gluonts.torch.model.patch_tst.module module#

class gluonts.torch.model.patch_tst.module.PatchTSTModel(prediction_length: int, context_length: int, patch_len: int, stride: int, padding_patch: str, d_model: int, nhead: int, dim_feedforward: int, num_feat_dynamic_real: int, dropout: float, activation: str, norm_first: bool, num_encoder_layers: int, scaling: Optional[str], distr_output=gluonts.torch.distributions.studentT.StudentTOutput(beta=0.0))[source]#

Bases: Module

Module implementing the PatchTST model for forecasting as described in https://arxiv.org/abs/2211.14730 extended to be probabilistic.

Parameters:

  • prediction_length – Number of time points to predict.

  • context_length – Number of time steps prior to prediction time that the model.

  • num_feat_dynamic_real – Number of dynamic real features in the data (default: 0).

  • distr_output – Distribution to use to evaluate observations and sample predictions. Default: StudentTOutput().

describe_inputs(batch_size=1) → InputSpec[source]#
forward(past_target: Tensor, past_observed_values: Tensor, past_time_feat: Optional[Tensor] = None, future_time_feat: Optional[Tensor] = None) → Tuple[Tuple[Tensor, ...], Tensor, Tensor][source]#

Define 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.

loss(past_target: Tensor, past_observed_values: Tensor, future_target: Tensor, future_observed_values: Tensor, past_time_feat: Optional[Tensor] = None, future_time_feat: Optional[Tensor] = None) → Tensor[source]#
class gluonts.torch.model.patch_tst.module.SinusoidalPositionalEmbedding(num_positions: int, embedding_dim: int)[source]#

Bases: Embedding

This module produces sinusoidal positional embeddings of any length.

forward(input_ids_shape: Size, past_key_values_length: int = 0) → Tensor[source]#

input_ids_shape is expected to be [bsz x seqlen x …].