Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Do you have saved models available? #2

Open
admk opened this issue Nov 5, 2020 · 2 comments
Open

Do you have saved models available? #2

admk opened this issue Nov 5, 2020 · 2 comments

Comments

@admk
Copy link

admk commented Nov 5, 2020

Could you provide us with saved models?

Thanks!

@htwang14
Copy link
Collaborator

htwang14 commented Nov 5, 2020

Our pretrained ATMC 8bit models are available at https://drive.google.com/drive/folders/1HDGkz8BKQGn6SFRPzaTt13hFOC8JbtUD?usp=sharing

The files are names as quant_closest_<dataset>_atmc_<cr>.pth. For example, the file quant_closest_svhn_atmc_0.005.pth is the ATMC 8-bit model on SVHN dataset with 0.005 compression ratio (0.5% non-zero model parameters).

You can evaluated them using {dataset}/inference_model.py.
For example, use the following command to evaluate robustness against PGD attacks with eps=4 (controlled by attack_eps parameter) and iter=7 (controlled by attack_iter parameter) on SVHN dataset. (Set loaddir and data_root as proper file paths on your own machine.)

python svhn/inference_model.py --gpu 0 --loaddir /ckpt/root/path/svhn/atmc_8 --model_name svhn_atmc_0.1.pth --prefix_name quant_closest_ --data_root /data/root/path/svhn --attack_algo pgd --attack_eps 4 --attack_iter 7 --prune_algo l0proj --prune_ratio 0.1 --quantize_algo kmeans_nnz_fixed_0_center --quantize_bits 8 --exp_logger svhn_atmc_8.txt --abc_special --abc_initialize

@admk
Copy link
Author

admk commented Nov 6, 2020

Sorry my question wasn't specific enough, do you have CIFAR-10 & MNIST models? Thank you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants