def test_save_updated_trial(self):
experiment, _ = self.init_experiment_and_generation_strategy(
save_generation_strategy=False
)
exp = _load_experiment(
experiment.name, decoder=self.with_db_settings.db_settings.decoder
)
trial = exp.new_trial()
_save_or_update_trials(
experiment=experiment,
trials=[trial],
encoder=self.with_db_settings.db_settings.encoder,
decoder=self.with_db_settings.db_settings.decoder,
)
self.assertEqual(trial.status, TrialStatus.CANDIDATE)
trial.mark_running(True)
saved = self.with_db_settings._save_or_update_trial_in_db_if_possible(
exp, trial
)
self.assertTrue(saved)
exp = _load_experiment(
experiment.name, decoder=self.with_db_settings.db_settings.decoder
)
self.assertEqual(len(exp.trials), 1)
self.assertEqual(exp.trials[0].status, TrialStatus.RUNNING)
def _save_updated_trials_to_db_if_possible(
self,
experiment: Experiment,
trials: List[BaseTrial],
suppress_all_errors: bool = False,
) -> bool:
"""Saves updated trials on given experiment if DB settings are set on this
`WithDBSettingsBase` instance.
Args:
experiment: Experiment, on which to save updated trials in DB.
trials: Newly updated trials to save.
suppress_all_errors: Flag for `retry_on_exception` that makes
the decorator suppress the thrown exception even if it
occurred in all the retries (exception is still logged).
Returns:
bool: Whether the trials were saved.
"""
if self.db_settings_set:
start_time = time.time()
_save_or_update_trials(experiment=experiment,
trials=trials,
encoder=self.db_settings.encoder)
logger.debug(
f"Updated trials {[trial.index for trial in trials]} in "
f"{_round_floats_for_logging(time.time() - start_time)} seconds."
)
return True
return False
def _save_or_update_trials_in_db_if_possible(
experiment: Experiment,
trials: List[BaseTrial],
encoder: Encoder,
decoder: Decoder,
suppress_all_errors: bool,
) -> None:
start_time = time.time()
_save_or_update_trials(
experiment=experiment,
trials=trials,
encoder=encoder,
decoder=decoder,
batch_size=STORAGE_MINI_BATCH_SIZE,
)
logger.debug(
f"Saved or updated trials {[trial.index for trial in trials]} in "
f"{_round_floats_for_logging(time.time() - start_time)} seconds "
f"in mini-batches of {STORAGE_MINI_BATCH_SIZE}.")
