DeepARModel

class DeepARModel(batch_size: int = 64, context_length: Optional[int] = None, max_epochs: int = 10, gpus: Union[int, List[int]] = 0, gradient_clip_val: float = 0.1, learning_rate: Optional[List[float]] = None, cell_type: str = 'LSTM', hidden_size: int = 10, rnn_layers: int = 2, dropout: float = 0.1, loss: Optional[pytorch_forecasting.metrics.DistributionLoss] = None, trainer_kwargs: Optional[Dict[str, Any]] = None, quantiles_kwargs: Optional[Dict[str, Any]] = None)

Bases: etna.models.nn.utils._DeepCopyMixin, etna.models.mixins.SaveNNMixin, etna.models.base.PredictionIntervalContextIgnorantAbstractModel

Wrapper for pytorch_forecasting.models.deepar.DeepAR.

Notes

We save pytorch_forecasting.data.timeseries.TimeSeriesDataSet in instance to use it in the model. It`s not right pattern of using Transforms and TSDataset.

Initialize DeepAR wrapper.

Parameters

• batch_size (int) – Batch size.

• context_length (Optional[int]) – Max encoder length, if None max encoder length is equal to 2 horizons.

• max_epochs (int) – Max epochs.

• gpus (Union[int, List[int]]) – 0 - is CPU, or [n_{i}] - to choose n_{i} GPU from cluster.

• gradient_clip_val (float) – Clipping by norm is using, choose 0 to not clip.

• learning_rate (Optional[List[float]]) – Learning rate.

• cell_type (str) – One of ‘LSTM’, ‘GRU’.

• hidden_size (int) – Hidden size of network which can range from 8 to 512.

• rnn_layers (int) – Number of LSTM layers.

• dropout (float) – Dropout rate.

• loss (Optional[DistributionLoss]) – Distribution loss function. Keep in mind that each distribution loss function might have specific requirements for target normalization. Defaults to pytorch_forecasting.metrics.NormalDistributionLoss.

• trainer_kwargs (Optional[Dict[str, Any]]) – Additional arguments for pytorch_lightning Trainer.

• quantiles_kwargs (Optional[Dict[str, Any]]) – Additional arguments for computing quantiles, look at to_quantiles() method for your loss.

Inherited-members

Methods

fit(ts)	Fit model.
forecast(ts[, prediction_interval, quantiles])	Make predictions.
get_model()	Get internal model that is used inside etna class.
load(path)	Load an object.
predict(ts[, prediction_interval, quantiles])	Make predictions.
save(path)	Save the object.
to_dict()	Collect all information about etna object in dict.

Attributes

context_size

fit(ts: etna.datasets.tsdataset.TSDataset) → etna.models.nn.deepar.DeepARModel

Fit model.

Parameters

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

Return type

DeepARModel

forecast(ts: etna.datasets.tsdataset.TSDataset, prediction_interval: bool = False, quantiles: Sequence[float] = (0.025, 0.975)) → etna.datasets.tsdataset.TSDataset

Make predictions.

This method will make autoregressive predictions.

Parameters

• ts (etna.datasets.tsdataset.TSDataset) – Dataset with features

• prediction_interval (bool) – If True returns prediction interval for forecast

• quantiles (Sequence[float]) – Levels of prediction distribution. By default 2.5% and 97.5% are taken to form a 95% prediction interval

Returns

TSDataset with predictions.

Return type

TSDataset

get_model() → Any

Get internal model that is used inside etna class.

Internal model is a model that is used inside etna to forecast segments, e.g. catboost.CatBoostRegressor or sklearn.linear_model.Ridge.

Returns

Internal model

Return type

Any

predict(ts: etna.datasets.tsdataset.TSDataset, prediction_interval: bool = False, quantiles: Sequence[float] = (0.025, 0.975)) → etna.datasets.tsdataset.TSDataset

Make predictions.

This method will make predictions using true values instead of predicted on a previous step. It can be useful for making in-sample forecasts.

Parameters

• ts (etna.datasets.tsdataset.TSDataset) – Dataset with features

• prediction_interval (bool) – If True returns prediction interval for forecast

• quantiles (Sequence[float]) – Levels of prediction distribution. By default 2.5% and 97.5% are taken to form a 95% prediction interval

Returns

TSDataset with predictions.

Return type

TSDataset