// Map from tensor ID to how many references still exist for this tensor in
  // the tape.
  std::unordered_map<int64_t, int64_t> tensor_usage_counts;

  // Maps from op ID to how many output tensors of this op still need to have
  // their gradients computed.
  std::unordered_map<int64_t, int64_t> op_missing_tensor;
};

// If `persistent_tape` is true, op_tape is not changed and none of the

  TapeVSpace vspace(ctx);
  std::vector<int64_t> target_tensor_ids = MakeTensorIDList(targets);
  std::vector<int64_t> source_tensor_ids = MakeTensorIDList(sources);
  tensorflow::gtl::FlatSet<int64_t> sources_set(source_tensor_ids.begin(),
                                                source_tensor_ids.end());
  std::unordered_map<int64_t, TapeTensor> sources_that_are_targets;

TFE_MonitoringCounter1* CreateCounter1(const char* counter_name,
                                       const char* label);
void IncrementCounter0(TFE_MonitoringCounter0* counter, int64_t delta = 1);
void IncrementCounter1(TFE_MonitoringCounter1* counter, const char* label,
                       int64_t delta = 1);

TEST(CAPI, MonitoringCellReader0) {
  auto counter_name = "test/counter0";
  auto* counter = CreateCounter0(counter_name);

namespace {

std::vector<int64_t> TensorShapeAsVector(const tensorflow::TensorHandle& handle,
                                         tensorflow::Status* status) {
  std::vector<int64_t> shape;
  int rank = -1;
  *status = handle.NumDims(&rank);
  if (!status->ok()) {
    return shape;
  }
  shape.reserve(rank);
  for (int i = 0; i < rank; ++i) {
    int64_t dim;
    *status = handle.Dim(i, &dim);

                       size_t length) override;
  Status SetAttrInt(const char* attr_name, int64_t value) override;
  Status SetAttrFloat(const char* attr_name, float value) override;
  Status SetAttrBool(const char* attr_name, bool value) override;
  Status SetAttrType(const char* attr_name, DataType value) override;
  Status SetAttrShape(const char* attr_name, const int64_t* dims,
                      const int num_dims) override;

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

TFE_MonitoringCounterReader* TFE_MonitoringNewCounterReader(const char* name) {
  auto* result = new TFE_MonitoringCounterReader(name);

  return result;
}

int64_t TFE_MonitoringReadCounter0(TFE_MonitoringCounterReader* cell_reader) {
  int64_t result = cell_reader->Read();

  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;
};

}
Status TapeOperation::SetAttrIntList(const char* attr_name,
                                     const int64_t* values, int num_values) {
  forward_op_.attrs.Set(
      attr_name, gtl::ArraySlice<const int64_t>(
                     reinterpret_cast<const int64_t*>(values), num_values));
  return parent_op_->SetAttrIntList(attr_name, values, num_values);
}

    return absl::OkStatus();
  }
  Status SetAttrShape(const char* attr_name, const int64_t* dims,
                      const int num_dims) override {
    PartialTensorShape shape;
    if (num_dims >= 0) {
      shape = PartialTensorShape(ArraySlice<int64_t>(
          reinterpret_cast<const int64_t*>(dims), num_dims));
    }
    op_->node_builder.Attr(attr_name, shape);
    return absl::OkStatus();
  }

    buf.push_back(i);
  }
  // The fetcher just fetches slices of the buffer.
  auto fetcher = [&buf](const string& filename, size_t offset, size_t n,
                        char* buffer, TF_Status* status) -> int64_t {
    int64_t bytes_transferred;
    if (offset < buf.size()) {
      size_t bytes_to_copy = std::min<size_t>(buf.size() - offset, n);
      memcpy(buffer, buf.data() + offset, bytes_to_copy);