Tune

class Tune(pipeline: etna.pipeline.base.BasePipeline, target_metric: etna.metrics.base.Metric, horizon: int, metric_aggregation: Literal['median', 'mean', 'std', 'percentile_5', 'percentile_25', 'percentile_75', 'percentile_95'] = 'mean', backtest_params: Optional[dict] = None, experiment_folder: Optional[str] = None, runner: Optional[etna.auto.runner.base.AbstractRunner] = None, storage: Optional[optuna.storages._base.BaseStorage] = None, metrics: Optional[List[etna.metrics.base.Metric]] = None, sampler: Optional[optuna.samplers._base.BaseSampler] = None, params_to_tune: Optional[Dict[str, etna.distributions.distributions.BaseDistribution]] = None)

Bases: etna.auto.auto.AutoBase

Automatic tuning of custom pipeline.

This class takes given pipelines and tries to optimize its hyperparameters by using params_to_tune.

Trials with duplicate parameters are skipped and previously computed results are returned.

Initialize Tune class.

Parameters

• pipeline (etna.pipeline.base.BasePipeline) – Pipeline to optimize.

• target_metric (etna.metrics.base.Metric) – Metric to optimize.

• horizon (int) – Horizon to forecast for.

• metric_aggregation (Literal['median', 'mean', 'std', 'percentile_5', 'percentile_25', 'percentile_75', 'percentile_95']) – Aggregation method for per-segment metrics. By default, mean aggregation is used.

• backtest_params (Optional[dict]) – Custom parameters for backtest instead of default backtest parameters.

• experiment_folder (Optional[str]) – Name for saving experiment results, it determines the name for optuna study. By default, isn’t set.

• runner (Optional[etna.auto.runner.base.AbstractRunner]) – Runner to use for distributed training. By default, LocalRunner is used.

• storage (Optional[optuna.storages._base.BaseStorage]) – Optuna storage to use. By default, sqlite storage is used with name “etna-auto.db”.

• metrics (Optional[List[etna.metrics.base.Metric]]) – List of metrics to compute. By default, Sign, SMAPE, MAE, MSE, MedAE metrics are used.

• sampler (Optional[optuna.samplers._base.BaseSampler]) – Optuna sampler to use. By default, TPE sampler is used.

• params_to_tune (Optional[Dict[str, etna.distributions.distributions.BaseDistribution]]) – Parameters of pipeline that should be tuned with corresponding tuning distributions. By default, pipeline.params_to_tune() is used.

Inherited-members

Methods

fit(ts[, timeout, n_trials, initializer, ...])	Start automatic pipeline tuning.
objective(ts, pipeline, params_to_tune, ...)	Optuna objective wrapper.
summary()	Get trials summary.
top_k([k])	Get top k pipelines with the best metric value.

fit(ts: etna.datasets.tsdataset.TSDataset, timeout: Optional[int] = None, n_trials: Optional[int] = None, initializer: Optional[etna.auto.auto._Initializer] = None, callback: Optional[etna.auto.auto._Callback] = None, **kwargs) → etna.pipeline.base.BasePipeline

Start automatic pipeline tuning.

Parameters

• ts (etna.datasets.tsdataset.TSDataset) – TSDataset to fit on.

• timeout (Optional[int]) – Timeout for optuna. N.B. this is timeout for each worker. By default, isn’t set.

• n_trials (Optional[int]) – Number of trials for optuna. N.B. this is number of trials for each worker. By default, isn’t set.

• initializer (Optional[etna.auto.auto._Initializer]) – Object that is called before each pipeline backtest, can be used to initialize loggers.

• callback (Optional[etna.auto.auto._Callback]) – Object that is called after each pipeline backtest, can be used to log extra metrics.

• **kwargs – Additional parameters for optuna optuna.study.Study.optimize().

Return type

etna.pipeline.base.BasePipeline

static objective(ts: etna.datasets.tsdataset.TSDataset, pipeline: etna.pipeline.base.BasePipeline, params_to_tune: Dict[str, etna.distributions.distributions.BaseDistribution], target_metric: etna.metrics.base.Metric, metric_aggregation: Literal['median', 'mean', 'std', 'percentile_5', 'percentile_25', 'percentile_75', 'percentile_95'], metrics: List[etna.metrics.base.Metric], backtest_params: dict, initializer: Optional[etna.auto.auto._Initializer] = None, callback: Optional[etna.auto.auto._Callback] = None) → Callable[[optuna.trial._trial.Trial], float]

Optuna objective wrapper.

Parameters

• ts (etna.datasets.tsdataset.TSDataset) – TSDataset to fit on.

• pipeline (etna.pipeline.base.BasePipeline) – Pipeline to tune.

• params_to_tune (Dict[str, etna.distributions.distributions.BaseDistribution]) – Parameters of pipeline that should be tuned with corresponding tuning distributions.

• target_metric (etna.metrics.base.Metric) – Metric to optimize.

• metric_aggregation (Literal['median', 'mean', 'std', 'percentile_5', 'percentile_25', 'percentile_75', 'percentile_95']) – Aggregation method for per-segment metrics.

• metrics (List[etna.metrics.base.Metric]) – List of metrics to compute.

• backtest_params (dict) – Custom parameters for backtest instead of default backtest parameters.

• initializer (Optional[etna.auto.auto._Initializer]) – Object that is called before each pipeline backtest, can be used to initialize loggers.

• callback (Optional[etna.auto.auto._Callback]) – Object that is called after each pipeline backtest, can be used to log extra metrics.

Returns

function that runs specified trial and returns its evaluated score

Return type

objective

summary() → pandas.core.frame.DataFrame

Get trials summary.

There are columns:

• hash: hash of the pipeline;

• pipeline: pipeline object;

• metrics: columns with metrics’ values;

• state: state of the trial.

Returns

dataframe with detailed info on each performed trial

Return type

study_dataframe

top_k(k: int = 5) → List[etna.pipeline.base.BasePipeline]

Get top k pipelines with the best metric value.

Only complete and non-duplicate studies are taken into account.

Parameters

k (int) – Number of pipelines to return.

Returns

List of top k pipelines.

Return type

List[etna.pipeline.base.BasePipeline]