search

GeraldHan / GGE

Code for Greedy Gradient Ensemble for Visual Question Answering （ICCV 2021, Oral）
MIT License
25 stars 2 forks source link

'weight' is undefined for self.debias_loss_fn (with --debias gradient) #7

Closed nihirv closed 1 year ago

nihirv commented 2 years ago

Hi

If I run the code with the following command: python main.py --dataset cpv2 --mode gge_iter --debias gradient --topq 1 --topv -1 --qvp 5 --output []

(i.e. --mode gge_iter and --debias gradient), then I get an error on line 81 of base_model.py: loss = self.debias_loss_fn(None, logits, ref_logits, labels, weight)

TypeError: forward() takes 5 positional arguments but 6 were given

To fix this, I've added weight as a parameter of GreedyGradient in vqa_debias_loss_functions.py, and multiplied the output of the BCE loss by weight: loss = F.binary_cross_entropy_with_logits(logits, y_gradient) * weight

I couldn't find your weighting/scaling factor in the paper, so please let me know if that's correct

GeraldHan commented 2 years ago

Sorry for the bug. The weight parameter is only for improved version GGD (with --debias decompose and base_model_sfce). You can just fix this bug by deleting weight in base_model.py as our new update or adding weight=1 as default. It has nothing to do with loss = F.binary_cross_entropy_with_logits(logits, y_gradient)