gluonts.torch.model.tft.layers module

class gluonts.torch.model.tft.layers.FeatureEmbedder(cardinalities: List[int], embedding_dims: List[int])

Bases: gluonts.torch.modules.feature.FeatureEmbedder

forward(features: torch.Tensor) → List[torch.Tensor]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class gluonts.torch.model.tft.layers.FeatureProjector(feature_dims: List[int], embedding_dims: List[int], **kwargs)

Bases: torch.nn.modules.module.Module

forward(features: torch.Tensor) → List[torch.Tensor]

Parameters

features – Numerical features with shape (…, sum(self.feature_dims)).

Returns

List of project features, with shapes [(…, self.embedding_dims[i]) for i in self.embedding_dims]

Return type

projected_features

training: bool

class gluonts.torch.model.tft.layers.GatedLinearUnit(dim: int = - 1, nonlinear: bool = True)

Bases: torch.nn.modules.module.Module

forward(x: torch.Tensor) → torch.Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class gluonts.torch.model.tft.layers.GatedResidualNetwork(d_hidden: int, d_input: Optional[int] = None, d_output: Optional[int] = None, d_static: Optional[int] = None, dropout: float = 0.0)

Bases: torch.nn.modules.module.Module

forward(x: torch.Tensor, c: Optional[torch.Tensor] = None) → torch.Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class gluonts.torch.model.tft.layers.TemporalFusionDecoder(context_length: int, prediction_length: int, d_hidden: int, d_var: int, num_heads: int, dropout: float = 0.0)

Bases: torch.nn.modules.module.Module

forward(x: torch.Tensor, static: torch.Tensor, mask: Optional[torch.Tensor] = None) → torch.Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class gluonts.torch.model.tft.layers.TemporalFusionEncoder(d_input: int, d_hidden: int)

Bases: torch.nn.modules.module.Module

forward(ctx_input: torch.Tensor, tgt_input: Optional[torch.Tensor] = None, states: Optional[List[torch.Tensor]] = None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class gluonts.torch.model.tft.layers.VariableSelectionNetwork(d_hidden: int, num_vars: int, dropout: float = 0.0, add_static: bool = False)

Bases: torch.nn.modules.module.Module

forward(variables: List[torch.Tensor], static: Optional[torch.Tensor] = None) → Tuple[torch.Tensor, torch.Tensor]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool