def test_get_all_study_summaries_with_no_trials(storage_mode: str) -> None:
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage)
summaries = get_all_study_summaries(study._storage)
summary = [s for s in summaries if s._study_id == study._study_id][0]
assert summary.study_name == study.study_name
assert summary.n_trials == 0
assert summary.datetime_start is None
def test_get_all_study_summaries(storage_mode: str) -> None:
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage)
study.optimize(Func(), n_trials=5)
summaries = get_all_study_summaries(study._storage)
summary = [s for s in summaries if s._study_id == study._study_id][0]
assert summary.study_name == study.study_name
assert summary.n_trials == 5
def test_get_all_study_summaries_with_no_trials(storage_mode, cache_mode):
# type: (str, bool) -> None
with StorageSupplier(storage_mode, cache_mode) as storage:
study = optuna.create_study(storage=storage)
summaries = optuna.get_all_study_summaries(study._storage)
summary = [s for s in summaries if s.study_id == study.study_id][0]
assert summary.study_name == study.study_name
assert summary.n_trials == 0
assert summary.datetime_start is None
def test_get_all_study_summaries(storage_mode, cache_mode):
# type: (str, bool) -> None
with StorageSupplier(storage_mode, cache_mode) as storage:
study = optuna.create_study(storage=storage)
study.optimize(Func(), n_trials=5)
summaries = optuna.get_all_study_summaries(study._storage)
summary = [s for s in summaries if s.study_id == study.study_id][0]
assert summary.study_name == study.study_name
assert summary.n_trials == 5
def take_action(self, parsed_args):
# type: (Namespace) -> Tuple[Tuple, Tuple[Tuple, ...]]
storage_url = _check_storage_url(self.app_args.storage)
summaries = optuna.get_all_study_summaries(storage=storage_url)
rows = []
for s in summaries:
start = (s.datetime_start.strftime(self._datetime_format)
if s.datetime_start is not None else None)
row = (s.study_name, s.direction.name, s.n_trials, start)
rows.append(row)
return self._study_list_header, tuple(rows)
def take_action(self,
parsed_args: Namespace) -> Tuple[Tuple, Tuple[Tuple, ...]]:
storage_url = _check_storage_url(self.app_args.storage)
summaries = optuna.get_all_study_summaries(storage=storage_url)
rows = []
for s in summaries:
start = (s.datetime_start.strftime(_DATETIME_FORMAT)
if s.datetime_start is not None else None)
row = (s.study_name, tuple(d.name for d in s.directions),
s.n_trials, start)
rows.append(row)
return self._study_list_header, tuple(rows)
def test_get_all_study_summaries(storage_mode: str,
include_best_trial: bool) -> None:
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage)
study.optimize(func, n_trials=5)
summaries = get_all_study_summaries(study._storage, include_best_trial)
summary = [s for s in summaries if s._study_id == study._study_id][0]
assert summary.study_name == study.study_name
assert summary.n_trials == 5
if include_best_trial:
assert summary.best_trial is not None
else:
assert summary.best_trial is None
def take_action(self, parsed_args: Namespace) -> None:
storage_url = _check_storage_url(self.app_args.storage)
summaries = optuna.get_all_study_summaries(storage=storage_url)
records = []
for s in summaries:
start = (s.datetime_start.strftime(_DATETIME_FORMAT)
if s.datetime_start is not None else None)
record: Dict[Tuple[str, str], Any] = {}
record[("name", "")] = s.study_name
record[("direction", "")] = tuple(d.name for d in s.directions)
record[("n_trials", "")] = s.n_trials
record[("datetime_start", "")] = start
records.append(record)
print(
_format_output(records, self._study_list_header,
parsed_args.format, parsed_args.flatten))
import optuna # pip install optuna optuna.create_study() optuna.load_study() optuna.delete_study() optuna.copy_study() optuna.get_all_study_summaries() optuna.TrialPruned trial = Trial(study, trial_id) trial.datetime_start trial.distributions trial.number trial.params trial.system_attrs trial.user_attrs trial.report(value, step) trial.set_system_attr(key, value) trial.set_user_attr(key, value) trial.should_prune() trial.suggest_categorical(name, choices) trial.suggest_discrete_uniform(name, low, high, q) trial.suggest_float(name, low, high, *, ?step, ?log) trial.suggest_int(name, low, high, ?step, ?log) trial.suggest_loguniform(name, low, high) trial.suggest_uniform(name, low, high) #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ import tensorflow as tf from tf import keras
import sklearn.model_selection
import optuna
study = optuna.create_study(storage='sqlite:///example.db')
study.set_user_attr('contributors', ['Akiba', 'Sano'])
study.set_user_attr('dataset', 'MNIST')
###################################################################################################
# We can access annotated attributes with :attr:`~optuna.study.Study.user_attr` property.
study.user_attrs # {'contributors': ['Akiba', 'Sano'], 'dataset': 'MNIST'}
###################################################################################################
# :class:`~optuna.struct.StudySummary` object, which can be retrieved by
# :func:`~optuna.study.get_all_study_summaries`, also contains user-defined attributes.
study_summaries = optuna.get_all_study_summaries('sqlite:///example.db')
study_summaries[
0].user_attrs # {'contributors': ['Akiba', 'Sano'], 'dataset': 'MNIST'}
###################################################################################################
# .. seealso::
# ``optuna study set-user-attr`` command, which sets an attribute via command line interface.
###################################################################################################
# Adding User Attributes to Trials
# --------------------------------
#
# As with :class:`~optuna.study.Study`, a :class:`~optuna.trial.Trial` object provides
# :func:`~optuna.trial.Trial.set_user_attr` method.
# Attributes are set inside an objective function.
import argparse
import json
import re
import sys
import optuna
parser = argparse.ArgumentParser()
parser.add_argument("storage", type=str)
parser.add_argument("target_study_name", type=str)
args = parser.parse_args()
study_names = [
study.study_name for study in optuna.get_all_study_summaries(args.storage)
if re.match(args.target_study_name, study.study_name)
]
print("Target studies: {}".format(study_names), file=sys.stderr)
trials = []
for study_name in study_names:
study = optuna.load_study(study_name=study_name, storage=args.storage)
for trial in study.trials:
values = [
v if v is None or d == optuna.study.StudyDirection.MINIMIZE else -v
for v, d in zip(trial.values, study.directions)
]
distributions = {
name: json.loads(optuna.distributions.distribution_to_json(d))
for name, d in trial.distributions.items()
}
# Create a study for single-objective optimization.
study = optuna.create_study(storage=args.storage_url, study_name="single")
study.set_system_attr("c", 2)
study.set_user_attr("d", 3)
study.optimize(objective_test_upgrade, n_trials=1)
# Create a study for multi-objective optimization.
try:
optuna.create_study(storage=args.storage_url,
study_name="multi_empty",
directions=["minimize", "minimize"])
study = optuna.create_study(storage=args.storage_url,
study_name="multi",
directions=["minimize", "minimize"])
study.set_system_attr("c", 2)
study.set_user_attr("d", 3)
study.optimize(mo_objective_test_upgrade, n_trials=1)
except TypeError:
print(
f"optuna=={optuna.__version__} does not support multi-objective optimization."
)
# Create a study for distributions upgrade.
if version.parse(optuna.__version__) >= version.parse("2.4.0"):
study = optuna.create_study(storage=args.storage_url,
study_name="schema migration")
study.optimize(objective_test_upgrade_distributions, n_trials=1)
for s in optuna.get_all_study_summaries(args.storage_url):
print(f"{s.study_name}, {s.n_trials}")
def _get_study_summary(self) -> List[optuna.study.StudySummary]:
if self.summary_cache is not None:
return self.summary_cache
summary = optuna.get_all_study_summaries(storage=self.database_url)
self.summary_cache = summary
return summary
def optuna_run(
db_name: str,
n_gpus: int,
n_workers_per_gpu: int,
n_trials_per_worker: int,
n_startup_trials: int,
objective_func: Callable,
):
db_file = Path(db_name) / f'{db_name}.db'
db_file.parent.mkdir(parents=True, exist_ok=True)
db_url = f'sqlite:///{db_file.as_posix()}'
if db_file.exists():
print(f'Studies in storage: {db_url}')
for study in optuna.get_all_study_summaries(db_url):
print(f' Study {study.study_name} with {study.n_trials} trials')
storage = optuna.storages.RDBStorage(url=db_url, )
study = optuna.create_study(
study_name='study',
storage=storage,
load_if_exists=True,
direction='maximize',
)
# FAIL any zombie trials that are stuck in `RUNNING` state
trials = storage.get_all_trials(study._study_id,
deepcopy=False,
states=(optuna.trial.TrialState.RUNNING, ))
for trial in trials:
print(f'Setting trial {trial.number} with state {trial.state} to FAIL')
status = storage.set_trial_state(trial._trial_id,
optuna.trial.TrialState.FAIL)
print(f'Success?: {status}')
# launch workers
n_workers = n_gpus * n_workers_per_gpu
with concurrent.futures.ProcessPoolExecutor(
max_workers=n_workers) as executor:
# with concurrent.futures.ThreadPoolExecutor(max_workers=n_workers) as executor:
futures = []
for i_worker in range(n_workers):
i_gpu = i_worker % n_gpus
print(
f'Launching worker {i_worker+1} (of {n_workers}) on gpu {i_gpu} and running {n_trials_per_worker} trials'
)
future = executor.submit(
worker, # callable that calls study.optimize()
db_url,
db_file.parent.as_posix(),
n_trials_per_worker,
i_gpu,
n_startup_trials,
objective_func,
)
futures.append(future)
time.sleep(0.1)
concurrent.futures.wait(futures)
for i_future, future in enumerate(futures):
if future.exception() is None:
print(f'Future {i_future} returned {future.result()}')
else:
print(f'Future {i_future} exception:')
print(future.exception())
