version = 5 : i64
      } : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>, tensor<2xf32>) -> tensor<1x3x4x2xf32>
    %2 = "quantfork.stats"(%1) {layerStats = dense<[5.00000000e-6, 7.00000000e-1]> : tensor<2xf32>} : (tensor<1x3x4x2xf32>) -> tensor<1x3x4x2xf32>
    return %2 : tensor<1x3x4x2xf32>
  }
// CHECK: func.func private @quantize_conv_with_bias_1d_fn(%[[ARG_0:.+]]: tensor<1x3x4x3xf32>) -> tensor<1x3x4x2xf32> attributes {tf._original_func_name = "main_0"}

// CHECK-NEXT: %[[ADD:.*]] = "tf.AddV2"(%[[CONV2D]], %[[CONST_0]]) : (tensor<1x3x2x3xf32>, tensor<1x3x2x3xf32>) -> tensor<1x3x2x3xf32>
// CHECK-NEXT: return %[[ADD]] : tensor<1x3x2x3xf32>

// -----

func.func @fuse_conv2d_and_mul(%arg0: tensor<1x3x4x3xf32>) -> (tensor<1x3x2x2xf32>) {

    %1 = "tf.BiasAdd"(%0, %cst_0) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<3xf32>) -> tensor<*xf32>
    %2 = "tf.PartitionedCall"(%arg0, %cst_2) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_depthwise_conv2d_fn_1} : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<*xf32>

    %1 = "tf.Conv2D"(%0, %cst) {data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [], padding = "SAME", strides = [1, 1, 2, 1], use_cudnn_on_gpu = true} : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<1x3x2x2xf32>
    %2 = "tf.Relu"(%1) {device = ""} : (tensor<1x3x2x2xf32>) -> tensor<1x3x2x2xf32>

        : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<1x3x2x2xf32> loc(fused["Conv2D:", "test_opt_out"])
    %2 = "tf.IdentityN"(%1) {device = ""} : (tensor<1x3x2x2xf32>) -> tensor<1x3x2x2xf32>
    return %2 : tensor<1x3x2x2xf32>
  }

  func.func @serving_default(%arg0: tensor<1x3x4x3xf32>) -> (tensor<1x3x2x2xf32>) {

    %1 = "tf.BiasAdd"(%0, %cst) {data_format = "NHWC", device = ""} : (tensor<1x2x2x3xf32>, tensor<3xf32>) -> tensor<1x2x2x3xf32>
    %2 = "tf.Relu6"(%1) {device = ""} : (tensor<1x2x2x3xf32>) -> tensor<1x2x2x3xf32>
    %3 = "tf.Identity"(%2) {device = ""} : (tensor<1x2x2x3xf32>) -> tensor<1x2x2x3xf32>
    return %3 : tensor<1x2x2x3xf32>
  }
}

    %1 = "tf.BiasAdd"(%0, %cst_0) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<6xf32>) -> tensor<*xf32>
    func.return %1: tensor<*xf32>
  }
  func.func private @composite_depthwise_conv2d_fn(%arg0: tensor<1x3x4x3xf32>, %arg1: tensor<2x3x3x2xf32>) -> tensor<*xf32> attributes {tf_quant.composite_function} {

    %1 = "tf.BiasAdd"(%0, %cst_0) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<6xf32>) -> tensor<*xf32>
    func.return %1: tensor<*xf32>
  }
  func.func private @composite_depthwise_conv2d_fn(%arg0: tensor<1x3x4x3xf32>, %arg1: tensor<2x3x3x2xf32>) -> tensor<*xf32> attributes {tf_quant.composite_function} {

  %2 = "tf.Cast"(%1) {Truncate = false} : (tensor<1x3x2x2xbf16>) -> tensor<1x3x2x2xf32>
  %3 = "tf.IdentityN"(%2) {device = ""} : (tensor<1x3x2x2xf32>) -> tensor<1x3x2x2xf32>
  return %3 : tensor<1x3x2x2xf32>
// CHECK: tf.Conv2D
// CHECK-SAME: loc("Model/conv2d")
}

func.func @conv2d_with_valid_loc(%arg0: tensor<1x3x4x3xf32>) -> (tensor<1x3x2x2xf32>) {

    %0 = "quantfork.stats"(%arg0) {layerStats = dense<[1.27501142, 149.824783]> : tensor<2xf32>} : (tensor<1x3x4x3xf32>) -> tensor<1x3x4x3xf32>
    %1 = "tf.PartitionedCall"(%0, %cst_0, %cst) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", device = "", executor_type = "", f = @composite_conv2d_with_bias_and_relu6_fn_10} : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>, tensor<2xf32>) -> tensor<*xf32>