%3 = "tf.Mul"(%1, %2) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10xf32>
    return %3 : tensor<10xf32>
  }
}

// -----

// CHECK-LABEL: module
module attributes {tf_saved_model.semantics} {
  // CHECK: ml_program.global{{.*}}mutable{{.*}}@vars.v
  "tf_saved_model.global_tensor"() { is_mutable, sym_name = "v", type = tensor<10xf32>, value = dense<[0.,10.,2.,3.,4.,5.,6.,7.,8.,9.]> : tensor<10xf32> } : () -> ()

// PerChannel: %[[out_2:.*]] = "tf.PartitionedCall"(%arg0, %[[q_w]], %[[scale]], %[[zp]]) <{config = "", config_proto = "", executor_type = "",
// PerChannel-SAME: f = @quantized_conv2d_fn_0}> : (tensor<1x2x2x3xf32>, tensor<2x3x3x2xi8>, tensor<2xf32>, tensor<2xi32>) -> tensor<*xf32>
// PerChannel: return %[[out_1]], %[[out_2]]

}

// -----

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

func.func private @add(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>) -> tensor<*xf32>  {
  %0 = "tf.Add"(%arg0, %arg1): (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  func.return %0 : tensor<*xf32>
}

func.func private @mul(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>) -> tensor<*xf32>  {
  %0 = "tf.Multiply"(%arg0, %arg1): (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>

    %26 = "tf.Cast"(%25) {Truncate = false} : (tensor<2xi64>) -> tensor<2xf32>
    %27 = "tf.Equal"(%14, %26) {incompatible_shape_error = true} : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xi1>
    %28 = "tf.Cast"(%27) {Truncate = false} : (tensor<2xi1>) -> tensor<2xf32>
    %29 = "tf.Sum"(%28, %6) {keep_dims = false} : (tensor<2xf32>, tensor<1xi32>) -> tensor<f32>

module{
  func.func @main(%arg0: tensor<10xf32>) -> tensor<10xf32> {
    return %arg0 : tensor<10xf32>
  }
  func.func @func1(%arg0: tensor<10xf32>, %arg1: tensor<10xf32>) -> tensor<10xf32> {
    %0 = stablehlo.add %arg0, %arg1 : tensor<10xf32>
    %1 = stablehlo.add %arg0, %arg1 : tensor<10xf32>
    return %0 : tensor<10xf32>
  }
})mlir";

  %2 = "tf.Conv2D"(%1, %0) {data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [], padding = "SAME", strides = [1, 1, 2, 1], use_cudnn_on_gpu = true} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  %3 = "tf.BiasAdd"(%2, %cst) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<2xf32>) -> tensor<*xf32>

  %1:2 = "tfl.split"(%cst, %0) {num_splits = 2 : i32} : (tensor<i32>, tensor<4xf32>) -> (tensor<2xf32>, tensor<2xf32>)
  %2 = "tfl.quantize"(%1#0) {qtype = tensor<2x!quant.uniform<u8:f32, 1.0>>} : (tensor<2xf32>) -> tensor<2x!quant.uniform<u8:f32, 1.0>>
  %3 = "tfl.quantize"(%1#1) {qtype = tensor<2x!quant.uniform<u8:f32, 1.0>>} : (tensor<2xf32>) -> tensor<2x!quant.uniform<u8:f32, 1.0>>

// CHECK: return %[[cast_i32]] : !tfr.tensor
}

// CHECK-LABEL: decompose_output_type
func.func @decompose_output_type(%arg0: tensor<2xf32>) -> tensor<2xi32> {
  %0 = "tf.CastFloat"(%arg0) : (tensor<2xf32>) -> tensor<2xi32>
  func.return %0: tensor<2xi32>
// CHECK: %[[i32:.*]] = tfr.constant i32 -> !tfr.attr
// CHECK: tfr.call @tf__cast(%[[casted_arg:.*]], %[[i32]], %false) : (!tfr.tensor, !tfr.attr, i1) -> !tfr.tensor

}

// CHECK-DAG: name: "bar"
func.func @bar() -> tensor<10xf32> {
  %0 = tf_executor.graph {
    %1:2 = tf_executor.island wraps "tf.Const"() {device = "", dtype = "tfdtype$DT_FLOAT", value = dense<2.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32> loc("const_2")
    tf_executor.fetch %1#0 : tensor<10xf32>
  }
  func.return %0 : tensor<10xf32>

                             %arg2: tensor<5xf32> {tf._user_specified_name = "b1"},
                             %arg3: tensor<3920x10xf32> {tf._user_specified_name = "w2"},
                             %arg4: tensor<10xf32> {tf._user_specified_name = "b2"}) -> tensor<10xf32>