0bf56c5e8db53e152e7095087499d072018283e6,src/mjsynth.py,,threaded_input_pipeline,#,94

Before Change

        // Pad batches to max data size (bucketing it all into the same bucket)
        dataset = dataset.apply(tf.contrib.data.bucket_by_sequence_length
                                (element_length_func=_element_length_fn,
                                 bucket_batch_sizes=[batch_size, batch_size],
                                 bucket_boundaries=[0]))

        dataset = dataset.map(lambda image, 
                              width, label, 
                              length, text, filename: 
                              (image, width, 
                               tf.cast(tf.deserialize_many_sparse(label, tf.int64), 
                                       tf.int32),
                               length, text, filename))
        num_epochs = None
        // Repeat for num_epochs
        dataset = dataset.repeat(num_epochs)

    return dataset

def _element_length_fn(image, width, label, length, text, filename):

After Change

        // Hack -- probably a better way to do this! Just want dynamic padding!
        // Pad batches to max data size (bucketing it all into the same bucket)
        dataset = dataset.apply(tf.contrib.data.bucket_by_sequence_length
                                (element_length_func=_element_length_fn,
                                 bucket_batch_sizes=[0, 10],
                                 bucket_boundaries=[0]))

        dataset = dataset.map(lambda image, 
                              width, label, 
                              length, text, filename: 

Instances