Float32 truncated to 16 bits.

Double-precision complex.

Single-precision complex.

Data types that all computation devices are expected to be capable to support.

Do not use! These are only for parameters.

Not a legal value for DataType.

Quantized int16.

Quantized int32.

Quantized int8.

Quantized uint16.

Quantized uint8.

Arbitrary C++ data types.

Invalid/unknown specialized type.

"tensorflow::data::Optional" in the variant type registry.

"tensorflow::TensorList" in the variant type registry.

`MEAN`: Take the arithmetic mean ("average") of the updates across replicas.

`NONE`: This is the default, giving an error if you use a variable-update operation with multiple replicas.

`ONLY_FIRST_REPLICA`: This is for when every replica is performing the same update, but we only want to perform the update once.

`SUM`: Add the updates across replicas.

`AUTO`: Indicates that the synchronization will be determined by the current `DistributionStrategy` (eg.

`NONE`: Indicates that there will only be one copy of the variable, so there is no need to sync.

`ON_READ`: Indicates that the variable will be aggregated across devices when it is read (eg.

`ON_WRITE`: Indicates that the variable will be updated across devices every time it is written.

Enum value maps for DataType.

Enum value maps for DataType.

Enum value maps for SpecializedType.

Enum value maps for SpecializedType.

Enum value maps for VariableAggregation.

Enum value maps for VariableAggregation.

Enum value maps for VariableSynchronization.

Enum value maps for VariableSynchronization.

An allocation/de-allocation operation performed by the allocator.

Protocol buffer representing the value for an attr used to configure an Op.

LINT.IfChange.

Total cost of this graph, typically used for balancing decisions.

Inputs of this node.

Outputs of this node.

A function can be instantiated when the runtime can bind every attr with a value.

Attributes for function arguments.

A library is a set of named functions.

GradientDef defines the gradient function of a function defined in a function library.

Represents the graph of operations.

For memory tracking.

A list of attr names and their values.

Time/size stats recorded for a single execution of a graph node.

Output sizes recorded for a single execution of a graph node.

Defines an operation.

For describing inputs and outputs.

Description of the graph-construction-time configuration of this Op.

Information about version-dependent deprecation of an op.

A collection of OpDefs.

Protocol buffer representing a handle to a tensorflow resource.

Protocol buffer representing a pair of (data type, tensor shape).

Protocol buffer representing a tensor.

Dimensions of a tensor.

One dimension of the tensor.

Protocol buffer representing a Variable.

Protocol buffer representing the serialization format of DT_VARIANT tensors.

Version information for a piece of serialized data There are different types of versions for each type of data (GraphDef, etc.), but they all have the same common shape described here.

(== suppress_warning documentation-presence ==) LINT.IfChange.

For identifying the underlying type of a variant.

Indicates how a distributed variable will be aggregated.

Indicates when a distributed variable will be synced.