%weight = "tf.Const"() {value = dense<2.000000e+00> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>
    %1 = "tf.PartitionedCall"(%arg0, %weight) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_conv2d_fn_1} : (tensor<1x2x2x3xf32>, tensor<2x3x3x2xf32>) -> tensor<*xf32>

func.func @dont_inline_custom_on_duplicated_cases() -> tensor<2xi32> {
  // CHECK-NEXT: "tf.PartitionedCall"
  // CHECK-NEXT: "tf.PartitionedCall"
  // CHECK-NEXT: return
  %0 = "tf.PartitionedCall"() {config = "", config_proto = "", executor_type = "", f = @custom_callee} : () -> tensor<2xi32>
  %1 = "tf.PartitionedCall"() {config = "", config_proto = "", executor_type = "", f = @custom_callee} : () -> tensor<2xi32>

    %1 = "tf.PartitionedCall"(%0) <{
      config = "", config_proto = "", executor_type = "", f = @some_other_func
    }> {
      _collective_manager_ids = [], device = ""
    } : (tensor<1x2x2x3xf32>) -> tensor<1x2x2x3xf32>
    // CHECK-SKIP: tf.PartitionedCall
    // CHECK-NOSKIP: call @some_other_func
    // CHECK-NOSKIP-NOT: tf.PartitionedCall
    func.return %1: tensor<1x2x2x3xf32>
  }

  func.func @serving_default(%arg0: tensor<1x3x4x3xf32>) -> (tensor<1x3x2x2xf32>) {
    %0 = "tf.PartitionedCall"(%arg0) {config = "", config_proto = "", executor_type = "", f = @conv2d_with_inliner}
        : (tensor<1x3x4x3xf32>) -> tensor<1x3x2x2xf32>
    return %0 : tensor<1x3x2x2xf32>
  }

// CHECK-LABEL: func @serving_default
// CHECK: "tf.PartitionedCall"
// CHECK-SAME: f = @composite_conv2d_fn_1

module {
  func.func @matmul(%arg0: tensor<1x2x2x3xf32>) -> (tensor<*xf32>) {
    %cst_0 = "tf.Const"() {value = dense<0.000000e+00> : tensor<2x1024xf32>} : () -> tensor<2x1024xf32>
    %1 = "tf.PartitionedCall"(%arg0, %cst_0) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_matmul_fn} : (tensor<1x2x2x3xf32>, tensor<2x1024xf32>) -> tensor<*xf32>
    func.return %1: tensor<*xf32>
  }

module {
  func.func @matmul(%arg0: tensor<1x2x2x3xf32>) -> (tensor<*xf32>) {
    %cst_0 = "tf.Const"() {value = dense<0.000000e+00> : tensor<2x1024xf32>} : () -> tensor<2x1024xf32>
    %1 = "tf.PartitionedCall"(%arg0, %cst_0) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_matmul_fn} : (tensor<1x2x2x3xf32>, tensor<2x1024xf32>) -> tensor<*xf32>
    func.return %1: tensor<*xf32>
  }

// CHECK-NEXT: %[[V0:.*]] = "tf.PartitionedCall"(%[[ARG1]], %[[CST_1]], %[[CST_0]]) <{config = "", config_proto = "", executor_type = "", f = @quantize_i8}> : (tensor<2x3x3x2xf32>, tensor<f32>, tensor<i32>) -> tensor<2x3x3x2xi8>

// CHECK-NOT: f = @NoOp

      // CHECK: %[[CALL_OUTPUT_0:[0-9]*]] = "tf.PartitionedCall"(%[[A_OUTPUT]])
      // CHECK-SAME: f = @_func
      // CHECK-SAME: device = "/device:test_device:0"
      %3 = "tf_device.launch_func"(%2) {device = "/device:test_device:0", func = @_func} : (tensor<?xf32>) -> tensor<?xf32>

      // CHECK: %[[CALL_OUTPUT_1:[0-9]*]] = "tf.PartitionedCall"(%[[CALL_OUTPUT_0]])
      // CHECK-SAME: f = @_func

// CHECK: %[[PRESERVE_W:.*]] = "tf.Identity"(%[[W]]) : (tensor<2x1024xi8>) -> tensor<2x1024xi8>
// CHECK: %[[DEQUANTIZED:.*]] = "tf.PartitionedCall"(%[[PRESERVE_W]]) <{config = "", config_proto = "", executor_type = "", f = @composite_dequantize_uniform}> : (tensor<2x1024xi8>) -> tensor<2x1024xf32>