AbstractTensorHandle* loss;
  AbstractTensorHandle* backprop;
  TF_RETURN_IF_ERROR(ops::SparseSoftmaxCrossEntropyWithLogits(
      ctx, inputs[0], inputs[1], &loss, &backprop,
      "SparseSoftmaxCrossEntropyWithLogits"));
  // `gradient_checker` only works with model that returns only 1 tensor.
  // Although, `ops::SparseSoftmaxCrossEntropyWithLogits` returns 2 tensors, the
  // second tensor isn't needed for computing gradient so we could safely drop

#include "tensorflow/core/framework/types.h"
#include "tensorflow/core/platform/errors.h"

using tensorflow::tracing::MaybeSetOpName;

namespace tensorflow {
namespace ops {

// Op: SparseSoftmaxCrossEntropyWithLogits()
// Summary: Computes softmax cross entropy cost and gradients to backpropagate.
//
// Description:
//   Unlike `SoftmaxCrossEntropyWithLogits`, this operation does not accept

  TF_RETURN_IF_ERROR(op_ptr->AddInputList(input));
  int num_retvals = output.size();
  return op_ptr->Execute(output, &num_retvals);
}

// Op: SparseSoftmaxCrossEntropyWithLogits()
// Summary:
//
// Description:

  Mul \
  Conj \
  AddV2 \
  MatMul \
  Neg \
  Sum \
  Sub \
  Div \
  DivNoNan \
  Exp \
  Sqrt \
  SqrtGrad \
  Log1p

${generate} \
  --category=nn \
  SparseSoftmaxCrossEntropyWithLogits \
  ReluGrad \
  Relu \
  BiasAdd \
  BiasAddGrad

${generate} \
  --category=resource_variable \
  VarHandleOp \
  ReadVariableOp \
  AssignVariableOp \
  DestroyResourceOp

//
Status IdentityN(AbstractContext* ctx, absl::Span<AbstractTensorHandle* const> input, absl::Span<AbstractTensorHandle*> output, const char* name = nullptr, const char* raw_device_name = nullptr);

//
Status SparseSoftmaxCrossEntropyWithLogits(AbstractContext* ctx, AbstractTensorHandle* const features, AbstractTensorHandle* const labels, AbstractTensorHandle** loss, AbstractTensorHandle** backprop, const char* name = nullptr, const char* raw_device_name = nullptr);

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

namespace tensorflow {
namespace ops {

// Computes softmax cross entropy cost and gradients to backpropagate.
Status SparseSoftmaxCrossEntropyWithLogits(
    AbstractContext* ctx, AbstractTensorHandle* const features,
    AbstractTensorHandle* const labels, AbstractTensorHandle** loss,
    AbstractTensorHandle** backprop, const char* name = nullptr,

 MatMul 3
 MaxPool 1
 MaxPoolGrad 1
 Mean 1
 Mul 164
 Pad 1
 ReadVariableOp 646
 Relu 49
 ReluGrad 49
 Reshape 2
 ResourceApplyKerasMomentum 161
 ShapeN 50
 Softmax 1
 SparseSoftmaxCrossEntropyWithLogits 1
 Square 55
 Squeeze 1
 Sub 106
 Sum 57
 Tile 1

  string api_dirs = "";
  std::vector<string> ops = {
      "Neg",        // Simple unary Op
      "MatMul",     // 2 inputs & attrs with default values
      "IdentityN",  // Variadic input+output
      "SparseSoftmaxCrossEntropyWithLogits",  // 2 outputs
      "AccumulatorApplyGradient",             // 0 outputs
      "VarHandleOp",                          // type, shape, list(string) attrs
      "RestoreV2",  // Variadic output-only, list(type) attr

 MatMul 3
 MaxPool 1
 MaxPoolGrad 1
 Mean 1
 Mul 218
 Pad 2
 ReadVariableOp 538
 Relu 49
 ReluGrad 49
 Reshape 2
 ResourceApplyKerasMomentum 161
 Slice 1
 Softmax 1
 SparseSoftmaxCrossEntropyWithLogits 1
 Squeeze 1
 Sum 1
 Tile 1
 Transpose 1
cluster 1 size 815
 AddN 1
 AssignAddVariableOp 1
 AssignSubVariableOp 106
 Const 220
 DivNoNan 1
 Identity 1
 Mul 161
 ReadVariableOp 106

  func.return %0#0, %0#1 : tensor<?xf32>, tensor<?x?xf32>
}

// CHECK-LABEL: SparseSoftmaxCrossEntropyWithLogits
// CHECK-SAME: %[[FEATURES:.*]]: tensor<2x3xf32>, %[[SPARSE_LABELS:.*]]: tensor<2xi32>
func.func @SparseSoftmaxCrossEntropyWithLogits(%features: tensor<2x3xf32>, %labels: tensor<2xi32>) -> (tensor<2xf32>, tensor<2x3xf32>) {
  // Convert SPARSE_LABELS to dense LABELS.