Hang Yao

> i only achieved the 90.4% acc on AIR dataset, have anyone achieved 93%? i used same config as CUB, expect for swap_num=[2, 2]

good work, I achieve perforemance of CUB-200-2011(92.3%), but I only got 93.73% on stanfordcars. Is it any other different configure compared with CUB-200-2011 except for lr?

> In my experiment, lr had the greatest impact on the experimental results. Try `--lr 5e-3 --min-lr 5e-7 --warmup-lr 5e-8 --epochs 300 --warmup-epochs 20`? `python3 -m torch.distributed.launch --nproc_per_node 2 --master_port...

> > > In my experiment, lr had the greatest impact on the experimental results. Try `--lr 5e-3 --min-lr 5e-7 --warmup-lr 5e-8 --epochs 300 --warmup-epochs 20`? > > > >...

have you solved the problem？I also had this problem

> My problem solved. It was the non-proper installation of bitsandbytes package. I did an uninstall and then: pip install https://github.com/acpopescu/bitsandbytes/releases/download/v0.37.2-win.1/bitsandbytes-0.37.2-py3-none-any.whl > > Then the MiniGPT-4 can start without any...

> Check this in cmd `python -m pip install git+https://github.com/TencentARC/GFPGAN.git@8d2447a2d918f8eba5a4a01463fd48e45126a379 --prefer-binary` thank you! it works for me. But i compare this command and the command in the launch, and i...

> In my experiment, resnet50 got 88.23% accuracy, and resnet101 is worse than that. are you modified something in source code except for dataset path. i only achieved 87.9% on...