def testTrain(self):
"""Test the training loop is functional with basic hparams."""
nmt_parser = argparse.ArgumentParser()
nmt.add_arguments(nmt_parser)
flags, _ = nmt_parser.parse_known_args()
update_flags(flags, "nmt_train_test")
default_hparams = nmt.create_hparams(flags)
nmt.run_main(flags, default_hparams, estimator.train_and_eval_fn)
def testTrain(self):
"""Test the training loop is functional with basic hparams."""
nmt_parser = argparse.ArgumentParser()
nmt.add_arguments(nmt_parser)
FLAGS, unparsed = nmt_parser.parse_known_args()
_update_flags(FLAGS, "nmt_train_test")
default_hparams = nmt.create_hparams(FLAGS)
train_fn = train.train
nmt.run_main(FLAGS, default_hparams, train_fn, None)
def testTrainInputFn(self):
nmt_parser = argparse.ArgumentParser()
nmt.add_arguments(nmt_parser)
flags, _ = nmt_parser.parse_known_args()
update_flags(flags, "input_fn_test")
default_hparams = nmt.create_hparams(flags)
hparams = nmt.extend_hparams(default_hparams)
with self.test_session() as sess:
input_fn = make_input_fn(hparams, tf.contrib.learn.ModeKeys.TRAIN)
outputs = input_fn({})
sess.run(tf.tables_initializer())
iterator = outputs.make_initializable_iterator()
sess.run(iterator.initializer)
features = sess.run(iterator.get_next())
tf.logging.info("source: %s", features["source"])
tf.logging.info("target_input: %s", features["target_input"])
tf.logging.info("target_output: %s", features["target_output"])
tf.logging.info("source_sequence_length: %s",
features["source_sequence_length"])
tf.logging.info("target_sequence_length: %s",
features["target_sequence_length"])
def testInference(self):
"""Test inference is function with basic hparams."""
nmt_parser = argparse.ArgumentParser()
nmt.add_arguments(nmt_parser)
FLAGS, unparsed = nmt_parser.parse_known_args()
_update_flags(FLAGS, "nmt_train_infer")
# Train one step so we have a checkpoint.
FLAGS.num_train_steps = 1
default_hparams = nmt.create_hparams(FLAGS)
train_fn = train.train
nmt.run_main(FLAGS, default_hparams, train_fn, None)
# Update FLAGS for inference.
FLAGS.inference_input_file = ("nmt/testdata/" "iwslt15.tst2013.100.en")
FLAGS.inference_output_file = os.path.join(FLAGS.out_dir, "output")
FLAGS.inference_ref_file = ("nmt/testdata/" "iwslt15.tst2013.100.vi")
default_hparams = nmt.create_hparams(FLAGS)
inference_fn = inference.inference
nmt.run_main(FLAGS, default_hparams, None, inference_fn)
# Once in a while, we print statistics.
if global_step - last_stats_step >= steps_per_stats:
last_stats_step = global_step
is_overflow = train.process_stats(stats, info, global_step,
steps_per_stats, log_f)
train.print_step_info(" ", global_step, info,
train._get_best_results(hparams),
log_f)
if is_overflow:
break
# Reset statistics
stats = train.init_stats()
sess, num_workers, worker_id, num_replicas_per_worker = \
parallax.parallel_run(train_model.graph,
FLAGS.resource_info_file,
sync=FLAGS.sync,
parallax_config=parallax_config.build_config())
run(sess, num_workers, worker_id, num_replicas_per_worker)
if __name__ == "__main__":
import logging
logging.getLogger("tensorflow").setLevel(logging.DEBUG)
nmt_parser = argparse.ArgumentParser()
nmt.add_arguments(nmt_parser)
add_arguments(nmt_parser)
FLAGS, unparsed = nmt_parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
