// CHECK: func.func @init_func_restore_op
// CHECK-NOT: @tf_quant__save
}

// -----

// SaveV2 op not created when the initializer function doesn't exist.

module attributes {tf_saved_model.semantics} {
  "tf_saved_model.session_initializer"() {initializers = []} : () -> ()
// Test that the function for SaveV2 op is not created.
// CHECK-NOT: @tf_quant__save

    tf_executor.graph {
      tf_executor.fetch
    }
    return
  }
}
// Save function should be erased.
// CHECK-NOT: @tf_quant__save

// Test that the contents of @tf_quant__save are copied to @main.
// CHECK: func.func @main
// CHECK-SAME: %[[ARG_0:.*]]: tensor<!tf_type.string> {tf_saved_model.index_path = ["__tf_file_prefix"]}
// CHECK: tf_executor.graph

namespace mlir {
namespace quant {

// Name of the save function. The "tf_quant__" prefix is for avoiding conflict
// with existing function's name.
inline constexpr StringRef kTfQuantSaveFuncName = "tf_quant__save";

// Name of the TensorFlow Operation to be fetched to save the variables to
// checkpoint. This save op follows the SavedModel's load semantics, so it
// should return the file prefix of the checkpoint as a string tensor.

// reused if it already exist in @main. Otherwise a new file prefix argument
// will be created. @tf_quant__save function will be erased.
//
// Running this pass essentially has the effect of inlining the @tf_quant__save
// into the main graph. This is beneficial when we wish to find and fetch
// the node that saves the variables, after the ModuleOp has been exported as
// GraphDef.

constexpr StringRef kTfQuantSaveReturnOpName = "tf_quant__save_return";

// A pass that creates a new function that wraps the newly created SaveV2 op.
// The new function's name is "tf_quant__save". The function accepts a single
// string tensor as argument, which specifies the path to the checkpoint to
// which the variable's tensor values are saved. It finds