Hi:
Are there some methods to hack the code to implement the label smoothing for CrossEntropyLoss? Because the target must be torch.LongTensor to hinder the soft target (torch.FloatTensor).
6 Likes