`dtensor.shutdown_tpu_system`.

## Deprecations

*   The C++ `tensorflow::Code` and `tensorflow::Status` will become aliases of
    respectively `absl::StatusCode` and `absl::Status` in some future release.
    *   Use `tensorflow::OkStatus()` instead of `tensorflow::Status::OK()`.
    *   Stop constructing `Status` objects from `tensorflow::error::Code`.
    *   One MUST NOT access `tensorflow::errors::Code` fields. Accessing

// here.
void TF_SetTracingImplementation(const char* name, TF_Status*);

// Creates a new TensorFlow function. A Function is an execution context, and as
// such it can trace operations through TF_ExecuteOperation. After completing
// tracing, a function can be obtained by TF_FinalizeFunction.
TF_ExecutionContext* TF_CreateFunction(const char* fn_name, TF_Status* status);

// Creates a context for eager execution of operations.

  // When `reset_context` is true, initialize new cluster context state based
  // on cluster configurations provided in `server_def`; otherwise, update
  // existing context state with the provided `server_def`. Contexts created
  // on remote tasks will be considered stale and garbage collected after
  // `keep_alive_secs` of inactivity.
  virtual Status SetOrUpdateServerDef(const ServerDef& server_def,

                                                   TF_Status* status);

// Returns the status for the tensor handle. In TFRT, a tensor handle can carry
// error info if error happens. If so, the status will be set with the error
// info. If not, status will be set as OK.
TF_CAPI_EXPORT extern void TFE_TensorHandleGetStatus(TFE_TensorHandle* h,
                                                     TF_Status* status);

  if (h == nullptr) {
    status->status = tensorflow::errors::InvalidArgument("Invalid handle");
    return -1;
  }
  return tensorflow::unwrap(h)->DeviceId(&status->status);
}

TF_CAPI_EXPORT extern void TFE_TensorHandleGetStatus(TFE_TensorHandle* h,
                                                     TF_Status* status) {
  status->status = tensorflow::unwrap(h)->TensorHandleStatus();
}

- Provide links to your TensorFlow model and (optionally) TensorFlow Lite Model.
```

### 3. Failure after conversion
If the conversion is successful, but the generated model is wrong, then state what is wrong:

- Model produces wrong results and/or has lesser accuracy.
- Model produces correct results, but it is slower than expected.

### 4. (optional) RNN conversion support

  typedef std::pair<std::string, size_t> Key;

  /// \brief The state of a block.
  ///
  /// A block begins in the CREATED stage. The first thread will attempt to read
  /// the block from the filesystem, transitioning the state of the block to
  /// FETCHING. After completing, if the read was successful the state should
  /// be FINISHED. Otherwise the state should be ERROR. A subsequent read can

  }
  return pjrt_c_api_client;
}

absl::Status ResetPjRtClient(const DeviceType& device_type) {
  ResourceMgr* rmgr = tfrt_global::GetTFGlobalResourceMgr();
  PjRtState* pjrt_state;
  TF_RETURN_IF_ERROR(rmgr->Lookup(rmgr->default_container(),
                                  kPjRtStateResourceName, &pjrt_state));
  TF_RETURN_IF_ERROR(pjrt_state->MovePjRtClientToUnused(device_type));
  return absl::OkStatus();
}

Status SetAttrFloat(AbstractOperation*, const char* attr_name, float value,
                    ForwardOperation*);
Status SetAttrBool(AbstractOperation*, const char* attr_name, bool value,
                   ForwardOperation*);
Status SetAttrType(AbstractOperation*, const char* attr_name, DataType value,
                   ForwardOperation*);
Status SetAttrShape(AbstractOperation*, const char* attr_name,

//   // forward `op`.
//   return new AddGradientFunction;
// }
//
// Status RegisterGradients(GradientRegistry* registry) {
//   return registry->Register("Add", AddRegisterer);
// }
class GradientFunction {
 public:
  virtual Status Compute(AbstractContext* ctx,
                         absl::Span<AbstractTensorHandle* const> grad_outputs,