^bb0(%arg0: tensor<1x7x7x16xf32>):
  // expected-error @+1 {{requires attribute 'padding'}}
  %0 = "tf.AvgPool"(%arg0) {T = "tfdtype$DT_FLOAT", ksize = [1, 7, 7, 1], strides = [1, 1, 1, 1]} : (tensor<1x7x7x16xf32>) -> tensor<1x1x1x16xf32>
  func.return %0 : tensor<1x1x1x16xf32>
}

// -----

func.func @testAvgPoolWrongPadding(tensor<1x7x7x16xf32>) -> tensor<1x1x1x16xf32> {

func.func @testSelectV2With4DBroadcasting(%cond : tensor<1x1x3x1xi1>, %arg0 : tensor<1x1x1x4xf32>, %arg1 : tensor<1x2x1x1xf32>) -> tensor<1x2x3x4xf32> {
  // CHECK: "tfl.select_v2"(%arg0, %arg1, %arg2)
  %0 = "tfl.select_v2"(%cond, %arg0, %arg1): (tensor<1x1x3x1xi1>, tensor<1x1x1x4xf32>, tensor<1x2x1x1xf32>) -> tensor<1x2x3x4xf32>
  func.return %0 : tensor<1x2x3x4xf32>
}

// -----

  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>

    } : (tensor<1x6x6x16xf32>) -> tensor<1x1x1x16xf32>
  func.return %1 : tensor<1x1x1x16xf32>

// CHECK: %0 = "tfl.dequantize"(%arg0)
// CHECK: %1 = "tfl.average_pool_2d"(%0)
// CHECK: %2 = "tfl.quantize"(%1)
// CHECK: %3 = "tfl.dequantize"(%2)
// CHECK: return %3 : tensor<1x1x1x16xf32>
}

// CHECK-LABEL: QuantizeMaximum

    explicit_paddings = [],
    padding = "VALID",
    strides = [1, 1, 1, 1],
    use_cudnn_on_gpu = true
  } : (tensor<1x2x2x2xf32>, tensor<4xi32>, tensor<1x1x1x2xf32>) -> tensor<2x2x1x2xf32>
  func.return %result : tensor<2x2x1x2xf32>
}


// CHECK-LABEL: @conv3d_backprop_filter
func.func @conv3d_backprop_filter(%input: tensor<2x8x8x8x1xf32>, %out_backprop: tensor<2x8x8x8x6xf32>) -> tensor<3x3x3x1x6xf32> {