def _preprocess(self, trial: optuna.trial.Trial) -> None:
if self.pbar is not None:
self.pbar.set_description(self.pbar_fmt.format(self.step_name, self.best_score))
if "lambda_l1" in self.target_param_names:
self.lgbm_params["lambda_l1"] = trial.suggest_float("lambda_l1", 1e-8, 10.0, log=True)
if "lambda_l2" in self.target_param_names:
self.lgbm_params["lambda_l2"] = trial.suggest_float("lambda_l2", 1e-8, 10.0, log=True)
if "num_leaves" in self.target_param_names:
tree_depth = self.lgbm_params.get("max_depth", _DEFAULT_TUNER_TREE_DEPTH)
max_num_leaves = 2 ** tree_depth if tree_depth > 0 else 2 ** _DEFAULT_TUNER_TREE_DEPTH
self.lgbm_params["num_leaves"] = trial.suggest_int("num_leaves", 2, max_num_leaves)
if "feature_fraction" in self.target_param_names:
# `GridSampler` is used for sampling feature_fraction value.
# The value 1.0 for the hyperparameter is always sampled.
param_value = min(trial.suggest_float("feature_fraction", 0.4, 1.0 + _EPS), 1.0)
self.lgbm_params["feature_fraction"] = param_value
if "bagging_fraction" in self.target_param_names:
# `TPESampler` is used for sampling bagging_fraction value.
# The value 1.0 for the hyperparameter might by sampled.
param_value = min(trial.suggest_float("bagging_fraction", 0.4, 1.0 + _EPS), 1.0)
self.lgbm_params["bagging_fraction"] = param_value
if "bagging_freq" in self.target_param_names:
self.lgbm_params["bagging_freq"] = trial.suggest_int("bagging_freq", 1, 7)
if "min_child_samples" in self.target_param_names:
# `GridSampler` is used for sampling min_child_samples value.
# The value 1.0 for the hyperparameter is always sampled.
param_value = trial.suggest_int("min_child_samples", 5, 100)
self.lgbm_params["min_child_samples"] = param_value
def __call__(self, trial: optuna.trial.Trial) -> float:
data = TreeDataModule(
self._filename,
batch_size=trial.suggest_int("batch_size", 32, 160, 32),
)
kwargs = {
"lstm_size":
trial.suggest_categorical("lstm_size", [512, 1024, 2048]),
"dropout_prob":
trial.suggest_float("dropout", 0.1, 0.5, step=0.1),
"learning_rate":
trial.suggest_float("lr", 1e-3, 1e-1, log=True),
"weight_decay":
trial.suggest_float("weight_decay", 1e-3, 1e-1, log=True),
}
model = RouteDistanceModel(**kwargs)
gpus = int(torch.cuda.is_available())
pruning_callback = PyTorchLightningPruningCallback(
trial, monitor="val_monitor")
trainer = Trainer(
gpus=gpus,
logger=True, # become a tensorboard logger
checkpoint_callback=False,
callbacks=[pruning_callback], # type: ignore
max_epochs=EPOCHS,
)
trainer.fit(model, datamodule=data)
return trainer.callback_metrics["val_monitor"].item()
def objective2(trial: optuna.trial.Trial) -> float:
p1 = trial.suggest_float("p1", 50, 100,
log=True) # The range has been changed
p3 = trial.suggest_float("p3", 0, 9, step=3)
p5 = trial.suggest_float("p5", 0, 1)
return p1 + p3 + p5
def _objective_func(trial: optuna.trial.Trial) -> float:
x = trial.suggest_float("x", -1.0, 1.0)
y = trial.suggest_float("y", 20, 30, log=True)
z = trial.suggest_categorical("z", (-1.0, 1.0))
assert isinstance(z, float)
trial.set_user_attr("my_user_attr", "my_user_attr_value")
return (x - 2)**2 + (y - 25)**2 + z
def _multiobjective_func(trial: optuna.trial.Trial) -> Tuple[float, float]:
x = trial.suggest_float("x", low=-10, high=10)
y = trial.suggest_float("y", low=1, high=10, log=True)
first_objective = (x - 2)**2 + (y - 25)**2
second_objective = (x - 2)**3 + (y - 25)**3
return first_objective, second_objective
def objective0(trial: optuna.trial.Trial) -> float:
p0 = trial.suggest_float("p0", 0, 10)
p1 = trial.suggest_float("p1", 1, 10, log=True)
p2 = trial.suggest_int("p2", 0, 10)
p3 = trial.suggest_float("p3", 0, 9, step=3)
p4 = trial.suggest_categorical("p4", ["10", "20", "30"])
assert isinstance(p4, str)
return p0 + p1 + p2 + p3 + int(p4)
def objective_test_upgrade_distributions(trial: optuna.trial.Trial) -> float:
x1 = trial.suggest_float("x1", -5, 5)
x2 = trial.suggest_float("x2", 1e-5, 1e-3, log=True)
x3 = trial.suggest_float("x3", -6, 6, step=2)
y1 = trial.suggest_int("y1", 0, 10)
y2 = trial.suggest_int("y2", 1, 20, log=True)
y3 = trial.suggest_int("y3", 5, 15, step=3)
z = cast(float, trial.suggest_categorical("z", [-5, 0, 5]))
return x1**2 + x2**2 + x3**2 + y1**2 + y2**2 + y3**2 + z**2
def _objective_func_long_user_attr(trial: optuna.trial.Trial) -> float:
x = trial.suggest_float("x", -1.0, 1.0)
y = trial.suggest_float("y", 20, 30, log=True)
z = trial.suggest_categorical("z", (-1.0, 1.0))
assert isinstance(z, float)
long_str = str(list(range(5000)))
trial.set_user_attr("my_user_attr", long_str)
return (x - 2)**2 + (y - 25)**2 + z
def objective1(trial: optuna.trial.Trial) -> float:
# p0, p2 and p4 are deleted.
p1 = trial.suggest_float("p1", 1, 10, log=True)
p3 = trial.suggest_float("p3", 0, 9, step=3)
# p5 is added.
p5 = trial.suggest_float("p5", 0, 1)
return p1 + p3 + p5
def _multiobjective_func(trial: optuna.trial.Trial) -> Tuple[float, float]:
x = trial.suggest_float("x", low=-10, high=10)
y = trial.suggest_float("y", low=1, high=10, log=True)
z = trial.suggest_categorical("z", (-1.0, 1.0))
assert isinstance(z, float)
first_objective = (x - 2)**2 + (y - 25)**2 + z
second_objective = (x - 2)**3 + (y - 25)**3 - z
return first_objective, second_objective
def _objective_func(trial: optuna.trial.Trial) -> float:
u = trial.suggest_int("u", 0, 10, step=2)
v = trial.suggest_int("v", 1, 10, log=True)
w = trial.suggest_float("w", -1.0, 1.0, step=0.1)
x = trial.suggest_float("x", -1.0, 1.0)
y = trial.suggest_float("y", 20.0, 30.0, log=True)
z = trial.suggest_categorical("z", (-1.0, 1.0))
assert isinstance(z, float)
trial.set_user_attr("my_user_attr", "my_user_attr_value")
return u + v + w + (x - 2) ** 2 + (y - 25) ** 2 + z
def objective(trial: optuna.trial.Trial) -> Tuple[float, float]:
p0 = trial.suggest_float("p0", -10, 10)
p1 = trial.suggest_uniform("p1", 3, 5)
p2 = trial.suggest_loguniform("p2", 0.00001, 0.1)
p3 = trial.suggest_discrete_uniform("p3", 100, 200, q=5)
p4 = trial.suggest_int("p4", -20, -15)
p5 = cast(int, trial.suggest_categorical("p5", [7, 1, 100]))
p6 = trial.suggest_float("p6", -10, 10, step=1.0)
p7 = trial.suggest_int("p7", 1, 7, log=True)
return (
p0 + p1 + p2,
p3 + p4 + p5 + p6 + p7,
)
def _convert(trial: optuna.trial.Trial,
template: TuningParametersTemplate) -> Dict[str, Any]:
result: Dict[str, Any] = {}
for k, v in template.params_dict.items():
if isinstance(v, RandInt):
if v.log and v.q is not None:
value = trial.suggest_float(name=k, low=0, high=1.0)
result[k] = uniform_to_integers(
value,
low=v.low,
high=v.high,
q=v.q, # type: ignore
log=True,
include_high=v.include_high,
)
else:
_high: Any = v.high if v.include_high else v.high - 1
result[k] = trial.suggest_int(name=k,
low=v.low,
high=_high,
step=v.q,
log=v.log)
elif isinstance(v, Rand):
if v.log and v.q is not None:
value = trial.suggest_float(name=k, low=0, high=1.0)
result[k] = uniform_to_discrete(
value,
low=v.low,
high=v.high,
q=v.q,
log=True,
include_high=v.include_high,
)
else:
_high = v.high
if v.q is not None and not v.include_high:
_high -= _IGNORABLE_ERROR
result[k] = trial.suggest_float(name=k,
low=v.low,
high=_high,
step=v.q,
log=v.log)
elif isinstance(v, TransitionChoice):
result[k] = v.values[trial.suggest_int(name=k,
low=0,
high=len(v.values) - 1)]
elif isinstance(v, Choice):
result[k] = trial.suggest_categorical(name=k, choices=v.values)
else: # pragma: no cover
raise NotImplementedError
return result
def _objective(self, trial: optuna.trial.Trial):
cv = trial.suggest_int('cv', 2, 2**4)
opt_params = dict(
max_depth=trial.suggest_int("max_depth", 2, 2**4),
learning_rate=trial.suggest_float('learning_rate',
0.001,
1,
step=0.001),
# n_estimators=trial.suggest_int("n_estimators", 2, 2 ** 10, log=True),
gamma=trial.suggest_float('gamma', 1e-8, 1, log=True),
min_child_weight=trial.suggest_float('min_child_weight',
1e-8,
2**10,
log=True),
subsample=trial.suggest_float('subsample', 0.1, 1),
colsample_bytree=trial.suggest_float('colsample_bytree', 0.1, 1),
colsample_bylevel=trial.suggest_float('colsample_bylevel', 0.1, 1),
reg_alpha=trial.suggest_float('reg_alpha', 1e-8, 10, log=True),
reg_lambda=trial.suggest_float('reg_lambda', 1e-8, 10, log=True),
)
if self.params is not None:
opt_params.update(self.params)
clf_oof = XGBClassifierOOF(self.X,
self.y,
params=opt_params,
cv=cv,
feval=self.feval)
clf_oof.run()
return clf_oof.oof_score # todo: f1
def objective(trial: optuna.trial.Trial) -> Tuple[float, float]:
x = trial.suggest_float("x", 0, 5)
if trial._trial_id == n:
assert n in sampler._split_cache
assert n in sampler._weights_below
else:
assert n not in sampler._split_cache
assert n not in sampler._weights_below
y = trial.suggest_float("y", 0, 3)
v0 = 4 * x ** 2 + 4 * y ** 2
v1 = (x - 5) ** 2 + (y - 5) ** 2
return v0, v1
def objective(trial: optuna.trial.Trial) -> float:
# We optimize the number of layers, hidden units in each layer and dropouts.
n_layers = trial.suggest_int("n_layers", 1, 3)
dropout = trial.suggest_float("dropout", 0.2, 0.5)
output_dims = [
trial.suggest_int("n_units_l{}".format(i), 4, 128, log=True)
for i in range(n_layers)
]
model = LightningNet(dropout, output_dims)
datamodule = MNISTDataModule(data_dir=DIR, batch_size=BATCHSIZE)
trainer = pl.Trainer(
logger=True,
limit_val_batches=PERCENT_VALID_EXAMPLES,
checkpoint_callback=False,
max_epochs=EPOCHS,
gpus=-1 if torch.cuda.is_available() else None,
callbacks=[PyTorchLightningPruningCallback(trial, monitor="val_acc")],
)
hyperparameters = dict(n_layers=n_layers,
dropout=dropout,
output_dims=output_dims)
trainer.logger.log_hyperparams(hyperparameters)
trainer.fit(model, datamodule=datamodule)
return trainer.callback_metrics["val_acc"].item()
def _suggest(self, trial: optuna.trial.Trial) -> Suggestion:
suggestions: Suggestion = dict()
for name, config in self.api_config.items():
low, high = config["range"]
log = config["space"] == "log"
if config["space"] == "logit":
assert 0 < low <= high < 1
low = np.log(low / (1 - low))
high = np.log(high / (1 - high))
if config["type"] == "real":
param = trial.suggest_float(name, low, high, log=log)
elif config["type"] == "int":
param = trial.suggest_int(name, low, high, log=log)
else:
# TODO(xadrianzetx) Support `suggest_categorical` if benchmark is extended.
raise RuntimeError("CategoricalDistribution is not supported in bayesmark.")
suggestions[name] = param if config["space"] != "logit" else 1 / (1 + np.exp(-param))
return suggestions
def tcn(config: BaseConfig, trial: optuna.trial.Trial) -> pl.LightningModule:
"""Returns a tunable PyTorch lightning tcn module.
Args:
config (BaseConfig): the hard-coded configuration.
trial (optuna.Trial): optuna trial.
Returns:
pl.LightningModule: a lightning module.
"""
training_config = get_training_config(
lr=trial.suggest_loguniform('lr', 1e-3, 1e-0),
weight_decay=trial.suggest_loguniform('weight_decay', 1e-5, 1e-1),
max_epochs=config.MAX_EPOCHS)
tcn = TemporalConvNet(training_config=training_config,
num_inputs=config.NUM_INPUTS,
num_outputs=config.NUM_OUTPUTS,
num_hidden=trial.suggest_int('num_hidden', 1, 4),
kernel_size=trial.suggest_int('kernel_size', 2, 4),
num_layers=trial.suggest_int('num_layers', 1, 2),
dropout=trial.suggest_float('dropout', 0.1, 0.3))
return tcn
def mo_objective_test_upgrade(trial: optuna.trial.Trial) -> Tuple[float, float]:
x = trial.suggest_float("x", -5, 5)
y = trial.suggest_int("y", 0, 10)
z = cast(float, trial.suggest_categorical("z", [-5, 0, 5]))
trial.set_system_attr("a", 0)
trial.set_user_attr("b", 1)
return x, x ** 2 + y ** 2 + z ** 2
def objective(trial: optuna.trial.Trial, value: float) -> float:
trial.set_system_attr(
optuna.integration._lightgbm_tuner.optimize._STEP_NAME_KEY,
"step{:.0f}".format(value),
)
return trial.suggest_float("x", value, value)
def objective(trial: optuna.trial.Trial) -> float:
for i in range(N_REPORTS):
trial.report(i, step=i)
x = trial.suggest_float("x", -100, 100)
y = trial.suggest_int("y", -100, 100)
return x ** 2 + y ** 2
def feedforward(config: BaseConfig,
trial: optuna.trial.Trial) -> pl.LightningModule:
"""Returns a tunable PyTorch lightning feedforward module.
Args:
config (BaseConfig): the hard-coded configuration.
trial (optuna.Trial): optuna trial.
Returns:
pl.LightningModule: a lightning module.
"""
model = FeedForward(num_inputs=config.NUM_INPUTS,
num_outputs=config.NUM_OUTPUTS,
num_hidden=trial.suggest_int('num_hidden', 1, 4),
num_layers=trial.suggest_int('num_layers', 1, 2),
dropout=trial.suggest_float('dropout', 0.0, 0.5),
activation=trial.suggest_categorical(
'activation', ['relu', 'none']))
training_config = get_training_config(
lr=trial.suggest_loguniform('lr', 1e-3, 1e-0),
weight_decay=trial.suggest_loguniform('weight_decay', 1e-5, 1e-1),
max_epochs=config.MAX_EPOCHS)
pl_model = TemporalConvNet(training_config=training_config,
lr=trial.suggest_loguniform('lr', 1e-3, 1e-0),
weight_decay=trial.suggest_loguniform(
'weight_decay', 1e-5, 1e-1),
max_epochs=config.MAX_EPOCHS)
return pl_model
def objective(trial: optuna.trial.Trial) -> float:
num_units = trial.suggest_int("NUM_UNITS", 16, 32)
dropout_rate = trial.suggest_float("DROPOUT_RATE", 0.1, 0.2)
optimizer = trial.suggest_categorical("OPTIMIZER", ["sgd", "adam"])
accuracy = train_test_model(num_units, dropout_rate, optimizer) # type: ignore
return accuracy
def objective_test_upgrade(trial: optuna.trial.Trial) -> float:
x = trial.suggest_float("x", -5, 5) # optuna==0.9.0 does not have suggest_float.
y = trial.suggest_int("y", 0, 10)
z = cast(float, trial.suggest_categorical("z", [-5, 0, 5]))
trial.set_system_attr("a", 0)
trial.set_user_attr("b", 1)
trial.report(0.5, step=0)
return x ** 2 + y ** 2 + z ** 2
def _objective_func(trial: optuna.trial.Trial) -> float:
x = trial.suggest_uniform("x", -1.0, 1.0)
y = trial.suggest_loguniform("y", 20.0, 30.0)
z = trial.suggest_categorical("z", (-1.0, 1.0))
w = trial.suggest_float("w", -1.0, 1.0, step=0.1)
assert isinstance(z, float)
trial.set_user_attr("my_user_attr", "my_user_attr_value")
return (x - 2)**2 + (y - 25)**2 + z + w
def __call__(self, trial: optuna.trial.Trial):
classifier_name = trial.suggest_categorical("classifier", ["SVC", "RandomForest"])
if classifier_name == "SVC":
svc_c = trial.suggest_float("svc_c", 1e-10, 1e10, log=True)
classifier_obj = sklearn.svm.SVC(C=svc_c, gamma="auto")
else:
rf_max_depth = trial.suggest_int("rf_max_depth", 2, 32, log=True)
classifier_obj = ensemble.RandomForestClassifier(
max_depth=rf_max_depth, n_estimators=10
)
score = cross_val_score(classifier_obj, self.X, self.y, n_jobs=1, cv=self.cv)
accuracy = score.mean()
return accuracy
def _objective(self, trial: optuna.trial.Trial):
cv = trial.suggest_int('cv', 2, 2**4)
opt_params = dict(
objective=trial.suggest_categorical("objective",
["Logloss", "CrossEntropy"]),
boosting_type=trial.suggest_categorical("boosting_type",
["Ordered", "Plain"]),
bootstrap_type=trial.suggest_categorical(
"bootstrap_type", ["Bayesian", "Bernoulli", "MVS"]),
# used_ram_limit="3gb",
max_depth=trial.suggest_int("max_depth", 2, 2**4),
learning_rate=trial.suggest_discrete_uniform(
'learning_rate', 0.001, 1, 0.001),
# n_estimators=trial.suggest_int("n_estimators", 2, 2 ** 10, log=True),
colsample_bylevel=trial.suggest_float("colsample_bylevel", 0.01,
0.1),
reg_lambda=trial.suggest_float("reg_lambda", 1e-8, 100, log=True))
if opt_params["bootstrap_type"] == "Bayesian":
opt_params["bagging_temperature"] = trial.suggest_float(
"bagging_temperature", 0, 10)
elif opt_params["bootstrap_type"] == "Bernoulli":
opt_params["subsample"] = trial.suggest_float("subsample", 0.1, 1)
if self.params is not None:
opt_params.update(self.params)
clf_oof = CatBoostClassifierOOF(self.X,
self.y,
params=opt_params,
cv=cv,
feval=self.feval)
clf_oof.run()
return clf_oof.oof_score # todo: f1
def _objective(self, trial: optuna.trial.Trial):
cv = trial.suggest_int('cv', 2, 2**4)
opt_params = dict(
num_leaves=trial.suggest_int("num_leaves", 2, 2**8),
learning_rate=trial.suggest_float('learning_rate',
0.001,
1,
step=0.001),
min_child_samples=trial.suggest_int('min_child_samples', 2, 2**8),
min_child_weight=trial.suggest_float('min_child_weight',
1e-8,
1,
log=True),
min_split_gain=trial.suggest_float('min_split_gain',
1e-8,
1,
log=True),
bagging_fraction=trial.suggest_float('bagging_fraction', 0.4, 1),
bagging_freq=trial.suggest_int("bagging_freq", 0, 2**4),
feature_fraction=trial.suggest_float('feature_fraction', 0.4, 1),
lambda_l1=trial.suggest_float('lambda_l1', 1e-8, 10, log=True),
lambda_l2=trial.suggest_float('lambda_l2', 1e-8, 10, log=True),
)
if self.params is not None:
opt_params.update(self.params)
cv_result = lgb.cv(opt_params,
self.dtrain,
num_boost_round=10000,
nfold=cv,
stratified='reg' not in opt_params.get(
'application',
opt_params.get('objective', 'reg')),
feval=None,
early_stopping_rounds=100,
verbose_eval=100,
show_stdv=False,
seed=0,
eval_train_metric=False)
score = -1
self.best_num_boost_round = 0
for key in cv_result:
if 'mean' in key:
_ = cv_result[key]
score = _[-1]
self.best_num_boost_round = len(_)
print(
f'CV Score: {score if score != -1 else "cv_result donot contain mean-metric"}'
)
return score
def define_model(trial: optuna.trial.Trial) -> nn.Sequential:
n_layers = trial.suggest_int("n_layers", 1, 3)
dropout = trial.suggest_float("dropout", 0.2, 0.5)
input_dim = 28 * 28
layers = [nn.Flatten()]
for i in range(n_layers):
output_dim = trial.suggest_int("n_units_l{}".format(i),
4,
128,
log=True)
layers.append(nn.Linear(input_dim, output_dim))
layers.append(nn.ReLU())
layers.append(nn.Dropout(dropout))
input_dim = output_dim
layers.append(nn.Linear(input_dim, CLASSES))
return nn.Sequential(*layers)
def _objective(trial: optuna.trial.Trial) -> float:
p0 = trial.suggest_float("p0", -3.3, 5.2)
p1 = trial.suggest_float("p1", 2.0, 2.0)
p2 = trial.suggest_float("p2", 0.0001, 0.3, log=True)
p3 = trial.suggest_float("p3", 1.1, 1.1, log=True)
p4 = trial.suggest_int("p4", -100, 8)
p5 = trial.suggest_int("p5", -20, -20)
p6 = trial.suggest_float("p6", 10, 20, step=2)
p7 = trial.suggest_float("p7", 0.1, 1.0, step=0.1)
p8 = trial.suggest_float("p8", 2.2, 2.2, step=0.5)
p9 = trial.suggest_categorical("p9", ["9", "3", "0", "8"])
assert isinstance(p9, str)
return p0 + p1 + p2 + p3 + p4 + p5 + p6 + p7 + p8 + int(p9)
