input: "m"
  input: "v"
  input: "beta"
  input: "gamma"
  device: "/device:CPU:0"
  attr {
    key: "T"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "scale_after_normalization"
    value {
      b: false
    }
  }
  attr {
    key: "variance_epsilon"
    value {
      f: 0.001
    }
  }
}
library {
}

  (TF_AddOp
    (TF_MulOp $t, (TF_RsqrtOp:$rsqrt (TF_AddOp $v, (TF_ConstOp $variance_epsilon)))),
    (TF_SubOp $beta, (TF_MulOp $m, $rsqrt)))>;
def ConvertBatchNormWithGlobalNormalization_2 : Pat<
  (TF_BatchNormWithGlobalNormalizationOp
    $t, $m, $v, $beta, $gamma, F32Attr:$variance_epsilon, ConstBoolAttrTrue:$scale_after_normalization),
  (TF_AddOp

  %0 = "tf.BatchNormWithGlobalNormalization"(%t, %m, %v, %beta, %gamma) {T = "tfdtype$DT_FLOAT", variance_epsilon = 0.001 : f32, scale_after_normalization = false} : (tensor<1x10x10x3xf32>, tensor<3xf32>, tensor<3xf32>, tensor<3xf32>, tensor<3xf32>) -> (tensor<1x10x10x3xf32>)
  func.return %0 : tensor<1x10x10x3xf32>
// CHECK-LABEL: batchNormWithGlobalNormalization

If "scale_after_normalization" is true, this tensor will be multiplied
with the normalized tensor.}]>:$gamma,

    F32Attr:$variance_epsilon,
    BoolAttr:$scale_after_normalization
  );

  let results = (outs