ASSERT_TRUE(executed);
  auto handle_cleaner = tensorflow::gtl::MakeCleanup(
      [var_handle]() { TFE_DeleteTensorHandle(var_handle); });

  // Assign to the variable, copying to the custom device.
  std::unique_ptr<TFE_TensorHandle, decltype(&TFE_DeleteTensorHandle)> one(
      TestScalarTensorHandle(context.get(), 111.f), TFE_DeleteTensorHandle);

  tensorflow::Tensor tensor(allocator.get(), DT_FLOAT, {1});
  TF_ASSERT_OK_AND_ASSIGN(
      auto pjrt_client,
      xla::GetTfrtCpuClient(/*asynchronous=*/true, /*cpu_device_count=*/1));
  std::vector<int32_t> data(1, 0);
  xla::Shape shape = xla::ShapeUtil::MakeShape(xla::S32, {1});
  TF_ASSERT_OK_AND_ASSIGN(
      auto buffer,
      pjrt_client->BufferFromHostBuffer(

}

void Variable::AssignAdd(TFE_Context* context, TFE_TensorHandle* value,
                         TF_Status* status) {
  GeneralAssignment("AssignAddVariableOp", context, value, status);
}

void Variable::Assign(TFE_Context* context, TFE_TensorHandle* value,
                      TF_Status* status) {
  GeneralAssignment("AssignVariableOp", context, value, status);
}

    std::unique_ptr<CSession> expected_csession(
        new CSession(expected_graph_, s_));

    std::vector<TF_Output> grad_outputs_vec;
    grad_outputs_vec.assign(grad_outputs, grad_outputs + 2);
    csession->SetOutputs(grad_outputs_vec);
    csession->Run(s_);
    ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
    TF_Tensor* out0 = csession->output_tensor(0);

  if (av_tensor == nullptr) {
    TF_ASSIGN_OR_RETURN(
        *tensor, MakeTensorFromPjRtBuffer(tensor->dtype(), tensor->shape(),
                                          std::move(buffer)));
  } else {
    av_tensor->SetBuffer(std::move(buffer));
  }
  return absl::OkStatus();
}

absl::StatusOr<xla::PjRtCApiClient*> GetPjRtCApiClient(
    const DeviceType& device_type) {

TEST(CAPI_EXPERIMENTAL, IsStateful) {
  std::unique_ptr<TF_Status, decltype(&TF_DeleteStatus)> status(
      TF_NewStatus(), TF_DeleteStatus);
  int assign = TF_OpIsStateful("AssignAddVariableOp", status.get());
  ASSERT_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());
  EXPECT_EQ(assign, 1);
  int id = TF_OpIsStateful("Identity", status.get());
  ASSERT_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());
  EXPECT_EQ(id, 0);

  TFE_Execute(op, &var_handle, &num_retvals, status);
  if (TF_GetCode(status) != TF_OK) return nullptr;
  TFE_DeleteOp(op);
  if (TF_GetCode(status) != TF_OK) return nullptr;
  CHECK_EQ(1, num_retvals);

  // Assign 'value' to it.
  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);

  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
  TF_GraphCopyFunction(host_graph_, func_, grad_func2, s_);
  EXPECT_EQ(TF_INVALID_ARGUMENT, TF_GetCode(s_));
  EXPECT_EQ(string("Cannot assign gradient function 'MyGrad2' to 'MyFunc' "
                   "because it already has gradient function 'MyGrad1'"),
            string(TF_Message(s_)));

  TF_DeleteFunction(grad_func1);