wvangansbeke/revisiting-contrastive-ssl

Revisiting Contrastive Methods for Unsupervised Learning of Visual Representations. [NeurIPS 2021]

revisiting-contrastive-ssl's Issues

How to draw GradCAM without classifier?

Hi,
May I ask how did you get the GradCAM figure without classifier for these unsupervised methods?

Thx,

UserWarning: Argument interpolation should be of type interpolationmode

Brilliant work!

However when I run you command "sh scripts/ours_coco_200ep.sh", It print warning as follows:

......../torchvision\transforms\transforms.py:365: UserWarning: Argument interpolation should be of type InterpolationMode instead of int. Please, use InterpolationMode enum.

It seems that you use int number in the resize function of pytorch transforms and I guess it may not influence the performance , but I fail to locate the bug. So would you mind to give some suggestions?

Besides I use pytorch=1.8.0 with torchvision=0.9.0

Thanks~