return output_values

  def _read_tensor_array_file(self, file_path):
    tensor_protos = []
    for raw_record in tf_record.tf_record_iterator(file_path, options='ZLIB'):
      tensor_protos.append(
          tensorflow.make_ndarray(tensor_pb2.TensorProto.FromString(raw_record))
      )
    return np.array(tensor_protos)

  @parameterized.named_parameters(
      {
          'testcase_name': 'none',

  func.return %0 : tensor<4xi1>
}

func.func @complex64() -> tensor<4xcomplex<f32>> {
  // CHECK-LABEL: @complex64

        tensor_shape {
        }
        int_val: 7
      }
    }
  }
}
library {
}
versions {
  producer: 70
}

# CHECK: tf.Const
# CHECK-SAME: value = #tf_type<tensor_proto : "{{0[xX][0-9a-fA-F]*}}"> : tensor<!tf_type.quint8>

  func.return %0 : tensor<4xi1>
}

func.func @complex64() -> tensor<4xcomplex<f32>> {
  // CHECK-LABEL: @complex64

  %input_zps = "tf.Const"() { value = dense<4> : tensor<i32> } : () -> tensor<i32>

  // tensor_proto that points to dense<127> of type !tf_type.qint32.
  // CHECK-DAG: %[[RHS:.*]] = mhlo.constant() <{value = dense<127> : tensor<2xi32>}> : () -> tensor<2x!quant.uniform<i32:f32, 2.000000e+00:4>>

  tf_executor.graph {

// CHECK-NEXT:   } ],
// CHECK-NEXT:   signature_defs: [  ]
// CHECK-NEXT: }
func.func @main() -> tensor<!tf_type.variant<tensor<2xi32>>> {

}

// CHECK-LABEL: func @while_test_if
// CHECK-SAME: -> !tf_mlrt.tensor
func.func @while_test_if(%cond: tensor<i1>, %x: tensor<i1>, %y: tensor<i1>) -> (tensor<i32>) {
  // CHECK: [[CONST:%.*]] = tf_mlrt.constop {tensor_proto = "\08\03\12\00"}
  %cst = "tf.Const"() {__op_key = 2: i32, device = "/device:CPU:0", value = dense<0> : tensor<i32>} : () -> tensor<i32>
  // Predicate should be inlined.
  // CHECK-NEXT: tf_mlrt.predicate

  EXPECT_THAT(num_bytes, Eq(23));
}

TEST_F(GetSizeInBytesTest, ConstOpWithUnknownSizeAssumes4BytesPerElement) {
  constexpr absl::string_view kConstOpExpr = R"mlir(
    %cst = "tf.Const"() {value = #tf_type<tensor_proto : "0xDEADBAAD"> : tensor<!tf_type.variant>} : () -> tensor<!tf_type.variant>
  )mlir";

  Block block{};
  TF::ConstOp int_tensor_const_op = ParseConstOp(kConstOpExpr, block, ctx_);

        sample_proto = _RepresentativeDataSample.FromString(
            sample_bytes.numpy()
        )
        sample = {}
        for input_key, tensor_proto in sample_proto.tensor_proto_inputs.items():
          sample[input_key] = tensor_util.MakeNdarray(tensor_proto)
        samples.append(sample)
    return samples

  def load(self) -> RepresentativeDatasetMapping:
    """Loads the representative datasets.