LabelSmoothing

The LabelSmoothing is an instance of the pytorch nn.Module class. The LabelSmoothing class implements an adapted version of the label smoothing loss function [1].

class context_builder.loss.LabelSmoothing(*args: Any, **kwargs: Any)[source]

LabelSmoothing.__init__(size, smoothing=0.0)[source]

Implements label smoothing loss function

Parameters:

size (int) – Number of labels
smoothing (float, default=0.0) – Smoothing factor to apply

Forward

The forward() function takes actual output x and target output and computes the loss. This method is also called from the __call__ method, i.e. when the object is called directly.

LabelSmoothing.forward(x, target, weights=None, attention=None)[source]: Forward data

Reference

[1] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. Rethinking the Inception Architecture for Computer Vision. In Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR). [PDF]