%1 = "tfl.mul"(%0, %arg2) {fused_activation_function = "RELU6", per_device_costs = {CPU = 5.0 : f32, GPU = 1.0 : f32}, tac.device = "GPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>

  func.return
}

// CHECK-LABEL: func @basic
func.func @basic(
    %arg0: tensor<3x1xf32>,
    %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>>) -> (tensor<3x3xf32>) {

        func.func @main(%arg0 : tensor<?xf32>) -> tensor<?xf32> {
          %cst = "tf.Const"() {value = dense<1.0> : tensor<f32>} : () -> tensor<f32>
          %0 = "tf.Add"(%arg0, %cst): (tensor<?xf32>, tensor<f32>) -> tensor<?xf32>
          func.return %0 : tensor<?xf32>
        }
      }
    """
    module_bytecode = tensorflow_to_stablehlo.tensorflow_module_to_stablehlo(

// CHECK-NEXT:     "tf.IfrtRestoreVariableOp"([[PREFIX]], [[NAME]], [[SLICE]], [[HANDLEY]], [[HANDLEZ]])
// CHECK-SAME:        {restored_dtypes = [f32, f32], truncate_in_cast = array<i1: false, false>}
// CHECK-NOT:       "tf.RestoreV2"
// CHECK-NEXT:     return

module {
  func.func @restore_multiple() {

// CHECK: %[[out:.*]] = "tf.PartitionedCall"([[ARG0:%arg[0-9]+]], %[[q_cst]]) <{config = "", config_proto = "", executor_type = "", f = @composite_matmul_fn}> {_tfl_quant_trait = "fully_quantizable"} : (tensor<1x2x2x3xf32>, tensor<2x1024x!quant.uniform<i8<-127:127>:f32, 3.9370078740157481E-9>>) -> tensor<*xf32>

  // CHECK: tfrt_fallback_async.batch_function device("/device:CPU:0") @batched_function
  // CHECK-SAME: Tin = [f32]
  // CHECK-SAME: Tout = [f32]
  // CHECK-SAME: allowed_batch_sizes = [6]
  // CHECK-SAME: batch_timeout_micros = 100000 : i64
  // CHECK-SAME: batching_queue = ""
  // CHECK-SAME: container = ""

    %arg1: tensor<1x3xf32> {tf_saved_model.index_path = [0]})
      -> (tensor<3x3xf32> {tf_saved_model.index_path = []}) {
  // CHECK: {{%.*}} = corert.get_op_handler %arg0 "/device:GPU:0"
  %2 = "tf.MatMul"(%arg0, %arg1) {T = f32, _output_shapes = ["tfshape$dim { size: 3 } dim { size: 3 }"], device = "/device:GPU:0", transpose_a = false, transpose_b = false} : (tensor<3x1xf32>, tensor<1x3xf32>) -> tensor<3x3xf32>
  func.return %2 : tensor<3x3xf32>

// CHECK-NEXT: [[o_chain_0:%.*]], [[o1:%.*]] = tfrt_fallback_async.executeop.seq([[in_chain]]) key(1) cost({{.*}}) device("/job:localhost/replica:0/task:0/device:CPU:0") "tf.ReadVariableOp"([[o]]) {dtype = f32} : 1
// CHECK-NEXT: [[out_ch:%.*]] = tfrt.merge.chains [[o_chain]], [[o_chain_0]]
// CHECK-NEXT: tfrt.return [[out_ch]], [[o1]] : !tfrt.chain, !tfrt_fallback.tf_tensor

//  Tensor device_tensor(device_allocator, DT_FLOAT, TensorShape({2, 2}));
//  se::DeviceMemoryBase gpu_dst{device_tensor.data(), 4 * sizeof(float)};
//  xla::Shape shape(xla::F32, {2, 2}, {}, {})
//  tsl::AsyncValueRef<std::unique_ptr<se::Event>> done_event =
//      tsl::MakeConstructedAsyncValueRef<std::unique_ptr<se::Event>>(stream.parent());
//  done_event->Init();
//

//  stream_executor::Stream stream(executor);
//  Tensor device_tensor(device_allocator, DT_FLOAT, TensorShape({2, 2}));
//  se::DeviceMemoryBase gpu_dst{device_tensor.data(), 4 * sizeof(float)};
//  xla::Shape shape(xla::F32, {2, 2}, {}, {})
//  tsl::AsyncValueRef<std::unique_ptr<se::Event>> done_event =
//      tsl::MakeConstructedAsyncValueRef<std::unique_ptr<se::Event>>(stream.parent());
//  done_event->Init();
//  Tensor dest_cpu_tensor;
//