tensor_shape {
        }
        int_val: 2
      }
    }
  }
  experimental_debug_info {
    original_node_names: "n1"
    original_node_names: "n2"
    original_func_names: "f1"
    original_func_names: "f2"
  }
}
node {
  name: "unnamed1"
  op: "foo"
  experimental_debug_info {
  }
}
library {
  function {
    signature {
      name: "foo"
    }
    node_def {

    // We may retrieve the original function's name from the attribute.
    // Functions without this attribute are ignored.
    auto original_func_name =
        func_op->getAttrOfType<StringAttr>("tf._original_func_name");
    if (original_func_name) {
      if (auto alias_itr = function_aliases.find(original_func_name.str());
          alias_itr != function_aliases.end()) {
        const FunctionAlias alias = alias_itr->second;

};

std::string FindFunctionName(const GraphExportConfig& configs, FuncOp func) {
  if (auto original_func_name =
          func->getAttrOfType<mlir::StringAttr>("tf._original_func_name");
      configs.export_original_tf_func_name && original_func_name) {
    return original_func_name.str();
  }

  return func.getName().str();
}

absl::StatusOr<std::unique_ptr<NodeDef>> Exporter::GetArgumentNode(

};

std::string FindFunctionName(const GraphExportConfig& configs, FuncOp func) {
  if (auto original_func_name =
          func->getAttrOfType<mlir::StringAttr>("tf._original_func_name");
      configs.export_original_tf_func_name && original_func_name) {
    return original_func_name.str();
  }

  return func.getName().str();
}

absl::StatusOr<std::unique_ptr<NodeDef>> Exporter::GetArgumentNode(

  // MLIR @main function corresponds to the TF function "main_original".
  OwningOpRef<ModuleOp> module_op = ParseModuleOpString(R"mlir(
    func.func private @main(%arg: tensor<1x2xf32>) -> (tensor<1x2xf32>) attributes {tf._original_func_name = "main_original"} {
      return %arg : tensor<1x2xf32>
    }
  )mlir");
  ASSERT_TRUE(module_op);

  absl::flat_hash_map<FunctionName, FunctionAlias> function_aliases;

// between constant and XlaCallModule op with empty `weight_only_ptq` method
// and function name containing conv.

func.func @qdq_for_conv_weight_empty(%arg0: tensor<1x3x2x3xf32>) -> tensor<1x2x2x2xf32> attributes {tf._original_func_name = "main_0"} {
  %cst = "tf.Const"() {value = dense<3.000000e-01> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>
  %0 = "tf.XlaCallModule"(%arg0, %cst) {

  }
}

# CHECK-DAG: func private @custom_relu{{[0-9]*}}(){{.+}}tf._implements = #tf_type.func<@tensorflow.relu, {}>, tf._original_func_name = "custom_relu"}

# CHECK-SAME: else_branch = @"key/[[else:[a-zA-Z_0-9]+]]"
# CHECK-SAME: then_branch = @"key/[[then:[a-zA-Z_0-9]+]]"

# CHECK: func private @"key/[[else]]"(
# CHECK-SAME: tf._original_func_name
# CHECK: func private @"key/[[then]]"(
# CHECK-SAME: tf._original_func_name


def Test():
  data = tf.constant([1, 2, 3, 4, 5, 6])
  # Create placeholders to prevent constant folding.
  x_op = tf.placeholder(dtype=tf.int32)

  }
  // CHECK-LABEL: @funcB_renamed
  func.func private @funcB_renamed(%arg0: tensor<i64> {tf._user_specified_name = "args_0"}) -> tensor<!tf_type.variant> attributes {tf._original_func_name = "funcB", tf._tf_data_function = true, tf.signature.is_stateful} {
    %0 = "tf.Const"() {device = "/device:CPU:0", value = dense<0> : tensor<i64>} : () -> tensor<i64>

module attributes {tf_saved_model.semantics} {
// CHECK: func.func private @quantize_gather_fn(%[[ARG:.+]]: tensor<3x4x2xf32>) -> tensor<2x3x2x2xf32> attributes {tf._original_func_name = "main_0"}
  func.func private @quantize_gather_fn(%arg: tensor<3x4x2xf32>) -> tensor<2x3x2x2xf32> attributes {tf._original_func_name = "main_0"} {
    %cst = "tf.Const"() {value = dense<1> : tensor<2x3x2xi32>} : () -> tensor<2x3x2xi32>