Adding data for sustainability
Hello,
I have written a bachelor's thesis on the sustainability of (adversarial) training and inference. As part of my research, I trained your model, among others, multiple times—sometimes with modifications. Additionally, I measured the inference time of many models listed on the RobustBench website using CIFAR-100. If you're interested, I’d be happy to share my thesis and/or the results, though most of the data is currently in German.
I also think it would be beneficial to include resource efficiency data in your GitHub repository. This could cover training and/or inference metrics, such as runtime, power consumption, or a general sustainability label to compare models more easily.
Hi,
An interesting bachelor thesis. Could you send me an english-version by email ([email protected]) if possible?
Best, Jiequan