return context->GetContextId();
}

void TFE_MonitoringCounterCellIncrementBy(TFE_MonitoringCounterCell* cell,
                                          int64_t value) {
  cell->cell.IncrementBy(value);
}

int64_t TFE_MonitoringCounterCellValue(TFE_MonitoringCounterCell* cell) {
  return cell->cell.value();
}

TFE_MonitoringCounter0* TFE_MonitoringNewCounter0(const char* name,

// Atomically increments the value of the cell. The value must be non-negative.
TF_CAPI_EXPORT extern void TFE_MonitoringCounterCellIncrementBy(
    TFE_MonitoringCounterCell* cell, int64_t value);

// Retrieves the current value of the cell.
TF_CAPI_EXPORT extern int64_t TFE_MonitoringCounterCellValue(
    TFE_MonitoringCounterCell* cell);

// APIs for Counter without label.

//tensorflow/python/debug:grpc_tensorflow_server.par
//tensorflow/python/feature_column:vocabulary_testdata
//tensorflow/python/util:nest_test_main_lib
//tensorflow/lite/experimental/examples/lstm:rnn_cell
//tensorflow/lite/experimental/examples/lstm:rnn_cell.py
//tensorflow/lite/experimental/examples/lstm:unidirectional_sequence_lstm_test
//tensorflow/lite/experimental/examples/lstm:unidirectional_sequence_lstm_test.py
//tensorflow/lite/python:interpreter

                                   absl::Span<AbstractTensorHandle* const>,
                                   absl::Span<AbstractTensorHandle*>)>;

// Runs `model` maybe wrapped in a function call op. This can be thought as
// being equivalent to the following python code.
//
// if use_function:
//   outputs = tf.function(model)(inputs)
// else:
//   outputs = model(inputs)

        "tfe_monitoring_reader_internal.h",
    ],
    visibility = ["//tensorflow:internal"],
    deps = [
        ":c_api",
        "//tensorflow/c:c_api",
        "//tensorflow/core/lib/monitoring:cell_reader",
        "@com_google_absl//absl/memory",
    ],
    alwayslink = 1,
)

tf_cuda_cc_test(
    name = "c_api_experimental_test",
    size = "small",
    srcs = [

    *   tf.optimizers => tf.keras.optimizers
    *   Add tf.keras.layers.AbstractRNNCell as the preferred implementation of
        RNN cell for TF v2. User can use it to implement RNN cell with custom
        behavior.
    *   Adding `clear_losses` API to be able to clear losses at the end of
        forward pass in a custom training loop in eager.

                                            int keep_alive_secs) = 0;

  // Set up a multi-client distributed execution environment. Must be called
  // on all tasks in the cluster. This call internally coordinates with other
  // tasks to initialize the eager context and TF server for multi-client
  // execution.
  virtual Status EnableCollectiveOps(const ServerDef& server_def) = 0;

                          const void** buf, size_t* len);

// TODO(josh11b,mrry): Change Session to be able to use a Graph*
// directly, instead of requiring us to serialize to a GraphDef and
// call Session::Extend().
bool ExtendSessionGraphHelper(TF_Session* session, TF_Status* status) {
  if (session->graph != nullptr) {
    // Take the graph lock before the session lock to avoid deadlock. This is

}

// Note: this function looks up a thread local policy. So it should be called in
// the appropriate client thread. In particular, in async mode, it may not be
// safe to call this function from the async EagerExecutor threads.
extern TFE_ContextDevicePlacementPolicy TFE_ContextGetDevicePlacementPolicy(
    TFE_Context* ctx) {
  return static_cast<TFE_ContextDevicePlacementPolicy>(

      new GradientTape<Gradient, BackwardFunction, TapeTensor>(false));
  AccumulatorCallState& call_state = call_state_.top();
  call_state.backward_tape = tape.get();
  auto pop_backward_tape =
      gtl::MakeCleanup([&call_state] { call_state.backward_tape = nullptr; });
  std::vector<Gradient*> forwardprop_aids;
  std::vector<int64_t> sources;
  std::unordered_set<int64_t> sources_set;