func.func @torch_index_select(%arg0: tensor<2x1xf32>, %arg1: tensor<2xi32>) -> tensor<2x1xf32> {
  %0 = "mhlo.torch_index_select"(%arg0, %arg1) {
    batch_dims = 0 : i64, dim = 0 : i64
  } : (tensor<2x1xf32>, tensor<2xi32>) -> tensor<2x1xf32>
  func.return %0 : tensor<2x1xf32>
}

// CHECK-LABEL:   func @lowered_cumsum(
// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x12xf32>) -> tensor<4x12xf32> {

func.func @fuseTileWithBinaryOp1(%arg0: tensor<1x1xf32>, %arg1: tensor<1x128xf32>) -> tensor<1x128xf32> {
  %cst_0 = arith.constant dense<1.0> : tensor<f32>
  %cst_1 = arith.constant dense<[1, 128]> : tensor<2xi32>
  %0 = "tfl.add"(%arg0, %cst_0) {fused_activation_function = "NONE"} : (tensor<1x1xf32>, tensor<f32>) -> tensor<1x1xf32>
  %1 = "tfl.sqrt"(%0) : (tensor<1x1xf32>) -> tensor<1x1xf32>

  %3 = "tfl.pack"(%1, %2) {axis = 0 : i32, per_device_costs = {CPU = 2.0 : f32, GPU = -1.0 : f32}, values_count = 2 : i32, tac.device = "CPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
  func.return %3 : tensor<2x1xf32>
})";
  const std::string kExpectedFB = CreateRuntimeMetadata();
  mlir::DialectRegistry registry;
  registry.insert<mlir::TFL::TensorFlowLiteDialect, mlir::arith::ArithDialect,

  %w = arith.constant dense<127.0> : tensor<4x12xf32>
  %b = arith.constant dense<0.0> : tensor<4xf32>
  %fc = "tfl.fully_connected"(%arg0, %w, %b) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<1x224x224x3xf32>, tensor<4x12xf32>, tensor<4xf32>) -> tensor<1x112x112x4xf32>
  func.return %fc : tensor<1x112x112x4xf32>

  "tf.SomeOp2"() {device = "/device:CPU:0", _output_shapes = ["tfshape$"], f.Tin = [f32], f._read_only_resource_inputs = []} : () -> ()
  func.return
}

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

// CHECK-SAME: [[arg0:%.*]]: !tfrt_fallback.tf_tensor,
// CHECK-SAME: [[arg1:%.*]]: !tfrt_fallback.tf_tensor)
// CHECK-SAME: -> (!tfrt.chain, !tfrt_fallback.tf_tensor)
func.func @func_basic(
    %arg0: tensor<3x1xf32> {tf_saved_model.index_path = [0]},
    %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>> {tf_saved_model.index_path = [1]})
      -> (tensor<3x3xf32> {tf_saved_model.index_path = []})

  %outputs_6  =  "tf.Identity"(%outputs_2) {device = "/job:localhost/replica:0/task:0/device:CPU:0"} : (tensor<i32>) -> tensor<i32>
  %outputs_8  =  "tf.MatMul"(%arg5, %outputs_4) {device = "/job:localhost/replica:0/task:0/device:CPU:0", transpose_a = false, transpose_b = false} : (tensor<?x3xf32>, tensor<3x1xf32>) -> tensor<?x1xf32>

    ```mlir
      %0 = "tf.Const"() {value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
      %1 = "tf.Const"() {device = "", value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
      %2 = "tf.Const"() {device = "baz", value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
    ```

    then running this pass with 'default-device=foobar', we get:

    ```mlir

  // expected-error @+1 {{expect 'output' num_elements % 40 == 0, got 'tensor<1x41xf32>'}}
  %0 = "tfl.fully_connected"(%arg0, %arg1, %arg2) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<1x37xf32>, tensor<40x37xf32>, tensor<40xf32>) -> tensor<1x41xf32>
  func.return %0 : tensor<1x41xf32>
}

// -----

//                   (tensor<4x1xf32>, tensor<4x2xf32>, tensor<4x3xf32>)
//
// We will generate slices following slices:
// %0 = "mhlo.slice"(%input) {
//        limit_indices = dense<[4, 1]> : tensor<2xi64>,
//        start_indices = dense<0> : tensor<2xi64>,
//        strides = dense<1> : tensor<2xi64>} :
//        (tensor<4x6xf32>) -> tensor<4x1xf32>
// %1 = "mhlo.slice"(%input) {