def test_parity_problem():
main_loop = main(20, 1, 10, 10, 1)
assert main_loop.log.status['epochs_done'] == 1
assert main_loop.log.status['iterations_done'] == 10
"""This example shows how to train a simple RNN for the sequence classification
task: given a sequence of 0s and 1s, determine whether number of 1s in it
is odd or even
"""
import argparse
import logging
from parity_problem import main
if __name__ == "__main__":
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s: %(name)s: %(levelname)s: %(message)s")
parser = argparse.ArgumentParser("This example trains a simple LSTM "
"network that determines if number of 1s "
"in a given sequence of 0s and 1s is odd "
"or even")
parser.add_argument("--max-seq-length", type=int, default=100,
help="Max length of sequence to learn.")
parser.add_argument("--lstm-dim", type=int, default=1,
help="Number of hidden units in the LSTM layer.")
parser.add_argument("--batch-size", type=int, default=10,
help="Number of examples in a single batch.")
parser.add_argument("--num-batches", type=int, default=1000,
help="Number of batches in the training dataset.")
parser.add_argument("--num-epochs", type=int, default=30,
help="Number of epochs to do.")
args = parser.parse_args()
main(**vars(args))
