f72e30b86039b53dcced080dc999e1a9fb052c49,src/mjsynth.py,,bucketed_input_pipeline,#,28

Before Change

        keep_input = _get_input_filter(width, width_threshold,
                                       length, length_threshold)
        data_tuple = [image, label, length, text, filename]
        width,data_tuple = tf.contrib.training.bucket_by_sequence_length(
            input_length=width,
            tensors=data_tuple,
            bucket_boundaries=boundaries,

After Change

                              (element, width_threshold, length_threshold, training, data_tuples), 
                              num_parallel_calls=num_threads)
        
        dataset = dataset.filter(lambda image, 
                                 width, 
                                 label, 
                                 length, 
                                 text, 
                                 filename, 
                                 get_input: 
                                 get_input != None)

        dataset = dataset.apply(tf.contrib.data.bucket_by_sequence_length
                                (element_length_func=_element_length_fn,
                                 bucket_batch_sizes=np.full
                                 (len(boundaries) + 1, batch_size),
                                 bucket_boundaries=boundaries))

        iterator = dataset.make_one_shot_iterator()
        image, width, label, length, text, filename, get_input = iterator.get_next()

        label = tf.deserialize_many_sparse(label, tf.int64) // post-batching...
        label = tf.cast(label, tf.int32) // for ctc_loss

Instances