// https://www.tensorflow.org/api_docs/cc/class/tensorflow/ops/batch-mat-mul
  int64_t x_row_dim = x_shape[x_shape.size() - 2];
  int64_t x_col_dim = x_shape[x_shape.size() - 1];
  int64_t y_row_dim = y_shape[y_shape.size() - 2];
  int64_t y_col_dim = y_shape[y_shape.size() - 1];
  int64_t out_row_dim = output_shape[output_shape.size() - 2];
  int64_t out_col_dim = output_shape[output_shape.size() - 1];

          contracting_dims.add(c)
      x_signature = [
          None if c not in contracting_dims else x_shape[cidx]
          for cidx, c in enumerate(x_labels)
      ]
      y_signature = [
          None if c not in contracting_dims else y_shape[cidx]
          for cidx, c in enumerate(y_labels)
      ]
    return x_shape, y_shape, bias_shape, x_signature, y_signature

  def _create_einsum_model(
      self,

      x_signature = [
          None if c not in contracting_dims else x_shape[cidx]
          for cidx, c in enumerate(x_labels)
      ]
      y_signature = [
          None if c not in contracting_dims else y_shape[cidx]
          for cidx, c in enumerate(y_labels)
      ]
    return x_shape, y_shape, bias_shape, x_signature, y_signature

  def _create_einsum_model(
      self,

    x = Conj(scope, x);
    y = Conj(scope, y);
  }

  const auto x_shape = Shape(scope, x);
  const auto y_shape = Shape(scope, y);
  Output grad_x =
      EinsumGradWrt(scope, grad, y, x_shape, x_subs, y_subs, output_subs);
  Output grad_y =
      EinsumGradWrt(scope, grad, x, y_shape, y_subs, x_subs, output_subs);

  if (!absl::StrContains(output_subs, kEllipsis)) {

    Value result = reduction.getResult(0);

    // The mean op needs to divide by the product of the reduced dimensions.
    if (std::is_same<OpTy, TF::MeanOp>::value) {
      Value in_shape = rewriter.create<shape::ShapeOfOp>(loc, op.getInput());
      Value divisor_count = rewriter.create<arith::ConstantIndexOp>(loc, 1);
      for (size_t i = 0; i < input_shape.size(); ++i) {

//
//   Slice(op, begin, size <must be constant>) =>
//     Slice(op, begin, actual_size(op.shape(), size, begin));
//       _XlaCompileTimeConstantInputs={2}
//
// where
//
//   actual_size(op_shape, size, begin)[i] =
//     size[i] == -1 ? (op_shape[i] - size[i])
//                   : size[i]
//
// This pass, combined with jit/partially_decluster_pass, reduces the number of

#define TENSORFLOW_C_TF_SHAPE_INTERNAL_H_

#include "tensorflow/c/conversion_macros.h"
#include "tensorflow/core/framework/tensor_shape.h"

typedef struct TF_Shape TF_Shape;

namespace tensorflow {

DEFINE_CONVERSION_FUNCTIONS(tensorflow::PartialTensorShape, TF_Shape);

}

  // CHECK-DAG: %[[RHS:.+]] = mhlo.subtract %[[OFFSET]], %[[MUL_MEAN]] : tensor<?xf32>
  // CHECK-DAG: %[[X_SHAPE:.+]] = shape.shape_of %[[X]] : tensor<?x?x?x?xf32> -> tensor<4xindex>
  // CHECK-DAG: %[[MULTIPLIER_BCAST:.+]] = "mhlo.dynamic_broadcast_in_dim"(%[[MULTIPLIER]], %[[X_SHAPE]]) <{broadcast_dimensions = dense<1> : tensor<1xi64>}> : (tensor<?xf32>, tensor<4xindex>) -> tensor<?x?x?x?xf32>

  // The internal gradient implementation needs the shape of the input image.
  // x_shape = shape(x)[1:3]
  //         = slice(shape(x), {1}, {3 - 1})
  auto x_shape = Slice(scope, Shape(scope, op.input(0)), {1}, {2});
  grad_outputs->push_back(internal::ResizeNearestNeighborGrad(
      scope, grad_inputs[0], x_shape,
      internal::ResizeNearestNeighborGrad::AlignCorners(align_corners)

#define TENSORFLOW_C_EXPERIMENTAL_SAVED_MODEL_PUBLIC_TENSOR_SPEC_H_

#include <stddef.h>

#include "tensorflow/c/c_api_macros.h"
#include "tensorflow/c/tf_datatype.h"
#include "tensorflow/c/tf_shape.h"

#ifdef __cplusplus
extern "C" {
#endif  // __cplusplus

// An opaque type corresponding to TensorSpec
typedef struct TF_TensorSpec TF_TensorSpec;

// Returns the dtype associated with the TensorSpec.