func.return %1 : tensor<1x1x1x1x1xf32>

  // CHECK-DAG: %[[CST1:.*]] = "tfl.no_value"() <{value}> : () -> none
  // CHECK-DAG: %[[CST2:.*]] = arith.constant dense<2.000000e+00> : tensor<1x1x1x1x1xf32>

  %w = arith.constant dense<[[[[0.0]]], [[[127.0]]], [[[-127.0]]]]> : tensor<3x1x1x1xf32>
  %b = arith.constant dense<0.0> : tensor<3xf32>
  %conv = "tfl.conv_2d"(%arg0, %w, %b) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x5x5x1xf32>, tensor<3x1x1x1xf32>, tensor<3xf32>) -> tensor<1x5x5x3xf32>
  func.return %conv : tensor<1x5x5x3xf32>

// CHECK: %1 = "tfl.pad"(%0, %cst_0) : (tensor<1x1x8x1xf32>, tensor<4x2xi32>) -> tensor<1x1x11x1xf32>
// CHECK: %2 = "tfl.average_pool_2d"(%1) <{filter_height = 1 : i32, filter_width = 3 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 1 : i32, stride_w = 2 : i32}> : (tensor<1x1x11x1xf32>) -> tensor<1x1x5x1xf32>

    } : (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

}

// -----

func.func @sub_with_quantized_i8_five_dim_broadcasting(tensor<1x1x1x1x1xf32>, tensor<1xf32>) -> tensor<1x1x1x1x1x!quant.any<i8:f32>> {
^bb0(%arg0: tensor<1x1x1x1x1xf32>, %arg1: tensor<1xf32>):
  %0 = "tfl.sub"(%arg0, %arg1) {fused_activation_function = "RELU6"} : (tensor<1x1x1x1x1xf32>, tensor<1xf32>) -> tensor<1x1x1x1x1x!quant.any<i8:f32>>
  func.return %0#0 : tensor<1x1x1x1x1x!quant.any<i8:f32>>
}

// CHECK:           %[[VAL_2:.*]] = "tf.Reshape"(%[[VAL_0]], %[[VAL_1]]) : (tensor<3x1x16xf32>, tensor<4xi64>) -> tensor<3x1x1x16xf32>
// CHECK:           %[[VAL_3:.*]] = arith.constant dense<[3, 8, 8, 16]> : tensor<4xi64>
// CHECK:           %[[VAL_4:.*]] = "tf.BroadcastTo"(%[[VAL_2]], %[[VAL_3]]) : (tensor<3x1x1x16xf32>, tensor<4xi64>) -> tensor<3x8x8x16xf32>
// CHECK:           return %[[VAL_4]] : tensor<3x8x8x16xf32>
// CHECK:         }

  func.return %0 : tensor<256x9x7x16xf32>
}

// -----

// CHECK-LABEL: @conv2d_backprop_input_dynamic
func.func @conv2d_backprop_input_dynamic(%filter: tensor<2x2x1x16xf32>, %out_backprop: tensor<?x256x256x16xf32>) -> tensor<?x512x512x1xf32> {
  // CHECK: %[[REV_FILTER:.*]] = "mhlo.reverse"(%arg0) <{dimensions = dense<[0, 1]> : tensor<2xi64>}>