Hello,
I have written a bachelor's thesis on the sustainability of (adversarial) training and inference. As part of my research, I trained two models multiple times with various modifications. Additionally, I measured the inference time of many of the models listed on your page using CIFAR-100. If you're interested, I can share my thesis and/or the results with you - though most of the data is currently in German.
I also believe it would be valuable to add a column to your page indicating resource efficiency. This could include training and/or inference metrics, such as runtime, power consumption, or a general sustainability label for the models.
Thanks in advance :D
Hello,
I have written a bachelor's thesis on the sustainability of (adversarial) training and inference. As part of my research, I trained two models multiple times with various modifications. Additionally, I measured the inference time of many of the models listed on your page using CIFAR-100. If you're interested, I can share my thesis and/or the results with you - though most of the data is currently in German.
I also believe it would be valuable to add a column to your page indicating resource efficiency. This could include training and/or inference metrics, such as runtime, power consumption, or a general sustainability label for the models.
Thanks in advance :D