UnpicklingError: invalid load key, '<'.

RobustBench / robustbench

RobustBench: a standardized adversarial robustness benchmark [NeurIPS 2021 Benchmarks and Datasets Track]

https://robustbench.github.io

Other

669 stars 99 forks source link

UnpicklingError: invalid load key, '<'. #165

Closed senad96 closed 10 months ago

senad96 commented 10 months ago

Cannot download models with the current version

fra31 commented 10 months ago

Hi,

just tried here and it seems to work. Can you provide more context? Is it possible that you still have some older version or corrupted download around?

senad96 commented 10 months ago

hello, thanks for answering :)

Yes, I had a corrupted version of the library.

rishika2110 commented 10 months ago

@senad96 How did you fix it? I am facing the same error now. It doesn't work when I try to run model = load_model(model_name='Carmon2019Unlabeled', dataset='cifar10', threat_model='Linf') but works well when I do model = load_model('Singh2023Revisiting_ViT-B-ConvStem', dataset='imagenet', threat_model='Linf', model_dir='./models'). Update: Degrading timm to 0.6.7 works well.