def _define_step(self, name, batch_size, score, summary):
"""Combine operations of a phase.
Keeps track of the mean score and when to report it.
Args:
name: Name of the phase used for the score summary.
batch_size: Increment size of the global step.
score: Tensor holding the current scores.
summary: Tensor holding summary string to write if not an empty string.
Returns:
Tuple of summary tensor, mean score, and new global step. The mean score
is zero for non reporting steps.
"""
with tf.compat.v1.variable_scope('phase_{}'.format(name)):
score_mean = tools.StreamingMean((), tf.float32, 'score_mean')
score.set_shape((None, ))
with tf.control_dependencies([score, summary]):
submit_score = score_mean.submit(score)
with tf.control_dependencies([submit_score]):
mean_score = tf.cond(self._report, score_mean.clear, float)
summary = tf.cond(
self._report, lambda: tf.compat.v1.summary.merge([
summary,
tf.compat.v1.summary.scalar(
name + '/score', mean_score, family='trainer')
]), lambda: summary)
next_step = self._global_step.assign_add(batch_size)
with tf.control_dependencies([summary, mean_score, next_step]):
return (tf.identity(summary), tf.identity(mean_score),
tf.identity(next_step))
def print_metrics(metrics, step, every):
means, updates = [], []
for key, value in metrics:
mean = tools.StreamingMean((), tf.float32, 'mean_{}'.format(key))
means.append(mean)
updates.append(mean.submit(value))
with tf.control_dependencies(updates):
message = 'step/' + '/'.join(key for key, _ in metrics) + ' = '
gs = tf.train.get_or_create_global_step()
print_metrics = tf.cond(
tf.equal(step % every, 0),
lambda: tf.print(message, [gs] + [mean.clear() for mean in means]),
tf.no_op)
return print_metrics
def print_metrics(metrics, step, every):
means, updates = [], []
for key, value in metrics:
mean = tools.StreamingMean((), tf.float32, 'mean_{}'.format(key))
means.append(mean)
updates.append(mean.submit(value))
with tf.control_dependencies(updates):
# step/score/loss/zs_entropy/zs_divergence = [25000, -nan, 11413.0146, 37.0638504, 4.20143223]
message = 'step/' + '/'.join(key for key, _ in metrics) + ' = '
gs = tf.train.get_or_create_global_step()
print_metrics = tf.cond(
tf.equal(step % every, 0),
lambda: tf.print(message, [gs] + [mean.clear() for mean in means]),
tf.no_op)
return print_metrics
def print_metrics(metrics, step, every, name='metrics'):
means, updates = [], []
for key, value in metrics.items():
key = 'metrics_{}_{}'.format(name, key)
mean = tools.StreamingMean((), tf.float32, key)
means.append(mean)
updates.append(mean.submit(value))
with tf.control_dependencies(updates):
# message = 'step/' + '/'.join(metrics.keys()) + ' = '
message = '{}: step/{} ='.format(name, '/'.join(metrics.keys()))
gs = tf.train.get_or_create_global_step()
print_metrics = tf.cond(
tf.equal(step % every, 0),
lambda: tf.print(message, [gs] + [mean.clear() for mean in means]),
tf.no_op)
return print_metrics