def write_stats(self, category: str, values: Dict[str, StatsSummary],
step: int) -> None:
for val, stats_summary in values.items():
set_gauge(
GaugeWriter.sanitize_string(f"{category}.{val}.mean"),
float(stats_summary.mean),
)
def _write_summary(self, step: int) -> None:
"""
Saves training statistics to Tensorboard.
"""
is_training = "Training." if self.should_still_train else "Not Training."
stats_summary = self.stats_reporter.get_stats_summaries(
"Environment/Cumulative Reward"
)
if stats_summary.num > 0:
LOGGER.info(
" {}: {}: Step: {}. "
"Time Elapsed: {:0.3f} s "
"Mean "
"Reward: {:0.3f}"
". Std of Reward: {:0.3f}. {}".format(
self.run_id,
self.brain_name,
step,
time.time() - self.training_start_time,
stats_summary.mean,
stats_summary.std,
is_training,
)
)
set_gauge(f"{self.brain_name}.mean_reward", stats_summary.mean)
else:
LOGGER.info(
" {}: {}: Step: {}. No episode was completed since last summary. {}".format(
self.run_id, self.brain_name, step, is_training
)
)
self.stats_reporter.write_stats(int(step))
def write_summary(self, global_step: int,
delta_train_start: float) -> None:
"""
Saves training statistics to Tensorboard.
:param delta_train_start: Time elapsed since training started.
:param global_step: The number of steps the simulation has been going for
"""
if (global_step % self.trainer_parameters["summary_freq"] == 0
and global_step != 0):
is_training = ("Training." if self.is_training
and self.get_step <= self.get_max_steps else
"Not Training.")
step = min(self.get_step, self.get_max_steps)
stats_summary = self.stats_reporter.get_stats_summaries(
"Environment/Cumulative Reward")
if stats_summary.num > 0:
LOGGER.info(" {}: {}: Step: {}. "
"Time Elapsed: {:0.3f} s "
"Mean "
"Reward: {:0.3f}"
". Std of Reward: {:0.3f}. {}".format(
self.run_id,
self.brain_name,
step,
delta_train_start,
stats_summary.mean,
stats_summary.std,
is_training,
))
set_gauge(f"{self.brain_name}.mean_reward", stats_summary.mean)
else:
LOGGER.info(
" {}: {}: Step: {}. No episode was completed since last summary. {}"
.format(self.run_id, self.brain_name, step, is_training))
self.stats_reporter.write_stats(int(step))
def write_summary(self,
global_step: int,
delta_train_start: float,
lesson_num: int = 0) -> None:
"""
Saves training statistics to Tensorboard.
:param delta_train_start: Time elapsed since training started.
:param lesson_num: Current lesson number in curriculum.
:param global_step: The number of steps the simulation has been going for
"""
if (global_step % self.trainer_parameters["summary_freq"] == 0
and global_step != 0):
is_training = ("Training." if self.is_training
and self.get_step <= self.get_max_steps else
"Not Training.")
step = min(self.get_step, self.get_max_steps)
if len(self.stats["Environment/Cumulative Reward"]) > 0:
mean_reward = np.mean(
self.stats["Environment/Cumulative Reward"])
LOGGER.info(
" {}: {}: Step: {}. "
"Time Elapsed: {:0.3f} s "
"Mean "
"Reward: {:0.3f}"
". Std of Reward: {:0.3f}. {}".format(
self.run_id,
self.brain_name,
step,
delta_train_start,
mean_reward,
np.std(self.stats["Environment/Cumulative Reward"]),
is_training,
))
set_gauge(f"{self.brain_name}.mean_reward", mean_reward)
else:
LOGGER.info(
" {}: {}: Step: {}. No episode was completed since last summary. {}"
.format(self.run_id, self.brain_name, step, is_training))
summary = tf.Summary()
for key in self.stats:
if len(self.stats[key]) > 0:
stat_mean = float(np.mean(self.stats[key]))
summary.value.add(tag="{}".format(key),
simple_value=stat_mean)
self.stats[key] = []
summary.value.add(tag="Environment/Lesson",
simple_value=lesson_num)
self.summary_writer.add_summary(summary, step)
self.summary_writer.flush()
def decorated_func(x: int = 0, y: float = 1.0) -> str:
timers.set_gauge("my_gauge", x + y)
return f"{x} + {y} = {x + y}"
