mzweilin/robust-imagenet-models

All credits go to https://huggingface.co/madrylab/robust-imagenet-models

I simply extracted the model weights from their checkpoints, so they are loadable in PyTorch.

Verify the weights

You can verify that the files contain model weights that are identical to the checkpoints. The verification process will be slow, because you need to download 33GB of the original checkpoints.

python -m venv .venv
source .venv/bin/activate
pip install huggingface-hub torch torchvision tqdm
python verify.py

Usage

import torch
from torchvision.models import resnet18

model = resnet18()
model.load_state_dict(torch.load("resnet18_l2_eps1.pth", weights_only=True))