The operation was aborted, typically due to a concurrency issue like sequencer check failures, transaction aborts, etc.

Some entity that we attempted to create (e.g., file or directory) already exists.

The operation was cancelled (typically by the caller).

Unrecoverable data loss or corruption.

Deadline expired before operation could complete.

An extra enum entry to prevent people from writing code that fails to compile when a new code is added.

Operation was rejected because the system is not in a state required for the operation's execution.

Internal errors.

Client specified an invalid argument.

Some requested entity (e.g., file or directory) was not found.

Not an error; returned on success.

Operation tried to iterate past the valid input range.

The caller does not have permission to execute the specified operation.

Some resource has been exhausted, perhaps a per-user quota, or perhaps the entire file system is out of space.

The request does not have valid authentication credentials for the operation.

The service is currently unavailable.

Operation is not implemented or not supported/enabled in this service.

Unknown error.

No optimizations.

L1 is the default level.

The following settings turn on compilation, with higher values being more aggressive.

Enable some aggressive optimizations that use assumptions that TF graphs may break.

The default setting (SCHEDULING and SWAPPING HEURISTICS only).

Use any combination of swapping and recomputation heuristics.

Driven by manual op-level annotations.

Disabled in the meta-optimizer.

Recomputation heuristics will recompute ops (such as Relu activation) during backprop instead of storing them, reducing peak memory usage.

Scheduling will split big ops such as AddN and try to enforce a schedule of the new computations that decreases peak memory usage.

Swapping heuristic will move a tensor from the GPU to the CPU and move it back when needed to reduce peak memory usage.

Internal legacy format.

Deprecated format: tf.Saver() which works with tensorflow::table::Table.

Current format: more efficient.

An asset file def for a single file or a set of sharded files with the same name.

Describes the metadata related to a checkpointed tensor.

Special header that is associated with a bundle.

Defines a subgraph in another `GraphDef` as a set of feed points and nodes to be fetched or executed.

Cleanup the step state of a multi-device function (e.g.

Defines a TensorFlow cluster as a set of jobs.

CollectionDef should cover most collections.

AnyList is used for collecting Any protos.

BytesList is used for collecting strings and serialized protobufs.

FloatList is used for collecting float values.

Int64List is used for collecting int, int64 and long values.

NodeList is used for collecting nodes in graph.

Supplies one or more device names as members of the group identified by group_key.

Gives the complete membership of the group identified by group_key.

Supplies data about one collective op belonging to the instance identified by instance_key.

Confirms that every op in the instance has consistently declared itself.

Protocol buffer representing a CondContext object.

Session configuration parameters.

Everything inside Experimental is subject to change and is not subject to API stability guarantees in https://www.tensorflow.org/guide/version_compat.

Container for any kind of control flow context.

A convolution.

Protocol buffer representing a CriticalSection.

Protocol buffer representing a CriticalSection execution.

Options for initializing DebuggerState in TensorFlow Debugger (tfdbg).

Option for watching a node in TensorFlow Debugger (tfdbg).

Represents a Python dict keyed by `str`.

Options specific to the execution of a single step.

Represents `FunctionSpec` used in `Function`.

Request for next agreed-upon step_id for the specified graph_keys.

Next valid step_ids for one or more graph_keys.

Configuration for breaking down a visible GPU into multiple "virtual" devices.

Defines a single job in a TensorFlow cluster.

Represents a Python list.

Out-of-band request to begin or end logging, or to retrieve logs for particular steps.

Message for managing the response cache maintained on the sender side.

Some of the data from AllocatorStats.

NOTE: This protocol buffer is evolving, and will go through revisions in the coming months.

Meta information regarding the graph to be exported.

A pair of tensor name and tensor values.

Represents Python's namedtuple.

Records the creation of a new replay session.

Represents None.

A proto representation of an eager operation.

Options passed to the graph optimizer.

Represents a (key, value) pair.

Protocol buffer representing a QueueRunner.

Extra data needed on a non-RDMA RecvBufResponse.

Reset() allows misbehaving or slow sessions to be aborted and closed, and causes their resources eventually to be released.

Message to describe custom graph optimizer and its parameters.

Metadata output (i.e., non-Tensor) for a single Run() call.

Options for a single Run() call.

Everything inside Experimental is subject to change and is not subject to API stability guarantees in https://www.tensorflow.org/guide/version_compat.

A SavedAsset points to an asset in the MetaGraph.

Stores low-level information about a concrete function.

A function with multiple signatures, possibly with non-Tensor arguments.

SavedModel is the high level serialization format for TensorFlow Models.

A SavedResource represents a TF object that holds state during its lifetime.

A SavedUserObject is an object (in the object-oriented language of the TensorFlow program) of some user- or framework-defined class other than those handled specifically by the other kinds of SavedObjects.

Represents a Variable that is initialized by loading the contents from the checkpoint.

Protocol buffer representing the configuration of a Saver.

Defines the configuration of a single TensorFlow server.

Metadata about the session.

SignatureDef defines the signature of a computation supported by a TensorFlow graph.

`StructuredValue` represents a dynamically typed value representing various data structures that are inspired by Python data structures typically used in TensorFlow functions as inputs and outputs.

Defines a connection between two tensors in a `GraphDef`.

Information about a Tensor necessary for feeding or retrieval.

Generic encoding for composite tensors.

For sparse tensors, The COO encoding stores a triple of values, indices, and shape.

A protobuf to tf.TensorSpec.

Out-of-band request to configure distributed tracing.

Represents a Python tuple.

Represents a tf.TypeSpec.

UnimplementedEagerServiceServer can be embedded to have forward compatible implementations.

UnimplementedMasterServiceServer can be embedded to have forward compatible implementations.

UnimplementedWorkerServiceServer can be embedded to have forward compatible implementations.

Protocol buffer representing the values in ControlFlowContext.

The config for graph verifiers.

Protocol buffer representing a WhileContext object.

EagerServiceClient is the client API for EagerService service.

EagerServiceServer is the server API for EagerService service.

MasterServiceClient is the client API for MasterService service.

MasterServiceServer is the server API for MasterService service.

WorkerServiceClient is the client API for WorkerService service.

WorkerServiceServer is the server API for WorkerService service.

An enum indicating the endianness of the platform that produced this bundle.

The canonical error codes for TensorFlow APIs.

Control the use of the compiler/jit.

Optimization level.

Enum controlling the number of times to run optimizers.

TODO(pbar) Turn this into a TraceOptions proto which allows tracing to be controlled in a more orthogonal manner?.

A version number that identifies a different on-disk checkpoint format.