std::unique_ptr<TensorSliceReader::VarToShapeMap> var_to_shape_map_;
  std::unique_ptr<TensorSliceReader::VarToDataTypeMap> var_to_data_type_map_;

  CheckpointReader(const CheckpointReader&) = delete;
  void operator=(const CheckpointReader&) = delete;
};

}  // namespace checkpoint
}  // namespace tensorflow

class OpDeleter {
 public:
  void operator()(TFE_Op* to_delete) const { TFE_DeleteOp(to_delete); }
};

using OpPtr = std::unique_ptr<TFE_Op, OpDeleter>;

class StatusDeleter {
 public:
  void operator()(TF_Status* to_delete) const { TF_DeleteStatus(to_delete); }
};

using StatusPtr = std::unique_ptr<TF_Status, StatusDeleter>;

class ExecutorDeleter {
 public:
  void operator()(TFE_Executor* to_delete) const {

      read_only_memory_region_ops_;
  std::function<void*(size_t)> plugin_memory_allocate_;
  std::function<void(void*)> plugin_memory_free_;
  ModularFileSystem(const ModularFileSystem&) = delete;
  void operator=(const ModularFileSystem&) = delete;
};

class ModularRandomAccessFile final : public RandomAccessFile {
 public:
  ModularRandomAccessFile(const std::string& filename,

  TFE_OpAddInput(minOp, axis, status);
  CHECK_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);

  tensorflow::AttrValueMap attr_values = ExtractAttrs(minOp);
  tensorflow::AttrValueMap::const_iterator attr_found = attr_values.find("T");
  EXPECT_NE(attr_found, attr_values.cend());
  EXPECT_EQ(attr_found->second.type(), tensorflow::DataType::DT_FLOAT);
  attr_found = attr_values.find("Tidx");

limitations under the License.
==============================================================================*/

#include "tensorflow/c/c_api.h"

#include <algorithm>
#include <cstddef>
#include <iterator>
#include <memory>
#include <vector>

#include "tensorflow/c/c_api_internal.h"
#include "tensorflow/c/c_test_util.h"
#include "tensorflow/c/tf_buffer.h"
#include "tensorflow/c/tf_buffer_internal.h"

  return oper->node.attrs().size();
}

int TF_OperationGetAttrNameLength(TF_Operation* oper, int i) {
  auto attrs = oper->node.attrs();
  int count = 0;
  AttrValueMap::const_iterator it;
  for (it = attrs.begin(); it != attrs.end(); it++) {
    if (count == i) {
      return it->first.length();
    }
    count++;
  }
  return -1;
}

// it will be stopped and joined.
TF_CAPI_EXPORT extern void TF_DeleteServer(TF_Server* server);

// Register a listener method that processes printed messages.
//
// If any listeners are registered, the print operator will call all listeners
// with the printed messages and immediately return without writing to the
// logs.
TF_CAPI_EXPORT extern void TF_RegisterLogListener(
    void (*listener)(const char*));

      : AbstractContext(kind) {}
  ~ImmediateExecutionContext() override {}
};

namespace internal {
struct ImmediateExecutionContextDeleter {
  void operator()(ImmediateExecutionContext* p) const {
    if (p != nullptr) {
      p->Release();
    }
  }
};
}  // namespace internal

using ImmediateContextPtr =
    std::unique_ptr<ImmediateExecutionContext,

void ForwardAccumulator<Gradient, BackwardFunction, TapeTensor>::Watch(
    int64_t tensor_id, Gradient* tangent) {
  typename std::unordered_map<int64_t, Gradient*>::iterator existing =
      accumulated_gradients_.find(tensor_id);
  vspace_.MarkAsResult(tangent);
  if (existing == accumulated_gradients_.end()) {
    accumulated_gradients_.emplace(tensor_id, tangent);
  } else {

    *   Add NUMA-aware MapAndBatch dataset.
    *   Deprecate `tf.data.Dataset.make_one_shot_iterator()` in V1, removed it
        from V2, and added tf.compat.v1.data.make_one_shot_iterator()`.
    *   Deprecate `tf.data.Dataset.make_initializable_iterator()` in V1, removed
        it from V2, and added `tf.compat.v1.data.make_initializable_iterator()`.
    *   Enable nested dataset support in core `tf.data` transformations.