def plot_optimization_history(self, interactive=False):
'''
Plot parameters optimization history.
Parameters
----------
interactive : bool, optional
Create & save to current wd interactive html plot. The default is False.
Returns
-------
None.
'''
self._check_refit_status('plot_optimization_history()')
validate_plotting_interactive_argument(interactive)
if interactive:
from optuna.visualization import plot_optimization_history
fig = plot_optimization_history(self._study)
fig.write_html("optimization_history_plot.html")
try:
self._display_html("optimization_history_plot.html")
except Exception as e:
print(f'Display html error: {e}')
print(
f'Optimization History Plot is saved to {os.path.join(os.getcwd(), "optimization_history_plot.html")}'
)
else:
from optuna.visualization.matplotlib import plot_optimization_history
import matplotlib.pyplot as plt
plot_optimization_history(self._study)
plt.show()
def make_plots(logdir, study):
logdir = f'{logdir}/plots'
os.makedirs(logdir, exist_ok=True)
plot_optimization_history(study).write_image(f'{logdir}/history.svg')
plot_intermediate_values(study).write_image(f'{logdir}/intermediates.svg')
plot_parallel_coordinate(study).write_image(f'{logdir}/parallel_coordinates.png')
plot_slice(study).write_image(f'{logdir}/slices.svg')
plot_param_importances(study).write_image(f'{logdir}/importances.svg')
def test_plot_optimization_history_with_error_bar(direction: str) -> None:
n_studies = 10
# Test with no trial.
studies = [create_study(direction=direction) for _ in range(n_studies)]
figure = plot_optimization_history(studies, error_bar=True)
assert len(figure.data) == 0
def objective(trial: Trial) -> float:
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with trials.
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(studies, error_bar=True)
assert len(figure.data) == 4
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [1.0, 2.0, 0.0])
assert np.array_equal(figure.data[1].x, [0, 1, 2])
ydata = figure.data[1].y
if direction == "minimize":
assert np.array_equal(ydata, [1.0, 1.0, 0.0])
else:
assert np.array_equal(ydata, [1.0, 2.0, 2.0])
# Scatters for error bar don't have `name`.
legend_texts = [scatter.name for scatter in figure.data if scatter.name is not None]
assert sorted(legend_texts) == ["Best Value", "Objective Value"]
assert figure.layout.yaxis.title.text == "Objective Value"
# Test customized target.
with pytest.warns(UserWarning):
figure = plot_optimization_history(studies, target=lambda t: t.number, error_bar=True)
assert len(figure.data) == 1
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [0, 1, 2])
# Test customized target name.
custom_target_name = "Target Name"
figure = plot_optimization_history(studies, target_name=custom_target_name, error_bar=True)
legend_texts = [scatter.name for scatter in figure.data if scatter.name is not None]
assert sorted(legend_texts) == ["Best Value", custom_target_name]
assert figure.layout.yaxis.title.text == custom_target_name
# Ignore failed trials.
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(fail_objective, n_trials=1, catch=(ValueError,))
figure = plot_optimization_history(studies, error_bar=True)
assert len(figure.data) == 0
def test_plot_optimization_history_with_multiple_studies(direction: str) -> None:
n_studies = 10
# Test with no trial.
studies = [create_study(direction=direction) for _ in range(n_studies)]
figure = plot_optimization_history(studies)
assert len(figure.data) == 0
def objective(trial: Trial) -> float:
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with trials.
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(studies)
assert len(figure.data) == 2 * n_studies
assert figure.data[0].x == (0, 1, 2)
assert figure.data[0].y == (1.0, 2.0, 0.0)
assert figure.data[1].x == (0, 1, 2)
if direction == "minimize":
assert np.array_equal(figure.data[1].y, np.array([1.0, 1.0, 0.0]))
else:
assert np.array_equal(figure.data[1].y, np.array([1.0, 2.0, 2.0]))
assert figure.data[0].name == f"Objective Value of {studies[0].study_name}"
assert figure.layout.yaxis.title.text == "Objective Value"
# Test customized target.
with pytest.warns(UserWarning):
figure = plot_optimization_history(studies, target=lambda t: t.number)
assert len(figure.data) == 1 * n_studies
assert np.array_equal(figure.data[0].x, np.array([0, 1, 2], dtype=float))
assert np.array_equal(figure.data[0].y, np.array([0, 1, 2], dtype=float))
# Test customized target name.
figure = plot_optimization_history(studies, target_name="Target Name")
assert figure.data[0].name == f"Target Name of {studies[0].study_name}"
assert figure.layout.yaxis.title.text == "Target Name"
# Ignore failed trials.
def fail_objective(_: Trial) -> float:
raise ValueError
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(fail_objective, n_trials=1, catch=(ValueError,))
figure = plot_optimization_history(studies)
assert len(figure.data) == 0
def test_plot_optimization_history(direction: str) -> None:
# Test with no trial.
study = create_study(direction=direction)
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def objective(trial: Trial) -> float:
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with a trial.
study = create_study(direction=direction)
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(study)
assert len(figure.data) == 2
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [1.0, 2.0, 0.0])
assert np.array_equal(figure.data[1].x, [0, 1, 2])
ydata = figure.data[1].y
if direction == "minimize":
assert np.array_equal(ydata, [1.0, 1.0, 0.0])
else:
assert np.array_equal(ydata, [1.0, 2.0, 2.0])
legend_texts = [x.name for x in figure.data]
assert legend_texts == ["Objective Value", "Best Value"]
assert figure.layout.yaxis.title.text == "Objective Value"
# Test customized target.
with pytest.warns(UserWarning):
figure = plot_optimization_history(study, target=lambda t: t.number)
assert len(figure.data) == 1
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [0.0, 1.0, 2.0])
# Test customized target name.
custom_target_name = "Target Name"
figure = plot_optimization_history(study, target_name=custom_target_name)
legend_texts = [x.name for x in figure.data]
assert legend_texts == [custom_target_name, "Best Value"]
assert figure.layout.yaxis.title.text == custom_target_name
# Ignore failed trials.
def fail_objective(_: Trial) -> float:
raise ValueError
study = create_study(direction=direction)
study.optimize(fail_objective, n_trials=1, catch=(ValueError, ))
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def test_plot_optimization_history(direction: str) -> None:
# Test with no trial.
study = create_study(direction=direction)
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def objective(trial: Trial) -> float:
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with a trial.
study = create_study(direction=direction)
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(study)
assert len(figure.data) == 2
assert figure.data[0].x == (0, 1, 2)
assert figure.data[0].y == (1.0, 2.0, 0.0)
assert figure.data[1].x == (0, 1, 2)
if direction == "minimize":
assert figure.data[1].y == (1.0, 1.0, 0.0)
else:
assert figure.data[1].y == (1.0, 2.0, 2.0)
assert figure.data[0].name == "Objective Value"
assert figure.layout.yaxis.title.text == "Objective Value"
# Test customized target.
with pytest.warns(UserWarning):
figure = plot_optimization_history(study, target=lambda t: t.number)
assert len(figure.data) == 1
assert figure.data[0].x == (0, 1, 2)
assert figure.data[0].y == (0, 1, 2)
# Test customized target name.
figure = plot_optimization_history(study, target_name="Target Name")
assert figure.data[0].name == "Target Name"
assert figure.layout.yaxis.title.text == "Target Name"
# Ignore failed trials.
def fail_objective(_: Trial) -> float:
raise ValueError
study = create_study(direction=direction)
study.optimize(fail_objective, n_trials=1, catch=(ValueError, ))
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def __call__(self, study, trial):
import optuna.visualization as vis
self.exp.log_metric('run_score', trial.value)
self.exp.log_metric('best_so_far_run_score', study.best_value)
self.exp.log_text('run_parameters', str(trial.params))
if self.log_study:
pickle_and_log_artifact(study, 'study.pkl', experiment=self.exp)
if self.log_optimization_history:
log_chart(name='optimization_history',
chart=vis.plot_optimization_history(study),
experiment=self.exp)
if self.log_contour:
log_chart(name='contour',
chart=vis.plot_contour(study, params=self.params),
experiment=self.exp)
if self.log_parallel_coordinate:
log_chart(name='parallel_coordinate',
chart=vis.plot_parallel_coordinate(study,
params=self.params),
experiment=self.exp)
if self.log_slice:
log_chart(name='slice',
chart=vis.plot_slice(study, params=self.params),
experiment=self.exp)
def log_study_info(study, experiment=None, log_charts=True, params=None):
"""Logs runs results and parameters to neptune.
Logs all hyperparameter optimization results to Neptune. Those include best score ('best_score' metric),
best parameters ('best_parameters' property), the study object itself as artifact, and interactive optuna charts
('contour', 'parallel_coordinate', 'slice', 'optimization_history') as artifacts in 'charts' sub folder.
Args:
study('optuna.study.Study'): Optuna study object after training is completed.
experiment(`neptune.experiments.Experiment`): Neptune experiment. Default is None.
log_charts('bool'): Whether optuna visualization charts should be logged. By default all charts are logged.
params(`list`): List of parameters to be visualized. Default is all parameters.
Examples:
Initialize neptune_monitor::
import neptune
import neptunecontrib.monitoring.optuna as opt_utils
neptune.init(project_qualified_name='USER_NAME/PROJECT_NAME')
neptune.create_experiment(name='optuna sweep')
neptune_callback = opt_utils.NeptuneCallback()
Run Optuna training passing monitor as callback::
...
study = optuna.create_study(direction='maximize')
study.optimize(objective, n_trials=100, callbacks=[neptune_callback])
opt_utils.log_study_info(study)
You can explore an example experiment in Neptune:
https://ui.neptune.ai/o/shared/org/showroom/e/SHOW-1016/artifacts
"""
import optuna.visualization as vis
_exp = experiment if experiment else neptune
_exp.log_metric('best_score', study.best_value)
_exp.set_property('best_parameters', study.best_params)
if log_charts:
log_chart(name='optimization_history',
chart=vis.plot_optimization_history(study),
experiment=_exp)
log_chart(name='contour',
chart=vis.plot_contour(study, params=params),
experiment=_exp)
log_chart(name='parallel_coordinate',
chart=vis.plot_parallel_coordinate(study, params=params),
experiment=_exp)
log_chart(name='slice',
chart=vis.plot_slice(study, params=params),
experiment=_exp)
pickle_and_log_artifact(study, 'study.pkl', experiment=_exp)
def test_plot_optimization_history(direction):
# type: (str) -> None
# Test with no trial.
study = create_study(direction=direction)
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def objective(trial):
# type: (Trial) -> float
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with a trial.
study = create_study(direction=direction)
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(study)
assert len(figure.data) == 2
assert figure.data[0].x == (0, 1, 2)
assert figure.data[0].y == (1.0, 2.0, 0.0)
assert figure.data[1].x == (0, 1, 2)
if direction == "minimize":
assert figure.data[1].y == (1.0, 1.0, 0.0)
else:
assert figure.data[1].y == (1.0, 2.0, 2.0)
# Ignore failed trials.
def fail_objective(_):
# type: (Trial) -> float
raise ValueError
study = create_study(direction=direction)
study.optimize(fail_objective, n_trials=1, catch=(ValueError,))
figure = plot_optimization_history(study)
assert len(figure.data) == 0
def test_error_bar_in_optimization_history(direction: str) -> None:
def objective(trial: Trial) -> float:
return trial.suggest_float("x", 0, 1)
studies = [create_study(direction=direction) for _ in range(3)]
suggested_params = [0.1, 0.3, 0.2]
for x, study in zip(suggested_params, studies):
study.enqueue_trial({"x": x})
study.optimize(objective, n_trials=1)
figure = plot_optimization_history(studies, error_bar=True)
mean = np.mean(suggested_params)
std = np.std(suggested_params)
np.testing.assert_almost_equal(figure.data[0].y, mean)
np.testing.assert_almost_equal(figure.data[2].y, mean + std)
np.testing.assert_almost_equal(figure.data[3].y, mean - std)
def draw_results(study):
# 优化历史
plt.figure()
fig = pv.plot_optimization_history(study)
fig.write_image("./output/opt_his.png")
plt.close()
# 等高线图
plt.figure()
fig = pv.plot_contour(study)
fig.write_image("./output/opt_contour.png")
plt.close()
# 经验分布图
plt.figure()
fig = pv.plot_edf(study)
fig.write_image("./output/opt_edf.png")
plt.close()
# 高维参数
plt.figure()
fig = pv.plot_parallel_coordinate(study)
fig.write_image("./output/opt_coordinate.png")
plt.close()
def test_target_is_none_and_study_is_multi_obj() -> None:
study = create_study(directions=["minimize", "minimize"])
with pytest.raises(ValueError):
plot_optimization_history(study)
best_model_predict = best_model.predict(x_test)
print('\nBest Model performance at competition:')
print(
'RMSE: {:.4f} (should be lower than the trivial predictor using the mean MSE: {:.4f})'
.format(
math.sqrt(metrics.mean_squared_error(y_test, best_model_predict)),
math.sqrt(
metrics.mean_squared_error(
y_test, [y_test.mean() for i in range(len(y_test))]))))
print(
'R square: {:.4f} (should be higher than the trivial predictor using the mean: R square {:.4f})'
.format(
metrics.r2_score(y_test, best_model_predict),
metrics.r2_score(y_test, [y_test.mean() for i in range(len(y_test))])))
#3.8 Final model train
best_model.fit(x, y)
y_comp = [math.exp(i) for i in best_model.predict(x_comp)]
submission = pd.DataFrame(columns=['Id', 'SalePrice'])
submission['Id'] = pd.Series(range(1461, 2920))
submission['SalePrice'] = pd.Series(y_comp)
submission.to_csv('submission.csv', index=False)
#3.9 Optuna visualization
ov.plot_optimization_history(knn_optuna).show()
#ov.plot_parallel_coordinate(knn_optuna).show()
ov.plot_contour(knn_optuna).show()
ov.plot_slice(knn_optuna).show()
ov.plot_param_importances(knn_optuna).show()
#ov.plot_edf(knn_optuna).show()
def log_study_info(study, experiment=None,
log_study=True,
log_charts=True,
log_optimization_history=False,
log_contour=False,
log_parallel_coordinate=False,
log_slice=False,
params=None):
"""Logs runs results and parameters to neptune.
Logs all hyperparameter optimization results to Neptune. Those include best score ('best_score' metric),
best parameters ('best_parameters' property), the study object itself as artifact, and interactive optuna charts
('contour', 'parallel_coordinate', 'slice', 'optimization_history') as artifacts in 'charts' sub folder.
Args:
study('optuna.study.Study'): Optuna study object after training is completed.
experiment(`neptune.experiments.Experiment`): Neptune experiment. Default is None.
log_study('bool'): Whether optuna study object should be logged as pickle. Default is True.
log_charts('bool'): Deprecated argument. Whether all optuna visualizations charts should be logged.
By default all charts are sent.
To not log any charts set log_charts=False.
If you want to log a particular chart change the argument for that chart explicitly.
For example log_charts=False and log_slice=True will log only the slice plot to Neptune.
log_optimization_history('bool'): Whether optuna optimization history chart should be logged. Default is True.
log_contour('bool'): Whether optuna contour plot should be logged. Default is True.
log_parallel_coordinate('bool'): Whether optuna parallel coordinate plot should be logged. Default is True.
log_slice('bool'): Whether optuna slice chart should be logged. Default is True.
params(`list`): List of parameters to be visualized. Default is all parameters.
Examples:
Initialize neptune_monitor::
import neptune
import neptunecontrib.monitoring.optuna as opt_utils
neptune.init(project_qualified_name='USER_NAME/PROJECT_NAME')
neptune.create_experiment(name='optuna sweep')
neptune_callback = opt_utils.NeptuneCallback()
Run Optuna training passing monitor as callback::
...
study = optuna.create_study(direction='maximize')
study.optimize(objective, n_trials=100, callbacks=[neptune_callback])
opt_utils.log_study_info(study)
You can explore an example experiment in Neptune:
https://ui.neptune.ai/o/shared/org/showroom/e/SHOW-1016/artifacts
"""
import optuna.visualization as vis
_exp = experiment if experiment else neptune
_exp.log_metric('best_score', study.best_value)
_exp.set_property('best_parameters', study.best_params)
if log_charts:
message = """log_charts argument is depraceted and will be removed in future releases.
Please use log_optimization_history, log_contour, log_parallel_coordinate, log_slice, arguments explicitly.
"""
warnings.warn(message)
log_optimization_history = True
log_contour = True
log_parallel_coordinate = True
log_slice = True
if log_study:
pickle_and_log_artifact(study, 'study.pkl', experiment=_exp)
if log_optimization_history:
log_chart(name='optimization_history', chart=vis.plot_optimization_history(study), experiment=_exp)
if log_contour:
log_chart(name='contour', chart=vis.plot_contour(study, params=params), experiment=_exp)
if log_parallel_coordinate:
log_chart(name='parallel_coordinate', chart=vis.plot_parallel_coordinate(study, params=params), experiment=_exp)
if log_slice:
log_chart(name='slice', chart=vis.plot_slice(study, params=params), experiment=_exp)
def hyperparameters_optimization(self) -> None:
if self.verbose > 0:
print("Optimizing hyperparameters")
if self.storage is not None and self.study_name is None:
warnings.warn(
f"You passed a remote storage: {self.storage} but no `--study-name`."
"The study name will be generated by Optuna, make sure to re-use the same study name "
"when you want to do distributed hyperparameter optimization.")
if self.tensorboard_log is not None:
warnings.warn(
"Tensorboard log is deactivated when running hyperparameter optimization"
)
self.tensorboard_log = None
# TODO: eval each hyperparams several times to account for noisy evaluation
sampler = self._create_sampler(self.sampler)
pruner = self._create_pruner(self.pruner)
if self.verbose > 0:
print(f"Sampler: {self.sampler} - Pruner: {self.pruner}")
study = optuna.create_study(
sampler=sampler,
pruner=pruner,
storage=self.storage,
study_name=self.study_name,
load_if_exists=True,
direction="maximize",
)
try:
study.optimize(self.objective,
n_trials=self.n_trials,
n_jobs=self.n_jobs)
except KeyboardInterrupt:
pass
print("Number of finished trials: ", len(study.trials))
print("Best trial:")
trial = study.best_trial
print("Value: ", trial.value)
print("Params: ")
for key, value in trial.params.items():
print(f" {key}: {value}")
report_name = (
f"report_{self.env_id}_{self.n_trials}-trials-{self.n_timesteps}"
f"-{self.sampler}-{self.pruner}_{int(time.time())}")
log_path = os.path.join(self.log_folder, self.algo, report_name)
if self.verbose:
print(f"Writing report to {log_path}")
# Write report
os.makedirs(os.path.dirname(log_path), exist_ok=True)
study.trials_dataframe().to_csv(f"{log_path}.csv")
# Save python object to inspect/re-use it later
with open(f"{log_path}.pkl", "wb+") as f:
pkl.dump(study, f)
# Skip plots
if self.no_optim_plots:
return
# Plot optimization result
try:
fig1 = plot_optimization_history(study)
fig2 = plot_param_importances(study)
fig1.show()
fig2.show()
except (ValueError, ImportError, RuntimeError):
pass
def show_history(self):
return visualization.plot_optimization_history(self.study)
def ml_mlp_mul_ms(station_name="종로구"):
print("Start Multivariate MLP Mean Seasonality Decomposition (MSE) Model")
targets = ["PM10", "PM25"]
# targets = ["SO2", "CO", "O3", "NO2", "PM10", "PM25",
# "temp", "u", "v", "pres", "humid", "prep", "snow"]
# 24*14 = 336
#sample_size = 336
sample_size = 48
output_size = 24
# If you want to debug, fast_dev_run = True and n_trials should be small number
fast_dev_run = False
n_trials = 128
# fast_dev_run = True
# n_trials = 1
# Hyper parameter
epoch_size = 500
batch_size = 64
learning_rate = 1e-3
# Blocked Cross Validation
# neglect small overlap between train_dates and valid_dates
# 11y = ((2y, 0.5y), (2y, 0.5y), (2y, 0.5y), (2.5y, 1y))
train_dates = [(dt.datetime(2008, 1, 4, 1).astimezone(SEOULTZ),
dt.datetime(2009, 12, 31, 23).astimezone(SEOULTZ)),
(dt.datetime(2010, 7, 1, 0).astimezone(SEOULTZ),
dt.datetime(2012, 6, 30, 23).astimezone(SEOULTZ)),
(dt.datetime(2013, 1, 1, 0).astimezone(SEOULTZ),
dt.datetime(2014, 12, 31, 23).astimezone(SEOULTZ)),
(dt.datetime(2015, 7, 1, 0).astimezone(SEOULTZ),
dt.datetime(2017, 12, 31, 23).astimezone(SEOULTZ))]
valid_dates = [(dt.datetime(2010, 1, 1, 0).astimezone(SEOULTZ),
dt.datetime(2010, 6, 30, 23).astimezone(SEOULTZ)),
(dt.datetime(2012, 7, 1, 0).astimezone(SEOULTZ),
dt.datetime(2012, 12, 31, 23).astimezone(SEOULTZ)),
(dt.datetime(2015, 1, 1, 0).astimezone(SEOULTZ),
dt.datetime(2015, 6, 30, 23).astimezone(SEOULTZ)),
(dt.datetime(2018, 1, 1, 0).astimezone(SEOULTZ),
dt.datetime(2018, 12, 31, 23).astimezone(SEOULTZ))]
train_valid_fdate = dt.datetime(2008, 1, 3, 1).astimezone(SEOULTZ)
train_valid_tdate = dt.datetime(2018, 12, 31, 23).astimezone(SEOULTZ)
# Debug
if fast_dev_run:
train_dates = [(dt.datetime(2015, 7, 1, 0).astimezone(SEOULTZ),
dt.datetime(2017, 12, 31, 23).astimezone(SEOULTZ))]
valid_dates = [(dt.datetime(2018, 1, 1, 0).astimezone(SEOULTZ),
dt.datetime(2018, 12, 31, 23).astimezone(SEOULTZ))]
train_valid_fdate = dt.datetime(2015, 7, 1, 0).astimezone(SEOULTZ)
train_valid_tdate = dt.datetime(2018, 12, 31, 23).astimezone(SEOULTZ)
test_fdate = dt.datetime(2019, 1, 1, 0).astimezone(SEOULTZ)
test_tdate = dt.datetime(2020, 10, 31, 23).astimezone(SEOULTZ)
# check date range assumption
assert len(train_dates) == len(valid_dates)
for i, (td, vd) in enumerate(zip(train_dates, valid_dates)):
assert vd[0] > td[1]
assert test_fdate > train_dates[-1][1]
assert test_fdate > valid_dates[-1][1]
train_features = [
"SO2", "CO", "NO2", "PM10", "PM25", "temp", "wind_spd", "wind_cdir",
"wind_sdir", "pres", "humid", "prep"
]
train_features_periodic = [
"SO2", "CO", "NO2", "PM10", "PM25", "temp", "wind_spd", "wind_cdir",
"wind_sdir", "pres", "humid"
]
train_features_nonperiodic = ["prep"]
for target in targets:
print("Training " + target + "...")
output_dir = Path(
f"/mnt/data/MLPMSMultivariate/{station_name}/{target}/")
Path.mkdir(output_dir, parents=True, exist_ok=True)
model_dir = output_dir / "models"
Path.mkdir(model_dir, parents=True, exist_ok=True)
log_dir = output_dir / "log"
Path.mkdir(log_dir, parents=True, exist_ok=True)
_df_h = data.load_imputed(HOURLY_DATA_PATH)
df_h = _df_h.query('stationCode == "' +
str(SEOUL_STATIONS[station_name]) + '"')
if station_name == '종로구' and \
not Path("/input/python/input_jongno_imputed_hourly_pandas.csv").is_file():
# load imputed result
df_h.to_csv("/input/python/input_jongno_imputed_hourly_pandas.csv")
# construct dataset for seasonality
print("Construct Train/Validation Sets...", flush=True)
train_valid_dataset = construct_dataset(train_valid_fdate,
train_valid_tdate,
filepath=HOURLY_DATA_PATH,
station_name=station_name,
target=target,
sample_size=sample_size,
output_size=output_size,
transform=False)
# compute seasonality
train_valid_dataset.preprocess()
# For Block Cross Validation..
# load dataset in given range dates and transform using scaler from train_valid_set
# all dataset are saved in tuple
print("Construct Training Sets...", flush=True)
train_datasets = tuple(
construct_dataset(td[0],
td[1],
scaler_X=train_valid_dataset.scaler_X,
scaler_Y=train_valid_dataset.scaler_Y,
filepath=HOURLY_DATA_PATH,
station_name=station_name,
target=target,
sample_size=sample_size,
output_size=output_size,
features=train_features,
features_periodic=train_features_periodic,
features_nonperiodic=train_features_nonperiodic,
transform=True) for td in train_dates)
print("Construct Validation Sets...", flush=True)
valid_datasets = tuple(
construct_dataset(vd[0],
vd[1],
scaler_X=train_valid_dataset.scaler_X,
scaler_Y=train_valid_dataset.scaler_Y,
filepath=HOURLY_DATA_PATH,
station_name=station_name,
target=target,
sample_size=sample_size,
output_size=output_size,
features=train_features,
features_periodic=train_features_periodic,
features_nonperiodic=train_features_nonperiodic,
transform=True) for vd in valid_dates)
# just single test set
print("Construct Test Sets...", flush=True)
test_dataset = construct_dataset(
test_fdate,
test_tdate,
scaler_X=train_valid_dataset.scaler_X,
scaler_Y=train_valid_dataset.scaler_Y,
filepath=HOURLY_DATA_PATH,
station_name=station_name,
target=target,
sample_size=sample_size,
output_size=output_size,
features=train_features,
features_periodic=train_features_periodic,
features_nonperiodic=train_features_nonperiodic,
transform=True)
# convert tuple of datasets to ConcatDataset
train_dataset = ConcatDataset(train_datasets)
val_dataset = ConcatDataset(valid_datasets)
# num_layer == number of hidden layer
hparams = Namespace(num_layers=1,
layer_size=128,
learning_rate=learning_rate,
batch_size=batch_size)
def objective(trial):
model = BaseMLPModel(
trial=trial,
hparams=hparams,
input_size=sample_size * len(train_features),
sample_size=sample_size,
output_size=output_size,
station_name=station_name,
target=target,
features=train_features,
features_periodic=train_features_periodic,
features_nonperiodic=train_features_nonperiodic,
train_dataset=train_dataset,
val_dataset=val_dataset,
test_dataset=test_dataset,
scaler_X=train_valid_dataset.scaler_X,
scaler_Y=train_valid_dataset.scaler_Y,
output_dir=output_dir)
# most basic trainer, uses good defaults
trainer = Trainer(gpus=1 if torch.cuda.is_available() else None,
precision=32,
min_epochs=1,
max_epochs=20,
default_root_dir=output_dir,
fast_dev_run=fast_dev_run,
logger=True,
checkpoint_callback=False,
callbacks=[
PyTorchLightningPruningCallback(
trial, monitor="valid/MSE")
])
trainer.fit(model)
# Don't Log
# hyperparameters = model.hparams
# trainer.logger.log_hyperparams(hyperparameters)
return trainer.callback_metrics.get("valid/MSE")
if n_trials > 1:
study = optuna.create_study(direction="minimize")
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 4,
'layer_size': 8,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 4,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 4,
'layer_size': 64,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 4,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 8,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 1.3,
'num_layers': 12,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 0.7,
'num_layers': 4,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
study.enqueue_trial({
'sigma': 2.0,
'num_layers': 4,
'layer_size': 32,
'learning_rate': learning_rate,
'batch_size': batch_size
})
# timeout = 3600*36 = 36h
study.optimize(objective, n_trials=n_trials, timeout=3600 * 36)
trial = study.best_trial
print(" Value: ", trial.value)
print(" Params: ")
for key, value in trial.params.items():
print(" {}: {}".format(key, value))
print("sample_size : ", sample_size)
print("output_size : ", output_size)
# plot optmization results
fig_cont1 = optv.plot_contour(study,
params=['num_layers', 'layer_size'])
fig_cont1.write_image(
str(output_dir / "contour_num_layers_layer_size.png"))
fig_cont1.write_image(
str(output_dir / "contour_num_layers_layer_size.svg"))
fig_edf = optv.plot_edf(study)
fig_edf.write_image(str(output_dir / "edf.png"))
fig_edf.write_image(str(output_dir / "edf.svg"))
fig_iv = optv.plot_intermediate_values(study)
fig_iv.write_image(str(output_dir / "intermediate_values.png"))
fig_iv.write_image(str(output_dir / "intermediate_values.svg"))
fig_his = optv.plot_optimization_history(study)
fig_his.write_image(str(output_dir / "opt_history.png"))
fig_his.write_image(str(output_dir / "opt_history.svg"))
fig_pcoord = optv.plot_parallel_coordinate(
study, params=['num_layers', 'layer_size'])
fig_pcoord.write_image(str(output_dir / "parallel_coord.png"))
fig_pcoord.write_image(str(output_dir / "parallel_coord.svg"))
fig_slice = optv.plot_slice(study,
params=['num_layers', 'layer_size'])
fig_slice.write_image(str(output_dir / "slice.png"))
fig_slice.write_image(str(output_dir / "slice.svg"))
# set hparams with optmized value
hparams.num_layers = trial.params['num_layers']
hparams.layer_size = trial.params['layer_size']
dict_hparams = copy.copy(vars(hparams))
dict_hparams["sample_size"] = sample_size
dict_hparams["output_size"] = output_size
with open(output_dir / 'hparams.json', 'w') as f:
print(dict_hparams, file=f)
with open(output_dir / 'hparams.csv', 'w') as f:
print(pd.DataFrame.from_dict(dict_hparams, orient='index'),
file=f)
model = BaseMLPModel(hparams=hparams,
input_size=sample_size * len(train_features),
sample_size=sample_size,
output_size=output_size,
station_name=station_name,
target=target,
features=train_features,
features_periodic=train_features_periodic,
features_nonperiodic=train_features_nonperiodic,
train_dataset=train_dataset,
val_dataset=val_dataset,
test_dataset=test_dataset,
scaler_X=train_valid_dataset.scaler_X,
scaler_Y=train_valid_dataset.scaler_Y,
output_dir=output_dir)
# record input
for i, _train_set in enumerate(train_datasets):
_train_set.to_csv(
model.data_dir /
("df_trainset_{0}_".format(str(i).zfill(2)) + target + ".csv"))
for i, _valid_set in enumerate(valid_datasets):
_valid_set.to_csv(
model.data_dir /
("df_validset_{0}_".format(str(i).zfill(2)) + target + ".csv"))
train_valid_dataset.to_csv(model.data_dir /
("df_trainvalidset_" + target + ".csv"))
test_dataset.to_csv(model.data_dir / ("df_testset_" + target + ".csv"))
checkpoint_callback = pl.callbacks.ModelCheckpoint(os.path.join(
model_dir, "train_{epoch}_{valid/MSE:.2f}"),
monitor="valid/MSE",
period=10)
early_stop_callback = EarlyStopping(monitor='valid/MSE',
min_delta=0.001,
patience=30,
verbose=True,
mode='min')
log_version = dt.date.today().strftime("%y%m%d-%H-%M")
loggers = [ \
TensorBoardLogger(log_dir, version=log_version),
CSVLogger(log_dir, version=log_version)]
# most basic trainer, uses good defaults
trainer = Trainer(gpus=1 if torch.cuda.is_available() else None,
precision=32,
min_epochs=1,
max_epochs=epoch_size,
default_root_dir=output_dir,
fast_dev_run=fast_dev_run,
logger=loggers,
log_every_n_steps=5,
flush_logs_every_n_steps=10,
callbacks=[early_stop_callback],
checkpoint_callback=checkpoint_callback)
trainer.fit(model)
# run test set
trainer.test(ckpt_path=None)
shutil.rmtree(model_dir)
# Handle pruning based on the intermediate value.
if trial.should_prune():
raise optuna.TrialPruned()
return value
if __name__ == "__main__":
study = optuna.create_study(direction="maximize",
pruner=optuna.pruners.MedianPruner())
study.optimize(objective, n_trials=100, timeout=600)
# Visualize the optimization history.
plot_optimization_history(study).show()
# Visualize the learning curves of the trials.
plot_intermediate_values(study).show()
# Visualize high-dimensional parameter relationships.
plot_parallel_coordinate(study).show()
# Select parameters to visualize.
plot_parallel_coordinate(study, params=["lr_init", "n_units_l0"]).show()
# Visualize hyperparameter relationships.
plot_contour(study).show()
# Select parameters to visualize.
plot_contour(study, params=["n_units_l0", "n_units_l1"]).show()
return accuracy
###################################################################################################
study = optuna.create_study(
direction="maximize",
sampler=optuna.samplers.TPESampler(seed=SEED),
pruner=optuna.pruners.MedianPruner(n_warmup_steps=10),
)
study.optimize(objective, n_trials=100, timeout=600)
###################################################################################################
# Plot functions
# --------------
# Visualize the optimization history. See :func:`~optuna.visualization.plot_optimization_history` for the details.
plot_optimization_history(study)
###################################################################################################
# Visualize the learning curves of the trials. See :func:`~optuna.visualization.plot_intermediate_values` for the details.
plot_intermediate_values(study)
###################################################################################################
# Visualize high-dimensional parameter relationships. See :func:`~optuna.visualization.plot_parallel_coordinate` for the details.
plot_parallel_coordinate(study)
###################################################################################################
# Select parameters to visualize.
plot_parallel_coordinate(study, params=["bagging_freq", "bagging_fraction"])
###################################################################################################
# Visualize hyperparameter relationships. See :func:`~optuna.visualization.plot_contour` for the details.
def test_plot_optimization_history_with_multiple_studies(
direction: str) -> None:
n_studies = 10
# Test with no trial.
studies = [create_study(direction=direction) for _ in range(n_studies)]
figure = plot_optimization_history(studies)
assert len(figure.data) == 0
def objective(trial: Trial) -> float:
if trial.number == 0:
return 1.0
elif trial.number == 1:
return 2.0
elif trial.number == 2:
return 0.0
return 0.0
# Test with trials.
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(objective, n_trials=3)
figure = plot_optimization_history(studies)
assert len(figure.data) == 2 * n_studies
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [1.0, 2.0, 0.0])
assert np.array_equal(figure.data[1].x, [0, 1, 2])
ydata = figure.data[1].y
if direction == "minimize":
assert np.array_equal(ydata, [1.0, 1.0, 0.0])
else:
assert np.array_equal(ydata, [1.0, 2.0, 2.0])
expected_legend_texts = []
for i in range(n_studies):
expected_legend_texts.append(f"Best Value of {studies[i].study_name}")
expected_legend_texts.append(
f"Objective Value of {studies[i].study_name}")
legend_texts = [scatter.name for scatter in figure.data]
assert sorted(legend_texts) == sorted(expected_legend_texts)
assert figure.layout.yaxis.title.text == "Objective Value"
# Test customized target.
with pytest.warns(UserWarning):
figure = plot_optimization_history(studies, target=lambda t: t.number)
assert len(figure.data) == 1 * n_studies
assert np.array_equal(figure.data[0].x, [0, 1, 2])
assert np.array_equal(figure.data[0].y, [0, 1, 2])
# Test customized target name.
custom_target_name = "Target Name"
figure = plot_optimization_history(studies, target_name=custom_target_name)
expected_legend_texts = []
for i in range(n_studies):
expected_legend_texts.append(f"Best Value of {studies[i].study_name}")
expected_legend_texts.append(
f"{custom_target_name} of {studies[i].study_name}")
legend_texts = [scatter.name for scatter in figure.data]
assert sorted(legend_texts) == sorted(expected_legend_texts)
assert figure.layout.yaxis.title.text == custom_target_name
# Ignore failed trials.
def fail_objective(_: Trial) -> float:
raise ValueError
studies = [create_study(direction=direction) for _ in range(n_studies)]
for study in studies:
study.optimize(fail_objective, n_trials=1, catch=(ValueError, ))
figure = plot_optimization_history(studies)
assert len(figure.data) == 0