TF_Operation* neg = TF_GraphOperationByName(graph, "imported/neg");
  ASSERT_TRUE(scalar != nullptr);
  ASSERT_TRUE(feed != nullptr);
  ASSERT_TRUE(neg != nullptr);

  // Test basic structure of the imported graph.
  EXPECT_EQ(0, TF_OperationNumInputs(scalar));
  EXPECT_EQ(0, TF_OperationNumInputs(feed));
  ASSERT_EQ(1, TF_OperationNumInputs(neg));
  TF_Output neg_input = TF_OperationInput({neg, 0});

#ifndef TENSORFLOW_C_EAGER_TAPE_H_
#define TENSORFLOW_C_EAGER_TAPE_H_

// Language-agnostic gradient tape. Does not perform backpropagation, just
// maintains the data structures required to do so.

#include <stack>
#include <unordered_map>
#include <unordered_set>
#include <vector>

#include "tensorflow/core/config/flag_defs.h"
#include "tensorflow/core/config/flags.h"

        TensorList operations.
    *   `tf.custom_gradient` can now be applied to functions that accept nested
        structures of `tensors` as inputs (instead of just a list of tensors).
        Note that Python structures such as tuples and lists now won't be
        treated as tensors, so if you still want them to be treated that way,
        you need to wrap them with `tf.convert_to_tensor`.

package_group(name = "types_allowlist")

# Packages that use StructuredTensors.
# TODO(b/159007891) Remove this package once StructuredTensor is exported.
# LINT.IfChange
package_group(name = "structured_tensor_allowlist")
# LINT.ThenChange(copy.bara.sky)

filegroup(
    name = "intel_binary_blob",
    data = if_mkl_ml(
        [
            "@local_xla//xla/tsl/mkl:intel_binary_blob",
        ],

// does various checks while doing so. `input_nodes` will contain the same
// information as input_tensors just in a different structure to make
// following processing easier. TODO(iga): Simplify this nested structure.
Status ProcessInputs(
    const TF_Graph* fn_body, const char* fn_name, int ninputs,
    const TF_Output* inputs, std::vector<OutputTensor>* input_tensors,