search

CoinCheung / pytorch-loss

label-smooth, amsoftmax, partial-fc, focal-loss, triplet-loss, lovasz-softmax. Maybe useful
MIT License
2.17k stars 374 forks source link

about smooth label #4

Closed mk123qwe closed 4 years ago

mk123qwe commented 4 years ago

if i choose smooth label ,what loss function do i choose? in ptorch, celoss only support traget.long()

CoinCheung commented 4 years ago

Sorry for replying so late. Maybe you have figured it out yourself, you can use label smooth just like you use cross-entropy-loss of pytorch, only to be sure to add the args of smooth which controls the smooth label.