Tensorflow, try and except doesn't handle exception

Question

I'm new to tensorflow and I've been stuck at an annoying problem here.

I'm making a program that loads image "raw data" taken with tf.WholeFileReader.read(image_name_queue) from a tfrecord file and then decodes it with tf.image.decode_jpeg(raw_data, channels=3) and then passes it through a function that vectorizes it.

main code

logging.info('setting up folder')
create_image_data_folder()
save_configs()

logging.info('creating graph')
filename_queue = tf.train.string_input_producer([
                                             configs.TFRECORD_IMAGES_PATH],
                                             num_epochs=1)

image_tensor, name_tensor = read_and_decode(filename_queue)
image_batch_tensor, name_batch_tensor = tf.train.shuffle_batch(
                                        [image_tensor, name_tensor],
                                        configs.BATCH_SIZE,
                                        1000 + 3 * configs.BATCH_SIZE,
                                        min_after_dequeue=1000)
image_embedding_batch_tensor = configs.IMAGE_EMBEDDING_FUNCTION(image_batch_tensor)

init = tf.initialize_all_variables()
init_local = tf.initialize_local_variables()
logging.info('starting session')
with tf.Session().as_default() as sess:
    sess.run(init)
    sess.run(init_local)
    tf.train.start_queue_runners()

    logging.info('vectorizing')
    data_points = []
    for _ in tqdm(xrange(get_n_batches())):
        name_batch = sess.run(name_batch_tensor)
        image_embedding_batch = sess.run(image_embedding_batch_tensor)
        for vector, name in zip(list(image_embedding_batch), name_batch):
            data_points.append((vector, name))

logging.info('saving')
save_pkl_file(data_points, 'vectors.pkl')

read_and_decode function

def read_and_decode(tfrecord_file_queue):
    logging.debug('reading image and decodes it from queue')
    reader = tf.TFRecordReader()
    _, serialized_example = reader.read(tfrecord_file_queue)
    features = tf.parse_single_example(serialized_example,
        features={
            'image': tf.FixedLenFeature([], tf.string),
            'name': tf.FixedLenFeature([], tf.string)
        }
    )
    image = process_image_data(features['image'])

    return image, features['name']

The code is working, but eventually it comes across a bad, non-jpeg file and an error this is raised and the program stops running.

error

InvalidArgumentError (see above for traceback): Invalid JPEG data, size 556663

I want to skip these "errors". I tried to surround the code with try and except.

new code

for _ in tqdm(xrange(get_n_batches())):
    try:
        name_batch = sess.run(name_batch_tensor)
        image_embedding_batch = sess.run(image_embedding_batch_tensor)
        for vector, name in zip(list(image_embedding_batch), name_batch):
            data_points.append((vector, name))
    except Exception as e:
        logging.warning('error occured: {}'.format(e))

When I run the program again the same error occurs, the try and except doesn't seem to handle the error.

How can I handle these exceptions? Also, if you see that I've misunderstood the tensorflow "structure" please mention that.

Answer 1

Well, to solve your issue use tf.data.experimental.ignore_errors available in v2.8.0

Use below transformation to produce a dataset that contains the same elements as the input, but silently drops any elements that caused an error. For example:

dataset = tf.data.Dataset.from_tensor_slices([1., 2., 0., 4.])

# Computing `tf.debugging.check_numerics(1. / 0.)` will raise an
InvalidArgumentError.
dataset = dataset.map(lambda x: tf.debugging.check_numerics(1. / x, "error"))

# Using `ignore_errors()` will drop the element that causes an error.
dataset =
    dataset.apply(tf.data.experimental.ignore_errors())  # ==> {1., 0.5, 0.2}

Refer documentation here

Answer 2

I know this is not applicable to your example, but I stumbled upon a different scenario in which TensorFlow didn't seem to catch the exception, despite doing

try:
    # Run code that throw tensorflow error
except:
    print('This won't catch the exception...')

The thing that made the problem so hard to solve was that TensorFlow debugging pointed to the wrong line; it showed me that the error lied in the graph construction, rather than graph execution.

The concrete problem?

I tried to restore a model from a .meta file:

try:
    saver = tf.train.import_meta_graph('my_error_generating_model.meta') # tf throws err here
    graph = tf.get_default_graph()
except:
    print('This won't run')

with tf.Session() as sess:
    # This is where error is actually generated
    saver.restore(sess, tf.train.latest_checkpoint('./'))
    sess.run(...) # Propagating through graph generates a problem

The solution is, of course, to put your try-catch wrapper around the executing code instead!