logitorch.losses.unlikelihood_loss

Classes

CrossEntropyAndUnlikelihoodLoss	Calculates the cross-entropy and unlikelihood loss for a given prediction and target values.
UnlikelihoodLoss	Calculates the unlikelihood loss for a given prediction and target values.

Module Contents

class logitorch.losses.unlikelihood_loss.CrossEntropyAndUnlikelihoodLoss(ignore_index=-100, epsilon=1e-10)

Bases: torch.nn.Module

Calculates the cross-entropy and unlikelihood loss for a given prediction and target values.

Args:

ignore_index (int, optional): Index to ignore in the loss calculation. Defaults to -100. epsilon (float, optional): Small value added to the denominator to avoid division by zero. Defaults to 1e-10.

forward(pred_values, target_values, known_labels, unknown_labels)

Forward pass of the cross-entropy and unlikelihood loss calculation.

Args:

pred_values (torch.Tensor): Predicted values. target_values (torch.Tensor): Target values. known_labels (torch.Tensor): Known labels. unknown_labels (torch.Tensor): Unknown labels.

Returns:

torch.Tensor: Calculated loss.

epsilon = 1e-10

ignore_index = -100

class logitorch.losses.unlikelihood_loss.UnlikelihoodLoss(ignore_index=-100, epsilon=1e-10)

Bases: torch.nn.Module

Calculates the unlikelihood loss for a given prediction and target values.

Args:

ignore_index (int, optional): Index to ignore in the loss calculation. Defaults to -100. epsilon (float, optional): Small value added to the denominator to avoid division by zero. Defaults to 1e-10.

forward(pred_values, target_values)

Forward pass of the unlikelihood loss calculation.

Args:

pred_values (torch.Tensor): Predicted values. target_values (torch.Tensor): Target values.

Returns:

torch.Tensor: Calculated loss.

epsilon = 1e-10

ignore_index = -100