// Use other output
  %3:6 = "tf.FusedBatchNormV3"( %2#0, %arg1, %arg2, %arg3, %arg4) {T = "tfdtype$DT_FLOAT", U = "tfdtype$DT_FLOAT", data_format = "NHWC", epsilon = 0.001 : f32, is_training = false} : (tensor<8x8x8x8xf32>, tensor<8xf32>, tensor<8xf32>, tensor<8xf32>, tensor<8xf32>) -> (tensor<8x8x8x8xf32>, tensor<8xf32>, tensor<8xf32>, tensor<8xf32>, tensor<8xf32>, tensor<8xf32>)

  // return %[[SUM2]]
  %0 = "tf.AddN"(%arg0, %arg1, %arg2, %arg3) : (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  func.return %0 : tensor<*xf32>
}

// CHECK-LABEL: func @addN_5
func.func @addN_5(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>, %arg2: tensor<*xf32>, %arg3: tensor<*xf32>, %arg4: tensor<*xf32>) -> tensor<*xf32> {
  // CHECK: %[[SUM0:.*]] = "tf.AddV2"(%arg0, %arg1)

  %outputs_50 =  "tf.Mul"(%outputs_30, %outputs_48) {device = ""} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  %outputs_52 =  "tf.Reshape"(%outputs_50, %outputs_0) {device = ""} : (tensor<*xf32>, tensor<2xi32>) -> tensor<*xf32>
  %outputs_54 =  "tf.MatMul"(%outputs_40, %outputs_52) {device = "", transpose_a = false, transpose_b = false} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>

        tensor<2xf32>, tensor<2xf32>, tensor<2xf32>,
        tensor<2xf32>, tensor<2xf32>, tensor<2xf32>, tensor<2xf32>,
        tensor<4x2xf32>, tensor<4xf32>,
        tensor<1x4xf32>, tensor<1x2xf32>,
        tensor<2xf32>, tensor<2xf32>, tensor<2xf32>, tensor<2xf32>) -> tensor<*xf32>
  %24 = "quantfork.stats"(%23) {layerStats = dense<[-1.0, 2.0]> : tensor<2xf32>} : (tensor<*xf32>) -> tensor<*xf32>

    %cst = "tf.Const"() {value = dense<3.00000000e-1> : tensor<2x3xf32>} : () -> tensor<2x3xf32>
    %cst_0 = "tf.Const"() {value = dense<4.00000000e-1> : tensor<3xf32>} : () -> tensor<3xf32>
    %0 = "quantfork.stats"(%arg0) {layerStats = dense<[6.00000000e-6, 9.00000000e-1]> : tensor<2xf32>} : (tensor<?x2xf32>) -> tensor<?x2xf32>

  // CHECK: %[[VAL:.*]] = "tf.Const"() <{value = dense<[1.000000e+00, 2.000000e+00, 3.000000e+00]> : tensor<3xf32>}> : () -> tensor<3xf32>
  %value = "tf.Const"() {value = dense<[1.0, 2.0, 3.0]> : tensor<3xf32>} : () -> tensor<3xf32>
  // CHECK: %[[READ_VAR:.*]] = "tf.ReadVariableOp"(%[[VAR]])
  // CHECK: %[[UPDATE_SLICE:.*]] = "tf.Reshape"(%[[VAL]]
  // CHECK-SAME: -> tensor<1x3xf32>

    // CHECK-DAG: %[[SIGN:.*]] = "tf.Sign"(%[[PROX]]) : (tensor<4xf32>) -> tensor<4xf32>
    // CHECK-DAG: %[[ABS:.*]] = "tf.Abs"(%[[PROX]]) : (tensor<4xf32>) -> tensor<4xf32>
    // CHECK-DAG: %[[SCALED_L1:.*]] = "tf.Mul"(%[[ADAGRAD_LR]], %[[L1]]) : (tensor<4xf32>, tensor<f32>) -> tensor<4xf32>
    // CHECK-DAG: %[[PROX_NEW:.*]] = "tf.Sub"(%[[ABS]], %[[SCALED_L1]]) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>

    %0 = "tf.Div"(%arg1, %cst_0) : (tensor<?x?xf32>, tensor<f32>) -> tensor<?x?xf32>
    %1 = "tf.AddV2"(%0, %cst_1) : (tensor<?x?xf32>, tensor<f32>) -> tensor<?x?xf32>
    %2 = "tf.Floor"(%1) : (tensor<?x?xf32>) -> tensor<?x?xf32>
    %3 = "tf.ClipByValue"(%2, %cst_5, %cst_6) : (tensor<?x?xf32>, tensor<f32>, tensor<f32>) -> tensor<?x?xf32>
    %4 = "tf.Cast"(%3) {Truncate = false} : (tensor<?x?xf32>) -> tensor<?x?xi8>

}
// CHECK: func @if_else(%[[EARG0:.*]]: tensor<4xf32>, %[[EARG1:.*]]: tensor<4xf32>)
func.func @if_else(%arg0: tensor<*x!tf_type.resource<tensor<4xf32>>>, %arg1: tensor<*x!tf_type.resource<tensor<4xf32>>>)
    -> (tensor<*x!tf_type.resource<tensor<4xf32>>>, tensor<4xf32>) {
  %id = "tf.Identity"(%arg1) : (tensor<*x!tf_type.resource<tensor<4xf32>>>) -> tensor<*x!tf_type.resource<tensor<4xf32>>>

    %4 = "quantfork.stats"(%3) {layerStats = dense<[0.000000e+00, 0.18486841]> : tensor<2xf32>} : (tensor<*xf32>) -> tensor<*xf32>
    %5 = "tf.Identity"(%4) {device = ""} : (tensor<*xf32>) -> tensor<*xf32>
    %6 = "quantfork.stats"(%5) {layerStats = dense<[0.000000e+00, 0.18486841]> : tensor<2xf32>} : (tensor<*xf32>) -> tensor<*xf32>