}

TF_CAPI_EXPORT extern void TFE_AbortCollectiveOps(TFE_Context* ctx,
                                                  TF_Status* status) {
  tensorflow::EagerContext* context =
      tensorflow::ContextFromInterface(tensorflow::unwrap(ctx));
  auto collective_executor_handle = context->GetCollectiveExecutorHandle();
  collective_executor_handle->get()->StartAbort(status->status);
}

}

void TestConnectToCluster(bool keep_localhost_for_first_connect) {
  // Fail fast on GetStatus requests so we can get errors instead of timeout
  // when updating cluster with non-exsitent worker
  tensorflow::setenv("GRPC_FAIL_FAST", "TRUE", /*overwrite=*/1);

  const string first_name =
      keep_localhost_for_first_connect ? "localhost" : "abc";
  tensorflow::ServerDef server_def = GetServerDef(first_name, 1);

#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/c/eager/c_api_experimental.h"
#include "tensorflow/c/eager/parallel_device/parallel_device_testlib.h"
#include "tensorflow/c/eager/tfe_context_internal.h"
#include "tensorflow/core/common_runtime/eager/context.h"
#include "tensorflow/core/framework/function.h"
#include "tensorflow/core/lib/core/status_test_util.h"
#include "tensorflow/core/platform/test.h"

namespace tensorflow {

#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/math_ops.h"
#include "tensorflow/c/tf_status_helper.h"
#include "tensorflow/core/platform/tensor_float_32_utils.h"
#include "tensorflow/core/platform/test.h"

  TF_AddInput(desc, {input, 0});
  TF_SetAttrInt(desc, "num_split", 3);
  TF_SetAttrType(desc, "T", TF_INT32);
  // Set device to CPU since there is no version of split for int32 on GPU
  // TODO(iga): Convert all these helpers and tests to use floats because
  // they are usually available on GPUs. After doing this, remove TF_SetDevice
  // call in c_api_function_test.cc
  TF_SetDevice(desc, "/cpu:0");

#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;

  TFE_ContextOptions* opts = TFE_NewContextOptions();
  TFE_Context* context = TFE_NewContext(opts, status.get());
  TFE_DeleteContextOptions(opts);
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  bool arrived = false;
  bool executed = false;
  const char* name = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  RegisterLoggingDevice(context, name, /*strict_scope_placement=*/true,

  int64_t TensorId(AbstractTensorHandle* tensor) const override;

  // Converts a Gradient to a TapeTensor.
  TapeTensor TapeTensorFromGradient(AbstractTensorHandle* g) const override;

  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_;
};

  op = TFE_NewOp(ctx, "AssignVariableOp", status);
  if (TF_GetCode(status) != TF_OK) return nullptr;
  TFE_OpSetAttrType(op, "dtype", TF_FLOAT);
  TFE_OpAddInput(op, var_handle, status);

  // Convert 'value' to a TF_Tensor then a TFE_TensorHandle.
  std::unique_ptr<TF_Tensor, decltype(&TF_DeleteTensor)> t(
      TF_AllocateTensor(TF_FLOAT, nullptr, 0, sizeof(value)), TF_DeleteTensor);

    }
    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")) {