}

// CHECK-LABEL: bias_add_nchw
func.func @bias_add_nchw(%arg0: tensor<1x256x150x150xf32>, %arg1: tensor<256xf32>) -> tensor<1x256x150x150xf32> {
  // CHECK: (%[[ARG0:.*]]: tensor<1x256x150x150xf32>, %[[ARG1:.*]]: tensor<256xf32>)
  // CHECK: %[[CST:.*]] = "tf.Const"() <{value = dense<[0, 2, 3, 1]> : tensor<4xi64>}>
  // CHECK: %[[R0:.*]] = "tf.Transpose"(%[[ARG0]], %[[CST]])

    ```mlir
      func @test(%arg0: tensor<*xi32>) {
        "tf_device.cluster_func"(%arg0) {
            func = @func,
            step_marker_location = ""} : (tensor<*xi32>) -> tensor<*xi32>
        return
      }

      func @func(%arg0: tensor<*xi32>) -> tensor<*xi32> {
        %0 = "tf.XlaSharding"(%arg0) {_XlaSharding = "\01\02\03",

module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {

// CHECK-LABEL: conv_2d_bf16
func.func @conv_2d_bf16(%arg0 : tensor<256x32x32x3xbf16>, %arg1 : tensor<3x3x3x16xbf16>) -> tensor<256x8x7x16xbf16> {

func.func @reshape_float(%arg0: tensor<2x3x4xf32>) -> tensor<6x4xf32> {
  %0 = stablehlo.reshape %arg0 : (tensor<2x3x4xf32>) -> tensor<6x4xf32>
  return %0 : tensor<6x4xf32>
}
// CHECK-LABEL: reshape_float
// CHECK-NOT: tfl.reshape
// CHECK: stablehlo.reshape

// -----

// Tests that a quantized `stablehlo.select` is converted to `tfl.select_v2`.

func.func @select(
    %arg0: tensor<1x3xi1>,

  func.func private @f(%arg0: tensor<f32> {tf_saved_model.not_a_real_arg_attr = 1 : i32}) {
    func.return
  }

}

// -----

module attributes {tf_saved_model.semantics} {

  // expected-error@+1 {{'tf_saved_model.bound_input' attribute should be a FlatSymbolRefAttr}}
  func.func @f(
    %arg0: tensor<f32> {tf_saved_model.bound_input = 1 : i32}

// CHECK-NEXT: return [[B]]
func.func @transitive_control_dependencies(%arg0: tensor<i32>) -> tensor<i32> {
  %graph_result = tf_executor.graph {
    %output0, %control0 = tf_executor.island {
      %a = "tf.opA"(%arg0) : (tensor<i32>) -> tensor<i32>
      tf_executor.yield %a : tensor<i32>
    }
    %output1, %control1 = tf_executor.island(%control0) {
      %b = "tf.opB"(%arg0) : (tensor<i32>) -> tensor<i32>
      tf_executor.yield %b : tensor<i32>

// RUN: odml-to-stablehlo-opt %s -tfl-parse-stablehlo-ops | FileCheck -dump-input always %s

module {
  func.func @main(%arg0: tensor<8x8x1x207xf32>, %arg1: tensor<3x3x16x207xf32>) -> tensor<16x8x8x1xf32> {

func.func @main(tensor<40x37xf32>, tensor<40x37xf32>) -> tensor<40x40xf32> {
^bb0(%arg0: tensor<40x37xf32>, %arg1: tensor<40x37xf32>):
  %cst = arith.constant dense<1.0> : tensor<40xf32>

func.func private @sum_reducer(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>) -> tensor<*xf32> {
  %0 = "tf.AddV2"(%arg0, %arg1) {device = ""} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  func.return %0 : tensor<*xf32>
}

func.func @xla_func_0(%arg0: tensor<7xf32>, %arg1: tensor<f32>) -> tensor<10xf32> {
  %cst = "tf.Const"() {value = dense<0> : tensor<1x2xi32>} : () -> tensor<1x2xi32>

// CHECK-LABEL: @test_conv2d_float
func.func @test_conv2d_float(%arg0: tensor<1x32x32x8xf32>) -> tensor<1x32x32x16xf32> {
  // CHECK-DAG: %[[VAL0:.+]] = "tfl.pseudo_const"() <{value = dense<42> : tensor<16x1x1x8xi8>}>
  // CHECK-DAG: %[[VAL1:.+]] = "tfl.pseudo_const"() <{value = dense<1> : tensor<16x1x1x8xi8>}>