example usage (only compatible with python 2.x):
python convert_to_records.py
--train_directory=/Users/dgu/Documents/projects/tensorflow/rnn/data
--validation_directory=/Users/dgu/Documents/projects/tensorflow/rnn/data
--output_directory=/Users/dgu/Documents/projects/tensorflow/rnn/data
--labels_file=/Users/dgu/Documents/projects/tensorflow/rnn/data/label
origin file link: https://github.com/tensorflow/models/blob/master/inception/inception/data/build_image_data.py
reference link for how to write data to tfrecord and read data from tfrecord: https://www.tensorflow.org/versions/r0.10/how_tos/reading_data/index.html#file-formats
python -B main_train.py --data_dir=/Users/dgu/Documents/projects/tensorflow/rnn/data/sharded_data
This command will run the training subset by default