// the object. Using duplicate metric name will crash the program with fatal
// error.
TF_CAPI_EXPORT extern TFE_MonitoringCounter0* TFE_MonitoringNewCounter0(
    const char* name, TF_Status* status, const char* description);
// Deletes the Counter object.
TF_CAPI_EXPORT extern void TFE_MonitoringDeleteCounter0(
    TFE_MonitoringCounter0* counter);
// Retrieves the cell from the Counter object. The Counter object will manage

// read current values from streamz counters defined in other modules.
//
// The code under test will have created streamz counters like this:
// auto* streamz = tensorflow::monitoring::Counter<1>::New("name",
// "description", "label");
// and then incremented that counter for various values of label:
// streamz->GetCell("label-value")->IncrementBy(1);
//
// The test code can then read and test the value of that counter:
//

  // The devices operations run on.
  const std::vector<std::string>& underlying_devices() const {
    return underlying_devices_;
  }

  // Takes a description of a single operation being executed on the
  // ParallelDevice, and in turn runs one operation per component device with
  // its corresponding inputs from the input ParallelTensors. Wraps the

//                 names for outputs will be generated automatically.
//  opts - various options for the function, e.g. XLA's inlining control.
//  description - optional human-readable description of this function.
//  status - Set to OK on success and an appropriate error on failure.
//
// Note that when the same TF_Output is listed as both an input and an output,

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;
};

// The return value was current at the time of TFE_TensorDebugInfo creation.
TF_CAPI_EXPORT extern int64_t TFE_TensorDebugInfoOnDeviceDim(
    TFE_TensorDebugInfo* debug_info, int dim_index);

// Description of the TensorFlow op to execute.
//
// Assumes that the provided 'ctx' outlives the returned TFE_Op, i.e.,
// TFE_DeleteOp() is called before TFE_DeleteContext().
//

  // session has been run (this is detected in RecordMutation function). If the
  // string is empty, no mutation has occurred. Otherwise the string is a
  // description of the mutation suitable for returning to the user.
  //
  // Sessions are added to this map in TF_NewSession, and removed in
  // TF_DeleteSession.
  // TF_Graph may only / must be deleted when

      gtl::ArraySlice<Gradient*> output_gradients, absl::Span<Gradient*> result,
      bool build_default_zeros_grads = true);

  // Whether the tape is persistent. See ctor for detailed description.
  bool IsPersistent() const { return persistent_; }

 private:
  TensorTape tensor_tape_;
  OpTape<BackwardFunction, TapeTensor> op_tape_;
  int64_t next_op_id_{0};

      DataType dtype, absl::Span<const int64_t> dim_sizes) = 0;

  typedef void (*MemoryReleaser)(void* data, size_t len, void* arg);

  // Create a tensor instance from the given data buffer and description.
  // `memory_releaser` will be called on destruction, and it's responsible for
  // cleaning up the underlying buffer.
  virtual AbstractTensorInterface* CreateTensor(

  TF_Filesystem_Num_Option_Types,  // must always be the last item
} TF_Filesystem_Option_Type;

typedef struct TF_Filesystem_Option {
  char* name;                         // null terminated, owned
  char* description;                  // null terminated, owned
  int per_file;                       // bool actually, but bool is not a C type
  TF_Filesystem_Option_Value* value;  // owned
} TF_Filesystem_Option;