namespace tensorflow {
namespace saved_model {

// Utility functions for SavedModel reading and writing.

// Returns "WriteVersion" ("1" or "2") of the SavedModel protobuf. If the
// protobuf has exactly one MetaGraphDef, which contains a SavedObjectGraph, it
// is version 2. Else, the protobuf is version 1.
//
// NOTE: The "WriteVersion" does *not* equal the major version of TF.
std::string GetWriteVersion(const SavedModel& saved_model);

        buildFile << """
            task writeVersion {
                doLast {
                    file("version.txt").with {
                        createNewFile()
                        text = gradle.gradleVersion
                    }
                }
            }
        """

        when:
        def runner = this.runner('writeVersion')
        configurer.execute(runner)
        runner.build()

monitoring::CounterCell& SavedModelWriteCount(absl::string_view write_version) {
  return *saved_model_write_counter->GetCell(std::string(write_version));
}

monitoring::CounterCell& SavedModelReadCount(absl::string_view write_version) {
  return *saved_model_read_counter->GetCell(std::string(write_version));
}

monitoring::CounterCell& SavedModelWriteApi(absl::string_view api_label) {

// has 1 field "write_version", which is equal to the
// `tensorflow::libexport::GetWriteVersion` of the protobuf and should be
// incremented when a SavedModel has been successfully written.
monitoring::CounterCell& SavedModelWriteCount(absl::string_view write_version);

// Returns "/tensorflow/core/saved_model/read/count" cell. This metric
// has 1 field "write_version", which is equal to the

        return this;
    }

    public void generate(DefaultVersionCatalog model) {
        writeHeader();
        writeMetadata();
        writeVersions(model);
        writeLibraries(model);
        writeBundles(model);
        writePlugins(model);
    }

    private void writeVersions(DefaultVersionCatalog model) {
        List<String> versions = model.getVersionAliases();
        if (versions.isEmpty()) {
            return;

  y = variable_v1.VariableV1(constant_op.constant([0]), name='y_saved')
  math_ops.add(x, y, name='x_y_sum')

  init_op = variables.global_variables_initializer()
  saver = saver_lib.Saver(write_version=saver_pb2.SaverDef.V1)
  with session.Session() as sess:
    sess.run(init_op)
    sess.run(y.assign(y + 42))
    # Without the checkpoint, the variable won't be set to 42.

    (https://godoc.org/github.com/tensorflow/tensorflow/tensorflow/go)
*   New checkpoint format becomes the default in `tf.train.Saver`. Old V1
    checkpoints continue to be readable; controlled by the `write_version`
    argument, `tf.train.Saver` now by default writes out in the new V2 format.
    It significantly reduces the peak memory required and latency incurred
    during restore.