usage of tf.app.run() and argparse

Question

I have understood what a parser does, but I do not get its use when it is mingled with tf.app.run() in the following code:

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.register("type", "bool", lambda v: v.lower() == "true")

    parser.add_argument("--ps_hosts",
                        type=str,
                        default="",
                        help="Comma-seperated list of hostname:port pairs")
    parser.add_argument("--worker_hosts",
                        type=str,
                        default="",
                        help="Comma-seperated list of hostname:port pairs")
    parser.add_argument("--job_name",
                        type=str,
                        default="",
                        help="One of 'ps', 'worker'")
    parser.add_argument("--task_index",
                        type=int,
                        default=0,
                        help="Index of task within the job")
    parser.add_argument("--data_dir",
                        type=str,
                    -   default="/tmp/mnist_data",
                        help="Directory for storing input data")
    parser.add_argument("--log_dir",
                        type=str,
                        default="/tmp/train_logs",
                        help="Directory of train logs")
    FLAGS, unparsed = parser.parse_known_args()
    tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)

the main function in the program does not have any arguments as it is defined as def main(_). So what is the argv argument in tf.app.run() supposed to mean or do?

Thanks

Answer 1

The argv parameter is used in Tensorflow's built-in command-line flag parsing. It's mainly intended for demos. You can define flags like tf.flags.DEFINE_integer('batch_size', 128). You will then be able to access it with tf.flags.FLAGS.batch_size.

If you are parsing your arguments using ArgumentParser then you don't need to use tf.app.run.