TracingContext* tracing_ctx = dyn_cast<TracingContext>(unwrap(ctx));
  if (!tracing_ctx) {
    tsl::Set_TF_Status_from_Status(
        s, tensorflow::errors::InvalidArgument(
               "Only TracingContext can be converted into a function."));
    return nullptr;
  }
  tsl::Set_TF_Status_from_Status(s,
                                 tracing_ctx->Finalize(unwrap(outputs), &func));
  TF_DeleteExecutionContext(ctx);

      ->set_disable_optimize_for_static_graph(true);
}

void TF_TensorFromProto(const TF_Buffer* from, TF_Tensor* to,
                        TF_Status* status) {
  TF_SetStatus(status, TF_OK, "");
  tensorflow::TensorProto from_tensor_proto;
  status->status = BufferToMessage(from, &from_tensor_proto);
  if (!status->status.ok()) {
    return;
  }
  status->status =

==============================================================================*/

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

#include <vector>

#include "absl/strings/str_join.h"
#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,

     *
     *    dA = U
     *    dbias = reduceSum(U, dims = channel_dim)
     */

    AbstractTensorHandle* upstream_grad = grad_outputs[0];
    DCHECK(upstream_grad);

    // Recover data format from forward pass for gradient.
    std::string data_format;
    TF_RETURN_IF_ERROR(forward_attrs_.Get("data_format", &data_format));

    // Grad for A
    grad_inputs[0] = upstream_grad;
    grad_inputs[0]->Ref();

void TF_CreateDir(const char* dirname, TF_Status* status) {
  TF_SetStatus(status, TF_OK, "");
  ::tensorflow::Set_TF_Status_from_Status(
      status, ::tensorflow::Env::Default()->CreateDir(dirname));
}

void TF_DeleteDir(const char* dirname, TF_Status* status) {
  TF_SetStatus(status, TF_OK, "");
  ::tensorflow::Set_TF_Status_from_Status(
      status, ::tensorflow::Env::Default()->DeleteDir(dirname));
}

          ->LookUpOpDef(name, &op_def_ptr);
  if (!s.ok()) {
    tsl::Set_TF_Status_from_Status(status, s);
    return;
  }

  s = tensorflow::MessageToBuffer(*op_def_ptr, buf);
  if (!s.ok()) {
    tsl::Set_TF_Status_from_Status(status, s);
    return;
  }

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

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

#include <string>

#include "include/dlpack/dlpack.h"  // from @dlpack
#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/c/eager/c_api_experimental.h"
#include "tensorflow/c/eager/tfe_tensorhandle_internal.h"
#include "tensorflow/c/tf_status_internal.h"

    const TFE_OpAttrs* attributes, int expected_max_outputs,
    TF_Status* status) {
  absl::optional<std::vector<MaybeParallelTensorOwned>> result;
  // TODO(allenl): We should remove "TPU" from these op names at the very least,
  // or consider other ways of packing/unpacking parallel tensors.
  if (operation_name == std::string("TPUReplicatedInput")) {

  for (int i{}; i < inputs.size(); ++i) {
    theta_inputs[i] = inputs[i];
  }

  AbstractTensorHandle* theta =
      theta_inputs[input_index];  // parameter we are grad checking

  // Convert from AbstractTensor to TF_Tensor.
  TF_Tensor* theta_tensor;
  TF_RETURN_IF_ERROR(GetValue(theta, &theta_tensor));

  // Get number of elements and fill data.
  int num_elems = TF_TensorElementCount(theta_tensor);

                           bool* executed_flag, TF_Status* status) {
  TFE_CustomDevice custom_device;
  custom_device.copy_tensor_to_device = &CopyToLoggingDevice;
  custom_device.copy_tensor_from_device = &CopyTensorFromLoggingDevice;
  custom_device.delete_device = &DeleteLoggingDevice;
  custom_device.execute = &LoggingDeviceExecute;
  LoggingDevice* device = new LoggingDevice;
  device->arrived_flag = arrived_flag;