Expand criterion pool to support various types of losses

mlexchange / mlex_dlsia_segmentation_prototype

Other

3 stars 3 forks source link

Expand criterion pool to support various types of losses #30

Open TibbersHao opened 3 weeks ago

TibbersHao commented 3 weeks ago

Currently the build_criterion function in train.py only supports CrossEntropy loss, this needs to be expanded to support additional loss options such as MSE, Binary Entropy etc. As for a temporary solution, frontend will only display CrossEntropy loss for the default option until this feature upgrade is done.