func.func @testBiasAdd(%arg0: tensor<2x3x5x7xf32>, %arg1: tensor<5x7xf32>) -> tensor<2x3x5x7xf32> {
  // expected-error @+1 {{requires bias operand to have rank exactly one}}
  %0 = "tf.BiasAdd"(%arg0, %arg1) {data_format = "NHWC"} : (tensor<2x3x5x7xf32>, tensor<5x7xf32>) -> tensor<2x3x5x7xf32>
  func.return %0 : tensor<2x3x5x7xf32>
}

// -----

func.func @testBiasAdd(%arg0: tensor<2x3x5x7xf32>, %arg1: tensor<5xf32>) -> tensor<2x3x5x7xf32> {

}

// -----

// CHECK-LABEL: depthwiseconv_simple
func.func @depthwiseconv_simple(%arg0: tensor<?x4x5x3xf32>, %arg1: tensor<2x2x3x3xf32>) -> tensor<?x3x4x9xf32> {
  // CHECK: %[[RESHAPED_FILTER:.*]] = mhlo.reshape %arg1 : (tensor<2x2x3x3xf32>) -> tensor<2x2x1x9xf32>
  // CHECK: mhlo.convolution(%arg0, %[[RESHAPED_FILTER]])
  // CHECK-SAME: feature_group_count = 3

  %arg1: tensor<f32>
) -> tensor<2x4x3x3xf32> {
  %0 = "stablehlo.reduce_window"(%arg0, %arg1) ({
  ^bb0(%arg2: tensor<f32>, %arg3: tensor<f32>):
    %1 = stablehlo.maximum %arg2, %arg3 : tensor<f32>
    stablehlo.return %1 : tensor<f32>
  }) {window_dimensions = array<i64: 1, 3, 4, 1>, window_strides = array<i64: 1, 2, 3, 1>} : (tensor<2x9x10x3xf32>, tensor<f32>) -> tensor<2x4x3x3xf32>
  return %0 : tensor<2x4x3x3xf32>
}

// CHECK:           }) : (tensor<5x4x3x7xf32>, tensor<2x2xi32>, tensor<2x5x3xf32>) -> tensor<5x4x3x7xf32>
// CHECK:           return %[[VAL_3]] : tensor<5x4x3x7xf32>
// CHECK:         }
func.func @convert_scatter_update_to_non_trailing_operand_dimensions(
  %arg0: tensor<5x4x3x7xf32>,
  %arg1: tensor<2x2xi32>,
  %arg2: tensor<2x5x3xf32>) -> tensor<5x4x3x7xf32>
{
  %0 = "mhlo.scatter"(%arg0, %arg1, %arg2) ({

  // CHECK: %0 = "tfl.conv_2d"(%arg0, %arg1, %cst)
}

// CHECK-LABEL: fuse4DAddIntoConv2d
func.func @fuse4DAddIntoConv2d(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<2x3x3x3xf32>) -> tensor<256x32x32x2xf32> {
  %cst = arith.constant dense<[[[[1.0, 2.0]]]]> : tensor<1x1x1x2xf32>
  %cst_0 = arith.constant dense<[1.0, 2.0]> : tensor<2xf32>
  %0 = "tfl.conv_2d"(%arg0, %arg1, %cst_0) {