search

Spinozaaa / Federated-Long-tailed-Learning

7 stars 0 forks source link

MultiExpertTrainer without LADEloss #4

Open HongdaChen opened 7 months ago

HongdaChen commented 7 months ago

https://github.com/Spinozaaa/Federated-Long-tailed-Learning/blob/9944e7a55cc709cfd3d65741e49779425101a1ab/fedml_api/clsimb_fedavg/my_model_trainer_classification.py#L76

The parameter list in MultiExpertTrainer.init() method misses the criterion, so the actual criterion in MultiExpertTrainer used is not LADE, but is the "loss" assigned in *.yaml

HongdaChen commented 7 months ago

Also, the "ldae" may be "lade"

HongdaChen commented 7 months ago

"ldae" -> "bsm"