kaira.losses.text.CrossEntropyLoss

Inheritance diagram for CrossEntropyLoss

class kaira.losses.text.CrossEntropyLoss(weight=None, ignore_index=-100, label_smoothing=0.0)

Bases: BaseLoss

Cross Entropy Loss Module.

This module calculates the cross entropy loss for classification tasks.

Methods

__init__	Initialize the CrossEntropyLoss module.
forward	Forward pass through the CrossEntropyLoss module.

__init__(weight=None, ignore_index=-100, label_smoothing=0.0)

Initialize the CrossEntropyLoss module.

Parameters:

• weight (torch.Tensor, optional) – Class weights. Default is None.

• ignore_index (int) – Index to ignore. Default is -100.

• label_smoothing (float) – Label smoothing value. Default is 0.0.

forward(x: Tensor, target: Tensor) → Tensor

Forward pass through the CrossEntropyLoss module.

Parameters:

• x (torch.Tensor) – The input logits tensor.

• target (torch.Tensor) – The target tensor.

Returns:

The cross entropy loss.

Return type:

torch.Tensor