// expected-remark@above {{ID: 0}}

  %4 = "tf.Case"(%2, %arg0, %arg1) {branches = [@sub, @add], output_shapes = [#tf_type.shape<>], device = "noodle", is_stateless = false} : (tensor<i32>, tensor<f32>, tensor<f32>) -> tensor<f32>
  // CHECK: PartitionedCall
  // CHECK-SAME: f = @sub
  // CHECK-SAME: _cluster_launch = "not_ready"
  // CHECK-SAME: device = "noodle"

    %0:3 = "tf.While"(%cst_0, %cst_1, %arg0) {T = [i32, i32, f32], _lower_using_switch_merge = true, _num_original_outputs = 4 : i64, _read_only_resource_inputs = [], body = @while_body, cond = @while_cond, device = "", is_stateless = true, output_shapes = [#tf_type.shape<>, #tf_type.shape<>, #tf_type.shape<1x5x5x1024>], parallel_iterations = 10 : i64, shape_invariant} : (tensor<i32>, tensor<i32>, tensor<1x5x5x1024xf32>) -> (tensor<i32>, tensor<i32>, tensor<1x5x5x1024xf32>)

This means that if `rhs_scales` and `rhs_zero_points` are scalar `Tensor`s, `output_scales` and `output_zero_points` must be scalar `Tensor`s as well.}]>:$output_scales,
    Arg<TF_Int32Tensor, [{The int32 value(s) used as zero points when quantizing original data that output represents.
Same shape condition as `output_scales`.}]>:$output_zero_points,

    DefaultValuedOptionalAttr<I64ArrayAttr, "{}">:$window_strides,

    "tf.Yield"(%1) : (tensor<5xf32>) -> ()
  }) {device = "/job:tpu_host_worker/replica:0/task:0/device:CPU:0", metadata = "", operandSegmentSizes = array<i32: 1, 2, 1>, output_shapes = [#tf_type.shape<>], output_types = [!tf_type.string]} : (tensor<4xf32>, tensor<3xf32>, tensor<!tf_type.resource>, tensor<2xf32>) -> tensor<!tf_type.variant>
  return
}

// -----

  }

  auto output_shape = xla::ShapeInference::InferGatherShape(
      input_shape, start_indices_shape, gather_dim_numbers, slice_sizes);
  if (!output_shape.ok()) {
    op->emitError() << output_shape.status().message();
    return false;
  }

  auto refined_type = xla::ConvertShapeToType<RankedTensorType>(
      *output_shape, mlir::Builder(op));
  if (!refined_type.ok()) {

      %arg1: tensor<i64>
    ) {
      %1 = "tf.ReduceDataset"(%arg0, %arg1) {
        Targuments = [],
        Tstate = [i64], device = "",
        f = @__reduce_func_1, f._tf_data_function = true,
        output_shapes = [#tf_type.shape<>],
        output_types = [i64], use_inter_op_parallelism = true, _xla_compile_device_type="TPU"} :
          (tensor<!tf_type.variant>, tensor<i64>) -> (tensor<i64>)
      func.return
    }
    ```

    %arg1: tensor<i64>
  ) {
    %1 = "tf.ReduceDataset"(%arg0, %arg1) {
      Targuments = [],
      Tstate = [i64], device = "",
      f = @__reduce_func_1, f._tf_data_function = true,
      output_shapes = [#tf_type.shape<>],
      output_types = [i64], use_inter_op_parallelism = true, _xla_compile_device_type="TPU"} :
 (tensor<!tf_type.variant>, tensor<i64>) -> (tensor<i64>)
    func.return
 }
 ```

[i32, i32, i32, !tf_type.variant, !tf_type.variant, f32], _lower_using_switch_merge = true, _num_original_outputs = 6 : i64, _read_only_resource_inputs = [], body = @map_while_body_180, cond = @map_while_cond_170, device = "", is_stateless = true, output_shapes = [#tf_type.shape<>, #tf_type.shape<>, #tf_type.shape<>, #tf_type.shape<>, #tf_type.shape<>, #tf_type.shape<?>], parallel_iterations = 4 : i64, shape_invariant} : (tensor<i32>, tensor<i32>, tensor<i32>, tensor<!tf_type.variant<tensor<*xf32>>>,...

// RUN: tf-tfrt-opt -tf-to-tfrt %s | FileCheck %s --dump-input=fail

// _output_shapes and f.* attributes are removed during tf-to-tfrt lowering.
// CHECK-LABEL: func @remove_unused_attr
func.func @remove_unused_attr() {
  // CHECK: %out_op_chain = tfrt_fallback_async.executeop.seq(%arg0) key(0) cost({{.*}}) device("/device:CPU:0") "tf.SomeOp2"()