// CHECK: return %[[ADDV2_0]] : tensor<15x28x28x1xf32>
}

// CHECK-LABEL: conv2d_backprop_input_with_sub
func.func @conv2d_backprop_input_with_sub(%arg0: tensor<4xi32>, %arg1: tensor<3x3x1x32xf32>, %arg2: tensor<15x14x14x32xf32>) -> tensor<15x28x28x1xf32> {

  %1 = "tf.Const"() {value = dense<1.000000e+00> : tensor<1xf32>} : () -> tensor<1xf32>
  %2 = "tf.Sub"(%0, %1): (tensor<15x28x28x1xf32>, tensor<1xf32>) -> tensor<15x28x28x1xf32>
  func.return %2 : tensor<15x28x28x1xf32>
// CHECK: %[[CONST_0:.*]] = "tf.Const"() <{value = dense<1.000000e+00> : tensor<1xf32>}> : () -> tensor<1xf32>

  %2 = "tf.Add"(%0, %1): (tensor<15x28x28x1xf32>, tensor<15x28x28x1xf32>) -> tensor<15x28x28x1xf32>
  func.return %2 : tensor<15x28x28x1xf32>

  // CHECK-LABEL: conv2d_backprop_input
  // CHECK: %[[CST:.*]] = arith.constant dense<[2, 0, 1, 3]> : tensor<4xi32>

// CHECK: return %[[ADD]] : tensor<1x1x64xf32>
// CHECK: }

// -----

// CHECK-LABEL: @conv_with_bias_dynamic_fn(
// CHECK-SAME:                    %[[ARG_0:.*]]: tensor<?x28x28x1xf32>
func.func @conv_with_bias_dynamic_fn(%arg0: tensor<?x28x28x1xf32>) -> tensor<?x28x28x16xf32> {
  %0 = stablehlo.constant dense<2.000000e+00> : tensor<3x3x1x16xf32>
  %1 = stablehlo.constant dense<2.000000e+00> : tensor<16xf32>

  return model


class TestModule(tf.Module):

  def __init__(self):
    super(TestModule, self).__init__()
    self.model = mnist_model()

  # CHECK: func {{@[a-zA-Z_0-9]+}}(%arg0: tensor<1x28x28x1xf32> {tf._user_specified_name = "x", tf_saved_model.index_path = [0]}
  # CHECK: attributes {{.*}} tf_saved_model.exported_names = ["my_predict"]
  @tf.function(input_signature=[
      tf.TensorSpec([1, 28, 28, 1], tf.float32),

// CHECK-LABEL: convolution_3d
func.func @convolution_3d(%arg0: tensor<1x4x28x28x1xf32>, %arg1: tensor<2x3x3x1x16xf32>) -> tensor<1x3x26x26x16xf32> {
  %0 = stablehlo.convolution(%arg0, %arg1) dim_numbers = [b, 0, 1, 2, f]x[0, 1, 2, i, o]->[b, 0, 1, 2, f], window = {} {batch_group_count = 1 : i64, feature_group_count = 1 : i64} : (tensor<1x4x28x28x1xf32>, tensor<2x3x3x1x16xf32>) -> tensor<1x3x26x26x16xf32>
  return %0 : tensor<1x3x26x26x16xf32>
}

// RUN: tf-opt %s -allow-unregistered-dialect --tf-strip-tf-attributes --split-input-file | FileCheck %s

// CHECK-LABEL: strips_attributes
// CHECK-NOT: tf
func.func @strips_attributes(%arg0: tensor<32x28x28x1xf32> {tf._user_specified_name = "x"},
                             %arg1: tensor<3x3x1x5xf32> {tf._user_specified_name = "w1"},
                             %arg2: tensor<5xf32> {tf._user_specified_name = "b1"},

  return %3 : f32
}
// CHECK-DAG: a.x
// CHECK-DAG: b.x
// CHECK-DAG: c.x
// CHECK-NEXT: d.x

// -----

// CHECK-LABEL: @tf_and_mhlo
func.func @tf_and_mhlo(%arg0: tensor<32x28x28x1xf32>, %arg1: tensor<!tf_type.resource<tensor<3x3x1x5xf32>>>, %arg2: tensor<!tf_type.resource<tensor<5xf32>>>, %arg3: tensor<!tf_type.resource<tensor<3920x10xf32>>>, %arg4: tensor<!tf_type.resource<tensor<10xf32>>>) -> (tensor<32x10xf32>) {

  %1 = "tfl.div"(%0, %cst2) {fused_activation_function = "NONE"} : (tensor<1x28x23x1xf32>, tensor<f32>) -> tensor<1x28x23x1xf32>

  func.return %1 : tensor<1x28x23x1xf32>

    explicit_paddings = [],
    padding = "VALID",
    strides = [1, 1, 1, 1],
    use_cudnn_on_gpu = true
  } : (tensor<4xi32>, tensor<3x3x1x32xf32>, tensor<100x26x26x32xf32>) -> tensor<100x28x28x1xf32>
  func.return %result : tensor<100x28x28x1xf32>
}

// -----

// CHECK-LABEL: @conv2d_backprop_input_grouped
func.func @conv2d_backprop_input_grouped(
    %filter: tensor<2x2x5x21xf32>,