void TF_AddInput(TF_OperationDescription* desc, TF_Output input) {
  desc->node_builder.Input(&input.oper->node, input.index);
}

void TF_AddInputList(TF_OperationDescription* desc, const TF_Output* inputs,
                     int num_inputs) {
  std::vector<NodeBuilder::NodeOut> input_list;
  input_list.reserve(num_inputs);
  for (int i = 0; i < num_inputs; ++i) {
    input_list.emplace_back(&inputs[i].oper->node, inputs[i].index);
  }

  ]


@Composite(
    'NewFullyConnected',
    inputs=['input_: T', 'filter_: T', 'bias: T'],
    attrs=['act: {"", "RELU", "RELU6", "TANH"} = ""'],
    derived_attrs=['T: {float, int8}'],
    outputs=['o: T'])
def _composite_fully_connected(input_, filter_, bias, act):
  res = tf.raw_ops.MatMul(
      a=input_, b=filter_, transpose_a=False, transpose_b=True)
  res = tf.raw_ops.Add(x=res, y=bias)

}

// CHECK-LABEL: func @tensor_array_while_test
// CHECK-SAME: ([[in_chain:%.*]]: !tfrt.chain
func.func @tensor_array_while_test(%indices: tensor<?xi32>, %input_0: tensor<?x?x?xf32>, %input_1: tensor<?x?x?xf32>) -> (tensor<?x?x512xf32>, tensor<?x?x512xf32>) {
  %index = "tf.Const"() {device = "/device:CPU:0", value = dense<0> : tensor<i32>} : () -> (tensor<i32>)

  def test_mirror_pad(self, mode):
    input_ = tf.constant([[1, 2, 3], [4, 5, 6]], dtype=tf.float32)
    paddings = tf.constant([[
        1,
        1,
    ], [2, 2]])
    kwargs = {
        'input': input_,
        'paddings': paddings,
        'mode': mode,
    }
    kwargs_ = {
        'input_': input_,
        'paddings': paddings,
        'mode': mode,
    }

          num_split=2)

    input_ = tf.raw_ops.Concat(
        concat_dim=i, values=[left_padding, input_, right_padding])
  return input_


@tf.RegisterGradient('NewMirrorPad')
def _mirror_pad_grad(op, grad):
  mode = op.get_attr('mode')
  return [gen_array_ops.mirror_pad_grad(grad, op.inputs[1], mode=mode), None]


@Composite(
    'NewMirrorPadGrad',
    inputs=['input_: T', 'paddings: Tpaddings'],

      : inputs_(inputs.begin(), inputs.end()) {}

  InputList(const absl::Span<const Input>& inputs)  // NOLINT(runtime/explicit)
      : inputs_(inputs.begin(), inputs.end()) {}

  InputList(
      const std::initializer_list<Output>& out) {  // NOLINT(runtime/explicit)
    for (auto const& x : out) {
      inputs_.push_back(x);
    }
  }

    for (int i = 0; i < num_inputs; ++i) {
      inputs_.push_back(ops::Placeholder(scope_, dtype));
    }
  }

  void CreateLoop(const ops::CondGraphBuilderFn& cond,
                  const ops::BodyGraphBuilderFn& body,
                  const std::vector<Output>* inputs = nullptr) {
    if (inputs == nullptr) inputs = &inputs_;
    TF_ASSERT_OK(ops::BuildWhileLoop(scope_, *inputs, cond, body, "test_loop",

class MatMulGradientFunction : public GradientFunction {
 public:
  explicit MatMulGradientFunction(vector<AbstractTensorHandle*> f_inputs,
                                  AttrBuilder f_attrs)
      : forward_inputs_(f_inputs), forward_attrs_(f_attrs) {
    for (auto input : forward_inputs_) {
      if (input) {
        input->Ref();
      }
    }
  }

  Status Compute(AbstractContext* ctx,

    }
  }
}
node {
  name: "Case/branch0/_0/mul_0"
  op: "Mul"
  input: "Case/Case/input_0/_7"
  input: "Case/branch0/_0/mul/y"
  attr {
    key: "T"
    value {
      type: DT_FLOAT
    }
  }
}
node {
  name: "Case/branch1/_1/mul_0"
  op: "Mul"
  input: "Case/Case/input_0/_7:1"
  input: "Case/branch1/_1/mul/y"
  attr {
    key: "T"
    value {
      type: DT_FLOAT

                    (MHLO_CompositeOp:$old_val
                    (variadic $a_input), 
                    ConstantStrAttr<StrAttr, "aten.avg_pool2d.default">, $attrs, $_, $_),
                    (TFL_TransposeOp 
                        (TFL_AveragePool2DOp 
                            /*input*/ (TFL_TransposeOp $a_input,
                                (Arith_ConstantOp