FLYP
FLYP copied to clipboard
locuslab
Code for Finetune like you pretrain: Improved finetuning of zero-shot vision models
I ran the FLYP code to compare with "Masked Images Are Counterfactual Samples for Robust Fine-tuning, CVPR 2023", using **ViT-B/32 model.** I expect that FLYP can be competitive with other...
I run your code and got a similar result with the paper except ObjectNet. The top-1 acc on ObjectNet is only 54.7%. I found the images in my ObjectNet is...
https://github.com/locuslab/FLYP/blob/61275a47af1cc10aa1398cfce2301e265af6f4c4/src/datasets/common.py#L142 It seems the `noscale` argument is useless in this codebase, and the non-existence of `args.noscale` incurs errors. So it's better to remove all occurrences of `noscale` and `args.noscale` in...
