Python tensorflow.Session() Examples

def __init__(self, env, dueling, noisy, fname):
        self.g = tf.Graph()
        self.noisy = noisy
        self.dueling = dueling
        self.env = env
        with self.g.as_default():
            self.act = deepq.build_act_enjoy(
                make_obs_ph=lambda name: U.Uint8Input(
                    env.observation_space.shape, name=name),
                q_func=dueling_model if dueling else model,
                num_actions=env.action_space.n,
                noisy=noisy
            )
            self.saver = tf.train.Saver()
        self.sess = tf.Session(graph=self.g)

        if fname is not None:
            print('Loading Model...')
            self.saver.restore(self.sess, fname) 

def test_separate(test_file, configuration, backend):
    """ Test separation from raw data. """
    with tf.Session() as sess:
        instruments = MODEL_TO_INST[configuration]
        adapter = get_default_audio_adapter()
        waveform, _ = adapter.load(test_file)
        separator = Separator(configuration, stft_backend=backend)
        prediction = separator.separate(waveform, test_file)
        assert len(prediction) == len(instruments)
        for instrument in instruments:
            assert instrument in prediction
        for instrument in instruments:
            track = prediction[instrument]
            assert waveform.shape[:-1] == track.shape[:-1]
            assert not np.allclose(waveform, track)
            for compared in instruments:
                if instrument != compared:
                    assert not np.allclose(track, prediction[compared]) 

def test_feature_pairing(self):
        fgsm = FastGradientMethod(self.model)
        attack = lambda x: fgsm.generate(x)
        loss = FeaturePairing(self.model, weight=0.1, attack=attack)
        l = loss.fprop(self.x, self.y)
        with tf.Session() as sess:
            vl1 = sess.run(l, feed_dict={self.x: self.vx, self.y: self.vy})
            vl2 = sess.run(l, feed_dict={self.x: self.vx, self.y: self.vy})
        self.assertClose(vl1, sum([4.296023369, 2.963884830]) / 2., atol=1e-6)
        self.assertClose(vl2, sum([4.296023369, 2.963884830]) / 2., atol=1e-6)

        loss = FeaturePairing(self.model, weight=10., attack=attack)
        l = loss.fprop(self.x, self.y)
        with tf.Session() as sess:
            vl1 = sess.run(l, feed_dict={self.x: self.vx, self.y: self.vy})
            vl2 = sess.run(l, feed_dict={self.x: self.vx, self.y: self.vy})
        self.assertClose(vl1, sum([4.333082676, 3.00094414]) / 2., atol=1e-6)
        self.assertClose(vl2, sum([4.333082676, 3.00094414]) / 2., atol=1e-6) 

def test_override():
    # Make sure dropout_dict changes dropout probabilities successful

    # We would like to configure the test to deterministically drop,
    # so that the test does not need to use multiple runs.
    # However, tf.nn.dropout divides by include_prob, so zero or
    # infinitesimal include_prob causes NaNs.
    # For this test, random failure to drop will not cause the test to fail.
    # The stochastic version should not even run if everything is working
    # right.
    model = MLP(input_shape=[1, 1], layers=[Dropout(name='output',
                                                    include_prob=1e-8)])
    x = tf.constant([[1]], dtype=tf.float32)
    dropout_dict = {'output': 1.}
    y = model.get_layer(x, 'output', dropout=True, dropout_dict=dropout_dict)
    sess = tf.Session()
    y_value = sess.run(y)
    assert y_value == 1., y_value 

def test_drop():
    # Make sure dropout is activated successfully

    # We would like to configure the test to deterministically drop,
    # so that the test does not need to use multiple runs.
    # However, tf.nn.dropout divides by include_prob, so zero or
    # infinitesimal include_prob causes NaNs.
    # 1e-8 does not cause NaNs and shouldn't be a significant source
    # of test flakiness relative to dependency downloads failing, etc.
    model = MLP(input_shape=[1, 1], layers=[Dropout(name='output',
                                                    include_prob=1e-8)])
    x = tf.constant([[1]], dtype=tf.float32)
    y = model.get_layer(x, 'output', dropout=True)
    sess = tf.Session()
    y_value = sess.run(y)
    # Subject to very rare random failure because include_prob is not exact 0
    assert y_value == 0., y_value 

def main():
    rgb = False
    if rgb:
        kernels_list = [kernels.BLUR_FILTER_RGB,
                        kernels.SHARPEN_FILTER_RGB,
                        kernels.EDGE_FILTER_RGB,
                        kernels.TOP_SOBEL_RGB,
                        kernels.EMBOSS_FILTER_RGB]
    else:
        kernels_list = [kernels.BLUR_FILTER,
                        kernels.SHARPEN_FILTER,
                        kernels.EDGE_FILTER,
                        kernels.TOP_SOBEL,
                        kernels.EMBOSS_FILTER]

    kernels_list = kernels_list[1:]
    image = read_one_image('data/images/naruto.jpeg')
    if not rgb:
        image = tf.image.rgb_to_grayscale(image)
    image = tf.expand_dims(image, 0) # make it into a batch of 1 element
    images = convolve(image, kernels_list, rgb)
    with tf.Session() as sess:
        images = sess.run(images) # convert images from tensors to float values
    show_images(images, rgb) 

def create_session(config_dict=dict(), force_as_default=False):
    config = tf.ConfigProto()
    for key, value in config_dict.items():
        fields = key.split('.')
        obj = config
        for field in fields[:-1]:
            obj = getattr(obj, field)
        setattr(obj, fields[-1], value)
    session = tf.Session(config=config)
    if force_as_default:
        session._default_session = session.as_default()
        session._default_session.enforce_nesting = False
        session._default_session.__enter__()
    return session

#----------------------------------------------------------------------------
# Initialize all tf.Variables that have not already been initialized.
# Equivalent to the following, but more efficient and does not bloat the tf graph:
#   tf.variables_initializer(tf.report_unitialized_variables()).run() 

def setUp(self):
        super(TestRunnerMultiGPU, self).setUp()
        self.sess = tf.Session()

        inputs = []
        outputs = []
        self.niter = 10
        niter = self.niter
        # A Simple graph with `niter` sub-graphs.
        with tf.variable_scope(None, 'runner'):
            for i in range(niter):
                v = tf.get_variable('v%d' % i, shape=(100, 10))
                w = tf.get_variable('w%d' % i, shape=(100, 1))

                inputs += [{'v': v, 'w': w}]
                outputs += [{'v': v, 'w': w}]

        self.runner = RunnerMultiGPU(inputs, outputs, sess=self.sess) 

def savepb(self):
		"""
		Create a standalone const graph def that 
		C++	can load and run.
		"""
		darknet_pb = self.to_darknet()
		flags_pb = self.FLAGS
		flags_pb.verbalise = False
		
		flags_pb.train = False
		# rebuild another tfnet. all const.
		tfnet_pb = TFNet(flags_pb, darknet_pb)		
		tfnet_pb.sess = tf.Session(graph = tfnet_pb.graph)
		# tfnet_pb.predict() # uncomment for unit testing
		name = 'built_graph/{}.pb'.format(self.meta['name'])
		os.makedirs(os.path.dirname(name), exist_ok=True)
		#Save dump of everything in meta
		with open('built_graph/{}.meta'.format(self.meta['name']), 'w') as fp:
			json.dump(self.meta, fp)
		self.say('Saving const graph def to {}'.format(name))
		graph_def = tfnet_pb.sess.graph_def
		tf.train.write_graph(graph_def,'./', name, False) 

def cleverhans_attack_wrapper(cleverhans_attack_fn, reset=True):
    def attack(a):
        session = tf.Session()
        with session.as_default():
            model = RVBCleverhansModel(a)
            adversarial_image = cleverhans_attack_fn(model, session, a)
            adversarial_image = np.squeeze(adversarial_image, axis=0)
            if reset:
                # optionally, reset to ignore other adversarials
                # found during the search
                a._reset()
            # run predictions to make sure the returned adversarial
            # is taken into account
            min_, max_ = a.bounds()
            adversarial_image = np.clip(adversarial_image, min_, max_)
            a.predictions(adversarial_image)
    return attack 

def evaluate(mnist):
    with tf.Graph().as_default() as g:
        x = tf.placeholder(tf.float32, [
            mnist_train.BATCH_SIZE,
            mnist_inference.IMAGE_SIZE,
            mnist_inference.IMAGE_SIZE,
            mnist_inference.NUM_CHANNELS], name='x-input')
        y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE], name='y-input')

        validate_feed = {
            x: mnist.validation.images,
            y_: mnist.validation.labels}

        y = mnist_inference.inference(x, None)

        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

        variable_averages = tf.train.ExponentialMovingAverage(mnist_train.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.train.Saver(variables_to_restore)

        while True:
            with tf.Session() as sess:
                ckpt = tf.train.get_checkpoint_state(mnist_train.MODEL_SAVE_PATH)
                if ckpt and ckpt.model_checkpoint_path:
                    saver.restore(sess, ckpt.model_checkpoint_path)
                    global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
                    accuracy_score = sess.run(accuracy, feed_dict=validate_feed)
                    print("After %s training step(s), validation accuracy = %g" % (global_step, accuracy_score))
                else:
                    print("No Checkpoint file found")
                    return
            time.sleep(EVAL_INTERVAL_SECS) 

def test_save_and_load_var():
    """
    Tests that we can save and load a PicklableVariable with joblib
    """
    sess = tf.Session()
    with sess.as_default():
        x = np.ones(1)
        xv = PicklableVariable(x)
        xv.var.initializer.run()
        save("/tmp/var.joblib", xv)
        sess.run(tf.assign(xv.var, np.ones(1) * 2))
        new_xv = load("/tmp/var.joblib")
        assert np.allclose(sess.run(xv.var), np.ones(1) * 2)
        assert np.allclose(sess.run(new_xv.var), np.ones(1)) 

def train():
    # 学习率
    lr = 0.01

    x = tf.placeholder(tf.float32)
    y = tf.placeholder(tf.float32)

    y_ = inference(x)
    # 损失函数
    loss = tf.square(y_ - y)

    # 随机梯度下降
    opt = tf.train.GradientDescentOptimizer(lr)
    train_op = opt.minimize(loss)

    init = tf.global_variables_initializer()

    with tf.Session() as sess:
        sess.run(init)
        print("start training")
        for i in range(1000000):
            train_x, train_y = get_train_data()
            sess.run(train_op, feed_dict={x: train_x, y: train_y})

            if i % 10000 == 0:
                times = int(i / 10000)
                test_x_ndarray = np.arange(0, 2 * np.pi, 0.01)
                test_y_ndarray = np.zeros([len(test_x_ndarray)])
                ind = 0
                for test_x in test_x_ndarray:
                    test_y = sess.run(y_, feed_dict={x: test_x, y: 1})
                    np.put(test_y_ndarray, ind, test_y)
                    ind += 1
                draw_correct_line()
                pylab.plot(test_x_ndarray, test_y_ndarray, '--', label= str(times)+'times')
                pylab.show() 

def main():
    train_x = tf.placeholder(tf.float32)
    train_label = tf.placeholder(tf.float32)
    test_x = tf.placeholder(tf.float32)
    test_label = tf.placeholder(tf.float32)

    with tf.variable_scope("inference"):
        train_y = inference(train_x)
        tf.get_variable_scope().reuse_variables()
        test_y = inference(test_x)

    train_loss = tf.square(train_y - train_label)
    test_loss = tf.square(test_y - test_label)
    opt = tf.train.GradientDescentOptimizer(0.002)
    train_op = opt.minimize(train_loss)

    init = tf.global_variables_initializer()

    train_data_x, train_data_label = get_data(1000)
    test_data_x, test_data_label = get_data(1)

    with tf.Session() as sess:
        sess.run(init)
        for i in range(1000):
            sess.run(train_op, feed_dict={train_x: train_data_x[i],
                                          train_label: train_data_label[i]})
            if i % 10 == 0:
                test_loss_value = sess.run(test_loss, feed_dict={test_x:test_data_x[0],
                                                                 test_label:test_data_label[0]})
                print("step %d eval loss is %.3f" % (i, test_loss_value)) 

def train(data_dir, checkpoint_path, config):
    """Trains the model with the given data

    Args:
        data_dir: path to the data for the model (see data_utils for data
            format)
        checkpoint_path: the path to save the trained model checkpoints
        config: one of the above configs that specify the model and how it
            should be run and trained
    Returns:
        None
    """
    # Prepare Name data.
    print("Reading Name data in %s" % data_dir)
    names, counts = data_utils.read_names(data_dir)

    with tf.Graph().as_default(), tf.Session() as session:
        initializer = tf.random_uniform_initializer(-config.init_scale,
                                                    config.init_scale)
        with tf.variable_scope("model", reuse=None, initializer=initializer):
            m = NamignizerModel(is_training=True, config=config)

        tf.global_variables_initializer().run()

        for i in range(config.max_max_epoch):
            lr_decay = config.lr_decay ** max(i - config.max_epoch, 0.0)
            m.assign_lr(session, config.learning_rate * lr_decay)

            print("Epoch: %d Learning rate: %.3f" % (i + 1, session.run(m.lr)))
            train_perplexity = run_epoch(session, m, names, counts, config.epoch_size, m.train_op,
                                         verbose=True)
            print("Epoch: %d Train Perplexity: %.3f" %
                  (i + 1, train_perplexity))

            m.saver.save(session, checkpoint_path, global_step=i) 

def namignize(names, checkpoint_path, config):
    """Recognizes names and prints the Perplexity of the model for each names
    in the list

    Args:
        names: a list of names in the model format
        checkpoint_path: the path to restore the trained model from, should not
            include the model name, just the path to
        config: one of the above configs that specify the model and how it
            should be run and trained
    Returns:
        None
    """
    with tf.Graph().as_default(), tf.Session() as session:

        with tf.variable_scope("model"):
            m = NamignizerModel(is_training=False, config=config)

        m.saver.restore(session, checkpoint_path)

        for name in names:
            x, y = data_utils.name_to_batch(name, m.batch_size, m.num_steps)

            cost, loss, _ = session.run([m.cost, m.loss, tf.no_op()],
                                  {m.input_data: x,
                                   m.targets: y,
                                   m.weights: np.concatenate((
                                       np.ones(len(name)), np.zeros(m.batch_size * m.num_steps - len(name))))})

            print("Name {} gives us a perplexity of {}".format(
                name, np.exp(cost))) 

def train(mnist):
    x = tf.placeholder(tf.float32, [None, mnist_inference.INPUT_NODE], name='x-input')
    y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE], name='y-input')

    regularizer = tf.contrib.layers.l2_regularizer(REGULARAZATION_RATE)
    y = mnist_inference.inference(x, regularizer)
    global_step = tf.Variable(0, trainable=False)

    variable_averages = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
    variables_averages_op = variable_averages.apply(tf.trainable_variables())

    cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
    cross_entropy_mean = tf.reduce_mean(cross_entropy)
    loss = cross_entropy_mean + tf.add_n(tf.get_collection('losses'))
    learning_rate = tf.train.exponential_decay(
        LEARNING_RATE_BASE,
        global_step,
        mnist.train.num_examples / BATCH_SIZE,
        LEARNING_RATE_DECAY)
    train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)

    with tf.control_dependencies([train_step, variables_averages_op]):
        train_op = tf.no_op(name='train')

    # 初始化 TF 持久化类
    saver = tf.train.Saver()
    with tf.Session() as sess:
        tf.global_variables_initializer().run()

        for i in range(TRAINING_STEPS):
            xs, ys = mnist.train.next_batch(BATCH_SIZE)
            _, loss_value, step = sess.run([train_op, loss, global_step], feed_dict={x: xs, y_: ys})

            if i % 1000 == 0:
                print("After %d training step(s), loss on training batch is %g." % (step, loss_value))
                # 保存当前模型
                saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step=global_step) 

def evaluate(mnist):
    with tf.Graph().as_default() as g:
        x = tf.placeholder(tf.float32, [None, mnist_inference.INPUT_NODE], name='x-input')
        y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE], name='y-input')
        validate_feed = {
            x: mnist.validation.images,
            y_: mnist.validation.labels}

        y = mnist_inference.inference(x, None)

        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

        variable_averages = tf.train.ExponentialMovingAverage(mnist_train.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.train.Saver(variables_to_restore)

        while True:
            with tf.Session() as sess:
                ckpt = tf.train.get_checkpoint_state(mnist_train.MODEL_SAVE_PATH)
                if ckpt and ckpt.model_checkpoint_path:
                    saver.restore(sess, ckpt.model_checkpoint_path)
                    global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
                    accuracy_score = sess.run(accuracy, feed_dict=validate_feed)
                    print("After %s training step(s), validation accuracy = %g" % (global_step, accuracy_score))
                else:
                    print("No Checkpoint file found")
                    return
            time.sleep(EVAL_INTERVAL_SECS) 

def namignator(checkpoint_path, config):
    """Generates names randomly according to a given model

    Args:
        checkpoint_path: the path to restore the trained model from, should not
            include the model name, just the path to
        config: one of the above configs that specify the model and how it
            should be run and trained
    Returns:
        None
    """
    # mutate the config to become a name generator config
    config.num_steps = 1
    config.batch_size = 1

    with tf.Graph().as_default(), tf.Session() as session:

        with tf.variable_scope("model"):
            m = NamignizerModel(is_training=False, config=config)

        m.saver.restore(session, checkpoint_path)

        activations, final_state, _ = session.run([m.activations, m.final_state, tf.no_op()],
                                                  {m.input_data: np.zeros((1, 1)),
                                                   m.targets: np.zeros((1, 1)),
                                                   m.weights: np.ones(1)})

        # sample from our softmax activations
        next_letter = np.random.choice(27, p=activations[0])
        name = [next_letter]
        while next_letter != 0:
            activations, final_state, _ = session.run([m.activations, m.final_state, tf.no_op()],
                                                      {m.input_data: [[next_letter]],
                                                       m.targets: np.zeros((1, 1)),
                                                       m.initial_state: final_state,
                                                       m.weights: np.ones(1)})

            next_letter = np.random.choice(27, p=activations[0])
            name += [next_letter]

        print(map(lambda x: chr(x + 96), name)) 

def _add_to_tfrecord(data_filename, labels_filename, num_images,
                     tfrecord_writer):
  """Loads data from the binary MNIST files and writes files to a TFRecord.

  Args:
    data_filename: The filename of the MNIST images.
    labels_filename: The filename of the MNIST labels.
    num_images: The number of images in the dataset.
    tfrecord_writer: The TFRecord writer to use for writing.
  """
  images = _extract_images(data_filename, num_images)
  labels = _extract_labels(labels_filename, num_images)

  shape = (_IMAGE_SIZE, _IMAGE_SIZE, _NUM_CHANNELS)
  with tf.Graph().as_default():
    image = tf.placeholder(dtype=tf.uint8, shape=shape)
    encoded_png = tf.image.encode_png(image)

    with tf.Session('') as sess:
      for j in range(num_images):
        sys.stdout.write('\r>> Converting image %d/%d' % (j + 1, num_images))
        sys.stdout.flush()

        png_string = sess.run(encoded_png, feed_dict={image: images[j]})

        example = dataset_utils.image_to_tfexample(
            png_string, 'png'.encode(), _IMAGE_SIZE, _IMAGE_SIZE, labels[j])
        tfrecord_writer.write(example.SerializeToString()) 

def DecodeLoop(self):
    """Decoding loop for long running process."""
    sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True))
    step = 0
    while step < FLAGS.max_decode_steps:
      time.sleep(DECODE_LOOP_DELAY_SECS)
      if not self._Decode(self._saver, sess):
        continue
      step += 1 

def visualize(self, visual_fld, num_visualize):
        """ run "'tensorboard --logdir='visualization'" to see the embeddings """

        # create the list of num_variable most common words to visualize
        w2v_utils.most_common_words(visual_fld, num_visualize)

        saver = tf.train.Saver()
        with tf.Session() as sess:
            sess.run(tf.global_variables_initializer())
            ckpt = tf.train.get_checkpoint_state(os.path.dirname('data/checkpoints/checkpoint'))

            # if that checkpoint exists, restore from checkpoint
            if ckpt and ckpt.model_checkpoint_path:
                saver.restore(sess, ckpt.model_checkpoint_path)

            final_embed_matrix = sess.run(self.embed_matrix)

            # you have to store embeddings in a new variable
            embedding_var = tf.Variable(final_embed_matrix[:num_visualize], name='embedding')
            sess.run(embedding_var.initializer)

            config = projector.ProjectorConfig()
            summary_writer = tf.summary.FileWriter(visual_fld)

            # add embedding to the config file
            embedding = config.embeddings.add()
            embedding.tensor_name = embedding_var.name

            # link this tensor to its metadata file, in this case the first NUM_VISUALIZE words of vocab
            embedding.metadata_path = 'vocab_' + str(num_visualize) + '.tsv'

            # saves a configuration file that TensorBoard will read during startup.
            projector.visualize_embeddings(summary_writer, config)
            saver_embed = tf.train.Saver([embedding_var])
            saver_embed.save(sess, os.path.join(visual_fld, 'model.ckpt'), 1) 

def test_separate_to_file(test_file, configuration, backend):
    """ Test file based separation. """
    with tf.Session() as sess:
        instruments = MODEL_TO_INST[configuration]
        separator = Separator(configuration, stft_backend=backend)
        name = splitext(basename(test_file))[0]
        with TemporaryDirectory() as directory:
            separator.separate_to_file(
                test_file,
                directory)
            for instrument in instruments:
                assert exists(join(
                    directory,
                    '{}/{}.wav'.format(name, instrument))) 

def _get_session(self):
        if self._session is None:
            saver = tf.train.Saver()
            latest_checkpoint = tf.train.latest_checkpoint(get_default_model_dir(self._params['model_dir']))
            self._session = tf.Session()
            saver.restore(self._session, latest_checkpoint)
        return self._session 

def extract_vggish_features(paths, path2gt, model): 
    """Extracts VGGish features and their corresponding ground_truth and identifiers (the path).

       VGGish features are extracted from non-overlapping audio patches of 0.96 seconds, 
       where each audio patch covers 64 mel bands and 96 frames of 10 ms each.

       We repeat ground_truth and identifiers to fit the number of extracted VGGish features.
    """
    # 1) Extract log-mel spectrograms
    first_audio = True
    for p in paths:
        if first_audio:
            input_data = vggish_input.wavfile_to_examples(config['audio_folder'] + p)
            ground_truth = np.repeat(path2gt[p], input_data.shape[0], axis=0)
            identifiers = np.repeat(p, input_data.shape[0], axis=0)
            first_audio = False
        else:
            tmp_in = vggish_input.wavfile_to_examples(config['audio_folder'] + p)
            input_data = np.concatenate((input_data, tmp_in), axis=0)
            tmp_gt = np.repeat(path2gt[p], tmp_in.shape[0], axis=0)
            ground_truth = np.concatenate((ground_truth, tmp_gt), axis=0)
            tmp_id = np.repeat(p, tmp_in.shape[0], axis=0)
            identifiers = np.concatenate((identifiers, tmp_id), axis=0)

    # 2) Load Tensorflow model to extract VGGish features
    with tf.Graph().as_default(), tf.Session() as sess:
        vggish_slim.define_vggish_slim(training=False)
        vggish_slim.load_vggish_slim_checkpoint(sess, 'vggish_model.ckpt')
        features_tensor = sess.graph.get_tensor_by_name(vggish_params.INPUT_TENSOR_NAME)
        embedding_tensor = sess.graph.get_tensor_by_name(vggish_params.OUTPUT_TENSOR_NAME)
        extracted_feat = sess.run([embedding_tensor], feed_dict={features_tensor: input_data})
        feature = np.squeeze(np.asarray(extracted_feat))

    return [feature, ground_truth, identifiers] 

def _init_session(self):
        config = tf.ConfigProto(device_count={"gpu": 0})
        config.gpu_options.allow_growth = True
        return tf.Session(config=config) 

def setup_graph(self, input_audio_batch, target_phrase): 
        batch_size = input_audio_batch.shape[0]
        weird = (input_audio_batch.shape[1] - 1) // 320 
        logits_arg2 = np.tile(weird, batch_size)
        dense_arg1 = np.array(np.tile(target_phrase, (batch_size, 1)), dtype=np.int32)
        dense_arg2 = np.array(np.tile(target_phrase.shape[0], batch_size), dtype=np.int32)
        
        pass_in = np.clip(input_audio_batch, -2**15, 2**15-1)
        seq_len = np.tile(weird, batch_size).astype(np.int32)
        
        with tf.variable_scope('', reuse=tf.AUTO_REUSE):
            
            inputs = tf.placeholder(tf.float32, shape=pass_in.shape, name='a')
            len_batch = tf.placeholder(tf.float32, name='b')
            arg2_logits = tf.placeholder(tf.int32, shape=logits_arg2.shape, name='c')
            arg1_dense = tf.placeholder(tf.float32, shape=dense_arg1.shape, name='d')
            arg2_dense = tf.placeholder(tf.int32, shape=dense_arg2.shape, name='e')
            len_seq = tf.placeholder(tf.int32, shape=seq_len.shape, name='f')
            
            logits = get_logits(inputs, arg2_logits)
            target = ctc_label_dense_to_sparse(arg1_dense, arg2_dense, len_batch)
            ctcloss = tf.nn.ctc_loss(labels=tf.cast(target, tf.int32), inputs=logits, sequence_length=len_seq)
            decoded, _ = tf.nn.ctc_greedy_decoder(logits, arg2_logits, merge_repeated=True)
            
            sess = tf.Session()
            saver = tf.train.Saver(tf.global_variables())
            saver.restore(sess, "models/session_dump")
            
        func1 = lambda a, b, c, d, e, f: sess.run(ctcloss, 
            feed_dict={inputs: a, len_batch: b, arg2_logits: c, arg1_dense: d, arg2_dense: e, len_seq: f})
        func2 = lambda a, b, c, d, e, f: sess.run([ctcloss, decoded], 
            feed_dict={inputs: a, len_batch: b, arg2_logits: c, arg1_dense: d, arg2_dense: e, len_seq: f})
        return (func1, func2) 

def setUp(self):
        super(TestUtils, self).setUp()

        self.sess = tf.Session() 

def _init_session(self):
        # Set TF random seed to improve reproducibility
        self.rng = np.random.RandomState([2017, 8, 30])
        tf.set_random_seed(1234)

        # Create TF session
        self.sess = tf.Session(
            config=tf.ConfigProto(allow_soft_placement=True))

        # Object used to keep track of (and return) key accuracies
        if self.hparams.save:
            self.writer = tf.summary.FileWriter(self.hparams.save_dir,
                                                flush_secs=10)
        else:
            self.writer = None 

def main(argv):
    # Set TF random seed to improve reproducibility
    tf.set_random_seed(1234)

    input_shape = [FLAGS.batch_size, 224, 224, 3]
    x_src = tf.abs(tf.random_uniform(input_shape, 0., 1.))
    x_guide = tf.abs(tf.random_uniform(input_shape, 0., 1.))
    print("Input shape:")
    print(input_shape)

    model = make_imagenet_cnn(input_shape)
    attack = FastFeatureAdversaries(model)
    attack_params = {'eps': 0.3, 'clip_min': 0., 'clip_max': 1.,
                     'nb_iter': FLAGS.nb_iter, 'eps_iter': 0.01,
                     'layer': FLAGS.layer}
    x_adv = attack.generate(x_src, x_guide, **attack_params)
    h_adv = model.fprop(x_adv)[FLAGS.layer]
    h_src = model.fprop(x_src)[FLAGS.layer]
    h_guide = model.fprop(x_guide)[FLAGS.layer]

    with tf.Session() as sess:
        init = tf.global_variables_initializer()
        sess.run(init)
        ha, hs, hg, xa, xs, xg = sess.run(
            [h_adv, h_src, h_guide, x_adv, x_src, x_guide])

        print("L2 distance between source and adversarial example `%s`: %.4f" %
              (FLAGS.layer, ((hs-ha)*(hs-ha)).sum()))
        print("L2 distance between guide and adversarial example `%s`: %.4f" %
              (FLAGS.layer, ((hg-ha)*(hg-ha)).sum()))
        print("L2 distance between source and guide `%s`: %.4f" %
              (FLAGS.layer, ((hg-hs)*(hg-hs)).sum()))
        print("Maximum perturbation: %.4f" % np.abs((xa-xs)).max())
        print("Original features: ")
        print(hs[:10, :10])
        print("Adversarial features: ")
        print(ha[:10, :10])