def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
# pylint: disable=unused-variable
model_dir = os.path.expanduser(FLAGS.model_dir)
translations_dir = os.path.expanduser(FLAGS.translations_dir)
source = os.path.expanduser(FLAGS.source)
tf.gfile.MakeDirs(translations_dir)
translated_base_file = os.path.join(translations_dir, FLAGS.problems)
# Copy flags.txt with the original time, so t2t-bleu can report correct
# relative time.
flags_path = os.path.join(translations_dir, FLAGS.problems + "-flags.txt")
if not os.path.exists(flags_path):
shutil.copy2(os.path.join(model_dir, "flags.txt"), flags_path)
locals_and_flags = {'FLAGS': FLAGS}
for model in bleu_hook.stepfiles_iterator(model_dir, FLAGS.wait_minutes,
FLAGS.min_steps):
tf.logging.info("Translating " + model.filename)
out_file = translated_base_file + "-" + str(model.steps)
locals_and_flags.update(locals())
if os.path.exists(out_file):
tf.logging.info(out_file + " already exists, so skipping it.")
else:
tf.logging.info("Translating " + out_file)
params = (
"--t2t_usr_dir={FLAGS.t2t_usr_dir} --output_dir={model_dir} "
"--data_dir={FLAGS.data_dir} --problems={FLAGS.problems} "
"--decode_hparams=beam_size={FLAGS.beam_size},alpha={FLAGS.alpha} "
"--model={FLAGS.model} --hparams_set={FLAGS.hparams_set} "
"--checkpoint_path={model.filename} --decode_from_file={source} "
"--decode_to_file={out_file}").format(**locals_and_flags)
command = FLAGS.decoder_command.format(**locals())
tf.logging.info("Running:\n" + command)
os.system(command)
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
# pylint: disable=unused-variable
model_dir = os.path.expanduser(FLAGS.model_dir)
translations_dir = os.path.expanduser(FLAGS.translations_dir)
source = os.path.expanduser(FLAGS.source)
tf.gfile.MakeDirs(translations_dir)
translated_base_file = os.path.join(translations_dir, FLAGS.problem)
# Copy flags.txt with the original time, so t2t-bleu can report correct
# relative time.
flags_path = os.path.join(translations_dir, FLAGS.problem + "-flags.txt")
if not os.path.exists(flags_path):
shutil.copy2(os.path.join(model_dir, "flags.txt"), flags_path)
locals_and_flags = {"FLAGS": FLAGS}
for model in bleu_hook.stepfiles_iterator(model_dir, FLAGS.wait_minutes,
FLAGS.min_steps):
tf.logging.info("Translating " + model.filename)
out_file = translated_base_file + "-" + str(model.steps)
locals_and_flags.update(locals())
if os.path.exists(out_file):
tf.logging.info(out_file + " already exists, so skipping it.")
else:
tf.logging.info("Translating " + out_file)
params = (
"--t2t_usr_dir={FLAGS.t2t_usr_dir} --output_dir={model_dir} "
"--data_dir={FLAGS.data_dir} --problem={FLAGS.problem} "
"--decode_hparams=beam_size={FLAGS.beam_size},alpha={FLAGS.alpha} "
"--model={FLAGS.model} --hparams_set={FLAGS.hparams_set} "
"--checkpoint_path={model.filename} --decode_from_file={source} "
"--decode_to_file={out_file} --keep_timestamp"
).format(**locals_and_flags)
command = FLAGS.decoder_command.format(**locals())
tf.logging.info("Running:\n" + command)
os.system(command)
def main(_):
tf.logging._handler.setFormatter( # pylint: disable=protected-access
logging.Formatter("%(asctime)s:" + logging.BASIC_FORMAT, None))
tf.logging.set_verbosity(tf.logging.INFO)
model_dir = os.path.expanduser(FLAGS.model_dir)
output_dir = os.path.expanduser(FLAGS.output_dir)
out_base_file = os.path.join(output_dir, "model.ckpt")
# Copy flags.txt with the original time, so t2t-bleu can report correct
# relative time.
tf.gfile.MakeDirs(FLAGS.output_dir)
if not os.path.exists(os.path.join(output_dir, "flags.txt")):
shutil.copy2(os.path.join(model_dir, "flags.txt"),
os.path.join(output_dir, "flags.txt"))
models_processed = 0
queue = deque()
for model in bleu_hook.stepfiles_iterator(model_dir, FLAGS.wait_minutes,
FLAGS.min_steps):
if models_processed == 0:
var_list = tf.contrib.framework.list_variables(model.filename)
avg_values = {}
for (name, shape) in var_list:
if not name.startswith("global_step"):
avg_values[name] = np.zeros(shape)
models_processed += 1
tf.logging.info("Loading [%d]: %s" %
(models_processed, model.filename))
reader = tf.contrib.framework.load_checkpoint(model.filename)
for name in avg_values:
avg_values[name] += reader.get_tensor(name) / FLAGS.n
queue.append(model)
if len(queue) < FLAGS.n:
continue
out_file = "%s-%d" % (out_base_file, model.steps)
tf_vars = []
tf.logging.info("Averaging %s" % (out_file))
for (name, value) in six.iteritems(avg_values):
# TODO(martinpopel): dtype=var_dtypes[name]
tf_vars.append(tf.get_variable(name, shape=value.shape))
placeholders = [
tf.placeholder(v.dtype, shape=v.shape) for v in tf_vars
]
assign_ops = [tf.assign(v, p) for (v, p) in zip(tf_vars, placeholders)]
global_step = tf.get_variable("global_step",
initializer=tf.constant(model.steps,
dtype=tf.int64),
trainable=False)
saver = tf.train.Saver(tf.global_variables())
tf.logging.info("Running session for %s" % (out_file))
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for p, assign_op, (name, value) in zip(placeholders, assign_ops,
six.iteritems(avg_values)):
sess.run(assign_op, {p: value})
tf.logging.info("Storing to %s" % out_file)
saver.save(sess, out_base_file, global_step=global_step)
os.utime(out_file + ".index", (model.mtime, model.mtime))
tf.reset_default_graph()
first_model = queue.popleft()
reader = tf.contrib.framework.load_checkpoint(first_model.filename)
for name in avg_values:
avg_values[name] -= reader.get_tensor(name) / FLAGS.n
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
if FLAGS.translation:
if FLAGS.translations_dir:
raise ValueError(
"Cannot specify both --translation and --translations_dir.")
if FLAGS.bleu_variant in ("uncased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(FLAGS.reference, FLAGS.translation,
case_sensitive=False)
print("BLEU_uncased = %6.2f" % bleu)
if FLAGS.bleu_variant in ("cased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(FLAGS.reference, FLAGS.translation,
case_sensitive=True)
print("BLEU_cased = %6.2f" % bleu)
return
if not FLAGS.translations_dir:
raise ValueError(
"Either --translation or --translations_dir must be specified.")
transl_dir = os.path.expanduser(FLAGS.translations_dir)
if not os.path.exists(transl_dir):
exit_time = time.time() + FLAGS.wait_minutes * 60
tf.logging.info("Translation dir %s does not exist, waiting till %s.",
transl_dir, time.asctime(time.localtime(exit_time)))
while not os.path.exists(transl_dir):
time.sleep(10)
if time.time() > exit_time:
raise ValueError("Translation dir %s does not exist" % transl_dir)
last_step_file = os.path.join(FLAGS.event_dir, "last_evaluated_step.txt")
if FLAGS.min_steps == -1:
if tf.gfile.Exists(last_step_file):
with open(last_step_file) as ls_file:
FLAGS.min_steps = int(ls_file.read())
else:
FLAGS.min_steps = 0
if FLAGS.report_zero is None:
FLAGS.report_zero = FLAGS.min_steps == 0
writer = tf.summary.FileWriter(FLAGS.event_dir)
for transl_file in bleu_hook.stepfiles_iterator(
transl_dir, FLAGS.wait_minutes, FLAGS.min_steps, path_suffix=""):
# report_zero handling must be inside the for-loop,
# so we are sure the transl_dir is already created.
if FLAGS.report_zero:
all_files = (os.path.join(transl_dir, f) for f in os.listdir(transl_dir))
start_time = min(
os.path.getmtime(f) for f in all_files if os.path.isfile(f))
values = []
if FLAGS.bleu_variant in ("uncased", "both"):
values.append(tf.Summary.Value(
tag="BLEU_uncased" + FLAGS.tag_suffix, simple_value=0))
if FLAGS.bleu_variant in ("cased", "both"):
values.append(tf.Summary.Value(
tag="BLEU_cased" + FLAGS.tag_suffix, simple_value=0))
writer.add_event(tf.summary.Event(summary=tf.Summary(value=values),
wall_time=start_time, step=0))
FLAGS.report_zero = False
filename = transl_file.filename
tf.logging.info("Evaluating " + filename)
values = []
if FLAGS.bleu_variant in ("uncased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(FLAGS.reference, filename,
case_sensitive=False)
values.append(tf.Summary.Value(tag="BLEU_uncased" + FLAGS.tag_suffix,
simple_value=bleu))
tf.logging.info("%s: BLEU_uncased = %6.2f" % (filename, bleu))
if FLAGS.bleu_variant in ("cased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(FLAGS.reference, filename,
case_sensitive=True)
values.append(tf.Summary.Value(tag="BLEU_cased" + FLAGS.tag_suffix,
simple_value=bleu))
tf.logging.info("%s: BLEU_cased = %6.2f" % (transl_file.filename, bleu))
writer.add_event(tf.summary.Event(
summary=tf.Summary(value=values),
wall_time=transl_file.mtime, step=transl_file.steps))
writer.flush()
with open(last_step_file, "w") as ls_file:
ls_file.write(str(transl_file.steps) + "\n")
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
if FLAGS.translation:
if FLAGS.translations_dir:
raise ValueError(
"Cannot specify both --translation and --translations_dir.")
if FLAGS.bleu_variant in ("uncased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(
FLAGS.reference, FLAGS.translation, case_sensitive=False)
print("BLEU_uncased = %6.2f" % bleu)
if FLAGS.bleu_variant in ("cased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(
FLAGS.reference, FLAGS.translation, case_sensitive=True)
print("BLEU_cased = %6.2f" % bleu)
return
if not FLAGS.translations_dir:
raise ValueError(
"Either --translation or --translations_dir must be specified.")
transl_dir = os.path.expanduser(FLAGS.translations_dir)
if not os.path.exists(transl_dir):
exit_time = time.time() + FLAGS.wait_minutes * 60
tf.logging.info("Translation dir %s does not exist, waiting till %s.",
transl_dir, time.asctime(time.localtime(exit_time)))
while not os.path.exists(transl_dir):
time.sleep(10)
if time.time() > exit_time:
raise ValueError("Translation dir %s does not exist" %
transl_dir)
last_step_file = os.path.join(FLAGS.event_dir, "last_evaluated_step.txt")
if FLAGS.min_steps == -1:
if tf.gfile.Exists(last_step_file):
with open(last_step_file) as ls_file:
FLAGS.min_steps = int(ls_file.read())
else:
FLAGS.min_steps = 0
if FLAGS.report_zero is None:
FLAGS.report_zero = FLAGS.min_steps == 0
writer = tf.summary.FileWriter(FLAGS.event_dir)
for transl_file in bleu_hook.stepfiles_iterator(transl_dir,
FLAGS.wait_minutes,
FLAGS.min_steps,
path_suffix=""):
# report_zero handling must be inside the for-loop,
# so we are sure the transl_dir is already created.
if FLAGS.report_zero:
all_files = (os.path.join(transl_dir, f)
for f in os.listdir(transl_dir))
start_time = min(
os.path.getmtime(f) for f in all_files if os.path.isfile(f))
values = []
if FLAGS.bleu_variant in ("uncased", "both"):
values.append(
tf.Summary.Value(tag="BLEU_uncased" + FLAGS.tag_suffix,
simple_value=0))
if FLAGS.bleu_variant in ("cased", "both"):
values.append(
tf.Summary.Value(tag="BLEU_cased" + FLAGS.tag_suffix,
simple_value=0))
writer.add_event(
tf.summary.Event(summary=tf.Summary(value=values),
wall_time=start_time,
step=0))
FLAGS.report_zero = False
filename = transl_file.filename
tf.logging.info("Evaluating " + filename)
values = []
if FLAGS.bleu_variant in ("uncased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(
FLAGS.reference, filename, case_sensitive=False)
values.append(
tf.Summary.Value(tag="BLEU_uncased" + FLAGS.tag_suffix,
simple_value=bleu))
tf.logging.info("%s: BLEU_uncased = %6.2f" % (filename, bleu))
if FLAGS.bleu_variant in ("cased", "both"):
bleu = 100 * bleu_hook.bleu_wrapper(
FLAGS.reference, filename, case_sensitive=True)
values.append(
tf.Summary.Value(tag="BLEU_cased" + FLAGS.tag_suffix,
simple_value=bleu))
tf.logging.info("%s: BLEU_cased = %6.2f" %
(transl_file.filename, bleu))
writer.add_event(
tf.summary.Event(summary=tf.Summary(value=values),
wall_time=transl_file.mtime,
step=transl_file.steps))
writer.flush()
with open(last_step_file, "w") as ls_file:
ls_file.write(str(transl_file.steps) + "\n")
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
model_dir = os.path.expanduser(FLAGS.model_dir)
output_dir = os.path.expanduser(FLAGS.output_dir)
out_base_file = os.path.join(output_dir, "model.ckpt")
# Copy flags.txt with the original time, so t2t-bleu can report correct
# relative time.
tf.gfile.MakeDirs(FLAGS.output_dir)
if (not os.path.exists(os.path.join(output_dir, "flags.txt")) and
os.path.exists(os.path.join(model_dir, "flags.txt"))):
shutil.copy2(os.path.join(model_dir, "flags.txt"),
os.path.join(output_dir, "flags.txt"))
models_processed = 0
queue = deque()
for model in bleu_hook.stepfiles_iterator(model_dir, FLAGS.wait_minutes,
FLAGS.min_steps):
if models_processed == 0:
var_list = tf.contrib.framework.list_variables(model.filename)
avg_values = {}
for (name, shape) in var_list:
if not name.startswith("global_step"):
avg_values[name] = np.zeros(shape)
models_processed += 1
tf.logging.info("Loading [%d]: %s" % (models_processed, model.filename))
reader = tf.contrib.framework.load_checkpoint(model.filename)
for name in avg_values:
avg_values[name] += reader.get_tensor(name) / FLAGS.n
queue.append(model)
if len(queue) < FLAGS.n:
continue
out_file = "%s-%d" % (out_base_file, model.steps)
tf_vars = []
tf.logging.info("Averaging %s" % (out_file))
for (name, value) in six.iteritems(avg_values):
# TODO(martinpopel): dtype=var_dtypes[name]
tf_vars.append(tf.get_variable(name, shape=value.shape))
placeholders = [tf.placeholder(v.dtype, shape=v.shape) for v in tf_vars]
assign_ops = [tf.assign(v, p) for (v, p) in zip(tf_vars, placeholders)]
global_step = tf.get_variable(
"global_step",
initializer=tf.constant(model.steps, dtype=tf.int64),
trainable=False)
saver = tf.train.Saver(tf.global_variables())
tf.logging.info("Running session for %s" % (out_file))
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for p, assign_op, (name, value) in zip(
placeholders, assign_ops, six.iteritems(avg_values)):
sess.run(assign_op, {p: value})
tf.logging.info("Storing to %s" % out_file)
saver.save(sess, out_base_file, global_step=global_step)
os.utime(out_file + ".index", (model.mtime, model.mtime))
tf.reset_default_graph()
first_model = queue.popleft()
reader = tf.contrib.framework.load_checkpoint(first_model.filename)
for name in avg_values:
avg_values[name] -= reader.get_tensor(name) / FLAGS.n
