void** device_info) {
  TFE_CustomDevice* custom_device = new TFE_CustomDevice;
  custom_device->copy_tensor_to_device = &CopyToLoggingDevice;
  custom_device->copy_tensor_from_device = &CopyTensorFromLoggingDevice;
  custom_device->delete_device = &DeleteLoggingDevice;
  custom_device->execute = &LoggingDeviceExecute;
  *device = custom_device;
  LoggingDevice* logging_device = new LoggingDevice;

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

#ifndef TENSORFLOW_C_EAGER_CUSTOM_DEVICE_TESTUTIL_H_
#define TENSORFLOW_C_EAGER_CUSTOM_DEVICE_TESTUTIL_H_

// A simple logging device to test custom device registration.
#include <memory>

#include "tensorflow/c/c_api.h"
#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/c/eager/c_api_experimental.h"

    CHECK_EQ(grad_inputs.size(), 1);
    grad_inputs[0] = grad_outputs[0];
    if (grad_inputs[0]) {
      grad_inputs[0]->Ref();
    }
    return absl::OkStatus();
  }
};

// Computes:
//
// @tf.custom_gradient
// def f(input):
//   def grad(grads):
//     return grads[0]
//   return tf.exp(input), grad
// outputs = [f(inputs[0])]
Status ExpWithPassThroughGrad(AbstractContext* ctx,

  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());

  const char* custom0 = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  const char* custom1 = "/job:localhost/replica:0/task:0/device:CUSTOM:1";
  bool arrived = false;
  bool executed = false;
  RegisterLoggingDevice(context.get(), custom0,
                        /*strict_scope_placement=*/false, &arrived, &executed,

#include "tensorflow/c/c_api.h"
#include "tensorflow/c/experimental/grappler/grappler.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/grappler/optimizers/custom_graph_optimizer.h"
#include "tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.h"
#include "tensorflow/core/platform/status.h"
#include "tensorflow/core/protobuf/rewriter_config.pb.h"

namespace tensorflow {
namespace grappler {

-   type: input id: tfversion attributes: label: Tensorflow Version description:
    placeholder: ex,. tf 2.8 validations: required: true

-   type: dropdown id: Code attributes: label: Custom Code description:
    options: - "Yes" - "No" validations: required: true

-   type: input id: OS attributes: label: OS Platform and Distribution
    description: placeholder: e.g., Linux Ubuntu 16.04 validations: required:

  const char* first_device = "/job:worker/replica:0/task:1/device:CPU:0";
  const char* second_device = "/job:worker/replica:0/task:2/device:CPU:0";
  const char* device_name = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  std::array<const char*, 2> underlying_devices{first_device, second_device};
  RegisterParallelDevice(context.get(), device_name, underlying_devices,
                         status.get());

// on each underlying device.
//
// For example if `device_name` is
//   "/job:localhost/replica:0/task:0/device:CUSTOM:0"
// and `underlying_devices` is
//   {"/job:localhost/replica:0/task:0/device:GPU:0",
//    "/job:localhost/replica:0/task:0/device:GPU:1"}
// Then executing an operation on CUSTOM:0 will execute it on GPU:0 and GPU:1.
//
// Implicit copies onto `device_name` are allowed, replicating the value once

  // additional CPU.
  const char* second_device_name =
      "/job:localhost/replica:0/task:0/device:CUSTOM:1";
  std::array<const char*, 2> second_underlying_devices{
      "/job:localhost/replica:0/task:0/device:CUSTOM:0",
      "/job:localhost/replica:0/task:0/device:CPU:2"};
  RegisterParallelDevice(context.get(), second_device_name,

    id: tfversion
    attributes:
      label: TensorFlow version
      placeholder: e.g., tf 2.8
    validations:
      required: true
  - type: dropdown
    id: Code
    attributes:
      label: Custom code
      options:
        - "Yes"
        - "No"
    validations:
      required: true
  - type: input
    id: OS
    attributes:
      label: OS platform and distribution