credit.models.crossformer_ensemble

credit.models.crossformer_ensemble#

Attributes#

logger

Classes#

CrossFormerWithNoise

CrossFormer variant with pixel-wise noise injection in both encoder and decoder stages.

PixelNoiseInjection

A module that injects noise into feature maps, with a per-pixel and per-channel style modulation.

Module Contents#

class credit.models.crossformer_ensemble.CrossFormerWithNoise(noise_latent_dim=128, encoder_noise_factor=0.05, decoder_noise_factor=0.275, encoder_noise=True, freeze=True, **kwargs)#

Bases: credit.models.crossformer.CrossFormer

CrossFormer variant with pixel-wise noise injection in both encoder and decoder stages.

noise_latent_dim#

Dimensionality of the noise vector.

Type:

int

encoder_noise_factor#

Initial scaling factor for encoder noise injection.

Type:

float

decoder_noise_factor#

Initial scaling factor for decoder noise injection.

Type:

float

encoder_noise#

Whether to apply noise injection in the encoder.

Type:

bool

freeze#

Whether to freeze pre-trained model weights.

Type:

bool

noise_latent_dim = 128#
encoder_noise = True#
noise_inject1#
noise_inject2#
noise_inject3#
forward(x, noise=None, forecast_step=None)#

Forward pass through the CrossFormer with noise injection.

Parameters:

  • x (Tensor) – Input tensor of shape (batch_size, channels, height, width).

  • noise (Tensor, optional) – External noise tensor. If None, noise is sampled internally. Defaults to None.

Returns:

Output tensor after passing through the model.

Return type:

Tensor

class credit.models.crossformer_ensemble.PixelNoiseInjection(noise_dim, feature_channels, noise_factor=0.1)#

Bases: torch.nn.Module

A module that injects noise into feature maps, with a per-pixel and per-channel style modulation.

noise_transform#

A linear transformation to map latent noise to the feature map’s channels.

Type:

nn.Linear

modulation#

A learnable scaling factor applied to the noise.

Type:

nn.Parameter

noise_factor#

A scaling factor for controlling the intensity of the injected noise.

Type:

float

forward(feature_map, noise)#

Adds noise to the feature map, modulated by style and the modulation parameter.

noise_transform#
modulation#
noise_factor#
forward(feature_map, noise)#

Injects noise into the feature map.

Parameters:

  • feature_map (torch.Tensor) – The input feature map (batch, channels, height, width).

  • noise (torch.Tensor) – The latent noise tensor (batch, noise_dim), used for modulating the injected noise.

Returns:

The feature map with injected noise.

Return type:

torch.Tensor

credit.models.crossformer_ensemble.logger#
Contents