%0 = "tfl.pseudo_const" () {value = dense<1.0> : tensor<4xf32>} : () -> tensor<4xf32> loc("Const")
  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32> loc("mul")
  // tf.MyCustomOp is the result of conversion to a Custom op
  %2 = "tf.MyCustomOp"(%1, %0) {fused_activation_function = "RELU", int_attr = 2 : i32}  : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32> loc("MyCustomOp")

  %fill = "tf.Fill"(%cst_0, %cst) : (tensor<1xi64>, tensor<f32>) -> tensor<1xf32>
  tf_device.replicate([%0, %fill] as %arg_r0: tensor<1xf32>) {n = 2 : i32} {
    %1 = "tf_device.launch"() <{device = "TPU_REPLICATED_HOST_0"}> ({
      %2 = "tf.Identity"(%arg_r0) : (tensor<1xf32>) -> tensor<1xf32>
      tf_device.return %2 : tensor<1xf32>
    }) : () -> tensor<1xf32>

  %2:2 = "tf.RecvTPUEmbeddingActivations"() {_tpu_embedding_layer = "call1", config = "\0A\0B\0C\0D"} : () -> (tensor<2x2xf32>, tensor<4x4xf32>)
  func.return
}

// -----

  %0 = "tfl.pseudo_const" () {value = dense<0.0> : tensor<4xf32>} : () -> tensor<4xf32> loc("Const")
  %1 = "tfl.svdf"(%arg0, %arg1, %arg2, %arg3, %0) {fused_activation_function = "RELU", rank = 2 : i32} : (tensor<4xf32>, tensor<4xf32>, tensor<4xf32>, tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
  func.return %1 : tensor<4xf32>

  } : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<*xf32>
  %4 = "tf.BiasAdd"(%3, %cst) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<2xf32>) -> tensor<*xf32>

  func.return %2, %4 : tensor<*xf32>, tensor<*xf32>

// CHECK-DAG: %[[CONST_0:.*]] = "tf.Const"() <{value = dense<0.000000e+00> : tensor<2xf32>}> : () -> tensor<2xf32>

  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32> loc("mul")
  // tf.div is the result of conversion to a Flex TF op
  %2 = "tf.Div"(%1, %0)  : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32> loc("div")
  %3 = "tfl.exp"(%2)  : (tensor<4xf32>) -> tensor<4xf32> loc("exp")
  func.return %3 : tensor<4xf32>

// CHECK-LABEL: func @main
func.func @main(%arg0: tensor<1xf32>) -> tensor<1xf32>
attributes  {tf.entry_function = {inputs = "input", outputs = "output_node"}} {
  %0 = arith.constant dense<2.000000e+00> : tensor<f32>
  %1 = arith.constant dense<1.000000e+00> : tensor<f32>
  %2 = "tf.AddV2"(%arg0, %1) {T = "tfdtype$DT_FLOAT", device = "", name = "StatefulPartitionedCall/add"} : (tensor<1xf32>, tensor<f32>) -> tensor<1xf32>

  %2 = "tfl.add"(%arg0, %arg3) {fused_activation_function = "RELU6", per_device_costs = {CPU = 5.0 : f32, GPU = 1.0 : f32}, tac.device = "GPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %3 = "tfl.pack"(%1, %2) {axis = 0 : i32, per_device_costs = {CPU = 2.0 : f32, GPU = -1.0 : f32}, values_count = 2 : i32, tac.device = "CPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>

// CHECK: while_1([[ARG0:%[^ :]*]]: tensor<i32>, [[ARG1:%[^ :]*]]: tensor<1xf32>)
func.func @while_1(%arg0: tensor<i32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  // CHECK: [[CST:%[^ ]*]] = arith.constant dense<1> : tensor<i32>
  // CHECK: "tfl.while"([[ARG0]], [[ARG1]])
  // CHECK: (tensor<i32>, tensor<1xf32>) -> (tensor<i32>, tensor<1xf32>)
  %0:2 = "tfl.while"(%arg0, %arg1) (
    // cond
    {

    "tf.Yield"(%2) : (tensor<i1>) -> ()
  }, {
  ^body(%barg0: tensor<*xf32>):
    %2 = "tf.SomeOp"(%barg0) : (tensor<*xf32>) -> tensor<*xf32>
    "tf.Yield"(%2) : (tensor<*xf32>) -> ()
  }) {is_stateless = false, shape_invariant} : (tensor<4xf32>) -> (tensor<*xf32>)

  func.return %0, %1 : tensor<*xf32>, tensor<*xf32>
}

// Test that both passes drop the shape_invariant attribute from
// While/WhileRegion ops within a cluster.