TFE_NewContextOptions(), TFE_DeleteContextOptions);
  std::unique_ptr<TF_Status, decltype(&TF_DeleteStatus)> status(
      TF_NewStatus(), TF_DeleteStatus);
  std::unique_ptr<TFE_Context, decltype(&TFE_DeleteContext)> context(
      TFE_NewContext(opts.get(), status.get()), TFE_DeleteContext);
  tensorflow::ServerDef server_def = GetServerDef("worker", 3);

  // This server def has the task index set to 0.

#include "include/dlpack/dlpack.h"  // from @dlpack
#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/core/platform/test.h"

namespace tensorflow {
namespace {

void TestHandleFromDLPack(TF_Status* status, TFE_Context* ctx,
                          std::vector<int64_t> shape,
                          std::vector<int64_t> strides) {
  size_t num_elements = 1;
  for (int i = 0; i < static_cast<int32_t>(shape.size()); ++i) {

void TF_SetTracingImplementation(const char* name, TF_Status* s) {
  tsl::Set_TF_Status_from_Status(s, SetDefaultTracingEngine(name));
}

// Creates a new TensorFlow function, it is an execution context attached to a
// given tracing context.
TF_ExecutionContext* TF_CreateFunction(const char* fn_name, TF_Status* s) {
  return wrap(CreateTracingExecutionContext(fn_name, s));
}

#include "tensorflow/c/experimental/gradients/nn_grad.h"

#include "absl/types/span.h"
#include "tensorflow/c/eager/abstract_tensor_handle.h"
#include "tensorflow/c/eager/immediate_execution_context.h"
#include "tensorflow/c/eager/immediate_execution_tensor_handle.h"
#include "tensorflow/c/experimental/ops/array_ops.h"
#include "tensorflow/c/experimental/ops/math_ops.h"
#include "tensorflow/c/experimental/ops/nn_ops.h"

    }
    std::vector<MaybeParallelTensorOwned> result_content;
    result_content.reserve(1);
    result_content.push_back(ParallelTensor::FromTensorHandles(
        parallel_device, std::move(components), status));
    if (TF_GetCode(status) != TF_OK) return result;
    result.emplace(std::move(result_content));
    return result;
  } else if (operation_name == std::string("TPUReplicatedOutput")) {

#include "tensorflow/core/platform/logging.h"

namespace tensorflow {

namespace {

// Managing context for the DLManagedTensor, will manage the lifetime of
// DLManagedTensor. When calling DLManagedTensor::deleter, it will notify the
// original framework of destruction, and this context will be deleted also.
struct TfDlManagedTensorCtx {
  TensorReference reference;
  std::vector<int64_t> shape;

#include "tensorflow/c/eager/unified_api_testutil.h"
#include "tensorflow/c/experimental/gradients/grad_test_helper.h"
#include "tensorflow/c/experimental/gradients/tape/tape_context.h"
#include "tensorflow/c/experimental/ops/array_ops.h"
#include "tensorflow/c/tf_status_helper.h"
#include "tensorflow/core/platform/tensor_float_32_utils.h"
#include "tensorflow/core/platform/test.h"

  void MarkAsResult(AbstractTensorHandle* gradient) const override;

  void DeleteGradient(AbstractTensorHandle* gradient) const override;

 private:
  // The context where the aggregation op `Add` is to be created.
  AbstractContext* ctx_;
};

// Returns the number of elements in the gradient tensor.
int64_t TapeVSpace::NumElements(AbstractTensorHandle* tensor) const {

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

#include <memory>

#include "absl/container/flat_hash_set.h"
#include "absl/types/span.h"
#include "tensorflow/c/eager/abstract_context.h"
#include "tensorflow/c/eager/abstract_tensor_handle.h"
#include "tensorflow/c/eager/c_api_experimental.h"
#include "tensorflow/c/eager/c_api_test_util.h"
#include "tensorflow/c/eager/c_api_unified_experimental.h"

        op_outputs_ = std::vector<TensorHandlePtr>();
        Execute(context_, operation_name_, std::move(op_inputs_), attributes_,
                expected_max_outputs_, &op_outputs_, status_.get());
        execution_state_ = ExecutionState::kHasResult;
      }
    }
    finished_execute_.notify_one();
  }
}

void DeviceThread::StartExecute(TFE_Context* context,
                                const char* operation_name,