I am training a SegformerForSemanticSegmentation model for three classes. I trained the model using a normal trainer and got some okay results. I want to change the class weights to [0.2, 0.4, 0.4] since I have an unbalanced dataset. I searched using custom weights and found this code: loss_fct = n…

Class weights for Segformer loss function

mpham8 July 28, 2023, 7:41pm 2

I have the same issue

Topic		Replies	Views
Training with class weights 🤗Transformers	5	2994	November 18, 2023
Custom loss weight for train a different weight for validation 🤗Transformers	0	198	April 4, 2024
Maskformer loss , finetuning with weighted loss Models	0	285	January 27, 2024
Class weights in Trainer() instance Intermediate	1	717	September 10, 2021
Why my model doesn't learn anything? 🤗Transformers	0	834	July 29, 2021