train.py

import logging as log
import os, sys

import numpy as np

import reader
from flags import * # for FLAGS
from model import RNNLM

log.basicConfig(stream=sys.stderr, level=log.INFO,
    format='%(asctime)s [%(levelname)s]:%(message)s',  
    datefmt='%Y-%m-%d %H:%M:%S')                        

def main(_):
  train_data, dev_data, tok_id, id_tok = reader.get_train_data(
      FLAGS.train_data, FLAGS.dev_data, FLAGS.vocab_data, FLAGS.vocab_size)

  model = RNNLM(FLAGS, train_data, dev_data, id_tok)
  model.train()

  #predict_data, _, _ = reader.prepare_data(FLAGS.predict_data,
  #    FLAGS.vocab_data, FLAGS.vocab_size)
  #model.predict(predict_data)

if __name__ == '__main__':
  if FLAGS.output_mode == 'debug' or FLAGS.output_mode == 'verbose':
    log.getLogger().setLevel(log.DEBUG)
  elif FLAGS.output_mode == 'info':
    log.getLogger().setLevel(log.INFO)
  tf.app.run()