  SegmentMinV2<T extends Number> Computes the minimum along segments of a tensor.  SegmentProdV2<T> Computes the product along segments of a tensor.  SegmentSumV2<T> Computes the sum along segments of a tensor.  SelectV2<T>   Send Sends the named tensor from send_device to recv_device.  SendTPUEmbeddingGradients Performs gradient updates of embedding tables.  SetDiff1d<T, U extends Number> Computes the difference between two lists of numbers or strings.  SetSize Number of unique elements along last dimension of input `set`.  Shape<U extends Number> Returns the shape of a tensor.  ShapeN<U extends Number> Returns shape of tensors.  ShardDataset Creates a `Dataset` that includes only 1/`num_shards` of this dataset.  ShuffleAndRepeatDatasetV2   ShuffleDatasetV2   ShuffleDatasetV3   ShutdownDistributedTPU Shuts down a running distributed TPU system.  ShutdownTPUSystem An op that shuts down the TPU system.  Size<U extends Number> Returns the size of a tensor.  Skipgram Parses a text file and creates a batch of examples.  SleepDataset   Slice<T> Return a slice from 'input'.  SlidingWindowDataset Creates a dataset that passes a sliding window over `input_dataset`.  Snapshot<T> Returns a copy of the input tensor.  SnapshotDataset Creates a dataset that will write to / read from a snapshot.  SnapshotDatasetReader   SnapshotNestedDatasetReader   SobolSample<T extends Number> Generates points from the Sobol sequence.  SpaceToBatchNd<T> SpaceToBatch for N-D tensors of type T.  SparseApplyAdagradV2<T> Update relevant entries in '*var' and '*accum' according to the adagrad scheme.  SparseBincount<U extends Number> Counts the number of occurrences of each value in an integer array.  SparseCountSparseOutput<U extends Number> Performs sparse-output bin counting for a sparse tensor input.  SparseCrossHashed Generates sparse cross from a list of sparse and dense tensors.  SparseCrossV2 Generates sparse cross from a list of sparse and dense tensors.  SparseMatrixAdd Sparse addition of two CSR matrices, C = alpha * A + beta * B.  SparseMatrixMatMul<T> Matrix-multiplies a sparse matrix with a dense matrix.  SparseMatrixMul Element-wise multiplication of a sparse matrix with a dense tensor.  SparseMatrixNNZ Returns the number of nonzeroes of `sparse_matrix`.  SparseMatrixOrderingAMD Computes the Approximate Minimum Degree (AMD) ordering of `input`.  SparseMatrixSoftmax Calculates the softmax of a CSRSparseMatrix.  SparseMatrixSoftmaxGrad Calculates the gradient of the SparseMatrixSoftmax op.  SparseMatrixSparseCholesky Computes the sparse Cholesky decomposition of `input`.  SparseMatrixSparseMatMul Sparse-matrix-multiplies two CSR matrices `a` and `b`.  SparseMatrixTranspose Transposes the inner (matrix) dimensions of a CSRSparseMatrix.  SparseMatrixZeros Creates an all-zeros CSRSparseMatrix with shape `dense_shape`.  SparseSegmentSumGrad<T extends Number> Computes gradients for SparseSegmentSum.  SparseTensorToCSRSparseMatrix Converts a SparseTensor to a (possibly batched) CSRSparseMatrix.  Spence<T extends Number>   Split<T> Splits a tensor into `num_split` tensors along one dimension.  SplitDedupData<T extends Number, U extends Number> An op splits input deduplication data XLA tuple into integer and floating point tensors.  SplitV<T> Splits a tensor into `num_split` tensors along one dimension.  Squeeze<T> Removes dimensions of size 1 from the shape of a tensor.  Stack<T> Packs a list of `N` rank-`R` tensors into one rank-`(R+1)` tensor.  Stage Stage values similar to a lightweight Enqueue.  StageClear Op removes all elements in the underlying container.  StagePeek Op peeks at the values at the specified index.  StageSize Op returns the number of elements in the underlying container.  StatefulRandomBinomial<V extends Number>   StatefulStandardNormal<U> Outputs random values from a normal distribution.  StatefulStandardNormalV2<U> Outputs random values from a normal distribution.  StatefulTruncatedNormal<U> Outputs random values from a truncated normal distribution.  StatefulUniform<U> Outputs random values from a uniform distribution.  StatefulUniformFullInt<U> Outputs random integers from a uniform distribution.  StatefulUniformInt<U> Outputs random integers from a uniform distribution.  StatelessParameterizedTruncatedNormal<V extends Number>   StatelessRandomBinomial<W extends Number> Outputs deterministic pseudorandom random numbers from a binomial distribution.  StatelessRandomGammaV2<V extends Number> Outputs deterministic pseudorandom random numbers from a gamma distribution.  StatelessRandomGammaV3<U extends Number> Outputs deterministic pseudorandom random numbers from a gamma distribution.  StatelessRandomGetAlg Picks the best counter-based RNG algorithm based on device.  StatelessRandomGetKeyCounter Scrambles seed into key and counter, using the best algorithm based on device.  StatelessRandomGetKeyCounterAlg Picks the best algorithm based on device, and scrambles seed into key and counter.  StatelessRandomNormalV2<U extends Number> Outputs deterministic pseudorandom values from a normal distribution.  StatelessRandomPoisson<W extends Number> Outputs deterministic pseudorandom random numbers from a Poisson distribution.  StatelessRandomUniformFullInt<V extends Number> Outputs deterministic pseudorandom random integers from a uniform distribution.  StatelessRandomUniformFullIntV2<U extends Number> Outputs deterministic pseudorandom random integers from a uniform distribution.  StatelessRandomUniformIntV2<U extends Number> Outputs deterministic pseudorandom random integers from a uniform distribution.  StatelessRandomUniformV2<U extends Number> Outputs deterministic pseudorandom random values from a uniform distribution.  StatelessSampleDistortedBoundingBox<T extends Number> Generate a randomly distorted bounding box for an image deterministically.  StatelessShuffle<T> Randomly and deterministically shuffles a tensor along its first dimension.  StatelessTruncatedNormalV2<U extends Number> Outputs deterministic pseudorandom values from a truncated normal distribution.  StatsAggregatorHandleV2   StatsAggregatorSetSummaryWriter Set a summary_writer_interface to record statistics using given stats_aggregator.  StopGradient<T> Stops gradient computation.  StridedSlice<T> Return a strided slice from `input`.  StridedSliceAssign<T> Assign `value` to the sliced l-value reference of `ref`.  StridedSliceGrad<U> Returns the gradient of `StridedSlice`.  StringLower Converts all uppercase characters into their respective lowercase replacements.  StringNGrams<T extends Number> Creates ngrams from ragged string data.  StringUpper Converts all lowercase characters into their respective uppercase replacements.  Sum<T> Computes the sum of elements across dimensions of a tensor.  SwitchCond<T> Forwards `data` to the output port determined by `pred`.  SyncDevice Synchronizes the device this op is run on.  TPUCompilationResult Returns the result of a TPU compilation.  TPUCompileSucceededAssert Asserts that compilation succeeded.  TPUEmbeddingActivations An op enabling differentiation of TPU Embeddings.  TPUExecute Op that loads and executes a TPU program on a TPU device.  TPUExecuteAndUpdateVariables Op that executes a program with optional in-place variable updates.  TPUOrdinalSelector A TPU core selector Op.  TPUPartitionedInput<T> An op that groups a list of partitioned inputs together.  TPUPartitionedInputV2<T> An op that groups a list of partitioned inputs together.  TPUPartitionedOutput<T> An op that demultiplexes a tensor to be sharded by XLA to a list of partitioned outputs outside the XLA computation.  TPUPartitionedOutputV2<T> An op that demultiplexes a tensor to be sharded by XLA to a list of partitioned outputs outside the XLA computation.  TPUReplicateMetadata Metadata indicating how the TPU computation should be replicated.  TPUReplicatedInput<T> Connects N inputs to an N-way replicated TPU computation.  TPUReplicatedOutput<T> Connects N outputs from an N-way replicated TPU computation.  TPUReshardVariables Op that reshards on-device TPU variables to specified state.  TPURoundRobin Round-robin load balancing on TPU cores.  TemporaryVariable<T> Returns a tensor that may be mutated, but only persists within a single step.  TensorArray An array of Tensors of given size.  TensorArrayClose Delete the TensorArray from its resource container.  TensorArrayConcat<T> Concat the elements from the TensorArray into value `value`.  TensorArrayGather<T> Gather specific elements from the TensorArray into output `value`.  TensorArrayGrad Creates a TensorArray for storing the gradients of values in the given handle.  TensorArrayGradWithShape Creates a TensorArray for storing multiple gradients of values in the given handle.  TensorArrayPack<T>   TensorArrayRead<T> Read an element from the TensorArray into output `value`.  TensorArrayScatter Scatter the data from the input value into specific TensorArray elements.  TensorArraySize Get the current size of the TensorArray.  TensorArraySplit Split the data from the input value into TensorArray elements.  TensorArrayUnpack   TensorArrayWrite Push an element onto the tensor_array.  TensorListConcat<T> Concats all tensors in the list along the 0th dimension.  TensorListConcatLists   TensorListConcatV2<U> Concats all tensors in the list along the 0th dimension.  TensorListElementShape<T extends Number> The shape of the elements of the given list, as a tensor.  TensorListFromTensor Creates a TensorList which, when stacked, has the value of `tensor`.  TensorListGather<T> Creates a Tensor by indexing into the TensorList.  TensorListGetItem<T>   TensorListLength Returns the number of tensors in the input tensor list.  TensorListPopBack<T> Returns the last element of the input list as well as a list with all but that element.  TensorListPushBack Returns a list which has the passed-in `Tensor` as last element and the other elements of the given list in `input_handle`.  TensorListPushBackBatch   TensorListReserve List of the given size with empty elements.  TensorListResize Resizes the list.  TensorListScatter Creates a TensorList by indexing into a Tensor.  TensorListScatterIntoExistingList Scatters tensor at indices in an input list.  TensorListScatterV2 Creates a TensorList by indexing into a Tensor.  TensorListSetItem   TensorListSplit Splits a tensor into a list.  TensorListStack<T> Stacks all tensors in the list.  TensorMapErase Returns a tensor map with item from given key erased.  TensorMapHasKey Returns whether the given key exists in the map.  TensorMapInsert Returns a map that is the 'input_handle' with the given key-value pair inserted.  TensorMapLookup<U> Returns the value from a given key in a tensor map.  TensorMapSize Returns the number of tensors in the input tensor map.  TensorMapStackKeys<T> Returns a Tensor stack of all keys in a tensor map.  TensorScatterAdd<T> Adds sparse `updates` to an existing tensor according to `indices`.  TensorScatterMax<T> Apply a sparse update to a tensor taking the element-wise maximum.  TensorScatterMin<T>   TensorScatterSub<T> Subtracts sparse `updates` from an existing tensor according to `indices`.  TensorScatterUpdate<T> Scatter `updates` into an existing tensor according to `indices`.  TensorStridedSliceUpdate<T> Assign `value` to the sliced l-value reference of `input`.  ThreadPoolDataset Creates a dataset that uses a custom thread pool to compute `input_dataset`.  ThreadPoolHandle Creates a dataset that uses a custom thread pool to compute `input_dataset`.  Tile<T> Constructs a tensor by tiling a given tensor.  Timestamp Provides the time since epoch in seconds.  ToBool Converts a tensor to a scalar predicate.  TopKUnique Returns the TopK unique values in the array in sorted order.  TopKWithUnique Returns the TopK values in the array in sorted order.  TpuHandleToProtoKey Converts XRT's uid handles to TensorFlow-friendly input format.  TridiagonalMatMul<T> Calculate product with tridiagonal matrix.  TridiagonalSolve<T> Solves tridiagonal systems of equations.  Unbatch<T> Reverses the operation of Batch for a single output Tensor.  UnbatchGrad<T> Gradient of Unbatch.  UncompressElement Uncompresses a compressed dataset element.  UnicodeDecode<T extends Number> Decodes each string in `input` into a sequence of Unicode code points.  UnicodeEncode Encode a tensor of ints into unicode strings.  UniformDequantize<U extends Number> Perform dequantization on the quantized Tensor `input`.  UniformQuantize<U> Perform quantization on Tensor `input`.  UniformQuantizedAdd<T> Perform quantized add of quantized Tensor `lhs` and quantized Tensor `rhs` to make quantized `output`.  UniformQuantizedClipByValue<T> Perform clip by value on the quantized Tensor `operand`.  UniformQuantizedConvolution<U> Perform quantized convolution of quantized Tensor `lhs` and quantized Tensor `rhs`.  UniformQuantizedConvolutionHybrid<V extends Number> Perform hybrid quantized convolution of float Tensor `lhs` and quantized Tensor `rhs`.  UniformQuantizedDot<U> Perform quantized dot of quantized Tensor `lhs` and quantized Tensor `rhs` to make quantized `output`.  UniformQuantizedDotHybrid<V extends Number> Perform hybrid quantized dot of float Tensor `lhs` and quantized Tensor `rhs`.  UniformRequantize<U> Given quantized tensor `input`, requantize it with new quantization parameters.  Unique<T, V extends Number> Finds unique elements along an axis of a tensor.  UniqueDataset Creates a dataset that contains the unique elements of `input_dataset`.  UniqueWithCounts<T, V extends Number> Finds unique elements along an axis of a tensor.  UnravelIndex<T extends Number> Converts an array of flat indices into a tuple of coordinate arrays.  UnsortedSegmentJoin   Unstack<T> Unpacks a given dimension of a rank-`R` tensor into `num` rank-`(R-1)` tensors.  Unstage Op is similar to a lightweight Dequeue.  UnwrapDatasetVariant   UpperBound<U extends Number> Applies upper_bound(sorted_search_values, values) along each row.  VarHandleOp Creates a handle to a Variable resource.  VarIsInitializedOp Checks whether a resource handle-based variable has been initialized.  Variable<T> Holds state in the form of a tensor that persists across steps.  VariableShape<T extends Number> Returns the shape of the variable pointed to by `resource`.  Where Returns locations of nonzero / true values in a tensor.  Where3<T> Selects elements from `x` or `y`, depending on `condition`.  WindowOp   WorkerHeartbeat Worker heartbeat op.  WrapDatasetVariant   WriteRawProtoSummary Writes a serialized proto summary.  XlaConcatND<T> Concats input tensor across all dimensions.  XlaRecvFromHost<T> An op to receive a tensor from the host.  XlaRecvTPUEmbeddingActivations An op that receives embedding activations on the TPU.  XlaRecvTPUEmbeddingDeduplicationData Receives deduplication data (indices and weights) from the embedding core.  XlaSendTPUEmbeddingGradients An op that performs gradient updates of embedding tables.  XlaSendToHost An op to send a tensor to the host.  XlaSplitND<T> Splits input tensor across all dimensions.  Xlog1py<T> Returns 0 if x == 0, and x * log1p(y) otherwise, elementwise.  Zeros<T> An operator creating a constant initialized with zeros of the shape given by `dims`.  ZerosLike<T> Returns a tensor of zeros with the same shape and type as x.

A marker interface for all operation wrappers.

Operation wrappers provide strongly typed interfaces for building and execution operations without the use of literals and indexes, as required in the core classes.

This interface allows keeping references to any operation wrapper using a common type.

// All values returned by an Ops call can be referred as a Op
 Op split = ops.array().split(...);
 Op shape = ops.array().shape(...);

 // All operations could be added to an Op collection
 Collection<Op> allOps = Arrays.asList(split, shape);