gents.model package

Subpackages

• gents.model.diffeq package
  • Module contents
    • LS4
    • LatentODE
    • LatentSDE
    • SDEGAN
• gents.model.diffusion package
  • Module contents
    • CSDI
    • DiffusionTS
    • FIDE
    • FourierDiffusion
    • FourierDiffusionLSTM
    • FourierDiffusionMLP
    • ImagenTime
    • TMDM
    • VanillaDDPM
• gents.model.flow package
  • Module contents
    • FourierFlow
    • VanillaMAF
• gents.model.gan package
  • Module contents
    • COSCIGAN
    • GTGAN
    • PSAGAN
    • RCGAN
    • TimeGAN
    • VanillaGAN
• gents.model.vae package
  • Module contents
    • KoVAE
    • TimeVAE
    • TimeVQVAE
    • VanillaVAE

Submodules

gents.model.base module

class gents.model.base.BaseModel(seq_len: int, seq_dim: int, condition: str, **kwargs)

Bases: ABC, LightningModule

Base class for time series generative models in PyTorch Lightning.

Parameters:

• seq_len (int) – Target sequence length

• seq_dim (int) – Target sequence dimension, for univariate time series, set as 1

• condition (str) – Possible condition type, choose from [None, ‘predict’, ‘impute’, ‘class’, ‘super_resolution’]. None standards for unconditional generation.

• **kwargs – Additional arguments for the model

ALLOW_CONDITION = Ellipsis

sample(n_sample: int = 1, condition: Tensor | int = None, **kwargs) → Tensor

Sample time series from trained model in evaluation mode.

Parameters:

• n_sample (int, optional) – The number of samples. Defaults to 1.

• condition (torch.Tensor | int, optional) – Condition tensor in shape (batch_size, seq_len or obs_len, seq_dim). If the model is class label, then could be int. Defaults to None.

• kwargs – Additional arguments for the sampling process. E.g. data_mask, t, etc.

Returns:

sampled time series of shape (n_sample, seq_len, seq_dim) for unconditional generation, or (batch_size, seq_len, seq_dim, n_sample) for conditional generation.

Return type:

torch.Tensor