Is pytorch resnet101 output 7x7 or 8x8?

Hi. In the ResNet paper, the resnet output of a 224x224 image is 7x7, before the average pooling layer. However, PyTorch seems to give it as 8x8. Is there any different approach token? Attached is the image of demo.

It seems your image shape is 254x254, which is probably a typo.
Could you try it again with 224x224?

Oops. Sorry for my silly mistake. Most probably read it in the paper fast.