private:
  static_assert(!std::is_reference<F>::value, "F must not be a reference");

  bool released_ = false;
  F f_;
};

template <int&... ExplicitParameterBarrier, typename F,
          typename DecayF = typename std::decay<F>::type>
Cleanup<DecayF> MakeCleanup(F&& f) {
  return Cleanup<DecayF>(std::forward<F>(f));
}

}  // namespace tf_gcs_filesystem

  tensorflow::monitoring::GaugeCell<tensorflow::string> cell;
};
struct TFE_MonitoringBoolGaugeCell {
  tensorflow::monitoring::GaugeCell<bool> cell;
};

template <typename ValueType, int NumLabels>
struct TFE_MonitoringGauge {
  template <typename... LabelDesc>
  TFE_MonitoringGauge(const char* name, const char* description,
                      LabelDesc&&... label) {
    gauge = absl::WrapUnique(

TensorHandlePtr Multiply(TFE_Context* context, TFE_TensorHandle* first,
                         TFE_TensorHandle* second, TF_Status* status);

// Assert that `handle` is equal to `expected_value`.
template <typename value_type>
void ExpectScalarEq(TFE_TensorHandle* handle, value_type expected_value);

template <std::size_t num_devices>
void RegisterParallelDevice(
    TFE_Context* context, const char* device_name,

                                                       TF_Status* status) const;

  // Construct a parallel tensor consisting of the scalar values from `values`.
  template <typename DataType>
  std::unique_ptr<ParallelTensor> ScalarsFromSequence(
      absl::Span<const DataType> values, TFE_Context* context,
      TF_Status* status) const;

        return true;
      }
    }
  }
  return false;
}

template <typename Gradient, typename BackwardFunction, typename TapeTensor>
void GradientTape<Gradient, BackwardFunction, TapeTensor>::Watch(
    int64_t tensor_id) {
  tensor_tape_.emplace(tensor_id, -1);
}

template <typename Gradient, typename BackwardFunction, typename TapeTensor>
void GradientTape<Gradient, BackwardFunction, TapeTensor>::RecordOperation(

struct TFE_MonitoringCounterReader {
  explicit TFE_MonitoringCounterReader(const char* name) {
    counter = std::make_unique<
        ::tensorflow::monitoring::testing::CellReader<int64_t>>(name);
  }
  template <typename... LabelType>
  int64_t Read(const LabelType&... labels);
  std::unique_ptr<::tensorflow::monitoring::testing::CellReader<int64_t>>
      counter;
};

namespace tf_gcs_filesystem {

/// \brief An LRU cache of string keys and arbitrary values, with configurable
/// max item age (in seconds) and max entries.
///
/// This class is thread safe.
template <typename T>
class ExpiringLRUCache {
 public:
  /// A `max_age` of 0 means that nothing is cached. A `max_entries` of 0 means
  /// that there is no limit on the number of entries in the cache (however, if

==============================================================================*/

#include "tensorflow/c/eager/c_api_experimental_reader.h"

#include "tensorflow/c/eager/tfe_monitoring_reader_internal.h"

template <typename... LabelType>
int64_t TFE_MonitoringCounterReader::Read(const LabelType&... labels) {
  return counter->Read(labels...);
}

TFE_MonitoringCounterReader* TFE_MonitoringNewCounterReader(const char* name) {

//     probe for gadgets inside core TensorFlow. We can even protect the area
//     of memory where the copies reside to not allow any more writes to it
//     after all copies are created.
template <typename T>
static std::unique_ptr<const T> CopyToCore(const T* plugin_ops,
                                           size_t plugin_size) {
  if (plugin_ops == nullptr) return nullptr;