tensorflow::ServerDef GetServerDef(const std::string& job_name, int num_tasks) {
  tensorflow::ServerDef server_def;
  server_def.set_protocol("grpc");
  server_def.set_job_name(job_name);
  server_def.set_task_index(0);
  tensorflow::ClusterDef* cluster_def = server_def.mutable_cluster();
  tensorflow::JobDef* job_def = cluster_def->add_job();
  job_def->set_name(job_name);
  for (int i = 0; i < num_tasks; i++) {

  auto* reader = TFE_MonitoringNewCounterReader(counter_name);
  IncrementCounter1(counter, label_name);

  int64_t actual = TFE_MonitoringReadCounter1(reader, label_name);

  CHECK_EQ(actual, 1);
}

TFE_MonitoringCounter0* CreateCounter0(const char* counter_name) {
  TF_Status* status = TF_NewStatus();
  auto* counter =
      TFE_MonitoringNewCounter0(counter_name, status, "description");
  TF_DeleteStatus(status);

extern "C" {
#endif

// Resets `op_to_reset` with `op_or_function_name` and `raw_device_name`. This
// is for performance optimization by reusing an exiting unused op rather than
// creating a new op every time. If `raw_device_name` is `NULL` or empty, it
// does not set the device name. If it's not `NULL`, then it attempts to parse
// and set the device name. It's effectively `TFE_OpSetDevice`, but it is faster

                          const char* name) {
  unique_tensor_ptr tensor(BoolTensor(v), TF_DeleteTensor);
  return Const(tensor.get(), graph, s, name);
}

TF_Operation* ScalarConst(int32_t v, TF_Graph* graph, TF_Status* s,
                          const char* name) {
  unique_tensor_ptr tensor(Int32Tensor(v), TF_DeleteTensor);
  return Const(tensor.get(), graph, s, name);
}

  )

# Add Keras module aliases
_losses = _KerasLazyLoader(globals(), submodule="losses", name="losses")
_metrics = _KerasLazyLoader(globals(), submodule="metrics", name="metrics")
_optimizers = _KerasLazyLoader(
    globals(), submodule="optimizers", name="optimizers")
_initializers = _KerasLazyLoader(
    globals(), submodule="initializers", name="initializers")
setattr(_current_module, "losses", _losses)

    return 1;
  }

  char file_name[100];
  time_t t = time(NULL);
  snprintf(file_name, sizeof(file_name), "test-%d-%ld.txt", getpid(), t);

  size_t length = 2 + strlen(path) + strlen(file_name);
  char* full_path = malloc(length);
  snprintf(full_path, length, "%s/%s", path, file_name);

  TF_WritableFileHandle* h;
  TF_Status* status = TF_NewStatus();

};

template <int NumLabels>
struct TFE_MonitoringCounter {
  template <typename... LabelDesc>
  TFE_MonitoringCounter(const char* name, const char* description,
                        LabelDesc&&... label) {
    counter = absl::WrapUnique(tensorflow::monitoring::Counter<NumLabels>::New(
        name, description, label...));
  }

  std::unique_ptr<tensorflow::monitoring::Counter<NumLabels>> counter;
};

      << TF_Message(status.get());

  const char* name = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  RegisterLoggingDevice(context.get(), name, /*strict_scope_placement=*/true,
                        &arrived, &executed, status.get());
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  RegisterLoggingDevice(context.get(), name, /*strict_scope_placement=*/true,

    ],
)

filegroup(
    name = "pywrap_tf_session_hdrs",
    srcs = [
        "python_api.h",
    ],
    visibility = [
        "//tensorflow/core:__pkg__",
        "//tensorflow/python:__pkg__",
    ],
)

cc_library(
    name = "tf_attrtype",
    hdrs = ["tf_attrtype.h"],
    visibility = ["//visibility:public"],
)

cc_library(
    name = "c_api_macros_hdrs",
    hdrs = [

### How to update/add requirements

By default, the name of the base requirements file is `requirements.in`, but it
can be set using the `REQUIREMENTS_FILE_NAME` variable. \
For example:

```
export REQUIREMENTS_FILE_NAME=my_requirements.in
```

To specify the file via a Bazel command argument, use the following:

```
--repo_env=REQUIREMENTS_FILE_NAME=my_requirements.in
```

### How to run the updater

```