Function main

code/training_evaluation/train.py:311–661

(_)

def main(_):


    tf.logging.set_verbosity(tf.logging.INFO)

    graph = tf.Graph()
    with graph.as_default(), tf.device('/cpu:0'):
        ######################
        # Config model_deploy#
        ######################

        # required from data
        num_samples_per_epoch = train_data['mouth'].shape[0]
        num_batches_per_epoch = int(num_samples_per_epoch / FLAGS.batch_size)

        num_samples_per_epoch_test = test_data['mouth'].shape[0]
        num_batches_per_epoch_test = int(num_samples_per_epoch_test / FLAGS.batch_size)

        # Create global_step
        global_step = tf.Variable(0, name='global_step', trainable=False)

        #########################################
        # Configure the larning rate. #
        #########################################
        learning_rate = _configure_learning_rate(num_samples_per_epoch, global_step)
        opt = _configure_optimizer(learning_rate)

        ######################
        # Select the network #
        ######################
        is_training = tf.placeholder(tf.bool)

        network_speech_fn = nets_factory.get_network_fn(
            FLAGS.model_speech_name,
            num_classes=2,
            weight_decay=FLAGS.weight_decay,
            is_training=is_training)

        network_mouth_fn = nets_factory.get_network_fn(
            FLAGS.model_mouth_name,
            num_classes=2,
            weight_decay=FLAGS.weight_decay,
            is_training=is_training)

        #####################################
        # Select the preprocessing function #
        #####################################

        # TODO: Do some preprocessing if necessary.

        ##############################################################
        # Create a dataset provider that loads data from the dataset #
        ##############################################################
        # with tf.device(deploy_config.inputs_device()):
        """
        Define the place holders and creating the batch tensor.
        """