LibMTL.loss

class AbsLoss[source]

Bases: object

An abstract class for loss functions.

compute_loss(self, pred, gt)[source]

Calculate the loss.

Parameters

  • pred (torch.Tensor) – The prediction tensor.

  • gt (torch.Tensor) – The ground-truth tensor.

Returns

The loss.

Return type

torch.Tensor

class CELoss[source]

Bases: AbsLoss

The cross-entropy loss function.

compute_loss(self, pred, gt)[source]
class KLDivLoss[source]

Bases: AbsLoss

The Kullback-Leibler divergence loss function.

compute_loss(self, pred, gt)[source]
class L1Loss[source]

Bases: AbsLoss

The Mean Absolute Error (MAE) loss function.

compute_loss(self, pred, gt)[source]
class MSELoss[source]

Bases: AbsLoss

The Mean Squared Error (MSE) loss function.

compute_loss(self, pred, gt)[source]