mask in test_in_any_resolution.py

[IDKiro]

Maybe you can padding the image first, otherwise invalid information will be fused in the large scale token:

def expand2square(timg, factor=128):
	# padding first
	_, _, h, w = timg.size()
	mod_pad_h = (factor - h % factor) % factor
	mod_pad_w = (factor - w % factor) % factor
	timg = F.pad(timg, (0, mod_pad_w, 0, mod_pad_h), 'reflect')

	_, _, h, w = timg.size()
	X = int(math.ceil(max(h,w)/float(factor))*factor)

	img = torch.zeros(1,3,X,X).type_as(timg) 	# 3,h,w
	mask = torch.ones(1,1,X,X).type_as(timg)	# for -inf

	img[:, :, :h,:w] = timg
	mask[:, :, :h,:w].fill_(0.0)
	
	return img, mask