func.func @test5DAddWithImplicitBroadcast(%arg0: tensor<1x1x1x3x1xi32>, %arg1 : tensor<1x1x1x1x4xi32>) -> tensor<1x1x1x3x4xi32> {
  %0 = "tf.Add"(%arg0, %arg1): (tensor<1x1x1x3x1xi32>, tensor<1x1x1x1x4xi32>) -> tensor<1x1x1x3x4xi32>
  func.return %0 : tensor<1x1x1x3x4xi32>
// CHECK-LABEL: test5DAddWithImplicitBroadcast
// CHECK: %0 = tfl.add(%arg0, %arg1) <{fused_activation_function = "NONE"}> : (tensor<1x1x1x3x1xi32>, tensor<1x1x1x1x4xi32>) -> tensor<1x1x1x3x4xi32>

  %0 = mhlo.constant dense<[[[[0, 1, 2, 3]]]]> : tensor<1x1x1x4xi32>
  %1 = mhlo.constant dense<[[[[0, 1, 2, 3]], [[0, 1, 2, 3]]]]> : tensor<1x2x1x4xi32>
  %2 = "mhlo.broadcast_in_dim"(%0) <{broadcast_dimensions = dense<[0, 1, 2, 3]> : tensor<4xi64>}> : (tensor<1x1x1x4xi32>) -> tensor<1x2x1x4xi32>
  %3 = mhlo.multiply %1, %2 : tensor<1x2x1x4xi32>

func.func @add_with_i32_five_dim_broadcasting(tensor<1x1x1x1x1xi32>, tensor<1xi32>) -> tensor<1x1x1x1x1xi32> {
^bb0(%arg0: tensor<1x1x1x1x1xi32>, %arg1: tensor<1xi32>):
  // CHECK: tfl.add(%arg0, %arg1) <{fused_activation_function = "RELU6"}>
  %0 = "tfl.add"(%arg0, %arg1) {fused_activation_function = "RELU6"} : (tensor<1x1x1x1x1xi32>, tensor<1xi32>) -> tensor<1x1x1x1x1xi32>
  func.return %0#0 : tensor<1x1x1x1x1xi32>
}

// -----

  %cst = arith.constant dense<2.0> : tensor<1x1x1x1x2xf32>
  %shape = arith.constant dense<[1, 1, 1, 1, 2]> : tensor<5xi32>
  %1 = "tfl.reshape"(%arg0, %shape) : (tensor<2x1x1x1x1xf32>, tensor<5xi32>) -> tensor<1x1x1x1x2xf32>
  %2 = "tfl.add"(%1, %cst) {fused_activation_function = "NONE"} : (tensor<1x1x1x1x2xf32>, tensor<1x1x1x1x2xf32>) -> tensor<1x1x1x1x2xf32>

// CHECK-DAG: %[[CONST_1:.*]] = "tf.Const"() <{value = dense<-43> : tensor<i8>}> : () -> tensor<i8>
// CHECK-DAG: %[[CONST_2:.*]] = "tf.Const"() <{value = dense<-2322> : tensor<1x1x1x1x2xi32>}> : () -> tensor<1x1x1x1x2xi32>

// CHECK: %[[PAD:.*]] = "tf.PadV2"({{.*}}, %[[CONST]], %[[CONST_1]])
// CHECK: %[[CONV:.*]] = "tf.XlaConvV2"(%[[PAD]], %[[WEIGHT]]

func.func @conv_with_bias_and_relu_srq(%arg0: tensor<1x5x5x2x!quant.uniform<i8:f32, 2.000000e+00:0>>) -> (tensor<1x4x4x4x!quant.uniform<i8:f32, 8.000000e+00:-128>>) {
    %0 = stablehlo.constant() {value = dense<5> : tensor<1x1x1x4xi32>} : () -> tensor<1x1x1x4x!quant.uniform<i32:f32:3, {2.000000e+00, 2.000000e+00, 2.000000e+00, 2.000000e+00}>>

func.func @testAvgPoolWrongDataType(tensor<1x7x7x16xi32>) -> tensor<1x1x1x16xi32> {
^bb0(%arg0: tensor<1x7x7x16xi32>):
  // expected-error @+1 {{must be tensor of floating-point values}}
  %0 = "tf.AvgPool"(%arg0) {T = "tfdtype$DT_INT", data_format = "NHWC", ksize = [1, 7, 7, 1], padding = "VALID", strides = [1, 1, 1, 1]} : (tensor<1x7x7x16xi32>) -> tensor<1x1x1x16xi32>
  func.return %0 : tensor<1x1x1x16xi32>
}

// -----

  func.func @simple_folding(%arg0: tensor<1x1x1x1xi32>, %arg1: tensor<1x1x1x1xf32>) -> tensor<?x?x?x?xf32> {
    // CHECK: %[[SHAPE:.*]] = "tf.Shape"
    // CHECK: %[[CONV:.*]] = "tf.Conv2DBackpropInput"(%[[SHAPE]]
    // CHECK-SAME: (tensor<4xi32>, tensor<1x1x1x1xf32>, tensor<1x1x1x1xf32>) -> tensor<1x1x1x1xf32>
    // CHECK: return %[[CONV]] : tensor<1x1x1x1xf32>
    %0 = "tf.Shape"(%arg0) : (tensor<1x1x1x1xi32>) -> tensor<4xi32>